1e . IS :h h.t )f Lll iy ~d ~. l/6!~ INVESTIGATIONS OF SURGE-TYPE GLACIERS IN Sv ALBARD Gordon Stuart Hamilton A dissertation submitted for the degree of Doctor of Philosophy in the University of Cambridge Scott Polar Research Institute Cambridge May 1992 DECLARATION The analysis and interpretation of data presented in this thesis is the result of my own work unless otherwise stated. I declare that this dissertation does not exceed the agreed regulations on length and that it has not been submitted for a degree at any other university. ii INVESTIGATIONS OF SURGE-TYPE GLACIERS IN SVALBARD by Gordon Stuart Hamilton -The factors affecting the distribution of surge-type glaciers and the causes of the surge mechanism are not fully understood. Statistical analyses of glaciers in Svalbard have been used to identify characteristics which are related to surging. Field experiments were undertaken on Bjuvbreen, a small surge-type glacier in central Spitsbergen, to determine the dynamics and hydrology of the glacier during its quiescent phase. The probability that a glacier in the Svalbard sample population was surge-type was 36·4%. The probability varied spatially within the sample area. Glaciers resting on sedimentary tocks had a greater probability of being surge-type compared to those overlying igneous or metamorphic rocks. The probability of surging was also increased in glaciers with a two-layered thermal structure as indicated by radio echo sounding. Geometrical characteristics such as slope, orientation, curvature and the presence of tributaries were not related to surge-type behaviour, although long glaciers had a greater chance of being surge-type. Certain aspects of Kamb's and Fowler's theories of surging were not supported by the statistical analysis. Despite the associations between surging and certain characteristics, no single factor fully explained the observed geographical distribution of surge-type glaciers in Svalbard. Bjuvbreen is a small surge-type in its quiescent phase. Changes in the geometry of the glacier are occurring relatively slowly. This slow rate of change is a function the low accumulation rates on Svalbard glaciers. On the basis of a simple model, the next surge of Bjuvbreen is predicted to occur between .2022- 2055, indicating a comparatively long quiescent period of -90-130 years. Bjuvbreen is comprised of two dynamically distinct zones which are separated by a large bulge. The lower portion of the glacier is inactive and stagnating, in contrast to the active ice up.:.glacier from the bulge. The observed velocity of the active region was compared with various hydrological characteristics of the glacier. The inferred behaviour of water within the glacier seems to have some, although limited, influence on the ice motion. A spatially restricted drainage system is the probable reason for this limited influence of hydrology on glacier velocity. iii ACKNOWLEDGEMENTS I wish to express my thanks to the many individuals and organisations who have assisted me at various stages in the preparation of this thesis. Firstly, I would like to thank my parents who, with inestimable patience, have supported me at every stage of my education. Without their c0nsta~t encouragement, I might have spent the last three years in a less rewarding occupation. The work reported in this thesis was carried out under the supervision of Julian Dowdeswell. I wish to thank him for the opportunity to study the glaciology of Svalbard, and for reading and commenting on the work presented here. The Natural Environment Research Council provided me with a studentship. Fieldwork was supported financially by NERC, the Royal Society, the Royal Geographical Society, the Gino Watkins Fund and Norsk Polarinstitutt. Visits to Oslo were made possible by awards from NATO and Jesus College. Mike Dancer, Adrian Fox, Malcolm MacIntyre and Fiona Sorensen helped me in the field and brightened up the nightlife in Honk City. Julian Dowdeswell and Arne Sretrang also came to see the best bulge in Svalbard. I am particularly grateful to Arne for his help with the drilling and the radio echo sounding. Judith Maizels of the Department of Geography, University of Aberdeen, very kindly loaned me a pressure transducer and data logger for the 1989 field season and Ric Gard of the same department went to considerable lengths to make sure it would work. I am especially grateful to my great bunch of friends for sharing their friendly arguments, beer, accommodation, mushroom stuffing and humour with me during the last few years. In particular, I would like to thank Tavi Murray, Marianne Cromack, David Sexton, Glyn Hyett, Martin Siegert, Particia Gilmour and Alison and Adrian Fox, among many others. Finally, I am enormously grateful to Fiona Sorensen for all the fun, love and inspiration she has given me in our time together. Her enthusiasm to keep going when things often weren't looking too good helped me see this through to the end. And Fudge and Griffin graciously allowed me to stay at my computer when they would have preferred I played games instead! iv CONTENTS Title ........................................................................... i Declaration ................................................................... ii Summary ..................................................................... 111 Acknowledgements ............................................................ iv Contents ......................................... a •••••••••••••••••••••••••••• V List of Figures ............................................................... ix List of Tables ................................................................ xii List of Symbols .............................................................. xiii 1 Introduction: surge-type glaciers 1. 1 Introduction ....... .... .............................................................. 1 1. 1.1 Introduction and aims of the thesis ................................... 1 1.1.2 Definition and characteristics of surge-type glaciers ................ 2 1.1. 3 Distribution of surge-type glaciers .................................... 3 1.1.4 A short history of investigations ...................................... 5 1.2 A review of glacier motion research .............................. .. ............. 7 1.2.1 The flow and deformation of glaciers ................................ 7 1. 2. 2 The influence of water at the glacier bed ............................. 8 1.2.3 The nature of the ice-bed interface and substrate ................... 10 1. 3 Mechanisms of glacier surging ................................................... 11 1.3.1 Introduction .............................................................. 11 1.3.2 Hard bed mechanisms ........................ .......................... 16 1.3.3 Soft bed mechanisms .................... .. ............................. 21 1.4 Other types of fast glacier flow ................................................... 24 1.4.1 Ice streams and fast flowing outlet glaciers .......................... 24 1.4.2 Tidewater glaciers ....................................................... 26 1.5 Summary and structure of the thesis ............................................. 28 1.5.1 Summary ............................................... ..... ............. 28 1.5.2 Structure of the thesis .................................................. 29 2 The glaciology of Svalbard 2.1 Introduction ......................................................................... 30 2.2 The physical environment of Svalbard ..................... ........... ..... ..... 30 2.2.1 Climate ........................................ .. ......................... 30 2.2.2 Geology ................................................. .. ............... 33 2.2.3 Quaternary glacial history ............. , ................................. 34 2.3 Contemporary glaciology of Svalbard ........................................... 36 2. 3.1 A short history of glaciological exploration .......................... 36 2.3.2 Distribution and classification of Svalbard ice masses . ............ 37 2.3.3 Mass balance ............................................................. 39 2.J.4 Thermal regime .......................................................... 41 2.4 Surge-type glaciers in Svalbard ................................................. .43 2.4.1 Introduction ............................................... ........ ...... . 43 2.4.2 Observations of surge behaviour. .................................... .43 2.4.3 The long duration of the active phase on surge-type glaciers in Svalbard .................... ................................ .............. 49 2.5 Summary ...................... ... ................................................... 50 V 3 Environmental controls on glacier surging: a statistical analysis of Svalbard glaciers 3 .1 Introduction .......... .......... .. .................. ..... ..... ..... ... ..... ...... .. , .. 51 3 .1.1 Summary of previous .studies ......................................... 51 3. 2 Sources of data ...................................................................... 54 3.2.1 Introduction ..................... ................. .............. ....... ... 54 3.2.2 Selection of sample population ........................................ 55 3.2.3 Acquisition of data ...................................................... 57 3.2.4 The surge index ......................................................... 58 3.3 The primary data set and geographical analysis ................................ 60 3.3.1 Surge probability for the primary data set ............................ 60 3.3.2 Geographical analysis .................................................. 61 3.4 Morphometric and topographic controls ......................................... 64 3.4.1 Length analysis .......................................................... 64 3.4.2 Does glacier length influence the geographical distribution of ... . surge-type glaciers? ............................................... ...... 67 3.4.3 Influence of tributaries ....... ... ................... ... ..... ..... ....... 72 3.4.4 Influence of elevation and slope ....................................... 74 3.4.5 Influence of glacier orientation ........................................ 79 3.4.6 Glacier curvature and surge probability .............................. 83 3. 5 Geological controls on glacier surging .......................................... 85 3. 5 .1 General lithological classification and surge probability ........... 86 3.5.2 Surge probabilities associated with individual rock types .......... 89 3.6 Other environmental controls ......... . ....... ... .... ..... .. .. ........ ............ 95 3.6.1 Introduction .............................................................. 95 3.6.2 Internal reflecting horizons, glacier thermal structure and surge probability ............................................................... 95 3.6.3 Slope data and Kamb's stability parameter .......................... 99 3.6.4 A statistical test of Fowler's surging criterion ............... ... ..... 100 3.7 Discussion and comclusions ...................................................... 102 3. 7 .1 Discussion .................................. . ...... ...... .. .............. 102 3. 7 .2 Conclusions .................. ... ..... .. ............................. . ... 107 4. Field area and methods 4.1 Introduction ......................................................................... 108 4. 2 The field area .............. . ............................ . ...... . .. .................. 109 4.2.1 Kjellstromdalen ......................................................... 109 4.2.2 Bjuvbreen ................... .... ........... . ......................... .. .. 111 4.~.3 The programme of field investigations ............................... 114 4.3 Radio echo sounding ... .... .. .. .. .... .... ... . . . ........ .... ... . ...... ... .. ..... .. . 115 4.3.1 Introduction ................ .... ................. ...... ..... . .... . ........ 115 4.3.2 System specification and design ...................................... 116 4.3.3 Geographic location of radio echo soundings on .. .. .............. . Bjuvbreen ... .. .... ... .................. . .... . ................... . ........ 116 4. 3 .4 Depth of firn for correction of echo sounding data . . ............... 117 4.4 Ground survey network and glacier topography .. .. ............ . ...... . ... . ... 119 4.4.1 Introduction ... ... ........ ... ................... . .. .... ........ ........ .. .. 119 4.4.2 Survey control and local triangulation betwork .................. . .. 119 4.4.3 Glacier topography and longitudinal profile ......................... 122 4.5 Glacier motion and deformation measurements ...... . .. . .... . .. .. . .. ... .... .. . 123 4.5.1 Introduction ..... ......... .... .. .. .. .. .... ...... . .. .... .. .. . ... .. .. ..... .. 123 4.5.2 Field methodology ............................... .......... ..... . . . ..... 123 4.5.3 Local coordinate scheme .... ........... . ... . . .. ............ . ... . . .... . . 126 4.5.4 Reduction of data. 1: velocity measurements . . .. .. .. .. . .... .. ... . ... 126 4.5.5 Reduction of data. 2: strain rates .... ................... ............... 126 vi 4. 5. 6 Calculation of survey errors ........................................... 128 4.6 Glacier hydrological investigations .............................................. 131 4.6.1 Introduction .............................................................. 131 4.6.2 Proglacial stream discharge monitoring ...... ... ... . ....... .......... 131 4.6.3 Proglacial stream suspended sediment concentrations .............. 134 4.6.4 Hot water drilling ............................... . ....................... 136 4.6.5 Meteorological observations ........................................... 139 5 The geometry and evolution of Bjuvbreen 5 .1 Introduction ......................................................................... 140 5 .1.1 Geometric evolution of B juvbreen: data sources ............... . .... 141 5.1.2 Aims of this chapter ............... .. ... . .... .. ... ....... . .............. 142 5.2 Surface and bed topography ............... . ........................... . .. . ....... 143 5.2.1 Long profile changes ...................... . ............................ 143 5.2.2 Bed topography ......................................................... 146 5.3 Basal shear stresses ................................................................ 147 5.3.1 Introduction . ............................................................. 147 5.3.2 Cross-sectional shape factors .......................................... 147 5. 3. 3 Calculation of basal shear stresses .................................... 148 5. 3 .4 Longitudinal averaging of basal shear stess .............. .. ..... .. .. 151 5.3.5 Basal normal stress ............ ... ............................. . .... .. .. 156 5.4 Glacier volume change ..................... . ...................................... 156 5.4.1 Introduction .............................................................. 156 5.4.2 Calculation of glacier volume .......... ... ............................. 157 5.4.3 Volume change and a simple predictive model ...................... 159 5.4.4 Discussion .................................................. . ............ 162 5.5 Thermal regime ....................................... ... ........ . .................. 163 5.5.1 Introduction .............................................................. 163 5.5.2 The moving-column model ........................................... . 163 5. 5. 3 Thermal regime of B juvbreen ......................................... 164 5. 5. 4 Discussion ............................................................... 166 5.6 Summary ........................................ ... ... .............................. 167 6 The flow and deformation of Bjuvbreen 6 .1 Introduction ......................................................................... 168 6.2 Horizontal velocity .. .... ....... .................................................... 169 6.2.1 Introduction ................................................ ........ ...... 169 6.2.2 Seasonal velocity pattern ............................................... 170 6.2.3 Short period velocity behaviour ....................................... 174 6. 2. 4 Discussion of velocity behaviour of B juvbreen ..................... 177 6.3 Vertical velocities at the glacier surface ........ ......... .... ......... .... . .... .. . 181 6.3-.1 Seasonal vertical velocities ............ .. ............................... 181 6.3.2 Short period fluctuations in vertical velocity ......................... 183 6. 4 Surface strain rate ................................. ... .............................. 187 6.4.1 Introduction ........ ...................................................... 187 6.4.2 Strain rates 1989 ..................... : .................................. 187 6.4.3 Strain rates 1990 ......................... . .... .. ........................ 189 6.4.4 Comparison of strain rates with crevasse distribution .............. 191 6.5 Basal sliding and internal deformation ofBjuvbreen .......................... 194 6.5.1 Background ............................................................... 194 6.5.2 Calculation of the basal sliding velocity .............................. 194 6.5.3 The contribution of basal sliding to the total velocity of .......... . Bjuvbreen ................................................................ 195 6.6 Balance velocity and flux ........... ...... ..... .... ... ............................. 199 6.6.1 Introduction ............................................ ..... .... .... ..... 199 6. 6.2 Calculation of balance velocity ........................................ 199 vii 6.6.3 Comparison between balance velocity and actual velocity ......... 200 6. 7 Summary .. ..... .. .... ............. ...... ........... ... .. ............ .... ............ 202 7 The hydrology of Bjuvbreen 7 .1 Introduction .. ...................................................................... 204 7 .1.1 Background ........................ ...................................... 204 7.1.2 Fieldwork programme .................................................. 205 7 .2 Meltwater discharge and suspended sediment dynamics .... . .......... ... . ... 205 7 .2.1 Discharge rating curves ... . .............................. .. ..... .. .... . 205 7 .2.2 Characteristics of the 1989 discharge hydrograph ..... .. ........... 206 7 .2.3 Suspended sediment variations 1989 . . . .. ...... . . ... . . ... . ........... 212 7.2.4 Suspended sediment variations 1990 ........ . ....... . .. .... .......... 214 7.2.5 Relationshi2 between discharge and water quality ................. 215 7 .2.6 Discussion . .. ..... -............................... . ... .................... 223 7 .3 Water behaviour within Bjuvbreen ............ ... ... ....... ...... ............... . 224 7.3.1 Introduction .. ...... ... ........... ... . . ............... .. .................. 224 7 .3.2 Hot water drilling on Bjuvbreen ........... ... .......... . ............. 225 7.3.3 Water pressures beneath Bjuvbreen .......................... . ..... .. 231 7.4 Relationship between glacier motion and hydrology .......................... 234 7.4.1 Introduction .............................................................. 234 7.4.2 Steady-state water pressures and seasonal velocities . .... .......... 234 7.4.3 The infuence of glacier hydrology on the short-term velocity ... . behaviour of Bjuvbreen . ..... ........... ..... ...................... . ... 239 7 .4.4 Discussion ............................................................... 245 7.5 Discussicn and summary .... ...................................................... 246 7. 5 .1 The hydrology of B juvbreen compared with other glaciers ........ 246 7.5.2 Summary ................................................................. 248 8 Summary and further work 8.1 Introduction ......................................................................... 250 8.2 Summary of significant results .. . ................ .. .. ... ... ...................... 250 8. 2.1 The significance of surge-type behaviour on Svalbard glaciers ... 250 8.2.2 Factors influencing the occurrence of surge-type glaciers in .. ... . Svalbard ....... .......... .. ....... ................ ................ . ...... . 251 8.2.3 The dynamics of Bjuvbreen ...... . ....... .. ........ .... .. . ...... ...... 253 8.3 Suggestions for further work ............................. ..... ................... 257 References ....................................................................................... 259 Appendix A - ................................. .......... ......................................... .. . Al Appendix B ....................................................................... ......... ... .... A8 viii - LIST OF FIGURES Figure 1.1. Some characteristics of surge-type glaciers............................. 4 Figure 1.2. Relationship between surface velocity and meteorological conditions on Saskatchewan Glacier............................................. .... 9 Figure 1.3. Budd's diagram showing glacier flux rate versus ice the product of ice surface slope, velocity and thickness . . . . .. . . . . . .. . .. . . . . . . . . . . .. . . . 13 Figure 1.4 Linked cavity basal hydrological system................................ 18 Figure 2.1 Map of Svalbard . . . .... . .. . . . . . . .. .. . . . .. . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . 31 Figure 2.2 Monthly mean temperature for Longyearbyen and Sveagruva...................................................................... . . . . . . . . . 32 Figure 2.3 Map of Svalbard showing the glacierized areas........................ 38 Figure 2.4 Glacier equilibrium line altitudes on Svalbard................. ......... 40 Figure 2.5 Locations of observed glacier surges on Svalbard..................... 48 Figure 3.1 Location of sample areas.................................................. 56 Figure 3.2 Geographic variation in the concentrations of surge-type glaciers.................... ..... .......................................................... 63 Figure 3.3 Distribution of lacier lengths for the Svalbard sample population . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Figure 3.4 Influence of glacier length on the probability of surging..... . . . . . . . . . 66 Figure 3.5 Distribution of glacier lengths for individual map sheets.............. 68 Figure 3.6. Geographic variation of length-predicted surge probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 Figure 3.7 Geographic variation of surge probabilities with the length influence removed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 Figure 3.8 Median maximum and median minimum elevations for glaciers in each length bin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 6 Figure 3.9 Median slope for glaciers in each length bin ................... . . . . . . . . . 77 Figure 3.10a. Influence of glacier slope on the probability of surging for glaciers in subset}+....... . ........................................................ 78 Figure 3.10b Influence of glacier slope on the probability of surging for glaciers in subset}-......... ..................... .................................. 78 Figure 3.lla Probability that the upper zones of glaciers have a certain compass orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 Figure 3.llb Probability that the lower zones of glaciers have a certain compass orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 Figure 3.12a Surge probabilities by compass orientation for the upper zones of glaciers... .............. ..... .................................................. 82 Figure 3.12b Surge probabilities by compass orientation for the lower zones of glaciers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 Figure 3.13 Variation of surge probability with degree of glacier curvature...... ....... .. ......... .... ..................................................... 84 Figure 3.14 Geographic variation of geology-predicted surge probabilities................................................................... .......... 88 Figure 3.15 Geographic variation of surge probabilties with the geology influence removed..................................... .. ............... .... .. 90 Figure 3.16 Geographic vruiation of surge probabilities predicted by the revised geological classification scheme.... ..................................... 93 Figure 3.17 Geographic variation of surge probabilities with the revised geological influence removed................................................ 94 Figure 3.18 Locations of glaciers in Svalbard where internal reflecting horizons have been recorded.......................................................... 97 Figure 3.19 Variation of surge probability with Fowler's surge criterion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 ix Figure 4.1 Location of the field area in central Spitsbergen... ... .... ... .... ..... ... 110 Figure 4.2a Aerial photograph of Bjuvbreen taken in 1977 . ..... .. .... . ... .. . . . . . . 112 Figure 4.2b Photograph of the Bjuvbreen showing the bulge...................... 113 Figure 4.3 Location of radio echo sounding points............................... .. . 118 Figure 4.4 Locations of survey stations and control points......................... 120 Figure 4.5a Spatial coverage of velocity and strain markers during the 1989 field season....................................................................... 124 Figure 4.5b Spatial coverage of velocity and strain markers during the 1990 field season....................................................................... 125 Figure 4.6 Photograph of pro glacial stream........................................... 133 Figure 4.7 Location of hot water driling sites......................................... 138 Figure 5.1 Ice surface profi1es for 1936, 1977, 1986 and 1990, and bed profile. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 Figure 5.2 Curve of Nye's shape factor.... .. .. .... .. .. .. .... .. .. ..................... 148 Figure 5.3 Variation of basal shear stress with distance on B juvbreen calculated using equation 5.1 .... .. .... .. .... ...... .. .. .. .. .. .... .. .... ......... .... .. 151 Figure 5.4 Variation of basal shear stress with distance on Bjuvbreen calculated using equation 5.2, and change through time........................... 153 Figure 5.5 Variation on basal normal stess with distance on Bjuvbreen........... 157 Figure 6.la Seasonal velocity vectors on Bjuvbreen during the 1989 field season.............................................................................. 171 Figure 6.lb Seasonal velocity vectors on Bjuvbreen during the 1990 field season.............................................................................. 171 Figure 6.2 Longitudinal variation of seasonal velocities for 1989 and 1990................................... . ................ .. .... . .......... . ........... . .... 172 Figure 6.3a Short term horizontal velocities 1989 .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 17 5 Figure 6.3b Short term horizontal velocities 1990........ .. ......................... 178 Figure 6.4 Relationship between seasonal surf ace velocities and shear stress..................................................................................... 179 Figure 6.5a Vertical components of the seasonal surface velocity vectors 1989 .. .. . . .. . . .. . .. . . .. . .. . . . . . . . . . .. . . . . . . . . . . . . . . . . .. . . .. . .. . . .. . . . . . . . . . .. . .. 182 Figure 6.5b Vertical components of the seasonal surface velocity vectors 1990 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182 Figure 6.6a Short term vertical velocities 1989 .... .. .... .. .......... .... ... .. .... .. .. 184 Figure 6.6b Short term vertical velocities 1990.... .... ...... .. ..... .... .. .. .. .... ... 186 Figure 6.7a Strain rates 1989 t1 ...... .. .. ...... ...... ........ .... ........ .... .... .... .. 188 Figure 6.7a continued Strain rates 1989 t2 .. .... .. .... ...... .. .... .... .. .... ....... 188 Figure 6.7b Strain rates 1990.... .... .. .. ...... .. .. .. ................................... 190 Figure 6.8a Pattern of crevasses 1989........................ .. .... .. .... .. .. .. .. ..... 193 Figure 6.8b Pattern of crevasses 1990 .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 193 Figure 6.9a Proportion of basal motion as a percentage of the seasonal total 1989..................... .......... . ............. . . ... . .. .. . .................. . .. ... 196 Figure 6.9b Proportion of basal motion as a percentage of the seasonal total 1990.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 Figure 6.10 Comparison between balance velocity and actual velocity ...... . ... . 202 Figure 7.1 Stage-discharge relationships 1989 .. . :...... ...... .... .... .... .... .. .... 207 Figure 7 .2 Hourly discharge hydrograph 1989 .. .. .. .. .. .. . .. . .. .. .. .. .. .. . .. . . .. .. . 208 Figure 7.3 Stacked discharges.. ............ .... ......................... .. .... .... ..... 210 Figure 7.4 Meteorological conditions 1989 .............. ........ .. .. .. ............... 211 Figure 7.5a Suspended sediment concentrations 1989.. .... ........... .. .. .. .... .. . 213 Figure 7.5b Suspended sediment concentrations 1990.................. .. .......... 213 Figure 7.6 Stacked suspended sediment concentrations 1989 ........ .. ............ 214 Figure 7.7 Discharge- suspended sediment power law relationship 1989.................................... .. ................... .... .. .... . .. .... .. ......... . 216 Figure 7.8 Combined record of discharge and suspended sediment concentrations 1989...... ... . ................... . ..... . ................. . . ........... . . 217 Figure 7.9a Hysteresis rating loop for normal flow conditions ...... .. ............ 220 Figure 7.9b Hysteresis rating loop for storm runoff conditions ................... 220 X Figure 7.10 Running percentages of total discharge and suspended sediment transport 1989. .. . . .. . . . . .. .. .. . . . .. . . . . . . .. . ... . . . . . ... .. . . .. ....... .. ... . .. 222 Figure 7.11 Map of Bjuvbreen showing hydraulically-inactive bed areas. ....... .. ..... ... .... .. .. .. ... . .. .. . ..... ... .. ... .. .. ...... .. ... ...... ... ..... .... ... 229 Figure 7.12 Theoretical hydraulic grade line on Bjuvbreen... . . .. .. ... . ... . .. . . . . . . 233 Figure 7.13 Longitudinal profile of effective normal stress..... ... ...... ... ........ 235 Figure 7.14 Longitudinal profile of bed separation index............. . ........... .. 236 Figure 7 .15 Combined record of short term velocities, hydrological data and meteorological conditions 1989 . ... .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . .. . . . . . . . . 240 Figure 7.16 Combined record of short term velocities, hydrological data and meteorological conditions 1990.... .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . .. . . . . . . . . 243 xi LIST OFT ABLES Table 3.1 Surge index distribution and probability scheme........... . ... .......... 61 Table 3.2 Surge probability statistics arranged by map sheet....................... 62 Table 3.3 Length predicted surge probabilities arranged by map sheet . . . . . . . . . . . 70 Table 3.4 Various measures of surge probability arranged by map sheet . . . . . . . . . 72 Table 3.5 Surge index distribution and probability scheme for subsets NT and T ................................................................................ 73 Table 3.6 Various measures of surge probability for subset NT.................... 75 Table 3.7 Influence of glacier orientation on surge probability............... ...... 83 Table 3.8 Glacier curvature and surge probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 Table 3.9 Surge probability statistics for petrological categories . . . . . . . . . . . . . . . . . . . 86 Table 3.10 Surge probability statistics for petrological categories.................. 87 Table 3.11 Surge probability statistics for the revised geological classification scheme................................................................... 89 Table 3.12 Surge probability statistics for the revised geological classification scheme arranged by map sheet........................................ 91 Table 3.13 Surge probability statistics for the Soviet sample population.......... 96 Table 3.14 Surge probability statistics for Fowler's surge criterion............... 102 Table 4.1 Reference control points..................................................... 121 Table 4.2 Total survey errors for velocity and strain targets . ....................... 130 Table 5.1 Cross-section shape factor data.. ......................................... .. 148 Table 5.2 Results of shear stress calculations......................................... 149 Table 5.3 Input parameters for the moving column model.......................... 165 Table 6.1 Comparison of velocity data for the 1989 and 1990 field seasons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4 Table 6.2 Strain rate measurement intervals.. ..... .................................... 189 xii LIST OF SYMBOLS A flow law parameter Ac cross-section area As glacier surface area - B accumulation rate bn net mass balance C longitudinal coupling length F shape factor f* flux shape factor g gravitational constant H latent heat of melting h ice thickness I Dawson's integral K thermal difusivity contant for ice L glacier length Le distance between orifices Lo length of orifices /2 constant in moving-column model M Manning roughness n flow law constant Neff effective normal stress Pw water pressure Q water discharge Qt mean water discharge at time t Qb(X) balance flux at point x R material constant in Rothlisberger equation r roughness coefficient s bed separation index s orifice step height u mean velocity through ice column u total ice surface velocity Uci internal deformation velocity UB basal velocity Us observed centreline velocity u average cross-sectional velocity xiii vb balance velocity w rate of warming/cooling of ice column w glacier width a ice surface slope f3 bed slope .1 water film thickness Tl ice viscosity tJ network tortuosity A head concentration factor ;L air temperature lapse rate - Kamb' s stability parameter M ...., Pi density of ice Pw density of water O"B basal normal stress 'X'B basal shear stress rs local slope stress (J) Fowler's surging criterion xiv CHAPTER 1 INTRODUCTION: SUR.GE-TYPE GLACIERS 1.1 INTRODUCTION 1.1.1 Introduction and aims of the thesis Glacier surges are a well-documented type of fast glacier flow which occur periodically on certain glaciers at regular intervals. Sharp (1988) estimated that, worldwide, some four per cent of contemporary glaciers are surge-type. However, the geographical distribution of surge-type glaciers is distinctly non-random, suggesting that certain environmental controls are required for surges to take place (Raymond, 1987). Despite a considerable amount of research, both the controls on surging and the details of the trigger mechanism are inadequately understood. A number of large polar ice streams are in a quasi-permanent state of fast flow analogous to the active phase of surge-type glaciers. These ice streams play a fundamental role in the stability of polar ice sheets. Current scientific concern on the likely response of these ice sheets to climate change has prompted glaciologists to investigate what might cause ice streams to switch between fast and slow flow. However, the timescales in which these processes operate and the enormous geographical scales involved, make detailed study of ice streams difficult and costly. In contrast, surges occur on glaciers generally a tenth, or less, of the size of ice streams and over comparatively shorter periods of time. Therefore, surge-type glaciers are potentially useful analogues for the study of polar ice streams. The aim of this thesis is to investigate the phenomenon of surging on glaciers in the high Arctic archipelago of Svalbard. Within this broad objective there are two, more specific aims. The first of these aims is to identify the controls on the spatial distribution of surge-type glaciers in Svalbard. This ·objective is pursued using statistical analysis. The second aim is to study the dynamics of a small surge-type glacier in central Svalbard as it progresses through its quiescent phase. This study is based primarily on work carried out during two field seasons. CHAPTER 1 INTRODUCTION: SURGE-TYPE GLACIERS 1.1 INTRODUCTION 1.1.1 Introduction and aims of the thesis Glacier surges are a well-documented type of fast glacier flow which occur periodically on certain glaciers at regular intervals. Sharp (1988) estimated that, worldwide, some four per cent of contemporary glaciers are surge-type. However, the geographical distribution of surge-type glaciers is distinctly non-random, suggesting that certain environmental controls are required for surges to take place (Raymond, 1987). Despite a considerable amount of research, both the controls on surging and the details of the trigger mechanism are inadequately understood. A number of large polar ice streams are in a quasi-permanent state of fast flow analogous to the active phase of surge-type glaciers. These ice streams play a fundamental role in the stability of polar ice sheets. Current scientific concern on the likely response of these ice sheets to climate change has prompted glaciologists to investigate what might cause ice streams to switch between fast and slow flow. However, the timescales in which these processes operate and the enormous geographical scales involved, make detailed study of ice streams difficult and costly. In contrast, surges occur on glaciers generally a tenth, or less, of the size of ice streams and over comparatively shorter periods of time. Therefore, surge-type glaciers are potentially useful analogues for the study of polar ice streams. The aim of this thesis is to investigate the phenomenon of surging on glaciers in the high Arctic archipelago of Svalbard. Within this broad objective there are two, more specific aims. The first of these aims is to identify the controls on the spatial distribution of surge-type glaciers in Svalbard. This ,objective is pursued using statistical analysis. The second aim is to study the dynamics of a small surge-type glacier in central Svalbard as it progresses through its quiescent phase. This study is based primarily on work carried out during two field seasons. Chapter 1: Surge-type glaciers 1.1.2 Definition and characteristics of surge-type glaciers Surge-type glaciers experience extreme flow pulsations. They are characterised by a two-phase cyclical flow instability (Meier and Post, 1969) involving a period of accelerated velocities (surge) followed by a period of stagnation and build-up of mass (quiescence). The active surge phase represents only a relatively short portion of the total cycle, and usually lasts between one and ten years. Surges occur at regular intervals punctuated by much longer periods of quiescence, which may last from 15 to 100 years or longer (Meier and Po-st, 1969). Glaciers that surge do so repeatedly and at regular intervals, strengthening the hypothesis that the controls on surge behaviour are inherent within the glacier itself and are not climate-related. It is important to distinguish between the use of the terms 'surge-type' and 'surging', since their indiscriminate application may lead to unnecessary ambiguities. 'Surge-type' is a non-specific term which can be used to describe a characteristic glacier at any stage of its surge cycle. The term 'surging' has a more distinct meaning and should only be applied when describing a glacier in the active phase of its cycle. This convention will be used throughout this thesis. A number of recognizable characteristics are common to most surge-type glaciers. During the active phase these include the rapid advance of the snout, widespread and chaotic crevassing of the surf ace, large increases in ice velocities and flux, marked vertical displacements of the ice surface, and the release of large quantities of turbid meltwater at the glacier margins (Post, 1960). These events result in the fonnation of distinctive ice structures and the contortion of medial moraines. The quiescent phase is characterised by the virtual stagnation of considerable areas of ice. Because of this, surges rarely result in any long-term net advance of the glacier terminus. During the active phase of the cycle, ice flow velocities are much enhanced. Speeds of up to 5 m h-1 or 6 km a-1 may be reached (Meier and Post, 1969) although the mean surge velocities are usually lower. Kamb et al. (1985) recorded an average surge propagation speed of 23 m d-1 during the initial months of the 1982-83 surge of Variegated Glacier, Alaska. There is, however, a consider,able variation in propagation speed between different surge-type glaciers. For example, Drewry et al. (unpublished) reported substantially slower velocities of 4 m d-1 during the recent surge of Bakaninbreen in Svalbard. Surge-type glaciers are often described, broadly, as being composed of two distinct zones: an ice reservoir and an ice receiving area. These zones need not necessarily correspond with the normal accumulation and ablation areas. During the quiescent phase there is a build-up of mass in the reservoir zone accompanied by thinning and stagnation in the receiving zone (Paterson, 1981 ). This causes a Pagel Chapter 1: Surge-type glaciers steepening of the glacier long profile (Meier and Post, 1969). A surge results in the rapid transfer of a large volume of ice from the reservoir to the receiving area and a subsequent reduction in the surface slope (Paterson, 1981). Some of the characteristics of surge-type glaciers are illustrated in Figure 1.1. 1.1. 3 Distribution of surge-type glaciers The geographical distribution of surge-type glaciers appears to be highly specific (Raymond, 1987; Sharp,_ 1988). Surge-type glaciers have been observed in many mountainous and glacierized regions. These regions include the Wrangell- St. Elias Range in the Yukon and Alaska (Post, 1969), the Pamirs of central Asia (Dolgushin and Osipova, 1975), the Andes (Lliboutry, 1958), the Canadian Arctic archipelago (Hattersley-Smith, 1964), Greenland (Rutishauser, 1971), Iceland (Thorarinsson, 1969) and Svalbard (Liest0l, 1969; Dowdeswell et al., 1991). However, there are few, if any, reports of surge-type glaciers in the Alps, Scandinavia, the Rocky Mountains or New Zealand. The factors that control the occurrence of surge-type glaciers are unknown, but they are believed to be inherent in the glacier system. Climate is generally discounted as a possibility since adjacent glaciers within the same basin may experience both surge- type and normal behaviour. Clarke et al. (1986) statistically analysed a population of 2356 glaciers in the St. Elias Mountains, Yukon Territory, in an attempt to identify the potential influences on surge behaviour. Within the sample they estimated that 151 glaciers (6.4%) were of surge-type. Despite examining a large sample population their study failed to emphasize any particular environmental control , although they found that long glaciers (> 15 km) had a greater probability of being surge-type than short ones, as did glaciers with a higher overall elevation. Clarke et al. (1986) were able to draw two broad conclusions from their study. Firstly, they found a geographical concentration of surge-type glaciers in certain basins which have experienced rapid tectonic uplift since the Late Cenozoic. Such regions are subject to active surface processes and it is, therefore, reasonable to speculate that glaciers in these locations will be underlain by a greater amount of subglacial debris than those in more mature landscapes. A mechanism for ·glacier surging based on the deformation of sub glacial sediments has been proposed by Clarke et al. (1984) (section 1.3.3). Thus, it is possible that there exists a connection between high rates of recent geological uplift, or active geomorphological processes, and the distribution of surge- type glaciers. Clarke et al. (1986) make the second point that, in general, the larger a particular system the more likely it is to experience some kind of fai lure. One mechanism proposed for surges involves the destruction of a tunnel-dominated subglacial drainage system (Kamb et al., 1985: Section 1.3.2). Therefore, it is Page 3 VAR IEGATED GLACIER - 2000~~----,-- z 1500 0 I- § 1000 w _J w 'E w l'.) z 79 BO ~9/81 w _J -so w winler speed -7 0.4 tJ E 0 w 0.2 -w Q_ (f) ' · ·-~ 20 15 10 OiSTANC E 5 ( km) Chapter 1: Surge-type glaciers I ' ' 1 ' j 0 Figure 1.1 Some of the characteristics of surge-type glaciers. This figure, taken from Raymond (1987) illustrates changes in the long profile, elevation and winter velocity of Variegated Glacier prior to its 1982-83 surge. The top figure shows 'the steepening of the long profile during quiescence and the subsequent reduction in gradient following the surge. The changes in ice thickness occurring during quiescence in the · reservoir and receiving zones are shown in the middle figure. The bottom graph illustrates the progressive increases in velocity that take place in the reservoir zone prior to the active phase. Page4 Chapter 1: Surge-type glaciers VARI EGATED GLACIER 2000 -----,- ::~ t: longitudinal profile z 1500 9/61, I ' ' 0 ~;.,- I I- ~/ : ~ 1000 - , w __J ' w I 500 ·E 50 I ' w ' (.'.) z w __J -so 1 w winler speed I - ~79- i30 J -;- 0.4 u ~ I E 78 -79 I 77~ i 0 A,~,J1 w 0.2 -w [L (J) 73-74 I - - L J 20 15 10 5 0 DI STAf-.JC C: ( km) Figure 1.1 Some of the characteristics of surge-type glaciers. This figure, taken from Raymond (1987) illustrates changes in the long profile, elevation and winter velocity of Variegated Glacier prior to its 1982-83 surge. The top figure shows ·the steepening of the long profile during quiescence and the subsequent reduction in gradient following the surge. The changes in ice thickness occurring during quiescence in the · reservoir and receiving zones are shown in the middle figure . The bottom graph illustrates the progressive increases in velocity that take place in the reservoir zone prior to the active phase. Page4 Chapter 1: Surge-type glaciers suggested that the larger or longer a glacier is, the greater the probability that tunnel destruction can occur. Furthermore, long glaciers will have a greater likelihood of crossing deformable sediments. These conclusions are equivocal, however, and they may not apply to other areas with a similar concentration of surge-type glaciers. 1.1. 4 A short history of investigations References to events interpreted as glacier surges can be found in the legends of the native peoples of western North Am~rica (Cruickshank, 1990). Catastrophic glacier advances often affected the way of life of these peoples, for example by damming rivers which subsequently drained rapidly leading to widespread flooding, or by blocking salmon migration routes upriver. One oral tradition of the Athapaskan natives of the Yukon Territory tells the story of a surge of the Lowell Glacier. Legend has it that this glacier surge was caused by a shaman seeking revenge on the people of a Tlingit Indian village who had insulted one of his elders (Cruickshank, 1990). When the Lowell Glacier surges, it blocks the flow of the Alsek River. The subsequent failure of the ice dam causes widespread flooding downstream. On this particular occasion, the flood killed the entire population of the rival village. This story, and others like it, are interesting because they provide glaciologists with insight, beyond the limits of modern scientific observations, into the surge history of a number of glaciers. Some of the earliest scientific descriptions of glacier surges were those of Tarr and Martin (1914) in their important historical work on the glaciers of Yakutat Bay, Alaska. Ralph Tarr noted, during a return visit to the area in 1906, that a number of glaciers had changed their character substantially since his earlier visit the previous year. During later visits to the area by the authors in 1909 and 1910 they observed similar changes in a number of other glaciers. Among those glaciers which they noted had been transformed from smooth to crevassed surfaces was Variegated Glacier, the recent surge of which has been the subject of intensive study (Kamb et al. , 1985). Tarr and Martin (1914) recognised that these advances were unusual and were not related to climate, since other glaciers in the region were retreating. They speculated that the advances were caused by the release of additional mass received from snow avalanches, which had been induced by the earthquake of 1899. This ' earthquake hypothesis' of glacier surges remained in favour until work carried out by Austin Post on Alaskan glaciers following the 1963 tremor showed that there were no large-scale glacier advances as a result (Post, 1965). Further reports of unusual glacier advances in Alaska were made in the following decades by geologists , prospectors and engineers. One of the most thorough early accounts of a surge is that by Moffit (1942), with a detailed description of the 1936-37 surge of Black Rapids Glacier in the Alaska Range. Moffit did not attribute Page 5 Chapter 1: Surge-type glaciers the cause of the surge to the recent earthquakes in the area, since the region regularly experienced tectonic activity without any reaction from the glacier. Instead, he realised that the advance was due to a sudden release of mass, which had built up in the preceding years faster than it could be dissipated and that, furthermore, it was a cyclical event. However, Moffit considered that the build-up of mass was climatically induced and, in the case of Black Rapids Glacier, thought that the particularly high precipitation in the winters of 1932 and 1933 were the cause. Nevertheless, he recognised that this was not a full explanation since neighbouring glaciers had shown no abnormal behaviour. In the following years, reports of exceptional glacier advances became more frequent. Many of these surges were discovered during the comparison of photographs taken during repeated visits to the same area; for example, the surge of Trapridge Glacier in the 1940s (Wood, 1936; Sharp, 1947). Reports of glacier surges also came from areas outside North America, such as the Andes (Lliboutry, 1958), the Karakoram (Desio, 1954), Iceland (Thorarinsson, 1964) and Svalbard (Glen, 1937). However, most of these reports were qualitative and did not make any attempt to investigate the trigger mechanism. In order to understand the processes by which glaciers surge it is important to monitor such a glacier in the period prior to and during its active phase. The first such study was carried out on Medvezhiy Glacier in the Soviet Pamirs (Dolgushin et al., 1964). Following the 1963 surge, a programme of monitoring was initiated in anticipation of another surge some 12- 14 years later. In 1969, the National Research Council of Canada hosted a conference on the topic of surge-type glaciers. The proceedings, which were published in the Canadian Journal of Earth Sciences, contained papers reporting the occurrence of surge-type glaciers and also proposing theories of the triggering mechanism. Many of these contributions shaped the future of research into surging. In particular, it was recognised that there was a lack of detailed field measurements of such glaciers and, further, that more and better field data were essential in order for the surge mechanism to be explained (Meier, 1969). Since that conference there have been a number of important developments. Amongst these have been the initiation of long-term field projects aimed at studying a glacier through its cycle from quiescence to surge. Glaciologists from the University of British Columbia began monitoring Trapridge Glacier in the Yukon Territory, Canada, in 1969 (Collins, 1972). This programme is continuing as the expec ted surge approaches (Clarke, 1989). In 1973, a group from the Universities of Washington and Alaska, the California Institute of Technology, and the United States Geological Survey launched a programme on Variegated Glacier, Alaska (Bindschadler et al ., unpublished) which ended a short time after the 1982- 83 surge (Kamb et al., 1985). In Page6 Chapter 1: Surge-type glaciers addition, several conferences on the broad topic of glacier flow have served to stimulate interest in surging (e.g. the International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, Interlaken, 1985, and the Chapman Conference on Fast Glacier Flow, Whistler, 1986). There are now a number of quantitative models of the surge process. These will be discussed in section 1.3. 1.2 A REVIEW OF GLACIER MOTION RESEARCH 1. 2 .1 The flow and deformation of glaciers Glacier flow is driven by gravity and opposed by resistive forces, such as friction. Three components of motion are possible. The first component is caused by the ice in a glacier deforming under its own weight (internal deformation) and the second component results from the glacier sole moving over its substrate (basal sliding). Recent research has shown there is a third possible component of motion, i.e. the deformation of subglacial sediments. The proportion that each of the above contribute to the overall motion of a glacier varies between individual cases. Internal deformation consists of the mutual displacement of ice crystals relative to one another. Glen (1955) studied the deformation of polycrystalline ice in the laboratory and found that, when a constant stress was applied to the ice, the strain rate soon reached a steady value. This observation is known as Glen's law and was adapted for glaciers by Nye (1957). It can be expressed as: (1.1) where Exy is the shear strain rate, -rs is shear stress, n is a constant and A is a temperature-dependent ice hardness parameter. This relationship provides a reasonable approximation of glacier flow, although a number of experimental studies have demonstrated widely varying values of A and n (Hooke, 1981; Paterson, 1981). Values of n may vary from 1·5 and 4·2 (Weertman, 1973), although a mean of 3 is commonly used in glacier studies. Typical velocities predicted using Glen's flow law to calculate internal deformation are often well below those measured on ,glaciers. This difference in velocity may be accounted for by the occurrence of basal sliding which is a temperature-dependent process, since it can only take place in glaciers where there is an element of pressure melting at the bed (i.e. glaciers with a temperate or sub-polar thermal regime) . Basal sliding has been confirmed by observations made in boreholes (Engelhardt et al., 1978), subglacial tunnels and cavities (Kamb and LaChappelle, 1964; Vivian and Bocquet, 1973). Nevertheless, Shreve (1984) demonstrated theoretically that sliding at sub-freezing temperatures can occur, and this has been confirmed by observations (Hallet, 1986; Echelmeyer and Wang Zhongxiang, 1987). Page 7 Chapter 1: Surge-type glaciers The physical processes that govern sliding are not fully understood, which presents problems when attempting to model glacier flow and behaviour. However, there has been much research in recent years which has attempted to formulate the sliding mechanism. Two important aspects of this work, particularly relevant to studies of surging, will be discussed in the following sections: the influence of water at the glacier bed and the nature of the ice-bed interface and substrate. 1. 2. 2 The influence of water at the glacier bed The bulk of subglacial water is d~rived from surface melt or precipitation which reaches the bed either directly through moulins (Hooke, 1988) or by percolation in veins through the ice (Raymond and Harrison, 1975). Evidence from field studies suggests that, at the glacier bed, water flows in reasonably straight channels as opposed to a thin film. Dye-tracing experiments (e.g., Burkimsher, 1983; Seaberg et al., 1988) demonstrate relatively fast throughflow rates consistent with a tunnel system. These channels may either be incised up into the ice (Rothlisberger, 1972) or eroded into the substrate (Nye, 1973). Rothlisberger channels are thought to be more common, since flowing water will be able to erode channels into ice at a faster rate than into the glacier sole. Hooke et al. (1990) believe that they will have a broad and low cross-section and not a semi-circular geometry assumed by Rothlisberger (1972) . This would explain why water pressures beneath glaciers are frequently higher than those predicted from Rothlisberger's theory. Rothlisberger channels are maintained by a balance between the viscous dissipation of heat produced by flowing water and creep closure by the moving ice (Rothlisberger, 1972). In the past, sliding laws (e.g. Weertman, 1957) have assumed that a glacier flows over a relatively smooth rigid bedrock surface. Glacier geometry was used to prescribe velocity. A measure of the driving stress causing glaciers to flow is the shear stress. Given that Pi is the density of ice, g is acceleration due to gravity, h is the thickness of ice normal to the glacier bed and a is the surface slope, shear stress can be calculated from: (1.2). However, Meier (1968) was unable to find a relationship between sliding velocity and basal shear stress using data from Nisqually Glacier, Washington. Furthermore, it is known that glaciers exhibit variations in velocity on time scales shorter than those at which the geometry will change; there are seasonal and diurnal variations as well as rainfall-induced fluctuations (e.g. Meier 1960; Iken, 1973) (Figure 1.2). It is clear, therefore, that subglacial water has an important effect on glacier motion. High subglacial water pressures lead to enhanced sliding by eliminating local shear stress at the bed. Bindschadler (1983) attempted to quantify this process Page 8 Or.v1,1l1CH I !rnr1 1 111c,1n vcl0c1ly HI pCf(Clll S l,1hcs C, - 2., fi - 2 0 h-21l I G- t. lnd1r!-- nl llH! l(. ury f'1(' ( 1pl1,1 ! 1un C lmu luu:~!. Chapter 1: Surge-type glaciers - JOO 1- -- lSO - AUGUST S[f'T[MIJLn Figure 1.2 Relationship between surface velocity and meteorological conditions on Saskatchewan Glacier, Alberta, Canada, 1953. Note that the y-axis of the velocity measurements represents the percentage deviation from the mean velocity. Fluctuations in surface velocity are related to variations in temperature and precipitation, indicating the importance of water on the movement of glaciers. From Meier (1960). Page9 Chapter 1: Surge-type glaciers by introducing a parameter known as the bed separation index. This index correlated well with estimates of basal sliding velocities when tested with data from Variegated and Columbia glaciers and Ice Stream B. Iken and Bindschadler (1986) studied the relationship between subglacial water pressure and sliding velocity on Findelengletscher at the onset of the ablation season. Water pressures measured in boreholes fluctuated diurnally as well as responding to rainfall events. Corresponding to peaks in water pressure were peaks in sliding velocities. They proposed bed separation as the cause of this relationship. The theory developed by Rothlisberger (19721 predicted high water pressures when the discharge in winter is low. Therefore, for a short tlme at the start of the melt season, high water pressures are further increased by the addition of meltwater from the glacier surface (Iken and Bindschadler, 1986). This allows the growth of new cavities as predicted by Bindschadler (1983) and leads to bed separation. However, as the melt season progresses, flow tends to become channelised and consequently pressures drop. 1. 2. 3 The nature of the ice-bed interface and substrate The theories of glacier motion discussed in the sections 1.2.1 and 1.2.2 were based on the assumption that glaciers are underlain by a relatively smooth, rigid bedrock surface. Whilst this is undoubtedly true of a number of glaciers, recent research has demonstrated the importance of unlithified water-saturated sediments in contributing to the forward movement of some glaciers. Despite numerous reports of deformation structures in sediments formerly overlain by Quaternary ice sheets (e.g. Mathews and MacKay, 1960; Moran, 1971) it was not until Boulton and Jones (1979) observed active sedimentary deformation beneath Breidamerkurjokull, Iceland, that this process was considered an important component of glacier motion. The hydrology of sedimentary glacier beds will be substantially different from that of bedrock surfaces. Murray (1990) recognised. a number of processes which may transport water through a subglacial sediment layer. These include: bulk transport of water in the deforming sediment; sheet flow at the ice-substrate boundary; Darcian flow through a porous medium (this will be enhanced if inhomogeneities are present in the sediment); and pipe-routed flow through cavities, conduits and channels. If any or a combination of these processes are effectively transporting water through the sediment, then the material will remain relatively immobile. Once these drainage pathways become disrupted, water will begin to build up and at a critical level will alter the rheology of the sediment. It will then behave like a 'slurry' (Boulton and Jories, 1979). If the overlying glac"ier is coupled to this layer it will begin to move at enhanced velocities. The recent discovery of an approximately 6 m thick, water-saturated sedimentary layer beneath Ice Stream Bin West Antarctica (section 1.4.2) (Blankenship Page 10 Chapter 1 : Surge-type glaciers et al., 1986; Blankenship et al., 1987) using seismic methods, and the subsequent retrieval of a core of this material (Engelhardt et al., 1990), have highlighted the significance of defomling beds in the dynamics of ice sheets. The same processes may also be responsible for periodic surging in mountain glaciers (Clarke, 1987a). Deforming beds may have been important in determining the behaviour of outlet glaciers draining the major Quaternary ice sheets. Clayton et al. (1985) found extensive deformation structures in the sediments of Minnesota which they attributed to surging of the south western margin of the Laurentide Ice Sheet over an unlithified substrate. However, such deformation structures may not always be related to fast glacier flow. Brown et al. (1987) report similar features in the sediments in the Puget Sound area, Washington, but suggest that they were caused by high subglacial water pressures at the ice-bed interface rather than active sediment deformation. Menzies (1989) has classified glacier beds according to three types. "H'' beds are classic rigid bedrock surfaces. "M" beds are those composed of soft or mobile subglacial debris. "Q" beds are an intermediate type and are characterised by spatial and temporal variations in the extent of hard and soft material. This intermediate classification may be the most realistic description of actual glacier beds. 1.3 MECHANISMS OF GLACIER SURGING 1. 3 .1 Introduction Despite a considerable amount of research, glaciologists remain uncertain about what causes a glacier to surge. Two hypotheses are currently attracting most attention. However, the lack of suitably detailed observations on a large number of surge-type glaciers has prevented the general acceptance of any one theory. Fast glacier flow results from fast sliding, not from fast creep (Clarke, 1987a), although this view has been challenged by Echelmeyer and Harrison (1990) . As discussed in Section 1.2, the processes involved in internal deformation are better understood than those involved in basal motion. Therefore, studies aimed at explaining surging and basal motion are closely related, and the mutual exchange of ideas has benefited both fields . During the last thirty years a number of mechanisms have been proposed as possible causes of glacier surges. The majority of these have been shown to be unrealistic and are no longer considered. Most effort has been directed towards three hypotheses, namely, thermal instabilities, the mechanical damming of water, and the deformation of unlithified sediments. Each of these theories will be discussed in greater detail below. Page 11 Chapter 1: Surge-type glaciers In addition to the above mechanisms, there are a number of theories that have sought to explain surge behaviour on the basis of the large-scale dynamics of ice masses. An example of this .type was developed by Budd (1975), based on a critical parameter given by the product of the balance flux (calculated from the averaged centreline velocity and the average centreline depth) per unit width and mean ice surface slope. Budd defined this product as the rate of heat dissipation by loss of potential energy, or the lubrication factor. The greater the heat lost, the greater the amount of basal water that is produced, therefore enhancing velocity. Based on this theory, Budd -(1975) classified glaciers into three categories according to their large-scale dynamics. Ordinary glaciers have a flux which is low enough for them to transport at 'normal' velocities and therefore remain in equilibrium with their climate. Fast glaciers are those with a sufficiently high flux that they remain in a constant fast mode. Budd included in this category ice streams and outlet glaciers draining the large polar ice sheets. Surge- type glaciers are those with a flux which is sufficiently high to enable them to reach the fast mode periodically but which is not sufficient to maintain them in such a state indefinitely. This point is illustrated in Figure 1.3, with surge-type glaciers falling in the intermediate stage between fast and ordinary glaciers. However, since Budd (1975) used only seven surge-type glaciers in the plot, this classification must be treated with caution. Calculations made by Bindschadler (1984) indicated that the fast flowing Jakobshavn Isbne in West Greenland was in balance, and would fall into Budd's fast glacier category. However, not all fast moving ice streams fit into this grouping. For · example, Ice Stream B in West Antarctica is flowing 28% faster than its current mass balance suggests it should and, therefore, its catchment area is thinning whilst the lower glacier is thickening (Whillans and Bindschadler, 1988). Furthermore, an analysis by Wilbur (1986) of surge-type and non surge-type glaciers in Alaska found that flux, flux rate and the product of flux rate and surface slope (Budd, 1975) were poor discriminators between the two glacier types. This was because these parameters tended to increase with glacier size, whereas surge-type behaviour was found across a range of glacier sizes. The influence of temperature on the defom1ation rate of ice is well established in theory (Paterson, 1981). Not surprisingly, therefore, thermal instabilities have been proposed as a possible cause of glacier surges. Two principal mechanisms of this type have been suggested. The first, creep instability, supposes that a small increase in the deformation rate due to an increase in strain heating and the ice temperature will produce a further increase in ice deformation (Robin, 1955). This represents a positive feedback effect which, under favourable conditions, may lead to a continual increase in temperature. Robin suggested that, in a cold glacier frozen to its bed, creep instability could warm the basal ice to the pressure melting point, thus allowing the glacier to Page 12 ) versus ice surface slope, velocity and thickness (Z). Using selected glaciers, the mean centreline ice thicknesses are shown as a function of the surface slope and the mean centreline velocity. Individual points for some rapidly flowing ice sheet outlets are shown as circles in the }:i.igh velocity area. The crosses represent the flux/slope relationship for seven surge-type glaciers. The heavy broken line defines a constant flux/slope curve which suggests that this may . mark a transition zone between ordinary and surge-type glaciers. From Budd (1975). Page 13 Chapter 1: Surge-type glaciers slide. This model was developed further by Clarke et al. (1977) and Paterson et al. (1978). They were concerned that creep instability may not be related to surging since the period of time over which it develops is an order of magnitude longer than observed intervals between surges. Fowler and Larson (1980) later concluded that a catastrophic thermal instability was not possible, regardless of the time scale involved. They argued that Clarke et al. and Paterson et al. had used unrealistic boundary conditions in their modelling. Thus, when Fowler and Larson (1980) repeated the analysis with physically realistic boundary conditions, they reached a different conclusion. A second mechanism involvinta thermal instability was developed by Jarvis and Clarke (1975) and Clarke (1976). They postulated that surging would occur when the base of a cold glacier reached the pressure melting point as a result of the combined effects of surface accumulation and compressive flow. During the quiescent phase these processes would cause the glacier to thicken, thereby raising the temperature of the basal ice. This would allow enhanced sliding to occur. The duration of the surge is controlled by two competing processes: 1. the generation of frictional heat at the bed produces meltwater which helps sustain fast sliding velocities, and 2. as the glacier extends along its length the ice becomes thinner, thus allowing the temperature gradient to increase. Since heat is able to escape at a faster rate the basal ice becomes frozen once more to its bed and the surge terminates. The cycle then repeats itself as the reservoir area begins to thicken and the glacier builds up to another surge. Schytt (1969) proposed a surge mechanism based on a thermal instability. He assumed that the ice masses of Nordaustlandet in Svalbard exhibited a sub-polar thermal regime, where a central core of warm-based ice was surrounded by an annulus of colder ice frozen to its bed. Using the 1936-38 surge of the outlet glacier Brasvellbreen as an example, he suggested that surges may result when ice from the central core breaks through the barrier of cold ice. The precise mechanism, however, was not fully explained. Whilst models of thermal instabilities have attracted some attention, more recent work does not support the view that they are realistic explanations of surges. The major reservation is that they are capable of explaining surges only in glaciers where there is an element of freezing at the base (i.e. those with polar or sub-polar thermal regimes). They are not suitable for explaining surges in temperate glaciers (Paterson, 1981; Sharp, 1988). Therefore, not all surges can result from thermal instabilities. Some glaciologists (e.g. Paterson, 1986) consider it preferable to find one mechanism capable of explaining all surges. As a result, thermal instability models have been neglected in recent research. Page 14 Chapter 1: Surge-type glaciers The importance of water at the bed of a glacier was discussed in Section 1.2.2. A number of theories involving an increase in the amount of water at the glacier bed have been proposed to account for surges. Weertman (1969) suggested that the presence of a water film of thickness .1 would have the effect of drowning out bed obstacles smaller than .1'. The glacier bed would thus appear smoother and basal sliding would increase due to a reduction in friction. Weertman argued. that once the critical thickness of the water film has been reached it can be maintained without difficulty through the production of meltwater by frictional heat. Weertman (1969) calculated that glaciers draining a ~arge area would be most likely to accumulate enough water at their beds to drown the controllmg obstacle size. The mechanism seems unable to account for surges which have been observed to occur in small glaciers (see Tables 1 and 2 in Dowdeswell et al., 1991). Furthermore, Weertman's (1969) model relies on a number of unrealistic assumptions. Firstly, the glacier substrate must be impervious to the drainage of water and, secondly, the bed must be smoother than observations of typical beds suggest. Furthermore, the model is capable of explaining surges only in glaciers where the bed is at the pressure melting point. The trapping of water at the glacier bed was proposed by Robin and Weertman (1973) as a surge mechanism. During quiescence, the lower portion of a surge-type glacier becomes increasingly stagnant whereas its upper reaches become progressively more active as a result of increased basal shear stress (due to increasing ice thickness). At the junction between the stagnant and active ice, a gradient in basal shear stress develops. Robin and Weertman (1973) postulci!ed.that this gradient is sufficient to induce a reverse gradient in water pressure that causes water to become trapped up-glacier of the boundary. This allows the propagation of fast sliding velocities up-glacier and also down-glacier, because of the increased meltwater production due to frictional heat. Surge velocities would be maintained until the ice thickness and surf ace slope were reduced to such low values that fast sliding would not be permitted even over a well- lubricated bed. This mechanism is valid only if the substrate is impermeable and the water trapped does not escape through Rothlisberger channels. Similarly, water must not escape through Nye channels, although this may be permitted if there is an excess supply of water and the channels become saturated (Robin and Weertman, 1973). Paterson (1981) believed that the mechanical trapping of water was a likely trigger of surges, but he speculated that it may be a thermal barrier rather than a gradient in shear stress that acts as the dam. However, other studies have shown that the Robin- Weertman model is unrealistic . Bindschadler et al. (1977) found no evidence of a reverse basal"shear stress gradient on Variegated Glacier in the zone where the surge originated. Calculations made by Dowdeswell (1984) of shear stresses on Finsterwalderbreen, the glacier taken as an example by Robin and Weertman (1973) in Page 15 Chapter 1: Surge-type glaciers their original model, demonstrated that when realistic distances are used for averaging surface slopes in the calculation of shear stress, there is no reverse gradient. Furthermore, the damming of water at a thermal barrier is now considered unlikely. Such a barrier exists on Trapridge Glacier but it does not trap water (Clarke et al., 1984). A hydrologically-related trigger is now the most favoured cause of glacier surges. Two theories have arisen from the detailed field projects undertaken on Variegated Glacier and Trapridge Glacier. Both theories begin from the same premise: that the destruction of the subglacial ckainage system is responsible. However, the exact processes involved differ between the two theories. These models will now be discussed in the following sections. 1. 3. 2 Hard bed mechanisms The most detailed arguments for a hydrological initiation of surges have originated from the observational and experimental work undertaken on Variegated Glacier, Alaska (Kamb et al., 1985). This project collected an extensive amount of data prior to and during the 1982-83 surge and constitutes the most thorough investigation of a surge-type glacier yet undertaken. Measurements of borehole deformation made during the 1982- 83 event demonstrated that approximately 95% of the velocity was achieved through basal sliding (Kamb et al., 1985), emphasizing the importance of basal water in a surge. The link between subglacial water and basal sliding has been demonstrated effectively in a number of glaciers, of both surge-type (Engelhardt et al., 1986; Kamb and Engelhardt, 1987) and non-surge-type (Hodge, 1976; Iken and Bindschadler, 1986) (see also Section 1.2.2). Periods of high borehole water pressures coincide with glacier uplift (Iken et al., 1983; Aellen, 1986) and pulses of enhanced motion, which on surge-type glaciers are termed mini-surges (Harrison et al. , 1986a; Kamb and Engelhardt, 1987). Results from the Variegated Glacier studies have been incorporated in the development of a model of surge initiation based on the destruction of the subglacial drainage system (Kamb 1986, 1987). Water flow at the bas~ of a glacier in a normal (or quiescent) state is almost always in large tunnels. The detailed model described by Kamb (1987) involves the destruction of the basal tunnel system followed by the evolution of a linked cavity network. This linked cavity network serves to restrict the flow of water, therefore increasing the basal water pressure. The consequent reduction in bed friction leads to enhanced basal sliding. As a glacier slides, separation occurs at the ice-bed interface. This leads to the formation of cavities, approximately 1 m high and 10 m wide, which are widely Page 16 Chapter 1: Surge-type glaciers scattered over the glacier bed. Water passes between the cavities in a diffuse system of narrow conduits called orifices (Figure 1.4 ). Evidence for the existence of such a network beneath Variegated Glacier is derived from the results of dye tracing experiments (Kamb, 1987). The mean water throughflow rate during the quiescent phase was 0.7 m s-1 compared to a rate of 0.025 m s-1 during the surge (Brugman, 1986). Furthermore, during quiescence, water was discharged from one major outflow stream, compared to many smaller portals around the glacier margin during the surge. This indicated that water was distributed across the width of the glacier bed. Three physical processes act together to determine the hydraulic behaviour of the linked cavity system (Kamb, 1987): 1. Basal sliding-induced cavitation determines the orifice size and geometry, given the micro-relief of the bed, water pressure and cryostatic pressure, and the sliding velocity. 2. The water flow through the linked cavity system will depend on orifice and cavity geometry and the hydraulic gradient. 3. Enlargement of the orifices will occur due to viscous dissipation of heat produced by the flow of the water. This will result in a modification of process 1 and a consequent increase in the flow given by process 2. Cavities are formed by the sliding of a glacier sole over bedrock perturbations (Lliboutry, 1968). During periods of low fluid discharges, the normal tunnel system degenerates, trapping residual meltwater in the cavities. The timing of this change in the configuration of the basal drainage network is believed to be seasonal. If theoretical predictions are correct this should occur in winter (Raymond and Harrison, 1986). They suggest that there is a critical minimum discharge below which a tunnel system win not survive, because as water discharge decreases creep flow of the glacier becomes more effective at closing the large tunnels. Once a linked cavity configuration has been established, basal sliding is enhanced and surge velocities are reached. However, Kamb (1987) believed. that the system can experience periodic instabilities. Kamb et al. (1985) document motion slow- down events associated with pronounced water discharge floods during the 1982-83 surge of Variegated Glacier. Motion slow-down events represent instabilities within the linked cavity network. Kamb (1987) attempted to predict the stability of subglacial drainage systems by proposing a stability parameter, 3, which is dependent on the local hydraulic gradient, aA/iJ (where iJ is network tortuosity and A is a head concentration fac tor, defined as Lcf L0 if Le is the distance between orifices and L0 is the length of orifices, both parallel to water flow), Manning roughness (M) , a constant, D , with dimensions of length (D = PiHlpwg where His latent heat of melting and Pw is the density of Page 17 Chapter 1: Surge-type glaciers 1 2 - !Orn Figure 1.4 Schematic representation of a linked cavity basal water conduit system. In 1, shading indicates areas of ice contact with the bed. The blank areas are cavities and conduits. The large arrow indicates the direction of glacier movement and the small arrows show the direction of water flow through the system. The cross-section A- A' and B-B' are shown in 2. Heavy shading indicates bedrock and the lighter shading, ice. The arrow shows the direction of glacier movement. The formation of cavities as a result of sliding-induced separation in the lee of bedrocks bumps is clearly shown. Orifice height has been exaggerated for clarity. From Kamb (1987). Page 18 Chapter I: Surge-type glaciers water), viscosity of ice (77), the sliding velocity of the ice (uB), the excess of overburden pressure (Neff, where Neff = O"B - Pw, given that O"B and Pw are ice pressure and water pressure respectively), and the orifice step height (s). Using the equation: 3= 21/3 (aAhJ/13 ( 77 }1;2 s7/6 . nl/2 DM UB Neff (1.3) the linked cavity system is stable when 3 < 1 ·0. Above this value, a tunnel network will begin to develop. Slow-down ev~nts indicate that the critical value has been exceeded and that water will be discharged more efficiently from the bed. These events, however, can be short-lived and a return to a stable linked cavity network may be achieved relatively quickly. The same process may also lead to the complete tem1ination of a surge (Kamb, 1987). The mechanism involved is multiple conduit instability and it is a form of positive feedback reaction. A trigger or perturbation is required to initiate the instability. This may, for instance, be a substantial increase in water input to the bed. If such a perturbation is sufficiently large, it will cause unstable growth to occur simultaneously in a considerable number of orifices due to the increased effect of viscous dissipation. The linked cavity network then develops into a series of interconnected tunnels which in turn proceeds to degenerate into a single tunnel system by multiple conduit instability. Water can, therefore, be discharged more effectively. The accompanying drop in basal water pressure leads to a consequent and significant decrease in glacier velocity. The data obtained during the Variegated Glacier studies lend support to the mechanism outlined above. Raymond and Harrison (1986) stated.that basal sliding during the winter season began to occur for the first time during 1978-79 and continued each winter until the winter of 1981. This suggests that the complete discharge of the previous summer's meltwater had been prevented by tunnel collapse. In the very early-stages of the melt seasons 1978-81 a number of mini-surges were recorded (Raymond and Malone, 1986; Kamb and Engelhardt, 1987). During these events, ice velocities increased markedly and remained high for up to one day. They were presumably associated with increased basal water pressures that resulted from the discharge of early season meltwater through insufficiently large tunnels (Kamb and Engelhardt, 1987). The terminations of the mini-surges were followed by peak flood discharges of highly turbid water (Humphrey et al., 1986). These water peaks probably initiated unstable orifice growth which led to the abrupt cessation of these events. During the winters 197 8-81 the stability parameter did not drop below the critical value required to maintain a linked cavity network. Thus, a full surge was Page 19 r Chapter 1: Surge-type glaciers unable to get underway. In November 1981, however, tunnel collapse occurred early enough to trap a significant amount of the previous summer's meltwater. The trapped water was pumped to the glacier bed causing an expansion of the linked cavity system (Raymond and Harrison, 1985). According to Kamb et al. (1985), the glacier began to flow at increased velocities in January 1982. The above events agree with the model prediction that tunnel collapse occurs during periods of falling meltwater discharge. Further support for this hypothesis of a seasonal initiation of surges comes from observations made during the surges of Peters Glacier, Alaska, in 1986- 87 (Echelmeyer et al., 1987) and We~t Fork Glacier, Alaska, in 1987-88 (Echelmeyer and Harrison, 1989). Both surges were initiated during the winter months. In late June 1982, within the space of a few hours, the velocity of Variegated Glacier decreased abruptly (Kamb et al., 1985; Engelhardt et al., 1986). Throughout the following months pulses of velocity were recorded until, in October 1982, the flow velocity began to increase steadily before reaching a stable level in January 1983. Surge velocities remained high and reached a peak in June 1983 before they decreased suddenly a month later. During this period of fast motion one major slow-down event occurred, in February 1983 (Kamb et al., 1985). Immediately prior to this and similar, but smaller, events there were marked increases in the water pressures measured in boreholes. The slow-downs in glacier velocity were followed by abnormally large floods of turbid meltwater discharged in the outflow streams. In discussing this mechanism, Kamb (1987) does not account for the temporary halt in surge activity in June 1982 (Kamb et al., 1985), but it is reasonable to assume the following. As the summer melt season got underway at the beginning of June, there was a sudden increase in water input which led to unstable growth of the orifices. This did not signal the termination of the surge because the ice thickness in the reservoir zone was sufficiently large the following year to promote a stability factor capable of allowing the development and maintenance of a linked cavity system. According to Kamb (1987, equation 37), the stability parameter is inversely proportional to basal shear stress. Shear stress will be be greatest in the reservoir zone owing to the thickness of ice (equation 1.2). Hence, a linked cavity network is able to evolve once again. The complete termination of the surge occurred in July 1983 (Kamb et al., 1985). Associated with it were a marked increase in basal water pressures and a spectacular outburst flood . The mechanism which caused the surge to terminate was multiple conduit instability (Kamb, 1987). Unlike the previous sumrrier, the glacier was unable to restart its surge since ice mass had been redistributed during the months between. Reduced ice thickness in the upper glacier and the lower overall surf ace slope led to a reduction in shear stress such that the stability parameter increased and Page 20 Chapter 1: Surge-type glaciers exceeded the critical value required to create or maintain a linked cavity network. Instead, meltwater was discharged through a normal tunnel system. Any theory which seeks to explain the mechanics of periodic fast glacier flow rnust be able to distinguish between surge-type and non-surge-type glaciers. Kamb (1987) argued that surge behaviour is a possibility if a set of glacier parameters lie within certain limits. The relevant parameters include glacier length, slope, ice overburden pressure, the concentration of orifices, roughness of the bed, and cavity geometry. Should these parameters have values within reasonable limits then the glacier will exhibit surge-type behaviour. - It is possible at a qualitative level to assess the validity of Kamb's hypothesis . Clarke et al. (1986) found that there was a correlation between glacier length and the probability of surge-type behaviour in a statistical analysis of Yukon glaciers. There is a strong inverse relationship between glacier length and surface slope. Therefore, it is reasonable to postulate that there is a similar relationship between slope and the probability of surge-type behaviour. Ice surface slope, a, is one of the variables in Kamb's model (equation 1.3). The stability parameter varies as a3/2. Thus, longer glaciers will have a greater probability of reaching values low enough to stabilize a linked cavity system (Kamb, 1987). The above model was developed on the assumption that the bed of a surge-type glacier is "hard", i.e. composed of bedrock. Nevertheless, Kamb (1987) considered that the mechanism would also apply to glaciers underlain by "soft", unconsolidated sediments. The process of cavitation on soft beds is not fundamentally different from that on hard beds. On soft beds, however, cavitation can only occur if there are a sufficient number of roughness features that remain fixed during ice motion. These may be bedrock outcrops or large boulders that are lodged at depths exceeding that of the deforming layer of sediment (Kamb, 1987). This discussion has shown that the mechanism developed by Kamb can be used to explain the 1982- 83 surge of Variegated Glacier. However, the theory leaves a number of questions unanswered. For example, it is not possible to explain theoretically how fast velocities are reached in winter when the availability of water is at a minimum (Kan1b, 1987). Moreover, Echelmeyer et al. (1987) observed that the 1986-87 surge of Peters Glacier terminated in winter, ·in contrast to the events on Variegated Glacier and those predicted by the model. 1. 3. 3 Soft bed mechanisms Theories of glacier movement were initially developed on the assumption that ice rests on a rigid bedrock substrate. Boulton and Jones (1979) demonstrated that this assumption may not be true of all glacier beds. Instead, they argued that a significant Page 21 Chapter 1: Surge-type glaciers number of present and fom1er glacier beds may be composed of soft sediments and that the deformation of such materials may allow enhanced forward motion (section 1.2.3). Post (1969) was the first to suggest that bed permeability might be a factor involved in surge mechanisms. A similar .idea was proposed by Jones (1979). He argued that internal water storage in a substrate would lead to excess pore pressure that could transform an unlithified impermeable substrate into slurry, thereby triggering a surge. Substrate permeability remains constant before and during the surge: of most significance is the effect of wate! on the rheology of the material. However, Jones (1979) failed to explain how such a process may operate periodically to induce surges. Clarke et al. (1984) have developed further the idea that surge-type glaciers lie on beds of deformable sediments. Since 1969 they have studied the progressive changes of the surge-type Trapridge Glacier in the Yukon Territory, in the build-up to its next surge. The most spectacular change is the evolution of a wave-like bulge in the lower region of the glacier. A similar feature was observed and photographed by Wood (1942) prior to a surge sometime during the 1940s. The bulge marks the boundary between two distinct patterns of behaviour. Up- glacier of the bulge, both ice thickness and ice velocity have increased considerably since measurements began whereas, below the bulge, the glacier has stagnated (Clarke et al., 1984). Jarvis and Clarke (1975) demonstrated that the glacier has a sub-polar thermal regime. Furthermore, the bulge has formed at the boundary between the warm- based (upstream) ice and the cold-based (downstream) ice. The presence of cold-based ice downstream has not prevented the bulge from migrating down-glacier. During the period 1969- 89 the bulge has propagated down-glacier at approximately 30 m a-1 (Clarke and Blake, 1991). This geometric change has been accompanied by a thermal evolution. The zone of transition between warm-based ice and cold-based ice is also moving down-glacier, but at a slower speed than the propagation of the bulge (Clarke and Blake, 1991). However, a thermal instability is not considered to be the likely trigger of the predicted forthcoming surge (Clarke et al. , 1984). Fluctuating basal water pressures measured upstream of the bulge indicate that the thermal boundary does not act as a barrier to the flow of subglacial water (Clarke, 1986). Moreover, temperature measurements indicate that the area of the bed lying under the bulfe and downstream of it is composed of permafrost. Clarke et al . (1984) suggesfk that this material is permeable and that subglacial water is flowing through it and not along the ice-substrate contact. They also argue that this material is deformable. Indirect evidence in support of this hypothesis is found on the recently exposed glacier foreland . Most of this area is covered with till and there are very few bedrock outcrops. The surge mechanism proposed by Clarke et al. (1984) begins with the same principle used by Kamb (1987), i.e. that the destruction of the subglacial drainage Page 22 Chapter I: Surge-type glaciers system is the surge trigger. At this point, however, the two theories diverge. Clarke et al. (1984) believefthat the subglacial drainage system of Trapridge Glacier is contained within the permeable substrate lying beneath it. If the subglacial sediments were to be transformed into a slurry by the accumulation of water in the material, the glacier would be able to flow at enhanced velocities. For such a transformation to take place the inflow of water must exceed the outflow. Clarke et al. (1984) sugges~ihat a reduction in the effective permeability of the substrate represents the destruction of the sub glacial drainage system. According to Clarke (l 987b),- the permeability of till decreases as it is compressed. One cause of compression is through shear deformation produced by the deviatoric component of stress. On Trapridge Glacier, the obstruction to the downhill flow of ice causes a progressive thickening of the glacier in its upper reaches, thus increasing the surface slope. Therefore, shear stress at the base increases, resulting in an increase in sediment deformation. This destroys the drainage system which is composed of conduits within the substrate, and water begins to accumulate. As it does so, subglacial water pressures increase and the effective shear stress at the bed is reduced. Both the sliding velocity of the glacier and the substrate pore pressure will increase, promoting weakening of the sediment. The surge accelerates further as deformation continues to reduce the effective permeability (Clarke et al., 1984). According to this model, the surge terminates because the redistribution of ice leads to a lowering of the surface slope which, in turn, reduces the basal shear stress to a level which cannot sustain high sliding velocities. Drainage pathways through the substrate are then established at a rate which exceeds their destruction until the surge ends. During quiescence the tunnel drainage system is fully re-established. Clarke et al. (1984) suggested that the spatial distribution of surge-type glaciers is controlled geologically. They argued that the location of these glaciers is determined by the presence of deformable sediments, the occurrence of which is likely to be influenced by the lithology and maturity of the landscape. Such an hypothesis is attractive. There h·ave been a considerable number of documented glacier surges in the Wrangell-St. Elias Mountains of western North America (Horvath and Field, 1969; Post, 1969; Clarke et al., 1986). This region is characteri-sed by recent tectonic uplift and is thus able to provide abundant, readily erodible material. Furthermore, other areas of the world where surge-type glaciers appear to be concentrated are composed of rocks that are ee1.stly erodible, for example parts of Iceland (Thotarinsson, 1969), Svalbard (LiestS,11, in press), and the Pamirs (Dolgushin and Osipova, 1973). Harrison et al. ( 1986b) ·suggested that Variegated Glacier may be resting on a bed of deformable sediments (termed "an active subsole drift"). Moreover, characteristic landforms produced by surge-type glaciers, such as crevasse-fill ridges (Sharp, 1985) may also Page 23 Chapter 1: Surge-type glaciers indicate the presence of a slurry-like material at the base of such glaciers. During a surge, this material is forced up into bottom crevasses which later melt leaving a ridge- like sedimentary feature. One major drawback of the model proposed by Clarke et al. (1984) is that it has yet to be applied to an actively surging glacier. Since monitoring of Trapridge Glacier began in 1969, it has remained in the quiescent phase of its surge cycle and, therefore, no attempt can be made to assess the theory's validity on this glacier. However, if the observations of Harrison et al. (1986b) are confirmed, and similar observations are made on additional surge-type glacjers, then the hypothesis that the distribution of such glaciers is controlled by the presence of deformable beds may be strengthened. 1.4 OTHER TYPES OF FAST GLACIER FLOW 1. 4. 1 Ice streams and fast flowing outlet glaciers The large polar ice sheets of Greenland and Antarctica are drained by fast moving ice streams and outlet glaciers. An ice stream is defined as part of an inland ice sheet that flows more rapidly than the surrounding ice (Swithinbank, 1954; Bentley, 1987). Ice streams flow within ice bordered channels, although not necessarily in the same direction as the ice sheet. Outlet glaciers differ from ice streams in that their flow may be contained at one, or both, margins by a range of mountains. Ice streams and outlet glaciers transport ice from deep within the ice sheets to the coast. Here they form ice shelves or ice tongues which, in turn, gradually lose mass by calving. Ice shelves or tongues differ from ice streams in that they are floating and, hence, cannot support shear stresses. Ice shelf spreading is , therefore, due to longitudinal stresses (Bentley, 1987). Recent research on a number of West Antarctic ice streams has led to an increased understanding of the dynamics of fast glacier flow and may provide information relevant to those studying glacier surges (Clarke, 1987a). The most intensively studied ice streams are those which drain the West Antarctic Ice Shee! and feed into the Ross Ice Shelf, designated simply ice streams A- F. They are characterised by very low surface elevation profile s, beds of low slope, and very low driving stresses that decrease continually down-stream (Bentley, 1987). Two of these ice streams are of particular interest. Ice Stre·am Bis currently moving at velocities in excess of 800 m a-1 . In contrast, its neighbour Ice Stream C is moving at a mere 5 m a-1 (Whillans et al., 1987). However, using estimates based on the depth of buried surface crevasses, Shabtaie and Bentley (1987) suggested that the Ice Stream C was, until 250 years ago, flowing at velocities comparable to those of Ice Stream B at the present time. The cause of this switch from fast to slow flow is currently unclear, although Rose (1979) postulated that catchment capture between the ice streams was Page 24 Chapter 1: Surge-type glaciers responsible for the change in regime. There is clearly a non-equilibrium situation present, in that Ice Stream B has a negative mass balance and Ice Stream C has a strongly positive mass balance (Shabtaie et al., 1988). It is tempting to speculate that this on/off behaviour may represent surging on a large spatial and temporal scale. One of the most significant discoveries in recent glaciological research was made by University of Wisconsin geophysicists. Their seismic investigations have shown that Ice Stream B is underlain by a layer of till approximately 8 m thick (Blankenship et al., 1986) and, furthermore, that this material is unconsolidated, porous and saturated with water under very high pressures (Blankenship et al., 1987). Alley et al. (1987a) believe that this till layer is actively deforming and suggest that this allows Ice Stream B to achieve and sustain its high velocities. The retrieval of several cores of this material (Engelhardt et al., 1990) enabled Kamb (1991) to analyse the significance of sediment deformation on the rapid motion of Ice Stream B. He concluded that the rheology of the subglacial till was probably non-linear, as opposed to a linear flow law assumed in earlier modelling studies (e.g. Alley et al., 1987b). Sediment with a non-linear rheology was found to be sensitive to peturbations in basal water pressures, suggesting that flow instabilities would occur. However, the apparent absence of instabilities in the currently active ice streams was taken by Kamb (1991) to imply that rapid ice stream flow is not controlled by deforming basal sediments. Fastook (1987) used a numerical model to study the transient behaviour of Ice Stream C and the cause of its present low velocity. These experiments suggested that catchment capture by Ice Stream B was not the major cause, since the modelled decrease in catchment area was unable to account for the observed reduction in velocity. A more likely cause was a change in the conditions at the bed leading to a removal of basal sliding. The base of Ice Stream C has been investigated using seismic methods (Anandakrishnan and Bentley, 1989) which may provide some evidence to support Fastook's result. Micro-earthquake activity on Ice Stream C is much more widespread than on Ice Stream B. The cause of this increased seismicity is attributed to brittle fracture of the basal sediments, which would occur in the absence of water acting as a lubricant, as it does on Ice Stream B (Anandakrishnan and Bentley, 1989). However, the mechanism leading to the draining of liquid from the basal sediments is not clear. The world's fastest moving glacier, Jakobshavn Isbne, drains approximately 5% of the Greenland Ice Sheet (Bindschadler, 1984). The glacier can be traced 200 km into the ice sheet where it flows as a distinct ice stream. As the glacier approaches the coast, the flow becomes constrained by valley walls. The lowest 20 km form a floating ice tongue within the fjord. In this region the average surface velocity is approximately 7 km a-1 (Lingle et al., 1981). In a recent study, Echelmeyer and Harrison (1990) measured surface velocities in the lower glacier over a number of years but did not Page 25 Chapter 1: Surge-type glaciers detect any seasonal variations. However, they presented evidence which suggested that the input of surface meltwater to the fjord via a subglacial route was a seasonal event. The implication of this discovery is that seasonal variations in the input of meltwater to the base have a limited effect on the motion of the glacier. This raises the possibility that the fast velocities are achieved through internal deformation of the ice (Echelmeyer and Harrison, 1990). According to Echelrneyer et al. (1991), enhanced internal deformation is likely to occur because of the large driving stresses present (200--300 kPa). These extremely high shear stresses result from the huge ice thicknesses and relatively steep surface slope of the ice stream. A positive feedback process might then operate, with the rapid motion generating high rates of basal melt which would enhance basal sliding (Echelmeyer et al., 1991). Ice streams and outlet glaciers are the major route for the discharge of ice from the interiors of large ice sheets to the oceans. Therefore, fast glacier flow is believed to play an important role in ice sheet collapse. Paterson (1972) suggested that a large, fast moving glacier flowing into Hudson Bay, and unable to stabilise its terminus in deep water, was responsible for discharging a considerable portion of the Laurentide Ice Sheet to the ocean during the Late Wisconsin deglaciation. The presence of deforming subglacial sediments might also influence the stability of ice sheets (Boulton and Jones, 1979). Ice streams or outlet glaciers underlain by unlithified sediments would experience enhanced basal motion and would be more effective at drawing ice from the ice sheet interiors, therefore promoting their collapse. The stability of the West Antarctic Ice Sheet to changes in climate is currently of considerable scientific interest. Therefore, there has been much research into the causes and mechanics of ice stream motion and its influence on ice sheet stability (National Research Council, 1984). Surging of the West Antarctic Ice Sheet has also been proposed as a mechanism of ice age initiation (Wilson 1964). An increase in the ice sheet surface profile due to a greater accumulation of snow may lead to an instability and a large scale surge. This would result in an increase in the area of the Southern Ocean covered by jce. It is suggested that the subsequent increase in albedo would be sufficient to cause a reduction in global temperatures, capable of initiating an ice age. Recent numerical modelling experiments, however, have suggested that large-scale surging of the Antarctic Ice Sheet is not likely to occur (Radok, et al., 1986). 1. 4. 2 Tidewater glaciers A number of the world' s glaciers terminate in the ocean or in fjords, where they lose mass by calving icebergs. Some of these glaciers are similar to actively surging glaciers in that they flow at high velocities, only that they do so continuously as opposed to periodically (Meier and Post, 1987). Tidewater glaciers are grounded at Page 26 II Chapter 1: Surge-type glaciers their termini. Grounding typically takes place on topographic highs on the fjord bed, such as moraines or bedrock sills. A small retreat from the grounding position can initiate the drastic retreat of the terminus to a point where it is able to stabilise again, often a considerable distance up the fjord. Grounded tidewater glaciers are usually stable, since the flux of ice through the glacier is balanced by the loss of ice by calving at the terminus. They may advance, but it is a lengthy process since it depends on the rate at which the terminus can build a shoal to stabilize on. Meier et al. (1980) analysed 33 slowly advancing tidewater glaciers in south-central Alaska and found that the advance phase of a typical glacier may last approximately 1000 years. In contrast to lengthy periods of advance, tidewater glaciers can also experience catastrophically rapid retreat (Post, 1975). Such retreats are due to an increase in the rate of iceberg calving at the terminus (Meier and Post, 1987). A full explanation of the processes controlling calving has yet to be formulated making a thorough understanding of rapid retreat difficult. Calving varies directly with water depth at the terminus. Therefore, tidewater glaciers are extremely sensitive to changes in sea level, or a retreat back from the frontal shoal. Columbia Glacier in south-central Alaska has become the type example of a catastrophically retreating tidewater glacier (Meier and Post, 1987) following an intensive programme of field investigations by the United States Geological Survey. Numerical models successfully predicted that a drastic retreat would begin in 1982 (Bindschadler and Rasmussen, 1982; Sikonia, 1982). The terminus retreated back from its shoal, lying in 22 m of water, into a proximal basin where water depth exceeded 200 m (Krimmel and Vaughn, 1987). In 1983 there was a substantial increase in surface velocity due to an increase in iceberg calving. This initiated a feedback reaction, in that glacier velocity had to increase to maintain a sufficient flux of ice to stabilize the terminus, and calving increased because ice was being supplied to the terminus at a faster rate. The enhanced ice velocities (up to 10 m d-1) were most likely due to basal sliding (Krimmel and Vaughn, 1987), although there were temporary slow-downs in motion during periods of high tide (Walters and Dunlap, 1987) resulting from the increase in back pressure exerted on the glacier terminus. The cause of the drastic retreat is not fully understood (Krimmel, 1987), especially the mechanism causing the initial increase in velocity. Meier and Post (1987) speculate that a series of negative mass balance years occurring when the glacier is in an extended forward position may trigger drastic retreat. This would lead to a thinning of the glacier and a subsequent reduction in ice overburden pressure, and, therefore, a decrease in basal friction. This would enable velocities to increase. Page 27 . I J Chapter 1 : Surge-type glaciers A number of tidewater glaciers are known to have advanced rapidly. In 1986 the Hubbard Glacier in south-central Alaska advanced 500--600 m across Russell Fjord, damming the upper part of the inlet (Mayo, 1989). The glacier was already in an active state and had been steadily advancing since the beginning of the century. A weak surge of a tributary, Valerie Glacier, provided the extra input which enabled the Hubbard Glacier to advance (Harrison, unpublished manuscript). The terminus was able to move across the fjord by pushing up glacimarine sediments at its sole. The damming of the upper fjord lasted.only !1- few months before a submarine landslide on the terminal moraine weakened the glacier terminus and water was able to break through. However, bathymetric surveys suggest that the moraine is comparatively undamaged so that Hubbard Glacier may be expected to re-advance across the fjord within a few years (Mayo, 1989). A drastic retreat of the type observed at Columbia Glacier is not thought likely. Hubbard Glacier has a strongly positive mass balance (an accumulation area ratio, AAR, of 0·95) whereas Columbia Glacier had an AAR of 0·57 when it began to retreat. 1.5 SUMMARY AND STRUCTURE OF THE THESIS 1. 5 .1 Summary This chapter has discussed the key themes of research into surge-type glaciers. In doing so, the discussion has described the areas of research that are understood and has identified problems where further work is required. The main points to emerge from this chapter are outlined below: • Surges occur on a wide variety of glacier types and are not restricted to glaciers of particular sizes or thermal regimes. • The geographical distribution of surge-type glaciers throughout the world is very specific. However, the controls on the spatial pattern of surge-:_type glaciers are not known. • Surges occur at quasi-regular intervals on affected glaciers. The triggering mechanism is not related to external forcing factors , such as climate, but is inherent within the glacier systein. • Current theories of the surge mechanism require the accumulation of water at the glacier bed. The increased amount of water can either enhance sliding of the glacier sole over a rigid bed or cause the deformation of subglacial sediments. • Surging may be one aspect of a continuum of flow regimes, ranging from short-term velocity increases on normal glaciers to rapid flow exhibited by some tidewater glaciers and polar ice streams. Page 28 Chapter 1: Surge-type glaciers 1. 5. 2 Structure of the thesis Thus far, this thesis has introduced the topic of surge-type glaciers by summarizing the current state of scientific understanding of rapid glacier flow. The remainder of the thesis will focus specifically on surge-type glaciers in the high Arctic archipelago of Svalbard. The following chapter, Chapter 2, provides an introduction to the physical environment of the islands especially relevant to glaciology and proceeds to summarize the results of glaciological research on Svalbard. Chapter 3 investigates the characteristics of surge-type glaciers in Svalbard and the possible controls on the occu1:ence of surging in the archipelago. This study is based on the statistical analysis of a large sample of ice masses, containing both normal and surge-type glaciers. The following four chapters are concerned with the field study of Bjuvbreen, a small surge-type glacier in central Spitsbergen, Svalbard. Chapter 4 describes the field area and the reasons for classifying Bjuvbreen as a surge-type glacier. This chapter also discusses the methodology used in the field and in the reduction of the results obtained. Chapter 5 presents an analysis of the geometry of Bjuvbreen and how this evolves through the surge cycle. A simple model is developed to predict when the next surge of the glacier should occur. In addition, the thermal regime of Bjuvbreen is modelled, since the distribution of temperature within the glacier is an important influence on its dynamics. The dynamics of Bjuvbreen are discussed in detail in Chapter 6. Chapter 7 describes the hydrology of the glacier and discusses the influence of water on the motion of Bjuvbreen. The final chapter of this thesis, Chapter 8, summarizes the principal findings of this investigation into surge-type glaciers in Svalbard. This chapter also identifies important areas of future research on the topic and suggests a number of methods which might be employed in these studies. Throughout this thesis, a critical rationalist approach to research has been adopted. The basic tenet of this philosophy is to develop a series of hypotheses and proceed to falsify them on the basis of the available data. Page 29 :1 I 'I 1:1 CHAPTER 2 T HE GLACIOLOGY OF SVALBARD 2.1 INTRODUCTION Svalbard is the collective name for the group of islands located in the Barents Sea between latitudes 7 4 ° and 81 ° North and longitudes 10° and 35° East (Hisdal , 1985). The total area of the islands is approximately 63,000 1cm2. Spitsbergen is the largest island of the archipelago (39,000 km2). The other major islands are Nordaustlandet, Edge0ya, Barents0ya, Kvit0ya, Prins Karls Farland, Kong Karls Land and Bj0rn0ya (Figure 2.1). Approximately 60% of Svalbard is ice covered (Liest01, in press) , although on the eastern islands of Nordaustlandet and Kvit0ya the proportion of glacierization is greatest. A range of ice-mass types, from ice caps to small valley and corrie glaciers, can be found on Svalbard. A considerable number of glaciers in the archipelago have been observed to surge (Liest01, in press) and a substantial number more are believed to be of surge-type. The largest recorded surge known anywhere in the world, the 1936-38 surge of Brasvellbreen, took place on Nordaustlandet, Svalbard. This chapter will discuss the glaciology of Svalbard, with particular emphasis on the occurrence of surge-type behaviour. Certain aspects of the physical environment of the islands, namely climate and geology, have an important influence on the distribution and style of glacier behaviour. Therefore, an outline of their general characteristics will be given in the early sections of this chapter. 2.2 THE PHYSICAL ENVIRONMENT OF SVALBARD 2. 2.1 Climate The climate of Svalbard is generally classified as maritime Arctic. Despite its northerly location the climate is surprisingly mild, due to the moderating influence of the West Spitsbergen Current. This is a northerly branch of the North Atlantic Drift which flows north along the west coast of Spitsbergen (Hisdal, 1985). The mean air temperature at Isfjord Radio on the west coast for the period 1912- 75 was -4.7 °C (Steffensen, 1982). Continentality increases away from the west coast, reflected in the higher summer temperatures and lower winter temperatures recorded at stations in inner ,, lO Sv c1l bard K vii '-'Y O /) V .. " " .. - - - ·~ R J0111~0 y" llopul b ~~·v " " ,, Figure 2.1 Map of Svalbard showing the major islands of the archipelago. The · inset shows the location of Svalbard relative to Norway and other Arctic islands. Meteorological stations are indicated by initials: IR is Isfjord Radio, L is Longyearbyen, NA is Ny Alesund and S is Sveagruva. Source: Dowdeswell (1984). Page 31 Chapter 2 : The glaciology of Svalbard fjords. The mean annual air temperature (MAAT) for the period 1957-75 at Isfjord Radio was -4·9 °C in comparison to -5·9 °C for Longyearbyen, approximately 50 km to the east. The monthly temperature distribution for Longyearbyen (1957-75) is illustrated in Figure 2.2. Temperatures generally decrease towards the north-east of the archipelago. Simoes (1990) used the few measurements recorded by expeditions to calculate a lapse rate of 0-6-1 ·0 °C per 100 km from the west coast. Major fluctuations in temperature can occur over short periods, particularly in winter (Hisdal, 1985). Increases of 20--25 °Cover 2-3 days are not uncommon. These events are caused by the passage over the islands of cyglonic systems originating from the south-west. At other times the cold Arctic Front dominates the weather by deflecting the cyclones away from the archipelago, bringing low temperatures. Summer temperatures below O °Care not uncommon. The mean number of days in a year with sub-zero temperatures at Longyearbyen is 261 (Steffensen, 1969). J a r Feb Mar Apr May Jun J ul Aug Sep Oct Nov Dec Month Longyearbyen Sveagruva Figure 2.2 Monthly mean temperature for Longyearbyen (1951- 75) and Sveagruva (1980-89) . The slightly more continental location of Sveagruva accounts for the lower winter temperatures . Compiled from data presented in Hanssen-Bauer et al. (1990). Precipitation on the islands is low and is generally less than 400 mm a-1 (Hisdal, 1985) . At Isfjord Radio the long term mean (1951-75) is 435 mm a- 1 (Steffensen, 1982). Due to its more continental location Longyearbyen has a long term mean (1957- 76) of 208 mm a-1. Precipitation may fall as rain or sno.w in any season. Hisdal (1985) recognised three areas of Spitsbergen which received greater amounts of precipitation than the rest of the island. These areas, south-west and north-west Spitsbergen and Olav V Land, are all mountainous regions, suggesting that there is a strong orographic influence on precipitation. Winter precipitation comes mostly from Page 32 Iii Ii Chapter 2: The glaciology of Svalbard the south and south-east whereas summer precipitation originates from the south-west (Hisdal, 1985). The location of weather stations operating on Svalbard is limited to low-lying coastal areas (Figure 2.1) and the record of observations at these stations is sporadic. The earliest reliable observations were made at Isfjord Radio from 1916, although they were discontinued after 1976 (Steffensen, 1982). However, the records are sufficient to demonstrate a significant change in the climate of Svalbard since the early part of this century. Simoes (1990) recognised four distinct fluctuations in MAAT since records began. From 1916-23 there was a- periqd of rapid warming, during which the mean winter temperature (December- February) increased by 8 ·c. Between 1924-56 there was a period of relatively high temperatures, which reached a maximum of - 2 ·c in 1954. Following this, temperatures dropped sharply to a minimum of - 8 ·c in 1968. Since then, temperatures increased, and from 1972 until 1986 they have remained approximately constant. Throughout these trends there have been a number of anomalous years with notable temperature reversals. Simoes (1990) also found that there was a general trend of increased precipitation during periods of increased temperature. 2.2.2 Geology The stratigraphy of Svalbard extends from the lower Proterozoic to the Quaternary and is comprised of sedimentary, igneous and metamorphic rock types. The physical relief of the islands is generally mountainous. Spitsbergen gains its name from the rugged mountains which are found on its west coast. In this area the peaks rise to between 1000-1400 m above sea level. Flat-topped mountains and dissected plateaus are characteristic of the south-eastern part of the archipelago. Newtontoppen and Perriertoppen (both 1717 m) in Olav V Land are the highest points in Svalbard. The oldest rocks in Svalbard are traditionally grouped under the name of Hecla Hoek complex. These sequences are of middle/late Proterozoic to Silurian in age. The geological composition of Hecia Hoek rocks displays a marked difference between the western and northeastern regions of the archipelago due principally to a lack of metamorphism in the northeast. Thre~ periods are recognized within the Hecla Hoek sequences. The Lower Hecla Hoek is characterised by diam1ctites and metavolcanics in the west and schists, gneisses and migmatites in the northeast. Limestone and marble are found in the Middle Hecla Hoek sequences in the west in contrast to the carbonates, quartzites and volcanics in the northeast. The Upper Hecla Hoek sequences in the west are composed of schists, psammites and quartzite. The corresponding sequences in the northeast are characterised by carbonates, sandstones, shales and diamictites. The Caledonian Orogeny during the late Silurian tectonized the western and lower Page 33 Chapter 2:- The glaciology of Svalbard northeastern sequences of the Hecla Hoek rocks. Granites were intruded into the existing sequences during the late- and post-orogenic phases (Devonian). The main phase of the Caledonian Orogeny was followed by extensive denudation of the Middle and Upper Hecla Hoek rocks during the Devonian. The main products were Old Red Sandstone, sandstones and shales which are widespread in northwest Spitsbergen. Almost continuous deposition occurred from the Carboniferous to the early Early Tertiary. Sandstones, shales and carbonates were deposited in marine and shallow marine environments. Coal seams, which are the subject of commercial coal extraction at several sites in centra1 Spitsbergen, were formed during the Carboniferous and Tertiary periods. Widespread deformation of strata in the west of the island occurred during the West Spitsbergen Orogeny of the Early Tertiary. The youngest terrestrial rocks in the archipelago are the products of Tertiary and Quaternary explosive volcanism. It is believed that the most recent eruptions took place between 130,000 and 70,000 years ago in Bockfjorden, northern Spitsbergen (Worsley, 1986). Quaternary glaciations have moulded much of the present landscape of Svalbard. There is a variety of surficial deposits covering much of the ice-free areas of the islands (Kristiansen and Sollid, 1989). These include till and morainic material, glaciofluvial deposits, tidal and estuarine sediments and weathering deposits. Surface processes are particularly effective at producing large amounts of debris, owing to the lack of any substantial vegetation cover which might otherwise provide some stability. Permafrost is a widespread phenomenon and has been found to penetrate to depths between 100 and 460 m (Liest01, 1977; Landvik et al., 1988). The geology of Svalbard is of relevance to glaciological studies of the area following the recognition that sedimentary beds can have an important influence on glacier motion (Boulton and Jones , 1979). This is particularly relevant in the case of research into the causes and dynamics of glacier surges (e.g. Clarke et al., 1984). 2. 2. 3 Quaternary glacial history During the Last Glacial Maximum (Weichselian), the extent of ice cover on Svalbard was significantly greater than at present. There was considerable debate (Elverh0i and Solheim, 1983) as to whether ice extended off the present coastline of the archipelago and formed an ice sheet in the Barents Sea shelf. It is now generally accepted that a Barents Sea Ice Sheet did exist (Elverh0i and Solheim, 1987), on the basis of evidence such as moraine ridges, submarine troughs and the relatively thin covering of glaciogenic sediments on much of the Barents Sea floor. Nevertheless, the characteristics of the ice sheet and the timing of its deglaciation are still matters of considerable debate, owing to difficulties in interpreting the marine geological evidence. Page 34 Chapter 2: The glaciology of Svalbard The extent of glaciers on Svalbard has also fluctuated at various times during the Holocene. However, the remote nature of much of the archipelago has prevented a wide spatial coverage of investigations. As a result, the Holocene glacial history of the west coast of Spitsbergen is better known in comparison to other areas of the islands. Until recently, many of the techniques used to date glacial advances in the islands have required sophisticated laboratory analyses. Methods such as radiocarbon dating (Baranowski and Karlen, 1976) and thermoluminescence (Forman, 1988) have been used in Svalbard. Recent research by Werner (1990) has resulted in the construction of a lichen growth curve for Spitsbergen which permits more readily available dating of glacial deposits. Werner (1990) constructed a lichen growth curve using remains of human activity, such as whaling settlements and anchor points for a Zeppelin airship, as control surfaces. He mapped and dated more than 70 moraine sequences in west Spitsbergen. Based on the lichen growth rates, he recognised a number of periods of glacier advance. The oldest recognisable Holocene moraines in Spitsbergen were stabilised approximately 1500 yr BP and a younger group at 1000 yr BP. Moraines dating from the last few centuries, known as the Little Ice Age, are less than 650 yr BP old, and are the most extensive. As a result, many older deposits have been obliterated and are visible only as remnant lateral moraines (Werner, 1990). Cromack (1991) has demonstrated a similar chronology based on proglacial lacustrine sedimentary sequences in north-west Spitsbergen. The marine deposits of some west Spitsbergen fjords have been investigated with the aim of determining the Holocene behaviour of glaciers terminating in tide water. Sexton et al. (1992) obtained acoustic profiles of sediments in Krossfjorden using seismic methods. These profiles were interpreted in terms of periods of glacier advance and retreat since the end of the Late Weichselian glaciation. The results obtained were in broad agreement with the chronology proposed by Werner (1990), with the Little Ice Age deposits being the most extensive. Ice core data examined for palaeodimatic information have provided a proxy record of glacier behaviour on Svalbard (Simoes, 1990). A 200 m long ice core, recovered from the Lomonosovfonna, was found to span the last 800 years. Analysis of this core, particularly of the down-core variation in o~ygen isotope composition, clearly demonstrated two distinct periods of cooling in the last 800 years (Gordiyenko et al., 1981). These cooler periods can be correlated with glacier advances during the Little Ice Age, which have been dated by lichenometry (Werner, 1990). However, there have been a number of problems associated with the interpretation of ice core data from Svalbard. Recently, Fujii et al. (1990) presented the results of their analysis of an 85·6 m long core from the H0ghetta ice dome in north-east Spitsbergen. They estimated that Page 35 Chapter 2: The glaciology of Svalbard the core represented 6000 years of climatic data, although a noticeable feature was a 4000 year hiatus from 4000 years BP to AD 1770. Fujii et al. interpreted this time gap in terms of negative mass balance over the 4000 year period. However, a number of problems in this interpretation have been pointed out by Dowdeswell et al. (1990). The latter workers suggested that dating of the core by tritium content may have been confused by melting and refreezing of the snowpack. Furthermore, Dowdeswell et al. (1990) found the 4000 year time gap to be inconsistent with palaeoclimatic and glaciological data from other ice cores and geological sources in the archipelago. Fluctuations in climate are not the only cause of variations in the position of glacier fronts in Svalbard. A number of glaciers in the archipelago are known to make exceptional advances not related to any change in the environment. These glaciers are described as surge-type. Their character and distribution will be discussed in section 2.4. The presence of surge-type glaciers can be a source of error when inferring palaeoclimate conditions from glacier advances. Punning et al. (1976) dated advances of Paulabreen between 7800 and 8500 years BP and at 650 year BP. They suggested that these advances were related to climatic coolings at these times. However, Rowan et al. (1982) believed that the advances were actually surges and, therefore, did not attach any climatic significance to the events. 2.3 CONTEMPORARY GLACIOLOGY OF SVALBARD 2. 3 .1 A short history of glaciological exploration There are references to a group of islands, which is thought to be Svalbard, in Icelandic literature dating from the twelfth century. However, it was the Dutch who made the first modem discovery of the archipelago in 1596, while they were searching for a new north-eastern sea route to the Far East. During the seventeenth and eighteenth centuries there was a considerable amount of whaling activity by British and Dutch ships off the west coast of Spitsbergen. As whale stocks became depleted, shore-based hunting and trapping became important in the nineteenth century. These early explorations of the islands were confined to coastal areas. Nevertheless, the maps and charts that resulted from these activities provided much glaciological information (Brown, 1911). The Swedish explorer Nordenskiold made the first crossing of Nordaustlandet in 1873 by sledge (Nordenskiold, 1873) and confirmed the presence of a large ice cap. The extent of glacier cover on Spitsbergen remained unknown, however, until the first expeditions to its interior were made by Conway in the 1890s (Hutchins, 1952). These expeditions provided the first information that, except for north-eastern parts of the island, the presence of glaciers was not as extensive as on Nordaustlandet. A number of Swedish (N athorst, 1909) and Norwegian (Hoel, 1919) Page 36 Chapter 2: The glaciology of Svalbard expeditions made frequent visits to Svalbard during the late-nineteenth and early rwen tieth centuries. A series of expeditions organised by the Prince of Monaco in 1906, 1907 and 1909 succeeded in mapping the whole of Spitsbergen. The maps from those expeditions contain much information concerning the historical variation of glaciers (M. Cromack, personal communication). The first detailed glaciological research was carried out by the Norwegian- Swedish expedition to Isachsenfonna, north-west Spitsbergen, in 1931. Ahlmann (1933) and Sverdrup (1933) made detailed studies of the mass balance and thermal regime of the ice cap during this visit. Ahlmann's geophysical classification of ice masses was based, in part, on wor1c carried carried out during this visit. During the 1930s and 1950s a number of Oxford University expeditions carried out work on the glaciology of Nordaustlandet (Thompson,1953). The Norwegian Polar Research Institute (Norsk Polarinstitutt, NP) began systematic mass balance measurements on Finsterwalderbreen in 1950, which continued intermittently until 1968, and on Austre Bnziggerbreen and Midre Lovenbreen in 1966 (Liest~l, 1990). Scientific investigations on Svalbard have steadily increased in frequency since the 1950s. The Polish have established a year-round research station in Hornsund, southern Spitsbergen. A considerable amount of glaciological research has been undertaken in this area (Baranowski, 1977), particularly on the tidewater glacier Hansbreen (Jania, 1986). Soviet researchers have also carried out much glaciological work, including mass balance studies (Gus'kov, 1981), radio echo sounding (Macheret, 1981) and ice core drilling (Zagorodnov and Zotikov, 1981). Some of the most comprehensive glaciological measurements in Svalbard were collected during a programme of radio echo sounding carried out jointly by the Scott Polar Research Institute (SPRI) and NP (Dowdeswell, 1984; Dowdeswell et al., 1984; Dowdeswell et al., 1986; Bamber, 1986). 2. 3. 2 Distribution and classification of Svalbard ice masses Approximately 60% of the land surface of Svalbard is presently covered by ice (Liest!Zll, in press). The distribution of glacier cover is illustrated in Figure 2.3. A clear trend of increasing glacierization from west to east is apparent. Approximately 56% of the land area of Spitsbergen is ice-covered compared to 75% of Nordaustlandet and 99% of Kvit~ya (Liest!Z)l, in press). The increase in glacierization is reflected in the change in the height of the equilibrium line which decreases from south-west to north- east and increases with distance from the coast (Liest!Z)l and Roland, in Steffensen, 1982) (Figure-2.4). Steffensen (1982) believed.that the equilibrium line altitude (ELA) reflects the pattern of precipitation in the archipelago, since the summer temperature does not vary markedly from area to area. The dominant moisture-laden winds blow Page 37 ------,r----,,-------, ----, 12°[ 15°[ 10°[ n7.1°E . 2~''E 2"/°[ - 79°N l'fllNS t.:f,lll!, r ollLAND . 78°N 77"N SVALBARD STOllFJOllDf: N '";; ~-·o 10 o 20 ~o (;o 80 100 L,., , J_L...-L-..l-L- 1____1- J___J.__.j_....J kill ?-J . 7B"N- 77'N 21i 0 E I 15"E I:) lO"E 7. l "E - -----'---~________J _____ __J___ - -------- 1-------- Figure 2.3 Map of Svalbard showing the glacierized areas in white. Approximately 60% of the archipelago is ice-covered, although the distribution of this · proportion increases towards the north-east. Page 38 Chapter 2: The glaciology of Svalbard from the south-east, hence the greater amount of ice cover on the south-eastern sides of Edge0ya and Nordaustlandet. A greater amount of winter precipitation on the southern slopes of Stor0ya, a small island off the east coast of Nordaustlandet, has produced an ice-cap with a distinct asymmetric geometry (Jonsson, 1982). In S0rkapplandet the ELA is 150 m above sea level compared to 800 m a.s.l. in Ny Friesland (Liest01, in press). There is a wide variety of glacier types on Svalbard. This ranges from the large ice caps of Nordaustlandet, Edge0ya, Barents0ya, Kvit0ya and Spitsbergen to small (less than 1 km2) corrie glaciers. There are a number of local ice domes on Spitsbergen which act as the accumulation areas for valley glaciers. Other valley glaciers are self- contained. Dowdeswell and Collin (1990) identified a number of fast-flowing ice-cap outlet glaciers and ice streams. Idunbreen, Frazerbreen, Aldousbreen and Bodleybreen, draining the southern portion of Vestfonna on Nordaustlandet, flow as ice streams in their upper reaches, being separated from one another by ridges of crevasse-free ice. In their lower reaches the flow becomes confined within rock channels. Many Svalbard ice masses terminate in tidewater. Dowdeswell (1989) estimated that 20% of the coastline (1030 km) of Svalbard is characterised by a glacier-marine interface. A number of tidewater-terminating glaciers are fast flowing, e.g. Kongsvegen in Kongsfjorden (Voigt, 1965) and Hansbreen in Hornsund (Jania, 1986). There are no ice shelves fringing the coastline of Svalbard since all glaciers which terminate in tidewater are grounded (Dowdeswell et al., 1986; Dowdeswell, 1989). 2. 3. 3 Mass balance The first mass balance measurements of Svalbard glaciers were made by Ahlmann (1933, 1935) on Nordaustlandet and Isachsenfonna. He used an indirect method which estimated accumulation as the amount of snow between two successive ice crusts, interpreted to be frozen surfaces produced at the end of the summer seasons. A mean net mass balance of 60-70 kg m·2 a-1 was inferred for the accumulation areas of the Nordaustlandet ice caps. During the 1935- 36 Oxford University Expedition, Moss (1938) measured a total accumulation of 370 kg m-2 a-1 on the summit of Vestfonna. For the same location Glen (1941) calculated a mean net mass balance of 7 5 kg m-2 a-1, using an approximation for the ablation rate. ,Direct measurements of net mass balance were made by Schytt (1964) on the summit of Vestfonna. A net mass balance of 590 kg m-2 a-1 was determined using dye-stained surfaces as marker horizons. Schytt suggested that the considerable difference between this result and those of Ahlmann and Glen was due to the misinterpretation of the snow stratigraphy and use of an overestimated ablation factor. Page 39 300 300 ~D v'O &10; I', o,, \~GJ· oO ') ,) -i_() O 0 ' , .... ::z::. ( -- \ I \ . \ \ \ l \ ', / \ -~ -- --" / \ l I \ ., \ , __ -400- , I 350 250 J 1:/ \: ~ /) j Joo 300 I I I Figure 2.4 Glacier equilibrium line altitudes (in metres) in Svalbard, derived from · aerial-photographs, Landsat imagery and direct observations. Compiled by Liest!Zll and Roland. From: Steffensen (1982) Page40 Chapter 2: The glaciology of Svalbard Norsk Polarinstitutt began systematic mass balance measurements on Finsterwalderbreen, Van Keulenfjorden, in 1950. Measurements were made bi- annually until 1968. In 1966 mass balance investigations began on Austre Brpggerbreen and a year later on Midre Lovenbreen, both in north-west Spitsbergen (Liestpl, in press). Measurements on these glaciers have been made annually since then. The importance of superimposed ice formed by the melting and refreezing of early winter snow is an important component of accumulation. Liestpl (1975) estimated that it might account for 20% of the total accumulation. Hagen and Liestpl (1990) analysed the data from Finsterwalderbreen, -Brpggerbreen and Lovenbreen and found that each glacier had negative mean specific mass balances in almost all years. The net balance was positive in only one balance year, 1986-87, due to abnormally low summer temperatures in 1987. The cumulative net balances for Brpggerbreen and Lovenbreen correspond to a loss of approximately 10% of each glacier's volume since 1967-68. The magnitude of balance loss has shown a slight decrease in recent years. Hagen and Liestpl (1990) estimate that a zero net balance would be obtained if summer temperatures were reduced by 1 °C or winter precipitation was increased by 50%. Statistical analysis by Lefauconnier and Hagen (1990) suggested that a reduction in summer temperature would be more effective. Mass balance investigations made by Soviet workers display a similar trend to the Norwegian investigations. Negative net balances were obtained for Voringbreen (Nordenskiold Land), Bertilbreen (Dickson Land) and Daudbreen on the central east coast (Gus'kov, 1981), demonstrating that the net balances are negative throughout Spitsbergen. Hagen (1988) suggested that glaciers covering higher altitudes are closer to steady-state balances than at lower elevations. Recent measurements made on Kongsvegen showed that for the 1987-88 balance year it had a net balance of - 0·05 m. This was compared to-0·50 m on nearby Brpggerbreen. Hagen attributed this variation to differences in the altitude-area distribution of the two glaciers. The mean altitude of Brpggerbreen is 310 m a.s.l. compared to 560 m a.s.l. on Kongsvegen. 2. 3. 4 Thermal regime The distribution of temperature in a glacier exerts a fundamental influence on its motion. According to Paterson (1981), the thermal regime of a glacier is determined by the input of heat from sources at the surface and at the base. Since there is considerable geographic variation in the magnitudes of these sources we can expect glaciers in different areas to have contrasting thermal distributions. The three-fold classification of glacier thermal regime proposed by Ahlmann (1933) is commonly used. Glaciers where ice, other than a shallow surface layer of 10-15 m, is at the pressure melting point are Page 41 Chapter 2: The glaciology of Svalbard described as temperate. The term "pressure melting point" is somewhat misleading since the temperature of ice is not solely dependent on pressure but also on chemical impurities (Paterson, 1981). Polar glaciers are below the pressure melting point, except for a surface layer which may respond to summer warming. An intermediate class is represented by sub-polar glaciers. These are characterised by a compound thermal regime, where most of the glacier is at , the pressure melting point but the thinner marginal areas are frozen to their bed. Hutter (1983) introduced a classification based on mechanical boundary conditions. Temperate glaciers are characterised by a perfect- slip condition and polar glaciers by a no-slip condition, with basal ice adhered to its substrate. Hutter argued that temperature measurements in and at the base of glaciers are so scarce that we cannot accurately classify glaciers in to any one distinct group. Instead, he suggested that all glaciers should be described as polythermal. However, the classification scheme proposed by Ahlmann (1933) will be adopted throughout this thesis, since its terminology is the most widely understood, applied and accepted in glaciology. Liest0l (in press) states that the majority of glaciers in Svalbard have a sub-polar thermal regime. The data to support such an assertion are limited, however. Most measurements have been made on the ice caps of Nordaustlandet. The results of Moss (1938), Hollin (1956), Schytt (1964) and Dowdeswell (1984) reveal a pattern of temperatures consistently higher on the crests of Austfonna and Vestfonna than at their margins. This supports Schytt's (1969) contention that the central areas of the Nordaustlandet ice caps are temperate and surrounded by colder peripheral ice which is frozen to its bed. The presence of basal melting at some locations was confirmed by Pfirman and Solheim (1989) who observed turbid meltwater plumes being discharged from the southern margin of Brasvellbreen into the open ocean. Sediment plumes have also been observed from satellite images of the coast around Nordaustlandet and Kvit0ya (Dowdeswell and Drewry, 1989; Bamber and Dowdeswell, 1990). Sub-polar thermal regimes have been measured directly on the Ston11yaj0kulen (Jonsson, 1982) and on Finsterwalderbreen (Nixon et al., 1985). Baranowski (1977) found that the sub-polar glaciers of Svalbard fell into a number of different categories. Many corrie glaciers on Svalbard are believed to have polar thermal regimes, because of the complete penetration of the winter cold wave through the shallow depth of ice (Liest01, in press). Page42 Chapter 2: The glaciology of Svalbard 2.4 SURGE-TYPE GLACIERS IN SVALBARD 2.4.1 Introduction A relatively large proportion of ice masses in Svalbard are believed to experience surge-type behaviour. Some estimates (e.g. Liest1<1l, in press) suggest that as many as 90% of glaciers and ice caps may fall into this category, although the number of surges actually observed is considerably lower. Nevertheless, the concentration of surge-type glaciers in the archipelago is remarkable. The largest surge known to have occurred anywhere in the world took place on Svalbard. During the 1930s, the - southern outlet of Austfonna, Brasvelloreen, surged forward 20 km along a 30 km wide front (Section 2.4.2). 2. 4. 2 Observations of surge behaviour Liest01 (in press; table 5.5-4) gives details of 69 glaciers in Svalbard that have been observed to surge in the last century. The table was compiled using evidence based on field observations, aerial photographs and satellite images of rapid changes in the nature of glacier surfaces and tennini. It is a minimum estimate of the occurrence of surge-type glaciers. Morphological evidence, in the form of distinctive moraine patterns, suggests that a substantial number more behave in this manner, but have not been observed in their active phases. Croot (1988) has identified a further 45 surge- type glaciers based on the presence of characteristic moraine patterns (Meier and Post, 1969). The significance of surge-type behaviour on Svalbard glaciers was recognised during the early stages of scientific exploration in the islands. Lamplugh (1911, p. 221) remarked that several Svalbard glaciers were "subject to spasmodic fits of rapid and tumultuous advance, alternating with longer intervals of retreat and ablation during which they become relatively stagnant." The earliest report of a surge-type glacier in the archipelago was that of Recherchebreen, Spitsbergen, made by the 1838 French-Norwegian Recherche expedition (Liest01, in press). This expedition noted that a seventeenth century map of the area, compiled -by Dutch whalers, showed the glacier lying some distance up-valley from the fjord. A river, which was named on the map, flowed from the glacier terminus to the coast. In contrast, the French-Norwegian expedition maps of the same area showed the glacier covering the entire fjord, and sketches illustrated an extremely crevassed surface (Liest1<1l, in press). Detailed information on surge events on Svalbard glaciers is scarce and incomplete. Data on ice velocities, rates of propagation of surge fronts and duration of active phases are available for only a very small number of the total population of surge-type glaciers in the islands. Liest0l (1969) and Schytt (1969) have reported the Page43 Chapter 2: The glaciology of Svalbard occurrence of surge-type glaciers in Svalbard, although neither author provided much detailed information on rates of processes occurring during the surges. Dowdeswell et al. (1991) presented data from eight glaciers for which the greatest amount of information is available. It is clear, therefore, that glacier surges, although recognised as important events on Svalbard, have received little quantitative study. The characteristics of three glacier surges, which are relatively well known, will be discussed below. Bakaninbreen A recent surge of Bakaninbreen has been studied by Drewry et al. (unpublished). These data represent one of the most detailed studies of the active phase on a surge-type glacier in Svalbard. Bakaninbreen, in central Spitsbergen, is a 17 km long valley glacier which terminates on land close to the coast of Rindersbukta at the head of Van Mijenfjorden. A marked change in the character of the glacier was noted between successive field visits in 1985 and 1986. In 1985 a ramp with a gradient of 6° was present in the upper reaches of the glacier. A limited amount of crevassing was noticeable on the crest of the ramp. Velocities recorded on the crest were on the order of 0·24 m d-1. A maximum strain rate of 1·20 a-1, recorded at the base of the ramp, indicated a significant component of longitudinally compressive flow. The behaviour of the glacier had changed considerably when it was next visited in the spring of 1986. The position of the ramp had migrated down-glacier approximately 1800 m. The ramp marked the leading edge of the surge front. Velocity measurements made from March to June showed a steady acceleration from 0·9 m d-1 to l ·4 m d-1. Between June and August 1986 there was a marked increase in the velocity recorded on the ramp. Substantial lowering of the ice surface in the upper glacier was contrasted with thickening in the lower glacier as the surge front propagated downwards. Between May 1985 and May 1986 the velocity of the surge front was 4·7 m d-1 . The average velocity of a stake located on the crest of the ramp for the period March to December 1986 was 1·76 m d-1, or 3.3 m d-1 if the period of rapid summer motion is included. Ice at the leading edge of the surge front was moving between 1 ·4 and 2· 7 times faster than ice only short distance behind the ramp, suggesting some form of kinematic wave (Drewry et al. , unpublished) . However, Weertman and Birchfield (1983) suggested a value of 4-5 times faster for well developed kinematic waves. Drewry et al. questioned the applicability of kinematic wave processes to Bakaninbreen, but concluded that the data they had recorded were inadequate to resolve the problem. The surge of Bakaninbreen was monitored from reconnaissance flights in May 1987, August 1988, July 1989 and July 1990. Between these dates propagation distances of 1·3, 1·7 and 0·4 km were estimated (Dowdeswell et al., 1991). The surge Page44 Chapter 2: The glaciology of Svalbard front had advanced between 3 and 4·7 m d-1 down-glacier over this period and was predicted to reach the terminus between July 1990 and July 1991. However, measurements made by I. Frearson (personal communication) during the summer 1990 suggested that the surge has diminished and is nearing its termination. Velocities recorded just behind the surge front were typically between 0· 1 and 0·5 m d- 1 (Frearson, unpublished data). Bakaninbreen is, therefore, considered to have had a minimum active surge phase duration of 4 years. Drewry et al. (unpublished) did not discuss in detail the likely trigger of the Bakaninbreen surge. However, they speculated that a thermal trigger similar to that proposed by Schytt (1969) was inappropriate. They tentatively suggested that water saturation of subglacial sediments leading to deformation and rapid flow may have been the cause. Recent retreat of the glacier prior to its surge revealed large quantities of unlithified sediments which were formerly overlain by the glacier. Indirect evidence for the presence of similar material beneath Bakaninbreen was reported by Drewry (1987). He compared the results of seismic and radar investigations on the glacier and the adjacent Paulabreen and found that in 50% of the comparisons, ice depths were up to 15 m deeper when measured with seismic equipment. The differences were explained by the presence of a subglacial horizon with a low seismic velocity, such as water saturated till. Usherbreen Usherbreen is a 12 km long outlet of Nordmannsfonna, an ice cap in east Spitsbergen. Its most recent surge was been studied by Hagen (1987a, b) . The precise timing of the initiation of the surge is unknown, but analysis of satellite-derived images revealed that the terminus began to advance in 1978. By 1980, the terminus had moved forward 1 km, giving an average rate of advance of 1·3 m d-1. The forward motion continued until 1985, when velocities decreased markedly and the surge tem1inated (Hagen, 1987 b ). The duration of the active phase is taken to be 8 years. Hagen (1987) calculated basal shear stresses prior to and following the surge. Both values were -comparatively low. The mean shear stress before the surge of Usherbreen was 61 kPa, compared to 160 kPa at a similar stage in the cycle on Variegated Glacier (Raymond and Harrison, 1988). After tµe surge, the shear stress on Usherbreen had reduced to 33 kPa. These values were calculated using an estimated thickness of ice based on an empirical relationship of area to thickness for Svalbard glaciers derived by Liest01 and Roland (unpublished). It is possible, therefore, that the low values of calculated shear stress were a result of an underestimation of the ice thickness. Page 45 Chapter 2: The glaciology of Svalbard Hagen (1987) did not suggest any cause of the surge. However, there is abundant sedimentary material in the glacier forefield which might be expected to extend under the present ice mass. Therefore, the enhanced deformation of unlithified subglacial till may have initiated the surge. Brdsvellbreen The largest surge ever recorded anywhere in the world took place in the 1930s on the southern margin of S(<'1rdomen on Nordaustlandet. The part of the ice cap that advanced was given the name BrAsvellbreen, Norwegian for 'rapid-growth glacier' (Schytt, 1969). The timing of the initiation of the surge is unclear. Glen and Croft of the 1936 Oxford University Expedition crossed S(<'1rdomen by sledge that year and did not report features characteristic of a surge (Glen, 1937). However, sealers working in the waters south of Nordaustlandet in 1937 reported a significant increase in the number of icebergs in the area (Vinje, 1985). In 1938 the pilot and photographer of an NP aircraft flying along the southern coastline of Nordaustlandet observed and photographed a large new glacier (Schytt, 1969). Bdsvellbreen covered approximately 500 km2 of previously ice-free ocean and had a heavily crevassed surface. The ice edge had advanced some 20 km outside its previous limit along a 30 km long front. Dege (1948, 1949) reported observations, made from a German submarine in 1944, of substantial calving at the terminus. However, the Oxford University Expedition of 1949 noted that crevassing had largely dissipated (Hartog and Thompson, 1950). The expedition also made a visit to the ice divide of S0rdomen and reported a large depression in the ice-cap, showing that considerable down-draw of ice had occurred during the surge. Based on mass continuity considerations, Schytt (1969) estimated that ice in the reservoir zone would have thinned by 70-100 m as a result of the surge. Schytt (1969) also calculated, based on tentative estimates of the mass balance of S0rdomen, that the return period between surges of BrAsvellbreen would be in excess of 200 years. Improved calculations made by Solheim (1991) confirmed Schytt's estimate, indicating that the surge interval is on the order of 500 years. Schytt (1969) suggested that a thermal instability was the trigger of the Bdsvellbreen surge. No direct evidence of the thermal regime of Bdsvellbreen was available but there were existing data relating to other part_s of Nordaustlandet. Schytt proposed that the central core of the ice cap was temperate but the outer regions were cold-based. The ice cap was, therefore, sub-polar. A surge would be triggered by temperate ice breaking through the annulus of colder ice. The mechanics of such behaviour were not fully explained by Schytt, although Holdsworth (1977) proposed a similar mechanism based on studies of the Barnes Ice Cap in the Canadian Arctic Archipelago. This ice cap has a similar thermal regime to S0rdomen and a number of its Page46 Chapter 2: The glaciology of Svalbard drainage basins have experienced periodic surges (L0ken, 1969). Holdsworth (1977) suggested that the size of the inner core of ice at the pressure melting point was controlled by the accumulation of superimposed ice, which warmed the upper surface of the ice cap. This increase in temperature allowed basal sliding and also enhanced creep deformation to take place. A surge will be initiated when bed friction (shear stress) in the cold outer region of ice is unable to restrain the interior ice moving outwards by sliding. However, Clarke et al. (1984) reported a similar thermal regime on Trapridge Glacier but did not believe that a thermal trigger was the primary cause of surges. The work of Solheim (1991) and Solheim and Pfirman (1985) has provided data which may suggest an alternative trigger mechanism. Brasvellbreen has retreated some 5 km from its maximum surge extent leaving much of the former glacier sole exposed on the sea floor. Marine geological investigations revealed that the bed is composed of overconsolidated sediments, probably till from the more expansive Late Weichselian glaciation, covered with mud-rich material deposited after the surge. Overconsolidation resulted from the compaction of unlithified sediments during the surge. Distinct sedimentary structures characteristic of ice-push features were also found. Radio echo sounding data reported by Dowdeswell et al. (1986) demonstrated that much of the bed of Brasvellbreen is currently below sea level. There is a marked step in the subglacial topography where the bed rapidly changes from 200 m above sea level to 500-100 m below sea level. This produces a non-equilibrium surface profile. The southern drainage basins of S0rfonna receive a greater amount of precipitation and are thus able to advance beyond the subglacial step and onto the sea floor. Drewry and Liest0l (1985) speculated that the ice cap would be at the pressure melting point at its lowest elevations, enabling it to slide over deformable sediments saturated by meltwater. The surge would continue until ice could no longer be supplied from the accumulation area. As a result, sliding would cease and the base of the glacier would become frozen to its bed. Other Surge-Type Glaciers in Svalbard Surges have· been reported for a number of glaciers in Svalbard (Figure 2.5). Those observed to have surged this century are listed in Table 5.5-4 in Liest01 (in press). Much of the data on these events are purely qualitative. An exception was the description of the surge of Sszire Franklinbreen, an outlet of Vestfonna, in the 1950s (Blake, 1962). During his visit to the glacier in 1957-58 the surface was heavily crevassed and had a maximum velocity in the summer of 2 m d-1. The mean annual velocity near the glacier terminus was 270 m a-1 . By 1957-58 the glacier had advanced 1 km from its position shown on maps made by Moss and Glen (1939) in 1938. The surge was accompanied by the copious production of icebergs which were discharged Page47 . I ,I I 2' E socr.J 77~r..J 1 S'E I c< -u C 10 9 ::o t.0 21' E I 2i'E l ' G I t r~ ~>.A NORDAUS TLANDET 1 0 & ~~ .ff~k, STORdYA ~ u < "'G•E'([H '\'.J 0 V .t..USTFO '-::-:;; 60 80 100 ,,,·,0 1 [KVITOYA/ I ~ I ' ,,. "J i /E't,1 - ; 20:'E I / ,; I, /I f I I i i I I I 77' :~ Figure 2.5 Map of Svalbard giving the locations of glaciers (in black) which have been observed to surge. Based on data from Liest~l (in press). Modified from Dowdeswell et al. (1991). Page48 I ' I Chapter 2: The glaciology of Svalbard into Lady Franklinfjorden. A similar event was observed in the fjord by Parry (1828) in 1827, suggesting that S0re Franklinbreen was surging at that time. 2 . 4 . 3 The long duration of the active phase on surge-type glaciers in Svalbard Surge-type glaciers typically remain in their active phase for between 1- 3 years (Meier and Post, 1969). A recent study by Dowdeswell et al. (1991) has shown that the duration of the active phase on surge-type glaciers in Svalbard is much longer relative to active phase durations on glaciers elsewhere for which data exist. This implies that - either a different surge mechanism is operating or that the same mechanism is operating at a different rate. The reasons for this difference are not yet clear since detailed information concerning surge processes on Svalbard glaciers is not available. Most of the data used in the Dowdeswell et al. comparison was drawn from northwestern North America and the Pamirs. No information was available for surge-type glaciers in other high Arctic environments, such as the Canadian Arctic archipelago. However, surge-type glaciers in those locations are likely to behave in much the same way as Svalbard surge-type glaciers. Dowdeswell et al. (1991) examined documented reports of glacier surges where the duration of the active phase was well defined. There were less than fifty glaciers worldwide for which data on the length of the active phase were available. For surge- type glaciers in areas other than Svalbard, 85% surged for less than two years. The longest reported surge was of four years duration. This occurred on only one glacier, Walsh Glacier in Alaska. In Svalbard the minimum duration of the surge phase was three to four years. This was reported for Hinloppenbreen and Osbornebreen. However, it is almost certain that these glaciers surged for longer, although a longer time series of observations was not available. Bakaninbreen, Bodleybreen and Hessbreen surged for more than five years. Fyrisbreen and Usherbreen both surged for eight years and Hyllingebreen surged for ten years. Velocities recorded during surges of Svalbard glaciers are lower than those recorded elsewhere. Therefore, mass is transferred down-gfacier at slower rates. The cause of the protracted active phase on surge-type glaciers in Svalbard is not clear. However, Dowdeswell et al. (1991) found that the length of the active phase was not dependant on parameters relating to glacier size. It is more likely that there is a difference in the physical processes of surging, or their rate of operation, between glaciers in Svalbard and those elsewhere. The glaciers of Svalbard are commonly characterised by a sub-polar thermal regime. Whilst it is not believed that a thermal instability is the primary cause of surging in the archipelago it may exert an important influence. For example, cold marginal ice may retard the flow of warmer ice in the Page49 I I ' I I 11 Chapter 2: The glaciology of Svalbard centre of the glacier. Such an effect may be detectable in lateral crevasse patterns (Chapter 3). Liest01 (1976) suggested that this was the case during the surge of Hessbreen in the mid-1970s. Recent research has demonstrated the importance of high subglacial water pressures in the surge mechanism (e.g. Clarke et al., 1984; Kamb, 1987). In Svalbard the summer melt season typically lasts from mid-June to early September. This is short in comparison with many other regions where surge activity is known to occur. It might be expected that the amount of water available at the glacier bed will be reduced. As a result, periods of high subglacial water pressures may not last long enough to be capable of promoting the continuously fast velocities recorded on surge-type glaciers in Alaska, for e£ample. Because the glaciers in Svalbard surge more slowly it will take longer to transport the mass built-up in the reservoir zone to the receiving area. The surface slope will remain steeper for longer and the ice in the reservoir zone thicker for longer. If the mechanism proposed by Clarke et al. (1984) is correct, then the prolonged higher shear stresses will continue to deform unlithified sediments allowing enhanced velocities. Alternatively, if Kamb's (1987) model is more realistic, and the shear stress remains high for a prolonged period, the stability parameter will be unable to exceed its critical value for multiple conduit instability (section 1.3.2). As a result the surge will be able to continue over a longer period. These ideas are speculative hypotheses since at present the data to test them are not available for a sufficient number of glaciers. 2.5 SUMMARY This chapter has summarized the results of previous glaciological research in Svalbard, in addition to highlighting several features of the physical environment of the islands relevant to its glaciology. Two aspects of the climate and geology can be emphasized: • the climate is characterised by low values of precipitation, and • the geological composition of parts of the archipelago is well suited for providing an abundant supply of debris to the glacier sole. The notable characteristics of the glaciology of Svalbard can be summarised as : • glaciers generally have a sub-polar thermal regime, • there is a conspicuous tendency for a substantic!,1 number of glaciers to exhibit surge-type behaviour, and • the duration of the active phase on surge-type glaciers in Svalbard is noticeably longer than on surge-type glaciers elsewhere in the world for which data on the active phase duration exist. The following chapters discuss surge-type behaviour in Svalbard in more detail, based upon data derived from remote sensing and field studies. Page SO , I 11 11 CHAPTER 3 ENVIRONMENTAL CONTROLS ON GLACIER SURGING: A STATISTICAL ANALYSIS OF SVALBARD GLACIERS 3.1 INTRODUCTION The geographical distribution of surge-type glaciers worldwide is highly specific (Section 1.1.3). On a global scale, there are noticeable concentrations of surge- type glaciers in certain mountain regions, such as the Wrangell- St. Elias, the Pamirs, Iceland and Svalbard, whereas in other areas, surge-type glaciers are rare or absent. An important characteristic of surging is its cyclical nature, implying, therefore, some repetitive instability within affected glaciers. However, the underlying factors promoting this instability have yet to be identified. A number of suggestions have been offered through earlier studies. Thorarinsson (1964) speculated that climate was important in triggering surges of Vatnajokull outlet glaciers. He suggested that these glaciers had geometries which made them particularly sensitive to variations in temperature and precipitation. Surge-type glaciers tended to be flat. Therefore, any changes occurring in their accumulation area would not be distributed along the glacier. Instead, there would be a build-up of mass in the upper reaches which would be released once a critical limit had been reached. Earthquakes were proposed as a surge trigger by Tarr and Martin (1914). They suggested that an earthquake could initiate a large avalanche which would cover a glacier and cause it to move forward. Both of the above suggestions have been discounted on the basis that surges occur on affected glaciers with a relatively regular periodicity, independent of changes in climate (Meier and Post, 1969) or seismic events (Post, 1965). Furthermore, adjacent glaciers often display asynchronous styles of behaviour, casting additional doubt on the importance of large-scale external forcings, such as climate and seismicity. 3 . 1.1 Summary of previous studies The non-random geographical distribution of surge-type glaciers has been recognised for some time. However, few studies have sought to explain this puzzling phenomenon. This section summarizes previous work on this topic. 11 11 I :II I CHAPTER 3 ENVIRONMENTAL CONTROLS ON GLACIER SURGING: A STATISTICAL ANALYSIS OF SVALBARD GLACIERS 3.1 INTRODUCTION The geographical distribution of surge-type glaciers worldwide is highly specific (Section 1.1.3). On a global scale, there are noticeable concentrations of surge- type glaciers in certain mountain regions, such as the Wrangell- St. Elias, the Pamirs, Iceland and Svalbard, whereas in other areas, surge-type glaciers are rare or absent. An important characteristic of surging is its cyclical nature, implying, therefore, some repetitive instability within affected glaciers. However, the underlying factors promoting this instability have yet to be identified. A number of suggestions have been offered through earlier studies. Thorarinsson (1964) speculated that climate was important in triggering surges of Vatnajokull outlet glaciers. He suggested that these glaciers had geometries which made them particularly sensitive to variations in temperature and precipitation. Surge-type glaciers tended to be flat. Therefore, any changes occurring in their accumulation area would not be distributed along the glacier. Instead, there would be a build-up of mass in the upper reaches which would be released orice a critical limit had been reached. Earthquakes were proposed as a surge trigger by Tarr and Martin (1914). They suggested that an earthquake could initiate a large avalanche which would cover a glacier and cause it to move forward. Both of the above suggestions have been discounted on the basis that surges occur on affected glaciers with a relatively regular periodicity, independent of changes in climate (Meier and Post, 1969) or seismic events (Post, 1965). Furthermore, adjacent glaciers often display asynchronous styles of behaviour, casting additional doubt on the importance of large-scale external forcings, such as climate and seismidty. 3 .1.1 Summary of previous studies The non-random geographical distribution of surge-type glaciers has been recognised for some time. However, few studies have sought to explain this puzzling phenomenon. This section summarizes previous work on this topic. I Chapter 3: Statistical analysis of Svalbard glaciers An early examination of the factors affecting the distribution of surge-type glaciers was made by Post (1969). He compiled data on 204 known surge-type glaciers in western North America and examined a number of attributes associated with them in a search for common factors. Although no statistical analyses were performed, Post (1969) identified several characteristics and events that were not related to surging. The wide variety of sizes and shapes of surge-type glaciers suggested that these parameters were not of primary importance. Orientation and topography were also found not to be important. On the basis of his earlier work (Post, 1964), earthquakes and other seismic activity were discounted as crucial factors in initiating surges. Post (1969) further found that the variety of climates and thermal regimes common to surge-type glaciers precluded those characteristics from being special factors influencing the distribution of such behaviour. The study pointed to a number of attributes which may have an influence on the occurrence of surging, although Post (1969) remarked that the available data were too inconclusive to draw any finn conclusions. One such possibility was anomalous geothermal heat flux beneath certain glaciers leading to an increase in the production of subglacial water. This might be caused by incipient volcanism, the presence of plutons at depth or friction along active faults. The heat generated by such sources could be transferred to the surface by groundwater convection. Post (1969) found a marked association between the presence of surge-type glaciers and the Denali Fault which runs through the Alaska Range. However, other fault valleys contained few or no surge-type glaciers. Post (1969) was the first to speculate that the nature of the glacier sole may play an important influence in surging. He compared the distribution of surge-type glaciers with geological maps but found that surging was not limited to any particular rock type. However, he noted that surge-type glaciers were rare in areas of predominantly granitic rocks, although more detailed geological mapping was required before any firm conclusions could be reached. Post (1969) went on to discuss the relationship between bedrock geology and the nature of the sub glacial material. The likelihood that different rock types could produce beds with markedly different roughness characteristics, therefore influencing the rate of basal slip (cf. Weertman, 1964), was considered unlikely to affect the distribution of surge-type glaciers. However, Post (1969) suggested that certain lithologies would be more likely to produce unconsolidated sedimentary beds. Surges would occur when these beds actively deformed (Clarke et al., 1984). Post speculated that bedrock with a high permeability would be subject to a greater amount of weathering, thus making it more suitable for the formation of sedimentary glacier beds. Lithologies with the highest permeabilities would be those situated close to faults, where tectonic movements would cause shattering of the rocks. This was suggested as an explanation for the association between surge-type glaciers Page 52 I i I i: 1 '; I II ',, ' I :.I , 11 I I 1' 1 I I,, , II Chapter 3: Statistical analysis of Svalbard glaciers and the Denali Fault. By the same argument, granitic rocks would have a low permeability, thus explaining why areas of predominantly granite contained few surge- type glaciers. Post (1969) stressed that much of the evidence used in his analysis was only of a preliminary nature. An extension of Post's· work was performed by Clarke et al. (1986), who used the glacier inventory of the Yukon Territory, Canada, as a database. Using statistical analysis, they compared populations of normal and surge-type glaciers by testing the physical behaviour of each population against several known parameters (Section 1.1.3). Clarke et al. (198-6) an_alysed 2356 glaciers, of which 151 were designated to be of surge-type. However, they were unable to isolate any obvious environmental control which might be responsible for surging (Section 1.1.3). Their analysis did suggest that long glaciers (> 15 km) had a greater likelihood of being surge-type, although the significance of this finding was not clear (Raymond, 1987). The suggestion made by Post (1969) concerning a geological control on surging was taken up by Clarke et al. (1986). They speculated that areas experiencing rapid rates of geological uplift, such as the St. Elias Mountains, were able to produce an abundant amount of debris which would promote the development of sedimentary glacier beds. Clarke et al. (1984) outlined a surge mechanism based on the deformation of unlithified subglacial sediments. However, Clarke et al. (1986) were unable to analyse this hypothesis statistically because of the lack of geological data for their study area. Clarke (1991) re-examined the correlation between length and surging in the Yukon data set. The original analysis (Clarke et al., 1986) suggested that surge-type glaciers were longer, wider and had lower slopes than normal glaciers. However, since length, width and slope are all inter-related with one another, it was not possible to establish which parameter exerted the dominant influence on the occurrence of surge- type behaviour. Clarke used multiple correlation analysis to demonstrate that the primary control was glacier length. This result had implications for theories of the surge mechanism developed by Kamb (1987) and Fowler (1989). These implications will be discussed later in thls chapter. There are two remaining studies which have examined the factors influencing the distribution of surge-type glaciers. Both of these studies have focused on glacier geometry and morphometry. Glazyrin (1978) worked with a sample of 62 glaciers (of which 69% were surge-type) from Central Asia and tried to predict their behaviour using a combination of 12 parameters. Using Bayesian statistics, glacier area was found to be the parameter most closely related to dynamic behaviour, although Glazyrin believed this was a chance association. Almost as successful at predicting surge-type behaviour were the ratios of accumulation zone area to ablation zone area, and Page 53 Chapter 3: Statistical analysis of Svalbard glaciers accumulation zone area to width of the glacier tongue. This last ratio represented the "damming" of the firn zone (Glazyrin, 1978). The second study which analysed the influence of glacier morphometry on dynamic behaviour was carried out by Wilbur (1988) using a sample of 146 glaciers in the Yukon and Alaska. He found that Budd's (1975) product of balance flux/width and slope was a poor discriminator between glacier types. Much better at characterising surge-type glaciers was glacier hypsometry. Hypsometry describes the distribution of glacier area with elevation. Wilbur found that surge-type and non-surge-type glaciers have different hypsometric curves:- 73% of the surge-type glaciers in his sample had 'bottom heavy' geometries, compared to 22% of non-surge-types. Bottom heavy geometries are characterised by a concave surface profile and a greater proportion of area at lower altitudes. Wilbur calculated the balance-related geometries for each glacier in his sample and discovered that the geometric conditions favouring surge-type behaviour commonly occurred in the upper basins of bottom heavy glaciers. If the association between glacier hypsometry and surging was real, Wilbur (1988) speculated that the ultimate control on the distribution of surge-type glaciers was geology. He asserted that glacier hypsometry was inherited from mountain geometry. Support for this hypothesis came from the fact that hypsometric curves appeared to be non-randomly distributed. Both of the studies discussed above have reached conclusions based on a certain amount of speculation. Neither have been able to identify conclusively the environmental controls involved in the surge mechanism (Raymond, 1987). Clearly, further statistical analyses of other regions containing surge-type glaciers are required to move closer to resolving the problem. However, the data available for such an exercise will probably be less detailed than that actually required. This chapter will describe the statistical analysis of a sample of ice masses on Svalbard which has the aim of elucidating factors that are common to surge-type glaciers. The methods and results , and the problems encountered, will be discussed below. 3.2 SOURCES OF DATA 3. 2 .1 Introduction For the purposes of statistical analysis it is useful to compare glaciers with contrasting dynamic behaviour within the same geographical region. The noticeable concentration_ of surge-type glaciers in Svalbard (Liest01, in press) makes the region suitable for a study of the fac tors associated with surging. Liestl2)1 (in press) identified approximately 2100 ice masses in the archipelago, of which 69 are known to have surged within the last 100 years. The true number of surge-type glaciers is likely to be Page 54 Chapter 3: Statistical analysis of Svalbard glaciers significantly greater however, since many surges are likely to have taken place unrecorded. Svalbard, therefore, is an area which will provide statistically viable sample populations of both normal and surge-type glaciers. 3. 2. 2 Selection of sample population Data for the study carried out by Clarke et al. (1986) were provided by the Canadian Glacier Inventory (CGI). This inventory contained information regarding glacier length, elevation and orientation, as well as some data on morphology and behaviour (Ommaney, 1970). A comparable inventory of Svalbard glaciers was not available. Therefore, the data required had to be collected specifically for this analysis. The principal sources of information were NP map sheets and aerial photographs. An initial aim of this study was to include the whole of the archipelago in the analysis. This would have required collecting data for the some 2100 identified ice masses. The scale of this task was realised in the early stages of the project and consequently the aims were revised. Instead of considering the entire archipelago as a sample, selected areas were chosen to form the data base. The method of selecting the sampled areas is described below. Much of the information on glacier characteristics was to be obtained from topographic map sheets. NP have produced complete map coverage of the islands at a scale of 1: 100,000. This represents 55 map sheets. The majority of the maps are produced in black and white only. On these sheets, glacier boundaries are often very indistinct, making it difficult to obtain accurate measurements of a number of glacier parameters. However, NP have published 18 map sheets in colour. The contrast in shading on these maps allows a greater distinction between glaciers and ice-free areas. As a result, measurements can be made with higher accuracy than on black and white maps. Colour maps were thus selected for sampling in order to obtain more accurate data. The distribution of the colour sheets is primarily restricted to Spitsbergen, although the northern and north eastern sections of the island are omitted from this coverage. The islands of Nordaustlandet, Edge0ya and Barents0ya are covered only by black and white maps. Of the 18 colour sheets, ten were used for the final data base (Figure 3.1). The remaining eight maps were excluded from the study because the areas they covered contained few glaciers (e.g. the two sheets covering Kong Karls Land and a further two covering the southern portion of Oscar II Land and southern half of Prins Karls Farland). Unfortunately, this selection procedure resulted in some restriction of the geographical coverage of the sample. The primary data set contained a total of 615 glaciers, which represents approximately 30% of the total glacier population of Svalbard. Page 55 Chapter 3: Statistical analysis of Svalbard glaciers 25° ~80° 78° 76° 76° 20° 25° Figure 3.1 Location and code numbers of the map sheets used to derive the primary _ data set. Page 56 Chapter 3: Statistical analysis of Svalbard glaciers 3. 2. 3 Acquisition of data The major source of data used in this study was the ten NP 1: 100,000 colour map sheets. With the exception of certain cases described below, all glaciers shown on these maps were sampled. Generally, bodies of ice less than 1·0 km in length were ignored, since most of these small ice masses are not true glaciers. Examination of several of these features on small scale aerial photographs (1: 18,000) showed that they were invariably remnant ice patches. Plateau ice caps were also excluded from the sample, principally because they lacked a distinct flow direction. Several glaciers were only partly covered by one map. These p~rtions were not sampled, unless the adjacent map was used, in which case the entire glacier could be measured using combined maps. The following data were obtained from the topographic maps: glacier length, width, area, orientation of the upper and lower glacier, and maximum and minimum elevations. The presence of tributaries was also noted. Glacier length, width and area were measured with a digitizer. The measurements were considered accurate to 0· 1 km. Length was measured along the imaginary central flow line of a glacier. The width was measured midway along the glacier's length, except when this was considered to be unrepresentative. Area measurements refer to the trunk of the glacier and do not include major tributaries which were catalogued separately. The digitized measurements are considered to be a fair representation of the true glacier dimensions, following favourable comparison of digitized values against published data for a number of ice masses. The frontal positions of glaciers do not remain constant through time and, thus, there is a problem in ensuring that the dimensions of the sampled ice masses are measured with respect to a constant date. This was possible with reasonable certainty only in the case of tidewater-terminating glaciers, which have historic frontal positions since 1936 annotated on recent map editions. The NP maps were originally compiled from 1936 and 1938 oblique aerial photographs and have been updated as more photographs have b.een taken (B. Lytskjold, personal communication). They have not, however, been redrawn. The positions of the land terminating glaciers are, therefore, taken to be those in 1936/38. The compass orientation of the upper and lower portions of each glacier was noted. The upper and lower portions were taken as an alternative for the accumulation and ablation zones. These zones could not be established since the precise altitude of the equilibrium line was unknown for the majority of sampled glaciers. The compass circle was divided into octants, i.e. N, NE, E, etc .. The elevations above sea level of the head and terminus of each glacier were derived from the maps. The contour interval of the maps was 50 m, therefore visual Page 57 Chapter 3: Statistical analysis of Svalbard glaciers interpolation was used to obtain heights to the nearest 10 m. In the case of glaciers which had their source region in the headwalls of steep valleys and carries, a certain amount of judgement was required to establish where the glacier actually began. Using these data, the slope of the glacier (a) was calculated from : a= tan-1[(zhi - z10 )/L] (3.1) where zhi is the elevation of the highest point, z10 is the elevation of the terminus and L is the length of the glacier. It was not possible to determine the elevations of the accumulation and ablation zones, and hence their slopes, since data on the equilibrium line altitude was missing for most glacier~ in the sample. For each glacier, a record was made of the presence of tributaries. If the tributaries were considered to actively contribute ice to the trunk, data were recorded for these separately as described above. An indication of the contribution provided by a tributary was provided by medial moraines annotated on the maps. The bedrock geology beneath each glacier was noted. This information was derived from two 1:500,000 geological maps (Flood et al., 1971; Hjelle and Lauritzen, 1982). At such a large scale the resolution of the geology over small areas was not as accurate as would be desired. However, the geological mapping of Svalbard at 1: 100,000 scale is not yet complete and, at present, only one sheet is available at the smaller scale (C9G covering the Adventdalen area). In the process of determining the subglacial lithology, it was assumed that the rock type shown at the glacier margins, i.e. the uppermost strata, was the same as that beneath the glacier. This might not always be the case. For example, a glacier situated in an incised valley or corrie may cut through the upper strata and actually rest on a lower rock type. This factor could be accounted for, because the thickness of individual stratigraphic units was given on the map legends. Therefore, in cases where the depth of the valley exceeded the thickness of uppermost strata, the geological cross-sections were used to identify the rock type more likely to be underneath a glacier. 3. 2. 4 The surge_ index The data recorded for each glacier were completed by assigning a value representing the likelihood that a particular glacier is of surge-type (the surge index). Clarke et al. (1986) used a six point surge index which described various degrees of certainty that a glacier was surge-type. In more recent work this index has been modified. Clarke (1991) simplified the six point index to a dichotomous scheme which classified glaciers as either surge-type or normal. Wilbur (1988) used the same six point scheme but included in the type 4 category, glaciers which have been observed to surge but whose surges were minor events. Thus, the type 5 category was reserved for glaciers which had experienced high magnitude surges. Eventually, Wilbur reduced the Page 58 1, Chapter 3: Statistical analysis of Svalbard glaciers index to a threefold classification comprising of normal, intermediate and surge-type categories. In the present study, it was considered that the differences between many of the classes used by Clarke et al. (1986) were too indistinct. Therefore, a modified version of their index, comprising four classes (0-3) was defined. The surge index (i) was thus classified as follows: i = 0: most likely to be normal (no features diagnostic of surging) i = l: possibly surge-type (1- 2 surge-type features) i = 2: probably surge-type (> 2 features, and/or historical report of surging) i = 3: most likely to be surge-type ('contemporary' observation). Several features were considered to be an indication of surge-type behaviour. These features included looped moraines, anomalous and/or widespread crevassing, pronounced ice ramps, stagnant ice (Meier and Post, 1969) and potholes (Sturm, 1987; Sturm and Cosgrove, 1990). The principal documentary sources of information used in the compilation of the surge index were Liest01 (in press) and Lefauconnier and Hagen (1991). Further references to the behaviour of individual glaciers were contained in scientific papers, theses and historical reports. However, most of the glaciers in sample population have not featured in the literature on Svalbard. Therefore, a substantial amount of information was obtained during a visit to the aerial photograph archive at Norsk Polarinstitutt, Oslo. The NP aerial photograph archive contains images of Svalbard from the following years: 1936, 1938, 1948, 1956, 1960, 1961, 1966, 1969, 1970, 1971, 1977 and 1990. The first four missions obtained oblique photography. The remaining photography is vertical, at scales ranging from 1: 15,000 to 1 :50,000. The whole of Svalbard was covered by the 1936 and 1938 imagery. For other years, photographic coverage is restricted to certain areas of the archipelago. Each glacier in the sample population was thus photographed in 1936 or 1938 and, in general, on at least four other occasions. Glaciers in the sample were assigned a surge index following an examination of the accessible data sources. Each glacier was examined on the available aerial photography for the-presence of surge-type features. For a number of the sampled ice masses, information concerning their behaviour was obtained from various reports and publications. For example, Liest0l (in press) lists 69 glaci_ers in Svalbard known to have surged in the last one hundred years. Some of the documented surges are attributed to historical reports from the latter half of the nineteenth century. However, it cannot be certain that these reports were actually describing surges, especially when this period represented the Neoglacial maximum of many glaciers on Svalbard (Werner, 1990; Cromack, 1991). In addition, Croot (1988) presented an expanded inventory of surge-type glaciers in the archipelago which included 45 ice masses not appearing in Page 59 11, Chapter 3: Statistical analysis of Svalbard glaciers Liest0l's list. All the additional entries were based on the presence of looped moraine patterns and composite terminal moraines. Croot (1988) stated that composite terminal ridges were associated only with surge-type glaciers. However, no statistical evidence was provided to support this statement. Therefore, Croot's interpretation of the dynamic behaviour of the additional glaciers in his inventory was not taken as given. Ice masses described in the above sources as surge-type were cross-examined using the photographs for confirmation of their dynamics. This ensured that the criteria used to assign the surge index were constant for each glacier. 3.3 THE PRIMARY DATA SET AND GEOGRAPHICAL ANALYSIS This section discusses the initial statistical analyses of the primary data set of Svalbard glaciers. The initial analyses involved calculating the surge probability, firstly for the entire sample population and then for individual geographical areas. Much of the methodology is taken from Clarke et al. (1986), who conducted a similar study in the Yukon Territory, Canada. The probability statistics are based on the classical work of Laplace (Stuart and Ord, 1987). The general expression for the probability of an event, x, occurring in sample population, N, is Px = n:xfN, where nx is the frequency of x. If the probability of an event, x, is certain, thenpx = 1 (or 100%). 3. 3 .1 Surge probability for the primary data set The primary data set was composed of 615 glaciers sampled from ten map sheets. The surge probability for this sample population was calculated from the surge index data noted for each glacier. The number of glaciers in each i class is denoted ni. The probability that a particular glacier is 'i-type ', Pi, was found from Pi = nJN, where N is the total number of glaciers in the data set. The surge index (section 3.2.4) is a qualitative description of the likelihood that a glacier is of surge-type. It was, therefore, necessary to quantify this index to provide a measure of the probability that a glacier surges, given that it is i-type, Psli. This was achieved by estimating the number of incorrect identifications associated with each i category. This estimate is subjective and was based on experiences obtained during the original classification. Therefore, this procedure cannot be mathematically defended. Furthermore, the estimate will most likely vary if another person performed the same task with similar data. However, the assigning of surge index values was carried out solely by myself and I am confident that the values of Psli obtained are reasonable. If each surge-type glacier in the sample was correctly identified, then Psli = 3 = 1 and Psli ~ 3 = 0. However, given the large sample population size it was likely that a number of glaciers were assigned to incorrect i categories. For the data set discussed Page 60 Chapter 3: Statistical analysis of Svalbard glaciers i 11i Pi Psli 11sli 0 393 0·639 0-051 20 1 141 0-229 0-894 126 2 26 0-042 0-923 24 3 55 0-089 0-980 54 Totals 615 1·000 224 Table 3.1. Surge index distribution and probability scheme for the Svalbard primary data set. Th~ surge probability for the data set is 36·4%. here, I was confident that all but one of the glaciers I had identified as type 3 (section 3.2.4) were surge-type. Therefore, Psli = 3 "" 0-980, or in other words, if a particular glacier is classified as type 3 there is a 98% probability that it is surge-type. Similarly, for type 2 glaciers I estimated that only two were wrongly classified, giving Psli = 2"" 0-923. For type 1 glaciers there was greater scope for error because of the increased ni and the lack of corroborating evidence suggesting surge-type behaviour. In this group I estimated 15 incorrectly classified glaciers. Thus, Psli = 1 "" 0·894. The largest errors were associated with type O glaciers. Again, this was partly a result of the increased ni. It was also a reflection of the long quiescent period length (> 100 years) of many surge- type glaciers in Svalbard (Dowdeswell et al., 1991) (section 2.4.3 and section 5.4.3). For example, a glacier with an extended quiescent phase may not develop surge-type features over the period for which aerial photography was available. Thus, it will have escaped identification as a surge-type glacier during the analysis of the photographs. For the type O category I estimated 20 possibly incorrect identifications, giving Psli = o"" 0-051. Miscl.a.sstficdions in types 1) t. ctl'lcl 3 were ii'\ -1:he o{ir~cti-or) a,f- type 0. Once Psli had been quantified, it was possible to calculate the number of surge- type glaciers in each category, nsli, from the product of ni and Psli · The surge probability for the primary data set was computed from the expression: 3 Ps= L PsliPi i=O (3.2) Table 3.1 shows the calculated values of ni, Psli and nsli for the primary data set. In the sample population, 224 glaciers were estimated to be sµrge-type. This represents a surge probability for the sample of 36-4%. 3. 3. 2 Geographical analysis The purpose of the geographical analysis was to indicate regions where surge- type glaciers are concentrated. In this case, regions were taken to be the ten map sheets used for sampling. By using discrete units, contrasts between areas can be illustrated Page 61 II I I Chapter 3: Statistical analysis of Svalbard glaciers Map Sheet nm nslm (%) Pslm (%) A6 Krossfjorden 56 7.9 15·8 A7 Kongsfjorden 36 16·3 45.3 B8 St. Jonsfjorden 56 26·2 46·8 Bll Van Keulenfjorden 76 22·0 28·9 C7 Dicksonfjorden 72 27·4 38· 1 C9 Adventdalen 98 40·0 40·8 ClO Braganzavagen - 98 36·7 37.4 - Cll Kvalvagen 50 13 .7 27•4 C 12 Markhambreen 42 19·5 46-4 D9 Agardhfjellet 31 13 ·1 42·3 Table 3.2. Surge probability statistics for the primary data set arranged by map sheet. The differences are portrayed graphically in Figure 3.2. with differences in shading. Clarke et al. (1986) chose to use drainage basins in their analysis of the spatial variation in surge-type glacier concentration. However, they did not elaborate on which order of drainage basin was chosen for sampling. In the analysis of Svalbard glaciers, individual map sheets provided a more convenient method of spatial discrimination, since major drainage basins often crossed a number of maps, some of which were produced in black and white and, thus excluded from the analysis (section 3.2.2). Figure 3.1 shows the distribution of the maps which were sampled. Map sheets were defined by the subscript m. The i-distribution of each sheet, n ilm, was determined and from this the probability that a particular glacier in map area, m, is i-type was calculated from Pilm = n ilmlnm. This enabled the surge probability on each map to be computed from: 3 Pslm = L, Pilm P sli i=O (3 .3). The map surge probabilities were expressed as percentages. Map sheets with higher than average concentrations of surge-type glaciers were those with surge probabilities greater than P s for the primary data set, i.e. 36·4%. Table 3.2 lists values of P slm in ranked order. The values range from 15·8% for sheet A6 (Krossfjorden) to 46·8% for sheet B8 (St. Jonsfjorden). The spatial variation of Pslm was represented graphically with the use of different shading of sampled areas (Figure 3.2). From the above computations, it was found that three map sheets had particularly high concentrations of surge-type glaciers (Figure 3.2) . These clusters are located in sheets A7, B8 and C12. High concentrations are also found in maps C9 and D9. The area with the greatest deficiency of surge-type glaciers was sheet A6. Low Page 62 I I Chapter 3: Statistical analysis of Svalbard glaciers ,50 20° 25° 78° 76° 15° 0 20 30 40 , 45 % Figure 3.2 Geographic variation in the concentrations of surge-type glaciers in the sampled map sheets. The degree of shading represents calculated values of Pslm, with the dark shades indicating areas where the probability of · surging is particularly high. The inset shows the code numbers of the individual map sheets. Page 63 Chapter 3: Statistical analysis of Svalbard glaciers concentrations were also found in maps B 11 and Cl 1. Thus, the clustering of surge- type glaciers observed at the global scale is reflected at the regional scale in Svalbard. The sample size was not constant between the various map sheets (see nm in Table 3.2). To test the null hypothesis that the i-distribution of each map sheet did not differ significantly from that of the primary data set, x2 calculations were made. At the 95% significance level, only map B 11 fell below the rejection limits. The x2 values for all other areas exceeded the critical limit at this level. Therefore, the null hypothesis is rejected and the differences in i-distribution between these map areas and the primary data set can be attributed to non-random variation with 95% certainty. 3.4 MORPHOMETRIC AND TOPOGRAPHIC CONTROLS This section examines various morphometric and topographic parameters associated with each glacier in the sample population to determine if any factors are related to the occurrence of surging. The analyses were essentially concerned with determining the probability of glacier surging given a variety of environmental characteristics. A deductive approach was adopted, in which a number of hypotheses were tested. In this way, factors common to surge-type glaciers could be identified. 3. 4 .1 Length analysis The purpose of this section is to examine the influence of glacier length on surge probability. Clarke et al. (1986) found that surge probability increased with glacier length for their sample population in the Yukon, although the glaciological significance of this discovery was not apparent (Raymond, 1987). A possible explanation was offered by Wilbur (1988) . He found that surge-type glaciers in the Yukon and Alaska have characteristic hypsometric geometries (section 3.2.1). Wilbur argued that long glaciers are not able to readily alter their hypsometry. Therefore, once a surge-type hypsometric curve has been reached, long glaciers will remain in this state, as opposed to short glaciers which may be able to switch in and out of surge-type hypsometries. The p_resent analysis tested the hypothesis that, in the Svalbard data set, long glaciers have an increased probability of being surge-type. Ice masses in the Svalbard sample population had lengths, L , in the range 0-9 < L ~ 35 km. These glaciers were organised into one of ten length bin s defined by various length limits. The probability that a glacier will fall into bin /, pz, was calculated from pz = nz JN, where nz is the number of glaciers in that length bin. Figure 3.3 illustrates that glaciers have the highest probability of falling into bin / = 3 (2 < L ~ 3 km). Page 64 ------------- - -------,-,,. Chapter 3: Statistical analysis of Svalbard glaciers Next, the influence of glacier length on the probability of surging was examined. The number of i-type glaciers in each length bin, nil/, was established and, given this, the probability that a glacier in bin l is i-type was found from Pill, = nil!ln1. 30 ~ ~ 20 g :.0 co .D e 0.. 10 0 2 3 4 5 Length Bins 6 7 • a A 8 Primary data set Subset T Subset NT 9 10 Figure 3.3 Distribution of glacier lengths for the Svalbard sample population. Plotted values represent the probability that a glacier has certain length limits. Results are shown for the primary data set and subsets T (tributaries) and NT (glaciers without a tributary). The data from subsets T and NT are discussed in section 3.3.5. The glacier surge probability for each length bin was then computed using the expression: 3 Psll = L, Pill Psli i = 0 (3.4 ). If there is no relationship between glacier length and surging, then Psi/ should be roughly equal for each length bin. However, as Figure 3.4 illustrates, this is clearly not the case. Before discussing the general implications of the result, it is worth explaining the exceptionally high value obtained for l = 1. This length bin contains glaciers in the range O < L :::;; 1 km. During the process of extracting data from the map sheets, a decision was taken· to exclude glaciers with lengths less than 1 ·0 km from the sample population, on the basis that most of these ice masses appeared to be remnant ice patches (section 3.2.3). Nevertheless, two glaciers with lengths just slightly less than one kilometer managed to become included in the primary data set. Moreover, both glaciers were classified as surge-type. Thus, glaciers in the length bin l = 1 would, on this basis, have an almost 100% probability of being surge-type. This is unlikely to be the case in reality, since a number of small glaciers and perennial ice masses observed Page 65 Chapter 3: Statistical analysis of Svalbard glaciers during the analysis of the aerial photographs were apparently normal. Therefore, this result will be ignored in the discussion of the influence of glacier length on surging. Returning to Figure 3.4, and beginning at l = 2, we observe an almost monotonic upward trend in the probability of surging with increasing glacier length. 100 80 ~ ~ 60 -~ ]5 ro .D 40 e Cl.. 20 0 0 2 3 4 • Primary data set -o-- Subset T 5 Length Bins 6 Subset NT 6 7 8 9 10 Figure 3.4 Influence of glacier length on the probability of surging. Results are shown for the primary data set and subsets T and NT. The results from subsets T and NT are discussed in section 3.3.5. For the primary data set, there is an almost monotonic increase in surge probability with glacier length, beginning at l = 2. Length bin l = l has an unusually high surge probability because only contains two glaciers and both were designated surge-type. Vaiues of Psi/ range from 20·4% for l = 2 to 75· 1 % for l = 10. The trend is disrupted only by a small drop in the value of Psi/ between bins l = 4 and l = 5. This trend suggests that, in the sample population, excluding bin l = l, long glaciers have a greater probability of surging than short ones. The analysis does not reveal distinct peaks of Psll . This may have been the case if short glaciers surge by one mechanism and long glaciers by another. :rhe results obtained using the Svalbard data set are similar to those obtained by Clarke et al. (1986) for the Yukon glacier population although Post (1969), using a much smaller dataset, found no qualitative relationship between glacier length and surging. Weertman (1969) demonstrated theoretically th·at surges were more likely in large glaciers (roughly those > 10 km long). This was predicted from his theory of surging based on water lubrication and the drowning of bed obstacles. The relationship between glacier length and surging may be a product of the sampling process. Clearly, surge-type features associated with long glaciers(-> 6 km) were unlikely to go unrecorded, although it is conceivable that similar characteristics on Page 66 Chapter 3: Statistical analysis of Svalbard glaciers short glaciers were missed during observations. However, if this was this case, we would expect at least 75· 1 % of the glaciers in the sample population to be surge-type. This figure is almost double that calculated from the original data. It is unlikely that such a considerable number of short surge-type glaciers went unrecorded. Therefore, it is concluded that length does have some influence on the probability of surging. 3. 4. 2 Does glacier length influence the geographical distribution of surge-type glaciers? If the conclusion reached in the preceding section is correct, it suggests that the concentration of surge-type glaciers 1n certain geographical areas could be a function of the distribution of long glaciers. To test this hypothesis, the following analysis was carried out. Employing the notation used earlier, n11m is defined as the number of glaciers in length bin I in map sheet m. If nm is the number of glaciers in a given map sheet, then the probability that a particular glacier in map m is in bin I is Pllm = nz1m !nm . The length-predicted surge probability for each map sheet can, thus, be calculated from: 10 P ;Im([) = L,. P /Im P sll l=l (3.5). Table 3.3 lists the values of length-predicted surge probabilities for each map sheet. It can be seen that the variation in p* slm{l) between the different areas is not great. The calculated values range from 31-4% for sheet C9, which has a large proportion of short glaciers, to 40-7% for map Cl 1, which has a number of long glaciers (Figure 3.5) . The length-predicted surge probability for the primary data set was 33-3%. The /- distribution for each map sheet was analysed using the x.2 statistic. This tested the null hypothesis that the /-distribution for a particular map sheet did not differ significantly from that for the primary data set. At the 95% significance level, sheets A6, Cl 1, C12 and D9 had different /-distributions from the primary sample. However, the above map sheets also differed from the primary data set for different reasons. Sheet A6 has a higher number of glaciers in bin l = 4 and a deficit in bin l = 5, Cl 1 has slightly more glaciers in bins l > 6 and C12 has a greater proportion of mid-length glaciers. Map D9 is characterised by a noticeable peak of glaciers in bin l = 9 (Figure 3.5). Maps with high x2 percentile values were not necessarily those with cqncentrations of surge-type glaciers. For example, sheet D9 is an area with an apparent deficit of surge-type glaciers (Figure 3.2), despite its /-distribution. In addition, map B8 has a high length- predicted concentration of surge-type glaciers but x2 calculations do ·not support the hypothesis that it has a significantly different /-distribution from the primary sample. Figure 3.6 illustrates the geographical distribution of p* s lm(l) · This figure shows that Page 67 40] SheetA6 40 Sheet A7 30 30 20 20 10 10 0 0 1 2 3 4 5 6 7 8 9 10 12345678910 40 Sheet B8 40 Sheet B11 30 30 20 20 10 10 0 0 0 12345678910 0 1 2 345678910 40 40 Sheet C7 Sheet C9 30 30 20 20 10 10 0 0 012345678910 0 1 2 345678910 40 40 Sheet C1 o Sheet C11 30 30 20 20 10 10 0 0 0 1 2 3 45678910 0 1 2345678 9 10 40 40 Sheet D9 Sheet C12 30 30 20 20 10 10 0 0 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Figure 3.5 Distribution of glacier lengths for individual map sheets. The horizontal · axes indicate the length bins. The vertical axes represent the probability that a glacier in a given map sheet falls into a certain length bin, Pllm· Page 68 I 1 ]I 111 I 20° 78° 76° 15° 20° 30 32 34 36 38 Figure 3.6 Geographic variation of length-predicted surge probabilities (p* slm(l)) for the primary data set. Long glaciers were shown to have a greater probability of being surge-type (section 3.4.1). The lengths of glaciers · on each map sheet were analyzed and the data used to predict the distribution of surge-type behaviour. Comparison of this figure with Figure 3.2 shows that the length-predicted pattern of surge-type glacier distribution does not match the observed pattern. This difference was confirmed by a low Spearrnan's rank correlation coefficient between the observed and length-predicted surge distributions. Page 69 11 Chapter 3: Statistical analysis of Svalbard glaciers Map sheet P*slm(/) % r* simm % P slm % A6 35.35 45 15·8 A7 34·17 133 45.3 B8 38·23 122 46·8 Bll 34-90 83 28 ·9 C7 34.93 109 38·1 C9 31·37 130 40·8 ClO 35.52 105 37.4 Cll 40·72 67 27 ·4 - C12 38·85 119 46·4 D9 35.92 118 42·3 Table 3.3 Length-predicted surge probabilities for the primary data set arranged by map sheet. The right-hand column gives the surge probability for each map sheet as a comparison. there is some re-organisation in the distribution of 'surge clusters' when length is used to predict probabilities, compared to the actual probabilities (Figure 3.2). The influence of length on the surge probability can be removed by computing the probability ratio: (3 .6) . According to Clarke et al. (1986), this ratio provides a measure of the difference between the observed surge probability and the length-predicted surge probability. These values are compared in Table 3.3. The ratios range from a high of 133% for map A7 which has a number of short glaciers which are surge-type, to a low of 45% for A6 which has several long glaciers which are not surge-type. W hen the ratios were represented by shading (Figure 3.7) there was a redistribution of the areas where surge- type glaciers are clustered, relative to Figure 3.6. This re-organisation produced a broadly similar pattern to that found when original values of Pslm were plotted (Figure 3.2). Thus, the areas with the highest concentrations of surge-type glaciers , calculated with the length influence removed, are A7 and C9. Surge probability also remains high in map sheets B8, C7, ClO, C12 and D9. If the various measures of surge tendency for each map are ranked (Table 3.4), the same map sheets appear in the top five of both Pslm and r* slm, although there is a slight difference in the order between the two lis ts. The Spearman's rank correlation coefficient for Pslm and r\1m(l) is 0·867, which demonstrates that these two measures give almost identical results. In contrast, the rank correlation coefficient for Pslm and p* slm(l) is only 0·091. This value indicates that thereis only a very weak correlation between the actual and length-predicted surge probabilities for each map sheet. Page 70 20° 25° 78° 76° 20° 0 50 75 100 125 % Figure 3. 7 Geographic variation of surge probabilities for the primary data set. The length influence has been removed by computing r* slm(l)· Comparison . of the above figure with Figure 3.2 shows a good agreement. The similarity between the two figures was confinned by a high Spearman' s rank correlation coefficient. This close agreement suggests that length does not influence the spatial distribution of surge-type glaciers. Page 71 Chapter 3: Statistical analysis of Svalbard glaciers Rank * * Pslm P slm(l) r slm(l) 1 B8 Cll A7 2 Cl2 C12 C9 3 A7 B8 B8 4 D9 D9 Cl2 5 C9 ClO D9 6 C7 A6 C7 7 ClO C7 ClO 8 Bll Bll Bll - 9 Cll A7 Cll 10 A6 C9 A6 Table 3.4 Various measures of surge probability, arranged by map sheet and ranked in order, with greatest first. Comparison of the columns headed Pslm and r* slm{l) shows that the same map sheets occur in the top five of both, although the order is slightly different. A Spearman' s rank correlation coefficient of 0-867 for these two columns indicates statistically that they are very similar. The conclusion of the above analysis is that length has some influence on the geographical distribution of surge-type glaciers in the Svalbard sample population. However, length cannot fully explain the concentration of surge-type glaciers in certain map sheets. 3.4.3 Influence of tributaries It was established in the above section that glacier length is not the sole factor responsible for the pattern of surge-type glacier distribution in Svalbard. In this section, an analysis of the influence of tributary glaciers on the probability of surging is made. Tributary glaciers were recorded as separate entries in the data set if they were considered to be actively contributing ice to the trunk into which they flowed. Glaciers which merged with other glaciers in their lower reaches and retained discrete flow units, e.g. Bakaninbreen and Paulabreen, were not considered tributaries. Using these data, two subsets of the original sample population were defined. Subset T contained those glaciers which actively contributed ice to a trunk ~hannel. Subset NT was composed of those glaciers that were not tributaries of another ice mass. The i- distribution for each subset and their surge probability statistics were determined using the same method that was applied to the primary data set (Table 3.5}. This analysis revealed that glaciers in subset NT have a greater probability of being surge-type (p}'IT = 39-3%) than glaciers in subset T (psT = 31 ·5%). These values compare with Ps = 36-4% for the primary data set. Page 72 Chapter 3: Statistical analysis of Svalbard glaciers Pi Psli llsli Subset NT 0 235 0·606 0·051 12 1 98 0·253 0·894 88 2 20 0·052 0·923 18 3 35 0·090 0·980 34 Totals 388 1·000 152 Subset T 0 158 -0·696 0·051 8 1 43 0·189 0·894 38 2 6 0·026 0·923 6 3 20 0·088 0·980 20 Totals 227 1·000 72 Table 3.5 Surge index distribution and probability scheme for subsets NT and T of the Svalbard data set. Subset T contains tributary glaciers. Only glaciers which actively contribute ice to another glacier are included in this category. The probability that a glacier in subset NT is surge-type is 39·3%. The probability that a glacier in subset T is surge-type is 31 ·5%. The analysis of surge probability statistics for subsets T and NT is not straightforward because glaciers in either set may influence the surge behaviour of glaciers in the other. For example, surges of tributary glaciers can be triggered internally or be induced by a surge of the trunk glacier, e.g. Osbornebreen (Dowdeswell et al., 1991). It is conceivable also, that a surge of a tributary can trigger a surge in its trunk. The analysis is further complicated because the /-distribution will differ for the two subsets. In Section 3.3.3 it was demonstrated that the probability of surging increases with increasing glacier length. The influence of glacier length on the surge probabilities of the two subsets was analysed statistically. For each subset, the probability that a glacier would fall in a certain length bin, p1T and p1NT, was fou!1d. The results are shown in Figure 3.3. The distribution of length bins for both subsets is similar to that for the primary data set. The main difference is that tributary glaciers (subset T) tend t~oncentrated in the medium length bins (/ = 3 and 4). In addition, there are fewer long glaciers in subset T compared to both the primary data set and subset NT. The approximate mean length of ice masses in each subset was computed. This demonstrated that tributaries are shorter than trunk glaciers ( £T = 4·6 km compared with £NT = 5·6 km). However, x2 calculations suggested that neither subset was significantly different in terms of /-distribution from the primary data set. Clarke et al. (1986) used the following expression: Page 73 Chapter 3: Statistical analysis of Svalbard glaciers 10 *T ~ NT T P s(/) = L,i P sl/ Pi l = I (3.7) to estimate the magnitude of the length influence on the surge probability of subset T. For the Svalbard data set, this calculation yielded a value of p* srzl = 38-7%. The actual surge probability was lower, at PsT = 31 ·5%. It is evident that the length-predicted surge probability exceeds the actual surge probability. Therefore, it is concluded that length does not have an influence on the probability of surging for either subset. This is confirmed when the length analysis is repeated for subset NT. Figure 3.4 shows a good similarity between the plotted values of Ps1zNT and Psll for the primary data set. Glaciers in subset NT have been shown to have a higher probability of being surge-type than those in subset T. Thus, the geographical concentrations of surge-type glaciers may be a product of the distribution of glaciers from subset NT. To address this possibility, estimates of PslmNT (actual and length-predicted) were obtained for each map sheet. From these estimates, the probability ratio r* slmNT was obtained for each unit. If the various measures of surge probability for subset NT are ranked (Table 3.6), the ordering is in close agreement with that for the primary data set. Spearman's rank correlation coefficients for the equivalent measures in subset NT and the primary data set were all greater than 0-8. The high correlation coefficients indicate that subset NT is, statistically, very similar to the primary sample population. It is concluded, therefore, that the presence of tributary glaciers does not have a notable influence on the probability of surging or the distribution of surge-type glaciers. 3 . 4. 4 Influence of elevation and slope The analysis of factors affecting the distribution of surge-type glaciers was continued by considering the influence of glacier elevation and slope on surge probability. Two hypotheses were tested. The first was that the elevation, or elevation range, of an ice mass was important in the surge mechanism. This could be due, for example, to the influence of altitude on glacier thermal regime or to mass balance effects. The second was that surge-type glaciers have certain slope requirements. Basal shear stress is primarily sensitive to variations in ice surface slope (Paterson, 1981). Meier and Post (1969) suggested that a surge is initiated when basal shear stress reaches and exceeds some critical value. The importance of critical shear stresses to surge initiation has also been discussed by Robin and Weertman (1973) and McMeeking and Johnson (1986). If certain shear stress conditions are important, surge-type glaciers will tend to have steep surface slopes as they approach their active phase. The analysis was conducted using the primary data set. Slopes were calculated using equation 3.1 (section 3.2.3). Due to the lack of information concerning the elevation of the equilibrium line on the majority of Svalbard Page 74 Chapter 3: Statistical analysis of Svalbard glaciers Rank PslmNT p* slm(l)NT * NT r slm(l) 1 B8 C12 C9 2 Cl2 Cll C7 3 C7 B8 A7 4 ClO D9 B8 5 D9 ClO ClO 6 A7 A6 Cl2 7 C9 C7 D9 8 Cll - A7 Bll 9 Bll Bll Cll 10 A6 C9 A6 Table 3.6 Various measures of surge probability for subset NT, arranged by map sheet and ranked in order, with greatest first. Comparison of the ranking orders of the above columns with equivalent columns in Table 3.4 (primary data set) shows a good similarity between the two data sets. Spearman's rank correlation coefficients for the respective columns in each population were all greater than 0·8, indicating that data sets are almost statistically identical. glaciers it was not possible to estimate the slope of the accumulation and ablation zones. Therefore, the analysis was concerned only with total slopes. The primary data set was arranged according to length bins. The median maximum and minimum elevations for glaciers in each length bin were calculated (Figure 3.8). The median slope for glaciers in each bin was also computed (Figure 3.9). These diagrams show that, for the Svalbard data set, as length increases there is an increase in elevation span but a decrease in median slope. Figure 3.9 demonstrates that there is not a simple relationship between median and maximum elevations and glacier length. Glaciers in length bin / = 1 have the highest median elevation (700 m). However, the lowest median maximum elevation is 568 m (l = 8) suggesting that there is not a marked difference in the highest altitudes between the various length bins. In comparison, the plotted values of median minimum elevation show a monotonic decrease from 400 m for l = 1 to 55 m for l = 10. Figure 3.8 suggested a slight trend of increasing range between maximum and minimum elevations with increasing glacier length. This trend is confinned in Figure 3.9. Median slope decreases monotonically with increasing length (Figure 3.9). The median slope for the primary data set is 6·3°. The demonstrated relationship between glacier slope ·and length may confuse an analysis of the influence of slope on surge probability, because surging is more likely in long glaciers. This length influence was removed using the procedures described by Clarke et al. (1986). For each length bin, Page 75 Chapter 3: Statistical analysis of Svalbard glaciers the glaciers were divided into two subsets. Those glaciers in subset j+ had surface slopes exceeding the median slope for that length bin. Glaciers with slopes below the median for that length bin were placed in subset f . The glaciers in the primary data set 800 -S 600 C: 0 -~ • Median maximum elevation > 90° 1-0 35.3 Table 3.8 Summary of various probability values for different degrees of glacier curvature. Calculations were made using the primary data set. The information used in this analysis was the orientation data collected for the accumulation and ablation zones of each glacier. A measure of the degree of curvature for a given glacier was found from the change in flow direction between the two zones. Four classes of curvature, b, were defined with b = l representing glaciers which were approximately straight (angular difference - 0°) and b = 4 containing glaciers with Page 84 I Chapter 3: Statistical analysis of Svalbard glaciers angular differences > 90° . The probability that a glacier has a certain curvature was found from Pb= nb/N, where nb is the number of glaciers in category b. Using the i- distribution for each category, the probability that a glacier is i-type and has a certain curvature was calculated as Pilb = nilbl nb, where nilb represents the number of i-type glaciers in that category. The surge probability for each degree of curvature is thus computed: Table 3.8 summarizes the results. 3 Pslb = L, Pi lb Psli i = 0 (3 .14). The calculated surge probabilities- remain reasonably similar between the different categories of curvature. Glaciers with channel curvatures between 45-90° (b = 3) are the most likely to surge (ps lb = 40-6%). The least likely to be surge-type are those with curvatures ranging from 0-45° , (pslb = 34·5%). It is noticeable that the variation in surge probability amongst the categories is not large (Figure 3.13). This is confirmed by a simple standard deviation calculation (s.d. = 2·4%). It is concluded from this analysis that channel curvature has little effect on the probability of glacier surging. The data used in the above analysis treated the curvature of glaciers as the difference in flow direction between two points. It did not take into account any curvature occurring between these two points. This can give a misleading impression of a glacier' s flowline. For example, Hyllingebreen in Kjellstromdalen was classified as a straight glacier because the orientations of both zones were in the same octant. In reality, however, this glacier has a sinuous flowline, changing direction by - 40° on two occasions. Thus, there may be a hidden channel curvature influence on surge probability that has escaped detection in this analysis because the data used did not fully describe each flowline. 3.5 GEOLOGICAL CONTROLS ON GLACIER SURGING This section examines the relationship between bedrock geology and the probability of glacier surging. There is general agreement amongst glaciologists that nature of the ice- bed interface is a significant component of the surge mechanism. However, there is a vigorous debate as to whether surging takes place on an essentially rigid and impermeable bed (e.g. Kamb, 1987) or if the deformation of subglacial sediments is the surge trigger (Clarke et al, 1984). The geological characteristics of bedrock bene.ath glaciers are, therefore, likely to be a key factor in determining the surge mechanism. Page 85 : I I I Chapter 3: Statistical analysis of Svalbard glaciers 3. 5 . 1 General lithological classification and surge probability Geological maps were available for all the sampled areas at a scale of 1 :500,000 and at 1: 100,000 for sheet C9 (Adventdalen). The data obtained from these maps were re-organised to find the i-distribution for each lithological type. For an individual glacier it was usually not possible to specify a single rock type on which it was resting. In most cases, glaciers were noted as being underlain by several rock types. There were two reasons for this type of classification. Firstly, the geological maps used in the analysis tended to assign several different, although lithologically similar, rock types to a given shaded area. This mapping procedure was used because there in insufficient detail available on the geology of Svalbard to enable a more precise classification. The second reason for the multiple rock types was that glaciers often crossed geological boundaries and were, therefore, underlain by more than one lithology. In total, 41 different combinations of rock type were noted. With such a high number of lithological combinations it was inevitable that some of these groups contained very few glaciers. In such cases, the sample populations would have been too small to permit a valid statistical analysis. Therefore, in order to obtain viable populations, the data were regrouped. This regrouping took the simplest form of classifying the rock types according to the major petrological categories, i.e. igneous, metamorphic and sedimentary rocks. Igneous rocks (g = 1) in the sample were granites, migmatite and gabbro. The metamorphic category (g = 2) contained quartzite, pelite, schist, gneiss and slate. Tillites, limestone, sandstone, Old Lithology N2 P2 % Psl2 % Igneous 18 3·0 5.1 Metamorphic 82 13·0 27·5 Sedimentary 515 84·0 39·8 Table 3.9 Surge probability statistics for the three petrological categories derived from the primary data set. Red sandstone, shale -and conglomerate were classified as sedimentary (g = 3). The i-distribution for each geological category was determined, from which the probability that a glacier overlying lithology g is i-type was cf].lculated as Pilg = nilglng. The surge probability of each petrological group was thus computed from : 3 Pslg = L Pilg P s li i=O (3 .15). Table 3.9 presents the results of the above calculations. If the rock type beneath glaciers does not influence surging, then Pslg should be roughly equal for each category. In the Svalbard sample population this is not the case. Igneous rocks have a surge probability Page 86 I, Chapter 3: Statistical analysis of Svalbard g laciers of 5· 1 %. This value is the saMe as the surge probability of type i = 0 glaciers because all glaciers overlying igneous rocks were 'normal '. Metamorphic rocks have a much greater surge probability (pslg = 27·5%). The highest surge probability was found for glaciers overlying sedimentary rocks (pslg = 39-8%). The above analysis demonstrated that subglacial geology has a marked influence on the probability of surging. There might, therefore, exist a geological control on the geographical distribution of surge-type glaciers. To test this hypothesis, the geological variability of each map sheet was examined. The probability that a glacier in map sheet m is resting on bedrock g was found frompglm = ng1mlnm, where ng1m is the number of glaciers above rock type gin map m. The geology-predicted surge probability for each map sheet was then calculated: 3 P\lm(g) = L, P glm Pslg g = l (3.16). The map surge probability predicted by geology was lowest for sheet A6 (19· 3% ). All the glaciers on this map were observed to be overlying igneous and metamorphic rocks . In comparison, maps C9, ClO, Cll C12 and D9 had ap\1m(g) of 39-8% (Table 3.10). These areas were entirely composed of sedimentary rocks. Figure 3.14 shows that this geological analysis is unable to identify those map sheets which have an observed concentration of surge-type glaciers (Figure 3.2) . Only sheet A6 stands out as an area with a low surge probability. x2 calculations showed that this map sheet was the only one to differ, at the 95% significance level, in terms of geological variability from the primary data set. Map Sheet Pslm % p\1m( P) % r* slm(P) % A6 15·8 19·3 82 A7 45.3 34·0 133 B8 46·8 38·7 121 Bl 1 28 ·9 37·0 78 C7 38· 1 35.9 106 C9 40·8 39·8 103 ClO 37 .4 39·8 94 C ll 27·4 39·8 69 C12 46·4 39·8 117 D9 42·3 39·8 106 Table 3.10 · Surge probability statistics predicted by the three-fold geological classification scheme, arranged by map sheet. The actual map surge probabilities are listed in the column headed p slm . Page 87 ,50 78° 76° 0 20 30 40 45 Figure 3.14 Geographic variation of geology-predicted surge probabilities for the primary data set. A threefold geological classification scheme was used in this prediction. Glaciers underlain by sedimentary rocks were found · to have the highest probability of being surge-type. However, this figure does not differentiate between regions of high and low concentrations of surge-type glaciers because large areas of sedimentary rocks were located on each map sheet. This unsuccessful prediction suggests that if surging is geologically controlled, individual rock types are likely to be more important than petrographic categories. Page 88 Chapter 3: Statistical analysis of Svalbard glaciers The probability measure r* slm(g) was computed for each map to assess the magnitude of the difference between actual and geology-predicted surge probabilities. The minimum ratio was obtained for sheet Cll (r\lm(g) = 68-8%). This map had an actual map surge probability of 27·4%. However, because all the glaciers on this map are underlain by sedimentary rocks, the probability of surging predicted by geology is much higher, at 39·8%. The maximum ratio was calculated for sheet A7 (r\im(g) = 133·2%). In the original calculation of Pslm, this map had the second highest surge probability, at 45-3%. When geology was used to predict the probability of surging, the value for A7 dropped to 34-0%. The comparatively low result was due to a large proportion of the glaciers on this mapoeing- based on metamorphic rocks. Figure 3.15 illustrates the map probability ratios calculated from the geology predictions. Comparison of this figure with Figure 3.2 shows some degree of similarity. 3. 5. 2 Surge probabilities associated with individual rock types From the above examination, it is concluded that the subglacial geology does have an influence on the probability of surging. However, geology cannot account for the geographical distribution of surge-type glaciers. The analysis performed above only considered subglacial geology in terms of three petrological groups. Thus, it is possible that by grouping many different rock types into just three categories some vital information was lost. Individual lithologies, and not petrological classes, probably play a more important role in influencing surge-type behaviour. An attempt was made to test this hypothesis, and is described below. It was demonstrated that glaciers overlying sedimentary rocks had the highest Lithology nlirh Plith % Psllith % Tillite 52 8·5 47.5 Limestone 9 1·5 62·5 Sandstone 61 9.9 50·6 Old Red Sandstone 36 5.9 47.5 Shale 337 54·8 36•7 Conglomerate 20 3.3 35·0 Metamorphic 82 13.3 27•5 Igneous 18 2·9 5· 1 Table 3.11 Surge probability statistics prepared using the revised geological classification scheme. The figures for the metamorphic and igneous · groups remain as before. The additional data were obtained by subdividing the lithologies comprising the original sedimentary group. Page 89 , I 78° I ' 15" 20' 2~~ 76° 15° 0 50 75 100 · 125 % Figure 3.15 Geographic variation of surg~ probabilities for the primary data set. The influence of the bedrock geology has been removed by computing the . ratio r* slm(g)· This figure is similar to the observed distribution of surge- type glaciers (Figure 3.2) (Spearman's rank correlation coefficient 0-93) . Page 90 11'1 I Chapter 3: Statistical analysis of Svalbard glaciers probability of being surge-type. It was also shown that many of the individual rock types occurred so infrequently that a statistically valid result could not be expected if they were tested separately. However, approximately 80% of the primary data set was composed of glaciers overlying sedimentary rocks. It was considered that the individual lithologies within this group would provide sufficient sample numbers to justify a statistical analysis of the influence of each. The sample sizes associated with the igneous and metamorphic groups were considered too small to permit a similar examination. The following lithologies were analysed individually: tillites, limestone, sandstone, Old Red sandstone, shale and-conglomerate. The number of glaciers underlain by each lithology was denoted n1irh and the probability that a glacier has a certain subglacial geology was Plith· Over half the glaciers in the sample were underlain by shale. Limestone was the least frequently occurring rock type. The i-distribution for each rock type was established. The number of i-type glaciers above that lithology was denoted nillith and the probability that a glacier with a certain subglacial geology is i- type was Pillith = nn1irhlnzith· The probability that a glacier surges given its subglacial geology was then found from: 3 Psllith = L, Pillith Psli i = 0 (3.17). Values ranged from 62·5% for limestone to 35·0% for conglomerate (Table 3.11) . The sedimentary petrology of each map sheet was examined in more detail and the frequency of individual rock types noted. New values of Pglm were calculated (pglmrr) . This cleared the way for new estimates of map surge probabilities to be predicted from Map Sheet Pslm % P\lm!oIT) % r* slm(oJT) % A6 15·8 19·3 81 ·9 A7 45.3 38·7 117 B8 46·8 45·6 103 Bll 28·9 43.7 66 C7 38· 1 43·2 88 C9 40·8 45·1 90 ClO 37-4 44.3 84 Cll 27•4 43.3 63 C12 46-4 41·3 112 D9 42·3 36·7 115 Table 3.12 Surge probability statistics predicted by the revised geological classification scheme, arranged by map sheet. The actual map surge probabilities are listed in the column headed Pslm· Page 91 Chapter 3: Statistical analysis of Svalbard glaciers the revised geological classification using a modified equation 3.16. Table 3.12 summarizes the results of this analysis and Figure 3.16 illustrates the differences between the map sheets graphically. New values of the probability ratio r* slm(gIT) were obtained in order to compare the revised geology-predicted surge probabilities with the actual probabilities. Table 3.12 lists the calculated values. The maximum and minimum ratios were found, once again, to occur on map sheets A7 and Cl 1 respectively. The probability ratios were plotted (Figure 3.17) and compared with other maps using different measures of surge probability. Figure 3.16 shows that using the revised geological classification scheme to predict surge-type behaviour results in a closer agreement to the pattern of actual surge probabilities (Figure 3.2) than does the original approach using geology (Figure 3.15). However, it tends to slightly overpredict the likelihood of surging in maps B 11 and Cll, and underpredict that for map A7. Nonetheless, it correctly identifies sheets A6 and B8 as areas with extremely low and high concentrations of surge-type glaciers, respectively. When the revised probability ratios are plotted (Figure 3.12) there is slight reorganization of the probability distribution to a pattern broadly similar to that of the actual surge probabilities (Figure 3.2). This ratio cannot, however, identify map sheets with extreme values. Figures 3.16 and 3.17 suggest that the revised geological classification scheme provides a better indicator of the probable distribution of surge-type glaciers than any of the previously analysed parameters. This hypothesis can be tested by comparing the standard deviations for the probability ratios calculated to remove the influence of length and subglacial geology (both the original and the revised classification). A ratio of 100% represents an exact match between the predicted and observed value. A lower standard deviation will suggest that more of the individual map sheet ratios are closer to the observed probability. Thus, it follows that the lower the standard deviation, the better that particular scheme is at predicting the distribution of surge-type glaciers. The calculated standard deviations are as follows: for the length-predicted probability, s.d. = ±27·5%; for the original geology-prediction, s.d. = ±19·3%; for the revised geology probability scheme, s.d. = ±18-4%. Therefore, the surge probabilities based on the more detailed geological classification are slightly more effective at predicting the distribution of surge-type glaciers. The statistical significance of this result cannot, however, be considered great. Page 92 20° 78° ' ~o·· :?5' 76° 15° 20° 0 20 40 % Figure 3.16 Geographic variation of surge probabilities for the primary data set predicted using the revised geological classification . scheme. The · comparison between this figure and the observed spatial distribution (Figure 3.2) of surge-type glaciers is qualitatively better that the comparison between Figure 3.14 and Figure 3.2. This improved comparison suggests that the more detailed lithological classification scheme is a better predictor of surge-type behaviour. However, a Spearman's rank correlation coefficient of 0·2 indicates only a weak agreement between this prediction and the observed pattern. Page 93 Figure 20° 78° 76° 15° 20° 0 50 75 100 125 3.17 Geographic variation of surge probabilities for the primary data set. The influence of the revised geological classification scheme has been - removed by computing the ratio r* slm(gII) · There is a reasonable agreement between this figure and the observed distribution of surge- type glaciers (Figure 3.2) (Spearman's rank correlation coefficient 0-855). However, this ratio cannot identify map sheets with abnormal concentrations of surge-type glaciers which suggests that the revised geological classification scheme was partly successful in predicting the spatial variation of surging. Page 94 Chapter 3: Statistical analysis of Svalbard glaciers 3.6 OTHER ENVIRONMENTAL CONTROLS 3. 6 .1 Introduction The previous sections presented the methods and results of a statistical analysis of Svalbard glaciers. It did not consider all the factors which might have an influence on glacier surging, but only those for which data were available. One environmental characteristic that was not tested was glacier thermal regime. There is, however, a limited amount of temperature data available for Svalbard ice masses. This permitted a simple statistical analysis of its importance on the probability of surging to be carried out. The results of that analysis are discussed-below. This section also contains an analysis of aspects of surge theories proposed by Kamb (1987) and Fowler (1989). These theories do not specifically indicate factors which might be related to surging. However, both Kamb and Fowler derived criteria which they suggested could discriminate between surge-type and normal glaciers. The Svalbard primary data set contains information which enables a statistical test of the validity of these criteria to be made. 3. 6. 2 Internal reflecting horizons, glacier thermal structure and surge probability The thermal instability mechanism of glacier surges has received some consideration in the past (Robin, 1955; Clarke, 1976; Paterson et al., 1978) (Section 1.3.1). Schytt (1969) called attention to the thermal regime of ice caps in eastern Svalbard. He speculated that portions of these ice caps surged when an inner core of ice at the melting point broke through an annulus of colder ice frozen to the bed (section 2.4.2) . A major criticism of thermal mechanisms, however, is that they can only explain surges in glaciers where there is an element of basal freezing (Paterson, 1981). Thus, since thermal instabilities cannot account for all surges, theories of this kind have not attracted much attention in recent studies. Schytt (1969) had only limited field data on which to base his hypothesis. Since then, however, more evidence has been obtained on the thermal characteristics of glaciers and ice masses in Svalbard. Much of the information concerning the thermal regime of Svalbard glaciers has been derived from the results of radio echo sounding programmes. These echo soundings have been carried out using a variety of instruments, operating at UHF (Macheret and Zhuravlev, 1982), VHF (Dowdeswell et al., 1984) and low frequency (Hagen and Sretrang, 1991). The quality of the glacier bed return signal varies according to the equipment used. However, a persistent feature on many echo returns has been the presence of internal reflecting horizons (IRH). Bamber (1987) reported the presence of continuous horizons, situated at a depth of Page 95 I I Chapter 3: Statistical analysis of Svalbard glaciers 3.6 OTHER ENVIRONMENTAL CONTROLS 3. 6 .1 Introduction The previous sections presented the methods and results of a statistical analysis of Svalbard glaciers. It did not consider all the factors which might have an influence on glacier surging, but only those for which data were available. One environmental characteristic that was not tested was glacier thermal regime. There is, however, a limited amount of temperature data available for Svalbard ice masses. This permitted a simple statistical analysis of its importance on the probability of surging to be carried out. The results of that analysis are discussed-below. This section also contains an analysis of aspects of surge theories proposed by Kamb (1987) and Fowler (1989). These theories do not specifically indicate factors which might be related to surging. However, both Kamb and Fowler derived criteria which they suggested could discriminate between surge-type and normal glaciers. The Svalbard primary data set contains information which enables a statistical test of the validity of these criteria to be made. 3. 6. 2 Internal reflecting horizons, glacier thermal structure and surge probability The thermal instability mechanism of glacier surges has received some consideration in the past (Robin, 1955; Clarke, 1976; Paterson et al., 1978) (Section 1.3.1). Schytt (1969) called attention to the thermal regime of ice caps in eastern Svalbard. He speculated that portions of these ice caps surged when an inner core of ice at the melting point broke through an annulus of colder ice frozen to the bed (section 2.4.2). A major criticism of thermal mechanisms, however, is that they can only explain surges in glaciers where there is an element of basal freezing (Paterson, 1981). Thus, since thermal instabilities cannot account for all surges, theories of this kind have not attracted much attention in recent studies. Schytt (1969) had only limited field data on which to base his hypothesis. Since then, however, more evidence has been obtained on the thermal characteristics of glaciers and ice masses in Svalbard. Much of the information concerning the thermal regime of Svalbard glaciers has been derived from the results .of radio echo sounding programmes. These echo soundings have been carried out using a variety of instruments, operating at UHF (Macheret and Zhuravlev, 1982), VHF (Dowdeswell et al., 1984) and low frequency (Hagen and Sretrang, 1991). The quality of the glacier bed return signal varies according to the equipment used. However, a persistent feature on many echo returns has been the presence of internal reflecting horizons (IRH). Bamber (1987) reported the presence of continuous horizons, situated at a depth of Page 95 Chapter 3: Statistical analysis of Svalbard glaciers 100-200 m below the ice surface, on approximately 60% of the glaciers echo sounded by SPRI in 1983. The distribution of glaciers which possess such horizons exhibits a clear geographical trend (Bamber, 1987; Macheret et al., in press) (Figure 3.18). There is a greater concentration of glaciers with IRHs on the western side of Spitsbergen than on north-east Spitsbergen or Nordaustlandet. Bamber (1987) attributed this distribution to the variation in climate over the archipelago. Hence, the IRHs are probably a function of the thermal regime of those glaciers. Radio echo sounding data do not exist for every glacier in the primary sample population. Thus, a complete statistic.al evaluation of the relationship of IRHs with surge-type glaciers cannot be made. However, Yu. Ya. Macheret (personal communication 1991) provided information, obtained during Soviet radio echo sounding programmes, for 136 glaciers in Svalbard. A number of the ice masses in this sample are included in the primary data set, although, in addition, the Soviet sample contains glaciers in north-east Spitsbergen and Nordaustlandet. The data were arranged by Macheret using a dichotomous surge classification scheme. In this scheme, glaciers were either surge-type or normal. Surge-type glaciers were identified using information given by Liest01 (in press). Independent checking confirmed that this list was accurate and, therefore, those glaciers identified by the Soviets as surge-type are taken to be correct. However, the Soviet sample does not take into account the probability that some of their 'normal' glaciers may be surge-type. Due to the differences between the Soviet sample and the primary data set, a thorough statistical analysis was not performed. The analysis described below was, therefore, necessarily a simple one. Using the data obtained from Macheret (personal communication), the probability that a glacier is surge-type is PslSoviet = 23·5%. This is somewhat less than the Ps = 36-4% for the primary data set, and is probably due to the Soviet sample containing surge-type glaciers incorrectly classified as normal. Macheret was able to Thermal regime N nsun1e Psltherm o/o Cold ]8 5 13·2 Two-layered 46 21 45.7 Relatively warm 23 1 4.3 Not specified 29 5 17·2 Table 3.13 Surge probability statistics for the Soviet sample (data from Macheret, personal communication). Thermal regime was derived from radio echo . sounding and borehole temperature data. Surge-type glaciers are those taken from Liest01 (in press) and only include glaciers where surges have been observed in the last one hundred years. Page 96 Glaci e rs poss es sing a co nt inuous IRH 0 50 100 km Figure 3.18 Location of glaciers in Svalbard where internal reflecting horizons . (IRH) have been recorded during radio echo sounding. The dashed lines show echo sounding flight paths . Based on Bamber (1987b) and Macheret et al. (in press). Page 97 C!wpter 3: Statistical analysis of Svalbard glaciers classify most of the glaciers in the sample according to thermal regime. This classification was achieved using data from radio echo sounding and temperature measurements in boreholes (Macheret, personal communication). The three categories were 'cold ', 'relatively warm' and 'two-layered'. These categories are equivalent to polar, temperate and sub-polar thermal regimes, respectively. Two-layered glaciers are sub-polar in the sense that their basal layers are at the pressure melting point and their surf aces are below the freezing point. For several glaciers, a thermal regime could not be specified, giving a fourth category. The probability that a glacier is surge-type given its thermal regime was estimated as Psltherm = nslthermfnrherm, where nsltherm is the number of surge-type glaciers, and nrherm is the total number of glaciers, in a particular thermal category. Table 3.13 summarises the available data and the results of this simple analysis. The interesting point to note is that two-layered glaciers have the greatest probability of being surge-type (psltherm = 45-7%). Cold glaciers had the second highest surge probability, but the figure was much lower (psltherm = 13·2%). The proportion of two-layered glaciers in the Soviet sample is 65·6% which is consistent with the proportion of two-layered glaciers in a sample reported by Bamber (1987). Two-layered glaciers display a thermal regime which is characterised by temperate basal ice and colder surf ace ice. The boundary between the two states is marked by an IRH. This has been confirmed by temperature measurements made in boreholes (Hagen and Sretrang, 1991). J.O. Hagen (personal communication 1991) found that the position of an IRH on Austre Brsziggerbreen, north west Spitsbergen, coincided with the position of the pressure melting isotherm measured in a borehole. Bamber (1987) suggested that the presence of water-saturated ice in the deeper layers of glaciers was the cause of the internal reflections. Macheret and Zhuravlev (1982) reported a water content of 1-2% in ice at a depth of 115 m in a core taken from Fridtjovbreen. This was slightly higher than a value of=== 1 % for temperate ice reported by Raymond and Harrison (1975) . Bamber (1987) adopted two approaches to modelling the position of an IRH above the bed. The first was based on classical groundwater principles and predicted a piezometric surface with a profile that did not match that of the IRHs. The second approach used Rothlisberger ' s (1972) theory to predict the height of the hydraulic head. The results of this model showed good agreement with recorded horizons (Bamber, 1987). The suggestion was that ice below the IRH was permeable, so that water could be transported through it, via small channels and conduits, to a Rothlisberger channel at the hydraulic grade line. An englacial Rothlisberger channel has been observed on Brsziggerbreen (Hagen et al., 1991) at the same depth which Hagen and Sretrang (1991) recorded an IRH. Thus, water pressures controlled the position of the IRHs. The influence of thermal regulation Page 98 Chapter 3: Statistical analysis of Svalbard glaciers on the position of an IRH was also modelled by Bamber (1987). However, he considered that to explain the freezing of water over the depth of an IRH required an unrealistic temperature gradient. Thus, the thermal structure of the glacier was believed to be of secondary importance in controlling the location of internal reflecting horizons. However, the fact that ice above the IRH was cold, and probably impermeable, may be important. The above discussion has shown that the presence of an IRH in a glacier suggests a particular thermal regime, although temperature is not the primary control on their formation. Nevertheless, we are concerned here with the possibility that glaciers with a certain thermal regime have an increased chance of surging. If indeed the connection between two-layered glaciers and surging is real, then it is interesting to speculate about the physical mechanism involved. The following hypothesis is suggested. In a two-layered surge-type glacier in its quiescent phase, water flows in an englacial Rothlisberger channel as envisaged by Bamber (1987). The number of Rothlisberger channels which might exist in such a situation is not known, although there are probably no more than a few in any one glacier. The channels are maintained by a balance between frictional heating of the ice walls by flowing water and the creep closure caused by internal deformation of the glacier. As ice overburden pressure increases, the amount of heat produced must increase if the channel is to remain open. In surge-type glaciers approaching the active phase, the overburden pressure will increase in the reservoir zone. Therefore, at periods of low water discharge the channel may no longer be able to stay open. If the conduit is closed, water will be driven down through the permeable lower layers of the glacier. This water could accumulate at the base where there are no efficient drainage pathways. A linked cavity drainage system may develop (Kamb, 1987), or subglacial sediments could become deformable (Clarke et al., 1984), both of which are thought be causes of surges. Alternatively, the water may accumulate in the lower ice strata, leading to an increase in creep deformation. The results discussed above do not advocate a thermal instability mechanism of glacier surges and, furthermore, there has been no implication that only glaciers with a two-layered structure are able to surge. However, if Bamber's model of water throughflow in these glaciers is correct, then we might expect their basal drainage system to be poorly developed. Therefore, once ice overburden closes the englacial hydrological network, water is diverted to the bed where it cannot be effectively discharged. Surge mechanisms proposed by Kamb or Clarke may then operate. 3 . 6. 3 Slope data and Kamb 's stability parameter The analysis of glacier slope and the probability of surging indicated that there was a very slight tendency for surge-type glaciers to have slopes steeper than the Page 99 Chapter 3: Statistical analysis of Svalbard glaciers median (section 3.4.4). Clarke et al. (1986) performed a similar analysis and reached corresponding conclusions. Their conclusion was used by Kamb (1987) to support his hypothesis of a linked-cavity model of glacier surging (section 1.3.2). This was because one of the variables determining the value of Kamb's stability parameter, 5, was the longitudinal hydraulic gradient, a, taken to be equal to the glacier surface slope. Data from both the Svalbard and the Yukon sample populations (Clarke et al., 1986) support Kamb's assertion that glacier slope varies inversely with glacier length. Kamb (1987) stated that lower slopes would promote a lower value of the stability parameter because 5 varies as a3/2. He predicted that a linked cavity system remains stable when 5 < 1 ·0 (section 1.3.2). However, in common with Yukon glaciers (Clarke et al., 1986), surge-type glaciers in Svalbard are not found to be more likely to have slopes lower than the median. Therefore, Kamb's (1987) model cannot be supported using either data set. Clarke (1991) questioned Kamb' s interpretation of their earlier statistical analysis of Yukon glaciers. Of greatest interest to Clarke was Kamb's assertion that glacier slope varies inversely with glacier length. Clarke ' s concern was that length and slope are influenced by one another and, hence, it was difficult to determine which variable was primarily associated with surging. Multiple correlation analysis was used to assess the relative importance of length, slope and width with surging. Greater than 99% of the multiple correlation between these three variables and surging was explained by the correlation between length and surging alone. In contrast, slope and surging only explained 60% of the multiple correlation. This analysis suggested that Kamb's (1987) prediction that low slopes favour surging, may not have any significance. 3 . 6. 4 A statistical test of Fowler's surging criterion A mathematical model of glacier surges was developed by Fowler (1989). The model is similar to Kamb's (1987), in that a surge is triggered by a change in the subglacial drainage system. As part of his analysis, Fowler introduced a surging parameter, OJ, which he used to distinguish between surge-type and normal glaciers. Surging is predicted when: OJ< w* where OJ* is some critical value of OJ, and OJ is defined as: CJJ = awf3 (3.18), (3.19). Here, a is the glacier surf ace slope, w is the glacier width and /3 is taken to be .:: 2. Equation 3.18 predicts that glaciers with a small value of OJ are more likely to be of surge-type than glaciers where the value of OJ is large. Fowler did not actually provide any limits for critical values of OJ and OJ* . However, since his criterion relies on Page 100 Chapter 3: Statistical analysis of Svalbard glaciers geometrical parameters to predict glacier behaviour, it is possible to use the Svalbard data set to test whether lower values of OJ are, in fact, associated with surge-type glaciers. Fowler (1989, p. 261) remarked that a statistical analysis of his surging criterion would be an interesting test. Values of the parameter OJ were computed for each glacier using width and slope data. The values obtained ranged from 0·8 to 53·8. To simplify the analysis, the OJ data were arranged into a number of categories defined by limits of OJ. The i- distribution for each OJ class was then determined. The probability that a glacier had a value of OJ which fell into a certain clas..s wasp w = nd N, where nw was the number of glaciers in that class. The probability that a glacier with a Fowler parameter in class OJ was i-type was found from Pilw = nndn(J), where nitw was the number of glaciers in that OJ class. From these derived values, the surge probability for that OJ class was calculated from: 3 Pst w= L PilwPsli i= 0 (3.20) . Table 3.14 summarizes the results of the above calculations. If the criterion introduced by Fowler is valid, then Pslw should be highest for low values of OJ. However, Figure 3.19 demonstrates that, for the Svalbard data set, the above statement does not fit. Instead, the probability of surging is at a minimum for glaciers with low values of OJ 70 60 . 50 ~ 0 40 :s:: Cl) 30 a.. 20 10 0 0 2· 3 4 5 6 7 8 9 10 w class Figure 3.19 Variation of surge probability with Fowler' s_ surging criterion, OJ. Calculated values of OJ were arranged into 10 classes. Fowler's (1989) theory predicts that glaciers with low values of OJ are more likely to be surge-type. The above data do not support that prediction. and there is a rough trend for Pslw to increase with larger values of OJ. The class containing the smallest value of OJ also has the lowest Pstw at 22·8%. In contrast, the Page IOI Chapter 3: Statistical analysis of Svalbard glaciers maximum calculated value of Pslw (69· 1 %) was found for glaciers in the second highest class of cv. Therefore, the Svalbard data set does not support Fowler's surging criterion. Clarke (1991) reached a similar conclusion using data from Yukon glaciers. The surging condition that this section has analysed was not the major component of Fowler's (1989) model. However, it was an hypothesis that was able to be tested easily and the results obtained suggest strongly that it is not a valid criterion. As a result, other parameters in Fowler's model may also be inapplicable to the surge mechanism. Category CV nw Pw% Ps1w% 1 0-2 19 3·1 22·8 2 2-4 163 26·5 30·8 3 4--6 167 27·2 40·6 4 6-10 156 25·4 34·8 5 10-15 71 11 ·6 41·6 6 15-20 13 2·1 47·0 7 20-25 14 2·3 36·0 8 25- 30 4 0·7 48·0 9 30-40 4 0·7 69·1 10 > 40 3 0·5 63·5 Table 3.14 Surge probability statistics for Fowler's parameter cv, arranged by category. The ranges of values of w contained in each category are indicated in the column marked cv. Results are for the primary data set. 3.7 DISCUSSION AND CONCLUSIONS 3. 7 .1 Discussion This chapter has presented the results of a statistical analysis of Svalbard glaciers. The analysis demonstrated that certain areas of the archipelago have higher than average concentrations of surge-type glaciers while in other areas surge-type glaciers are comparatively rare. This non-random geographical distribution is a local reflection of a global pattern. Probability statistics were used to determine associations between various environmental attributes and glacier behaviour. On the basis of this work, several"factors can be considered to have little influence on the occurrence of surging. Glacier slope, elevation, orientation and curvature do not have a significant influence on the probability of surging. Furthem10re, the width-slope product known Page 102 Chapter 3: Statistical analysis of Svalbard glaciers as the Fowler parameter was not a useful discriminator between types of glacier behaviour. These findings are in agreement with those from other studies. Clarke et al. (1986) could not establish any relationship between slope or orientation and surging of Yukon glaciers. They did not consider glacier curvature in their analysis . The non- statistical work of Post (1969) also suggested that elevation, slope and orientation were not related to surge-type behaviour. Wilbur (1988) examined different variables in his study of Yukon and Alaskan glaciers. Therefore, no comparisons can be drawn between the present study and his analysis. Glaciers without tributaries had-a sligl}tly higher probability of being surge-type than those with tributaries. This contradicted the findings of Clarke et al. (1986). They found that tributary glaciers had an increased chance of surging compared to those without tributaries. This was explained by tributaries having two possible trigger sources: internally triggered surges and surges induced by their trunk. The reverse case, however, is also possible, i.e. that a surge of a tributary can trigger a surge in a trunk. This scenario probably accounts for the higher surge probability calculated for subset NT of the Svalbard population. It is interesting to note that not all tributary glaciers surge when their trunk surges . During the examination of NP aerial photographs, a large surge of Hambergbreen in c. 1960 was observed. Surges were observed taking place in a number of its tributaries but there were notable exceptions. At the junction of two tributaries with the main channel, there were pronounced shear zones. Hambergbreen was heavily crevassed but both tributaries maintained a smooth upper surface. These glaciers were not observed to surge at any other time for which photography was available and were classified as type O in the surge index. Other tributary glaciers remained unaffected by the Hambergbreen surge, although shear zones at their junctions were not so evident. It is conceivable that surges only affect trunks and tributaries together when the tributary is actively contributing ice to the trunk. In other words, a tributary which appears to be physically attached to its trunk may not be an important component of the trunk' s flow. The data sources used in this study did not permit a full examination of the contribution of ice from tributaries to trunks. As a result, some glaciers may have been wrongly classified as tributaries when, in fact , they were essentially separate ice masses. The inclusion of non-surge- type glaciers wrongly interpreted as tributaries may have biased the composition of the two subsets. Glacier length was shown to have an influence on the probability for surging. Post (1969) reported that surge-type glaciers in western North America did not appear to have any length requirements. It is certainly true that very short glaciers can surge. However, the statistical analysis discussed here demonstrated that long glaciers have an increased probability of being surge-type but did not imply that only long glaciers can Page 103 Chapter 3: Statistical analysis of Svalbard glaciers surge. Post reached his conclusion probably because he did not examine the length distributions of glaciers in his samples statistically. Furthermore, he only concerned his analysis with surge-type glaciers and did not compare their characteristics with those of normal glaciers. Clarke et al. (1986) reported a statistical relationship between length and surging for a sample of Yukon glaciers, some of which were included in Post's data set. Recent work by Clarke (1991) confirmed that length was the dominant variable correlated with surging and that slope and width influences were derived from this association. Length was not completely succes~ful at predicting the location of surge clusters. The geographical distribution of length-predicted probabilities did not agree with the observed concentrations of surge-type glaciers. Moreover, the ratio computed to remove the length influence was a much better indicator of the true pattern of clustering. The fact that length cannot predict the distribution of surge-type glaciers suggests that length on its own is not a good discriminator between types of glacier dynamics. This must be the case, since not all long glaciers in Svalbard are surge-type. Furthermore, there are long glaciers in the Alps, the Cascades and the Coast mountains of British Columbia, amongst others, which are areas where surge-type glaciers do not occur. A second statistical relationship obtained was that between the occurrence of a two-layered structure and surging. The analysis upon which this result is based was not as rigorous as that carried out for other factors. Nevertheless, there was a striking difference between the relationship of surging with the presence of an IRH compared with other thermal regimes. Again, this examination did not imply that only two-layered glaciers can surge, but instead that they have an increased probability of being surge- type. A possible reason for this was discussed in section 3.4.2. There are no known reports of IRHs in other glaciers elsewhere in the world. Notwithstanding, it is interesting to speculate that two-layered glaciers may occur outside Svalbard and that there may be an association between them and surging. At present, however, the RES data required to test this_ hypothesis are not available. The third important statistical relationship demonstrated by this study was that between the bedrock geology and the probability of surging. This relationship was analysed in two separate tests. The first test arranged the various rock types according to whether they were igneous, metamorphic or sedimentary. Much the greatest probability of surging was calculated for glaciers overlying sedimentary rocks. Also noticeable was the extremely low surge probability for glaciers based on igneous rocks. Due to the markedly different surge probabilities for the three rock types, a test was conducted to determine how well lithology discriminated between regions with clusters of surge-type glaciers. This threefold geological classification was only able to predict Page 104 . I 1 1 Chapter 3: Statistical analysis of Svalbard glaciers the regions with low concentrations of surge-type glaciers. Those regions with surge clusters were not identified by this scheme, probably because five of the ten map sheets were composed entirely of sedimentary rocks. Therefore, all glaciers on these maps had identical geologically-predicted surge probabilities. For a number of these sheets this was an underestimation of their actual probabilities. This particular result implied that the threefold classification scheme was not detailed enough to discriminate these map sheets from one another. To investigate the possibility that surge probability was more sensitive to individual lithologies and not petrologic.al categories, a further analysis was made. Only the rocks in the sedimentary category provided a suitably large sample size to be analysed individually. Glaciers underlain by limestone had the highest surge probability, although this result was based on a population of only nine glaciers. Sandstone, Old Red sandstone and tillite also had high surge probabilities. When the revised geological categories were used to predict the map surge probabilities, a better representation of the actual pattern was obtained than was from the original threefold scheme. This suggested that detailed classification of geology was a more accurate discriminator of surge concentrations between regions. On the basis of the tests carried out, it appears that glaciers in Svalbard are more likely to be surge-type if they are resting on sedimentary rocks. This connection between lithology and surging was first proposed by Post (1969). Again, because of the lack of statistical analysis, Post concluded that surge-type glaciers were not confined to any particular rock type. However, he remarked that, in some areas, surging was more common where the bedrock was predominantly sedimentary, although surge-type glaciers were also found above plutons and metamorphic rocks. C. Raymond (personal communication, 1991) reported that a large number of surge-type glaciers in Alaska are underlain by metamorphic rocks. Post pointed out, perhaps significantly, that no surge-type glaciers were located in the largely granitic Coast Mountains or low-grade metamorphic Rocky Mountains. The association_ between surging and geology was suggested by Post (1969) to be due to differences in bedrock roughness or permeability. Bedrock roughness would influence the rate at which a glacier sole could slip on its base, although Post had difficulty imagining that roughness could be markedly different over the study area. More likely, he thought , were differences in bedrock permeability. In fact, Post considered not only bedrock permeability but also unconsolidated subglacial deposits (Post, 1969, p.238). The deformation of subglacial sediments has been proposed as a mechanism of glacier surges by Clarke et al. (1984). Bedrock permeability, it can be argued, may be a factor in determining the formation of sedimentary glacier beds. Rocks with a high permeability would allow easier penetration of water which would Page 105 Chapter 3: Statistical analysis of Svalbard glaciers increase the rate at which weathering processes take place. A greater rate of debris production would promote the formation of subglacial sediment deposits. The factors influencing the formation of sedimentary glacier beds have been investigated by Haeberli (1986). He considered that the principal influence was the relationship between the rate of debris input to a given glacier geometry and the rate at which debris is removed from a system by meltwater streams. This model predicted that unconsolidated sedimentary beds may be more common beneath glaciers situated in a rugged mountain environment with a dry continental climate compared to those in a heavily glacierized maritime environm~nt. Surge-type glaciers in Svalbard might, therefore, be expected to be underlain by sedimentary beds. Whether the same can be expected for surge-type glaciers near the coast of north western North America, such as Variegated Glacier, is debatable, given the conditions required by Haeberli's model. Not all sedimentary glacier beds may be potentially deformable. This potential to deform is probably controlled by lithology. The products of rock weathering differ between lithologies, so that some deposits may be more suited to deformation. Sediments which are poorly drained are likely to accumulate water. A number of theoretical and experimental studies have shown that water content is critical for the deformation potential of sediments (Clarke, 1987; Alley, 1989; Murray, 1990). Murray (1990) also demonstrated that particle size distribution has an effect on sediment deformation. For the same water content, clay rich sediments have a greater viscosity than silt rich material. However, silty sediments have higher permeabilities (Farmer, 1983), so there must be a balance between particle size and permeability. Therefore, the composition of the parent lithology will affect the properties of the derived material. The geological data used in this study were not detailed enough to justify a thorough discussion of why certain lithologies appear to be related to surge-type glaciers. Nevertheless, at a superficial level it cannot be ignored that sedimentary rocks have a greater weathering potential than metamorphic and igneous rocks (Farmer, 1983) . This could i1Y1ply that potentially deformable sediments are more likely to form from the disintegration of sedimentary lithologies. If this is the case, then theories of glacier surging based on deformable bed mechanisms (e.g. Jones, 1979; Clarke et al., 1984) seem more reasonable than those developed assuming a rigid impermeable substrate (e.g. Kamb, 1987; Fowler, 1987, 1989). Furthermore; regional differences in lithology may partly explain the non-random geographical distribution of surge-type glaciers. The fact that none of the factors analysed in conjunction with surging on Svalbard glaciers was able to explain the distribution of surge-type glaciers in the archipelago is noticeable. This finding implies that surging is probably a product of more than one environmental condition. Clarke et al. (1986) suggested that if the glacier Page 106 I I Chapter 3: Statistical analysis of Svalbard glaciers substrate influenced surging then ice overburden pressure might affect the mechanical behaviour of the sediment. In this way, the observed relationship between glacier length and surging could be explained by length being a proxy indicator of ice thickness . In order to elucidate those factors responsible for surging in Svalbard an analysis using multivariate statistical methods should be made. 3. 7 .2 Conclusions A statistical analysis of glaciers in Svalbard has revealed the following conclusions: • the geographic distribution of surge-type glaciers on Spitsbergen is non-random, with certain areas having noticeable clusters of surge- type glaciers and other areas having relatively few surge-type glaciers, • long glaciers have an increased probability of being surge-type compared to short glaciers, • other glacier characteristics, such as slope, elevation, orientation, curvature and the presence of tributaries do not influence the probability of surging, • glaciers overlying sedimentary rocks have a greater chance of being surge-type compared to those with metamorphic or igneous subglacial geologies, and, • using a smaller sample, two-layered glaciers have increased likelihood of being surge-type than cold or temperate glaciers. Furthermore, it was demonstrated that the results from this analysis do not support certain aspects of surge mechanisms proposed by Kamb (1987) and Fowler (1989). None of the factors with demonstrated links with surging were able to successfully predict the locations of surge clusters. A marginally better comparison between predicted and .observed clusters was obtained when the revised geological classification scheme was used as the discriminator in preference to the original geological categories or glacier length. However, the fact that no one factor could reproduce the observed distribution of surge-type glaciers probably suggests that glacier surging is a product of more than one control. Page 107 CHAPTER 4 FIELD AREA AND METHODS 4.1 INTRODUCTION Meier (1969) recognised that a major barrier to a full understanding of the physical processes of surging was the absence of detailed studies of surge-type glaciers through their surge cycles. Since that statement was made, both Medvezhiy Glacier in the Pamirs (Dolgushin and Osipova, 1975) and Variegated Glacier in Alaska (Kamb et al., 1985) have been studied in detail prior to and during their surges, and Trapridge Glacier in the Yukon (Clarke et al., 1984) is the subject of a similarly detailed project as it approaches its next surge. Much of the current understanding of surge mechanics has come from the research carried out on Variegated and Trapridge glaciers. Nevertheless, certain problems remain concerning aspects of current explanations of the surge mechanism. These problems will only be eliminated if the observational knowledge of surge-type glaciers through their surge cycles is expanded (Raymond, 1980). Bjuvbreen was selected as a suitable glacier for detailed field investigation for a number of reasons. Firstly, the glacier was considered to be in late-quiescence and, therefore, likely to surge sometime in the near future. The progressive changes occurring from year-to-year in late-quiescence are an important indication of the processes responsible for surge initiation (Bindschadler et al., 1977; Raymond, 1980). Secondly, the small size of the glacier was such that it could be studied by a small group of field workers. Access to the glacier was also comparatively easy. Finally, much of the current understanding of the surge mechanism has come from predominantly North American field studies. Thus, a field study of a surge-type glacier in Svalbard would serve as a useful geographical counterpoint to existing work. The different environmental conditions operating in Svalbard, for example climate and geology, may exert some influence on the nature of the surge mechanism. Noticeable contrasts in the behaviour of surge-type glaciers in Svalbard and other regions for which data are available have already been documented. Dowdeswell et al. (1991) described the considerably longer durations of the active phase on surge-type glaciers in Svalbard compared to elsewhere. Quiescent phase durations also appear to be protracted on Svalbard glaciers. Dowdeswell et al. were unable to explain these differences fully, owing to a lack of field data. Therefore, it is important that the Chapter 4: Field area and methods dynamics of surge-type glaciers in the archipelago are studied in the field. The work described in this chapter is the first detailed study of the dynamics of a surge-type glacier in Svalbard. This chapter outlines the methods of research employed in the field and discusses the procedures followed in the reduction of the data. To begin with, however, the field area is described. 4.2 THE FIELD AREA 4. 2 .1 Kjellstromdalen Bjuvbreen (77° 55' N, 17° 15' E) is located on the southern side of Kjellstromdalen in central Spitsbergen, Svalbard (Figure 4.1 ). Kjellstromdalen, and its eastern extension Agardhdalen, form a major through-valley between Van Mijenfjorden on the west and Agardhbukta on the east coast of Spitsbergen. The upper reaches of Kjellstromdalen are underlain by Jurassic sedimentary rocks while further down the valley, Cretaceous deltaic sediments, primarily sandstones, claystones and shales are dominant (Flood et al., 1971). Conglomerates are also common. There are a number of Tertiary outcrops in the area which are the subject of commercial coal exploration at the small mining settlement of Sveagruva at the mouth of Kjellstromdalen. A short distance to the east of Bjuvbreen there is a major north- south trending fault line. Kjellstromdalen is a wide, flat-bottomed valley with a low gradient. The 5 m contour crosses the valley floor 9 km inland from the coast. The valley is occupied by a large, braided river which is nourished by melt streams from a number of glaciers. It flows into Braganzav~gen, a tidal lagoon separated from Van Mijenfjorden by a large morainic complex. The surficial sediments of Kjellstromdalen are characterised by fine estuarine and fluvial deposits and coarser material carried down from glaciers by CA.re, proglacial streams. Along the valley sides theree number of features interpreted as alluvial fans and landslide deposits (Costner, 1925). Permafrost is present in the area. Measurements made during mining operations in Sveagruva have shown that frozen ground penetrates to a depth of at least 280 m (Liest01, 1980). There are approximately 20 glaciers of various sizes terminating in Kjellstromdalen. Corrie glaciers, less than 3 km long, are the most common glacier type. A number of glaciers receive their nourishment from plateau ice caps. On the northern side of the valley, both Hoganasbreen and Helsingborgbreen are fed by ice from Gruvfonna. The largest glacier in Kjellstromdalen is Edvardbreen (43 km2), which terminates in a lake at the eastern end of the valley. The glaciers of Kjellstromdalen have received little attention in the scientific literature. One of the only studies was that made by Costner (1925) during the Swedish Page 109 5 km Bragan za - vagen Van Mi jenfjorden Figm·e 4.1 Map showing the location of the field area in central Spitsbergen, Svalbard. Chapter 4: Field area and metlwds Spitsbergen Expedition of 1924. Most of his remarks were based on incidental observations. However, he commented that a number of glaciers in the valley experienced "a sudden and violent push" (Costner, 1925, p.115) . During his visit, Costner found the terminus of Hoganasbreen to be rounded and absent of any surface crevassing. This contrasted markedly with the appearance of its neighbour, Helsingborgbreen, which had an almost vertical lateral margin and a heavily crevassed surface. This suggests that Helsingborgbreen surged in 1924 or in the years immediately prior to that date. The only other glacier in Kjellstromdalen known to have surged is Hyllingebreen, the eastern neighbour of Bjuvbreen. Dowdeswell et al. (1991) reported the occurrence of a surge on Hyllingebreen, beginning in 1968. This surge is thought to have lasted a minimum of 10 years. - 4 . 2.2 Bjuvbreen Bjuvbreen is 2·7 km long and occupies an area of 1·34 km2. It has a simple form and no tributaries. The glacier originates on a corrie backwall, where it receives mass input from avalanches. It flows in a north- north- easterly orientation from an altitude of approximately 650 m a.s.l.. The elevation of the terminus is difficult to determine because ice stagnation makes the lower limit indistinct. The estimated minimum altitude is close to 100 m a.s.l.. Based on the analysis of aerial photographs of Bjuvbreen taken in August 1977 and 1990, the equilibrium line was estimated to lie at an altitude of -500 m a.s.l.. The most conspicuous feature present on Bjuvbreen is a large, wave-like bulge, located 1 km from the head of the glacier (Figure 4.2). The presence of this bulge suggested that Bjuvbreen is a surge-type glacier. A remarkably similar feature has been described on the surge-type Trapridge Glacier by Clarke et al. (1984). Further evidence that Bjuvbreen is surge-type was obtained from a series of aerial photographs. The NP archive contains photographs of Bjuvbreen for 1936, 1956, 1960, 1970, 1977 and most recently 1990. In the 1936 oblique images, the glacier appears to occupy a much greater volume than in subsequent photographs. The terminus extends as far as the large arcuate end mornine, some 200 m further forward than its present position. The lower reaches of the glacier are considerably thicker than at any other observed time. Crevassing is present in this area, although these crevasses do not have a "fresh" appearance. There is no indication that a bulge is present at th~ location observed in future years. By 1956 the lower portion of Bjuvbreen had entered a stage of stagnation. Ice thicknesses in this region had decreased considerably since 1936, shown by the height difference between the glacier surface and the lateral moraines. The glacier terminus no longer extends to the end moraine. There are no crevasses in this lower region, which had begun to accumulate debris. Unfortunately, snow cover obscures Page 111 I 1 Figure 4.2a Norsk Polarinstitutt aerial photograph S77 0902 taken on 16 August 1977. The scale is approximately 1:18,000 at sea level. North is towards the bottom of the page. Bjuvbreen is shown on the right of the photograph. The large bulge is clearly visible as is the stagnating portion of the lower glacier. The point where the proglacial stream bisects the terminal moraine is also seen. The glacier on the left, Hyllingebreen, is ln the late stages of a surge. Page 112 Figure 4.2b Photograph of Bjuvbreen taken from the stagnant lower portion of the glacier. The bulge is about 500 m distant. Page 113 Chapter 4: Field area and methods much of the detail on the 1960 photograph. Continued terminal recession and an increase in debris cover are visible, but no features in the upper portions of the glacier can be discerned. The 1970 photographs show Bjuvbreen continuing in its quiescent phase. A boundary is observed to be developing between the stagnant lower reaches of the glacier and thickening ice up-glacier. The wave-like bulge is seen clearly for the first time in the 1977 photographs. The feature has an arcuate transverse profile, in that the base of the bulge has extended further down-glacier in the centre than at the margins. Down-glacier of the bulge, the ice continues to stagnate and accumulate a mantle of debris. The 1990 aerial photographs did not become available until after both field seasons had been completed and were, therefore, not used in the original classification of Bjuvbreen as a surge-type glacier. The continuing development of the bulge, however, is clearly observed. A thorough discussion of the evolution of Bjuvbreen using combined evidence from aerial photographs and field work is presented in Chapter 5. The above observations of Bjuvbreen on this series of aerial photographs are interpreted in terms of a surge cycle. The 1936 photograph illustrates the glacier in the late stages of its active phase, or shortly after a surge. The photographs from 1956 to 1990 show the evolution of Bjuvbreen through its quiescent phase. The bulge marks the boundary between active ice up-glacier and stagnant ice down-glacier. Thus, the bulge separates the reservoir zone from the receiving zone. There are no recorded observations of a surge on Bjuvbreen other than the 1936 aerial photographs. Therefore, it is not possible to state the length of the surge return period. However, if it is assumed that Bjuvbreen surged during the ten years prior to 1936, its surge cycle is in excess of 60 years. This is longer than the observed return periods of surge-type glaciers in other glacierized areas (Meier and Post, 1969; Post, 1969; Dolgushin and Osipova, 1975). Nevertheless, a long return period for surges on Bjuvbreen is consistent with the observations of Dowdeswell et al. (1991) of protracted quiescent phases on other surge-type glaciers in Svalbard. 4. 2. 3 The programme of field investigations Two field seasons· were undertaken on Bjuvbreen. The first season took place between 7 July-26 August 1989, during which time a series of baseline data were collected by a group of four persons. Three survey stations were established from which a number of measurements were conducted. Ice surface velocities and strain rates at a number of points on the glacier were monitored from fixed locations as often as weather permitted, usually every three or so days. In addition, the elevation of the glacier was surveyed in order to produce a contour map (Fox, 1989). This map provided an extension to the record of topographic changes of the glacier obtained from Page 114 Chapter 4: Field area and metlwds the aerial photographs. The longitudinal profile of the glacier was also surveyed. Bjuvbreen is drained by a single meltwater stream. A gauging station was established on the stream and an hourly record of discharge was collected using automatic loggers. At the same location, an automatic water sampler was sited to collect hourly samples of meltwater for suspended sediment analysis. A second field visit was made between 23 May- 8 July 1990 by another group of four persons, joined for a short period by two additional workers. Several of the investigations made during the initial field season were repeated. Ice surface velocities and strain rates were monitored, and the glacier long profile was re-surveyed. Suspended sediment samples were again collected on an hourly basis from the proglacial stream, once the melt season was underway. It was intended to, again, measure discharge, but problems with equipment prevented these data from being obtained. With the assistance of Norsk Polarinstitutt we were able to carry out radio echo sounding of ice thicknesses and hot water drilling through the glacier to its bed. A considerable amount of logistical planning was required prior to each field season. Access to the glacier was by helicopter. Movements in the field were made on foot and skis, although a snowmobile was used during the 1990 season. The following sections describe the techniques used in the field and the methodology employed in the reduction of the data. 4.3 RADIO ECHO SOUNDING 4. 3 .1 Introduction A important variable in any glaciological study is ice thickness. It is required for the calculation of a number of geophysical characteristics of glaciers, including basal shear stress. In the case of Bjuvbreen it was also important to determine the subglacial topography in order to confirm that the wave-like bulge is a glaciological feature and not the surface expression of the subglacial relief. The technique used was radio echo sounding. The basic principle of radio echo sounding requires that a short electromagnetic pulse is emitted by a transmitter mounted on a moving or stationary platform over the glacier surface (Bogorodsky et al., 1985). As the pulse penetrates the ice it is reflected by inhomogeneities in the glacier and the glacier substrate. These reflections produce a series of echos which return to the surface and are detected by a receiver. If the speed of electromagnetic wave propagation through ice is known, the distance travelled by the pulse, and henc::e the thickness of ice, can be determined by the time delay of the echo. The velocity of electromagnetic waves through a glacier is dependent on the dielectric properties of ice. The velocity of radio waves in ice is equal to the velocity of Page 115 Chapter 4: Field area and methods radio waves in a vacuum divided by the square root of the dielectric constant, or refractive index, of ice (Paterson, 1981). This gives a velocity of 168 m µs- 1 which is similar to a value of 167 ·7 ± 0·3 m µs-1 obtained from an interferometric field experiment carried out by Robin (1975). Calculations made in this thesis will assume a velocity of 168 m µs-1. A substantial amount of radio echo sounding has been undertaken on Svalbard glaciers using a variety of instrument types. Most of this work has been carried out by the Soviets using 440-620 MHz equipment (Macheret et al., 1984) and a joint SPRI- NP programme using a 60 MHz radar unit (Dowdeswell et al., 1984). An interesting feature of many echo returns from Svalbard glaciers is the presence of an internal reflecting horizon (Bamber, 1987; Macheret et al., in press) (section 3.4.2). Bamber explained these horizons as meltwater channels flowing en glacially. A common problem with both the Soviet and SPRI-NP radio echo sounding experiments was the lack of a bed return from the accumulation zones of many glaciers (Dowdeswell et al., 1984). The accumulation areas of many Svalbard glaciers are probably at the melting point during the summer. Therefore, absorption and scattering of radio waves by meltwater and soaked firn will occur, thus obscuring the bed return signal (Bamber, 1987). Hagen and S~trang (1991) recently obtained bed profiles from many accumulation zones where the Soviets and SPRI- NP had been unsuccessful. This was achieved using a low frequency radar unit operating at 8 MHz. 4. 3. 2 System specification and design Measurements of the thickness of Bjuvbreen were made during the spring of 1990 using the Norsk Polarinstitutt portable radio echo sounder. The instrument was similar to the MARK II monopulse radar described by Sverrison et al. (1980). This radar was a broad band instrument operating at a centre frequency of 8 MHz (A. S~trang, personal communication). The frequency window of the receiver was 0· 1-10 MHz. The transmitting unit was attached to the centre of a 14 m antenna. Power for the transmitter was supplied by a 12 V car battery. The transmitter was seperated from the receiver by 30 m. The return signal was received by a second 14 m long antenna and was displayed on an oscilloscope. The antennae were placed transversely across the glacier. The radio echo sounder was capable of measuring depths between 30 and 400 m with o.. V"e6olulion, oF ~ YV\. 4. 3. 3 Geographic location of radio echo soundings on Bjuvbreen The system described above was operated at several locations ori. the glacier surf ace. Therefore, the results obtained represent point soundings of ice thickness and Page 116 Chnpter 4: Field area and metlwds not a continuous profile. The instrument was sufficiently compact that it could be easily transported between successive locations. The radio echo sounding investigations had two objectives. The first was to determine the thickness of ice at a number of locations on the glacier for use in the calculation of various glaciological parameters. The second was to test the hypothesis that the wave-like bulge is a glaciological feature and not a manifestation of the subglacial topography. A series of eight points was sounded, at approximately equally- spaced intervals up the centreline of Bjuvbreen. These soundings began on the upper slopes of the bulge and continued as far towards the corrie headwall as crevasses would permit for safety considerations. In addition _to these locations, two points were sounded either side of the centreline, midway between it and the glacier margins. Again, the number of measurements made was restricted by the presence of crevasses. The spatial coverage of investigations is illustrated in Figure 4.3. The location of the echo sounding points were marked with 2 m long stakes placed in the glacier which enabled them to be surveyed at a later date. An attempt was made to measure the thickness of ice at the base of the bulge. However, the minimum thickness detectable by the instrument was 30 m. No clear reflection was obtained at this site because the glacier thickness was believed to be outside the operating range of the echo sounder. This was later confirmed when hot water drilling was carried out at this location. The depth of the drill hole at this location was 18 m. In view of this result, no further soundings were made down-glacier from this point. 4. 3. 4 Depth of firn for correction of echo sounding data The velocity of electromagnetic waves through a glacier is dependent on the density of ice. In true glacier ice the velocity remains constant, but in firn the propagation speed is higher (Robin, 1975) due to a greater proportion of air. For measurements made over the large polar ice sheets of Greenland and Antarctica, the effects of firn layers up to several hundred metres thick must be accounted for by applying a correction factor to the original data. However, for fim depths of only a few metres it is not necessary to correct the data. To determine the firn depths at the echo sounding locations on Bjuvbreen, measurements were made at each site using an avalanche probe. A hard, impenetrable layer was assumed to mark the upper surface of true glacier ice. The excavation of several snow pits confirmed this assumption. The recorded firn depths did not exceed 2·5 m at any location and, therefore, it was not necessary to correct the echo sounding data. Smith and Evans (1972) considered that rain or meltwater soaked firn may affect radio wave velocity and, hence, lead to errors in ice thickness calculations. Page 117 I I \ \ \ \ \ \ \ Figure 4.3 \ \ \ \ \ \ \ ' \ .,/ / \ \ \ ' \ / ,," / / ' \ ' \ \ \ ' \ ' / \ / ' ' / / \ \ \ ,,.---------------------------, \ 0 * R1 crest of bulge 100 200 m ' -. -- - R 51C R 6 1C * R12 / / / - - - .,.,... / / / / I I I I \L._ _______ _ _ _ ______ __ ___. Map of Bjuvbreen showing the spatial coverage of radio echo sounding points. Point R12 is located approximately 100 m from the crest of the . bulge. Page 118 Chapter 4: Field area and metlwds However, Dowdeswell (1984) believed that soaking of the firn layer on Svalbard glaciers was unlikely to occur during spring-time before the start of the melt season. In a drill core taken from Lomonosovfonna, Spitsbergen, in summer, only the top 2 m of fim was found to be soaked (Kotlyakov et al., 1980). Therefore, in view of the above, no corrections were made to the echo sounding measurements made in this thesis. 4.4 GROUND SURVEY NETWORK AND GLACIER TOPOGRAPHY 4. 4 .1 Introduction Much of the work undertaken cm Bjuvbreen involved ground surveying techniques. Ice surface velocities and strain rates were measured repetitively over short intervals of time. These measurements were made to fixed points on the glacier surface. In addition, the three dimensional coordinates of a large number of points on the surface were obtained, in order to produce a contour map (Fox, 1989). The glacier longitudinal profile was surveyed during each field season to detect changes in surface slope as the quiescent phase progressed. This section gives an overview of the local survey network and describes the measurement of the glacier long profile. The following section deals with the measurement and reduction of velocity and strain data, and will include a discussion of the methodology specific to that study. 4. 4. 2 Survey control and local triangulation network A reconnaissance field survey of Bjuvbreen was carried out in 1986 by J.O. Hagen (NP). The station established for that survey remained in good condition and was used again during the 1989 and 1990 field seasons. This station (Station 1) was located on the highest point of the terminal moraine (Figure 4.4) and was aligned approximately with the central longitudinal axis of the glacier. It afforded a good view of most of the lower glacier, the bulge and a significant portion of the upper glacier. Unfortunately, however, it was not possible to survey a large area of the glacier towards the backwall, due to curvature of the glacier surface. Furthermore, morainic material obstructed much of the lateral margins lower down the glacier. Two additional survey stations were, therefore, established so that almost the whole of the glacier surface was visible. The design of each survey station was similar. A 1 m long wooden stake was secured in the ground and stabilised with a cairn of rocks. A cross notch was carved in the upper end of each stake to enable centering of the survey instrument. Fluorescent marker tape was wrapped around the stakes to aid their visibility from other stations. The two additional survey stations were located on the eastern and western lateral margins, approximately opposite each other (Figure 4.4). There were very few Page 119 T R I A N G U L A T I O N N E T W O R K 0 ( ! ) N a v i g a t i o n m a r k e r 5 k i l o m e t e r s N L a n g s t a k k e n 8 2 2 m 8 H u t c h i m n e y S t a t i o n 2 , F i g u r e 4 . 4 L o c a t i o n s o f s u r v e y s t a t i o n s a n d c o n t r o l p o i n t s . M o s t o f t h e s u r v e y i n g w a s u n d e r t a k e n f r o m S t a t i o n 1 w h i c h w a s l o c a t e d o n t h e h i g h e s t p o i n t o f t h e t e r m i n a l m o r a i n e o f B j u v b r e e n . T h e c o n t r o l p o i n t s w e r e u s e d t o c h e c k t h e a c c u r a c y o f e v e r y c o m p l e t e s u r v e y . L a _ n ~ s t a k k e n ( 8 2 2 m ) a n d t h e h u t a r e l o c a t e d i n F i g u r e 4 . 1 . T h e n a v i g a t i o n Chapter 4: Field area and methods locations that were stable enough for the siting of a survey station. A large proportion of the lateral margins was composed of ice-cored moraines. Areas that were not ice- cored were generally high-angled talus slopes. The locations where the stations were established were chosen because they were level and appeared to have been stable for some time, judging by the degree of lichen colonisation. It was necessary that the survey work was carried out with greatest possible accuracy. Therefore, a series of control points were defined that would enable the accuracy of each set of measurements to be determined (Figure 4.4). The objects chosen for this purpose had to satisfy a number of conditions. Firstly, they had to be visible on every day of surveying. Coastal fog and persistent low cloud characterised much of the weather during the field seasons. This meant that objects in the valley had a greater chance of being visible on every occasion than points on mountain tops. Secondly, the points had to be visible from at least two of the three survey stations to allow inter-comparison between measurements. Thirdly, the control points had to be located in a wide horizontal arc to avoid a high degree of colinearity. Finally, the objects had to have a sufficient resolution that would enable precise surveying on each occasion. Explicit directions were given of the exact location to be surveyed on each point, to ensure consistency. The points used to control the surveys from each station are shown in Table 4.1 . Survey Station Station 1 Station 2 Station 3 Table 4.1 Control Points Used Station 2 Station 3 Summit cairn of Langstakken Hut Chimney (mouth of Lundstromdalen) Station 1 Station 3 Summit cairn of Langstakken Hut Chimney (mouth of Lundstromdalen) Station 2 Station 3 Summit cairn of Langstakken Control points used to reference measurements from each survey station. Refer to Figure 4.4 for locations of control points All surveying was carried out using standard survey equipment. Identical instruments were used during both field seasons. A Wild tripod was positioned above the stakes at each survey station. At Station 1, the tripod was secured in the moraine and left in position for the duration of each field season. All the velocity and strain measurements were made from Station 1. By maintaining the tripod at a constant height, reduction of the data was simplified. At stations 2 and 3, tripods were erected when they were required. Mounted on the tripod and centered above each stake was a Wild T-2 optical Page 121 \ Chapter 4: Field area and methods theodolite. Horizontal and vertical angles were measured using this instrument, to the nearest second of arc. An EDM, electromagnetic distance measurer (Wild DI-3000), was attached to the top of the theodolite and was powered by a rechargeable battery. The DI-3000 measured the slope distance from the instrument to a Wild GPH3Z prism reflector which was stationed at a marker stake or a topographic survey point. The EDM was always used in the "normal" mode. When used in this way, the instrument takes a number of measurements, usually about nine, over a time interval of 3.5 seconds. If the infra-red beam is interrupted at any stage, the number of readings taken and the time interval used will be greater. The distance recorded is the cumulative mean of all the measurements. This mode provide~ a method of reducing the standard deviation between readings and, therefore, increases the accuracy of the measurements. When operated in the normal mode, the EDM has a manufacturer's rated accuracy of 5 mm± 1 ppm (Wild Technical Specifications). All measurements were made with the survey instruments in the forward theodolite position only. During survey work it is preferable to repeat all measurements in the reverse theodolite position, in order to reduce errors. However, a design limitation of the instruments prevented forward and reverse measurements being made without removing the EDM from the theodolite. As a check on the accuracy of the surveys, horizontal and vertical angles to all the control points were measured in forward and reverse positions at the start and completion of every round of targets without the EDM in position. Survey errors are discussed in more detail in section 4.5.6. 4. 4. 3 Glacier topography and longitudinal profile The topographic survey of the glacier surface undertaken in 1989 was described by Fox (1989). Approximately 500 points were surveyed, giving a good spatial coverage of most of the glacier. Some areas of the upper glacier were not surveyed because crevasses made safe movement difficult. Fox (1989) used the data collected to produce a contour map of the glacier surface. The longitudinal profile of the glacier was surveyed during each field season. A similar survey was made from Station 1 by J.O. Hagen (NP) in 1986. In order that comparisons could be made between each survey, the bearings to the control points used during the 1986 survey were reproduced as closely as possible in the 1989 and 1990 surveys. Therefore, the alignment of the axis that was surveyed should be almost identical on each occasion. Once the line of sight had been established by the observer, the person carrying the reflector was directed to within acceptable limits of it by radio communication. The reflector was never permitted to deviate more than 30' in the horizontal plane from the specified line. This represents a maximum deviation from the Page 122 \ Chapter 4: Field area and methods desired line of less than 25 m for the poi.nls -furthest up-glacier. Raymond and Harrison (1988) considered a deviation of 50 m either side of the centreline acceptable for their long profiles of Variegated Glacier. On Bjuvbreen a reading was generally taken every 25 m up-glacier, although this interval was shortened if there were noticeable changes in slope. 4.5 GLACIER MOTION AND DEFORMATION MEASUREMENTS 4. 5 .1 Introduction Variations in velocity on a numj)er of glaciers have been observed over timescales ranging from hours and days to between seasons (e.g. Meier, 1960; Iken, 1973; Hooke et al., 1989). Such variations have been observed on normal and surge- type glaciers. Research on Variegated Glacier, prior to its 1982- 83 surge, demonstrated the occurrence of both low-amplitude, short-term velocity fluctuations (Harrison et al., 1986) and also more dramatic events, known as mini-surges (Raymond and Malone, 1986; Kamb and Engelhardt, 1987). These variations in the motion of a surge-type glacier during late-quiescence are likely to be related to the initiation of a full surge (Raymond, 1987). During fieldwork on Bjuvbreen, the velocity and deformation of the glacier surface were measured over short intervals to elucidate the nature of: (i) any short-term fluctuations over the course of the field seasons, (ii) their change as the melt seasons progressed, and, (iii) any variations in their character between field seasons. 4. 5. 2 Field methodology A description of the equipment used to measure glacier velocities and strain rates was given in Section 4.4.2. The measurements were made from Station 1 to a series of markers fixed in the glacier surface. At the end of the 1989 field season, the markers that had been used were removed to comply with environmental regulations. Therefore, the locations of the markers used during the second field season were not identical. However, the markers that were established in 1990 were in locations and a pattern similar to that used in the previous year. The marker locations are illustrated in Figure 4.5. The markers consisted of 3 m long wooden stakes, with an upper surface diameter of approximately 30 mm. At each marker location the surface snow was excavated, and a hole in the ice was hand-augured to a depth of 1 m, in which a stake was placed. A mixture of salt and water-saturated snow was added to each hole to aid the freezing-in of the stake. Unfortunately, only one prism was available during fieldwork and, therefore, each marker had to be visited in turn to obtain a reading. Page 123 ' .,. / \ \ \ \ / / \ \ \ / / \ \ \ ,."' ,."' \ \ \ \ \ \ / \ \ / ,."' ,."' \ \ \ \ \ \ / / \ \ \ / / / / \ 0 \ \ \ \ \ 10 0 ,I I I J /8 b()se of bul"e;' , _ - 5' A:_-,;• 'j - - - o, - - C?....§ .lll_o 1•_?~_3 4e/ Diamond 3 Diamond f'if'-b 200 m Figure 4.Sa Map of Bjuvbreen showing the spatial coverage of velocity and strain . markers during the 1989 field season. Velocity markers are indicated by solid circles and strain markers by empty circles. Velocity markers 30- 44 were also used for strain measurements. Markers 1-8 and diamonds 1-3 are located at the base of the bulge. For clarity, only selected strain diamonds and triangles are identified individually. Page 124 ' \ \ \ \ ,, • \ \ \ . \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 0 D i a m o n d 3 )?;-::o o, cf-e-0 Ys / ~~/ . 6 Diamond 2 crest o - - bu~e. -- -------- / baseof liulg-- Diamond 1p-:_-. 4 a• -v 2. 't{3 10 0 200 m / I I / L---------------'1----- - - Figure 4.Sb Spatial distribution of velocity and strain markers on Bjuvbreen during . the 1990 field season. Velocity markers 1-4 are located at the base of the bulge. Page 125 \ I Chapter 4: Field area and metlwds The following survey data were recorded for each stake: horizontal and vertical angles, slope distance and the time of measurement. In addition, horizontal and vertical angles were taken to the control points (section 4.2.2), in face left and face right positions, at the start and completion of every complete round of all the markers. The EDM was corrected for the prevailing barometric pressure and air temperature, which were measured at camp, before the start of each survey. 4.5.3 The local coordinate scheme In order to calculate the velocity and strain rate at various points on the glacier surf ace, the data were first . · reduced to coordinates. The coordinates of the markers were calculated with respect to a locally-defined coordinate scheme. The grid system consisted of coordinates expressed as (x, y, z). The grid axis was defined arbitrarily, with Station 2 representing the reference coordinate. The coordinates of this point were set so that x = 5693-504 ., y = 5000 and z = 1298·200 . All other coordinates were calculated relative to this point. 4. 5. 4 Reduction of data. 1: velocity measurements Horizontal and vertical velocities and marker trajectories were calculated from the difference of successive positions of the markers with respect to the coordinate scheme described above. The data required to calculate the velocities were horizontal and vertical angles, slope distance and the time of measurement. The time was specified in Julian days and decimal parts thereof. The Julian clock was given an arbitrary origin at 0000 hours, Alpha time (GMT +1) on January 1 at the beginning of each year of fieldwork. Thus, readings began on Julian day 194 in 1989 and on Julian day 165 in 1990. Once the marker readings had been converted to coordinates, the next stage of the reduction process was to express successive coordinates of the same stake as horizontal and vertical displacements relative to the previous readings. These transformations were carried out using a series of standard trigonometric functions. Velocities were calculated-as displacements per unit of time since the preceding reading. The angle of the velocity vector was also computed. 4. 5. 5 Reduction of data. 2: strain rates In a detailed study of glacier flow, it is important to determine not only the absolute velocity of various points, but also strain rates, or velocity gradients (Paterson, 1981). The strain rate is defined as the change in the distance between two stakes in a unit of time, expressed as a fraction of the original length. It can be expressed logarithmically as: Page 126 Chapter 4: Field area and methods (4.1) (Nye, 1959), where L1 and L2 are the initial and final lengths and L1t is the elapsed time. A number of workers (e.g. Meier, 1960; Hodge, 1972; Meier et al., 1974) have used standard velocity data to calculate velocity gradients and hence strain rates. Van der Veen (1989) presented a numerical scheme which calculates strain rates from ice velocities and glacier geometries. However, the above methods only give an indication of strain components in terms of x and y. A proper understanding of the controls on glacier flow requires the magnitudes of all principal components of strain at the glacier surface (i.e. x, y, and z) to be known (Paterson, 1981). A method of obtaining these values based on field data was described by Nye (1959). In this study, two standard methods were used to obtain strain rate values on the surface of Bjuvbreen. The first method followed the procedures described by Nye (1959). Arrays of stakes were set out on the glacier surface in a diamond pattern with a further stake at the centre, thus producing a series of triangles. The stakes were arranged so that the triangle sides were parallel, perpendicular and diagonal to the assumed direction of flow (taken to be down-glacier). The locations of these strain diamonds in the 1989 and 1990 field seasons are shown in Figure 4.5. Reduction of the field data to strain rates followed the method discussed in Nye (1959). Using equation 4.1, logarithmic strain rates were calculated individually for all eight triangle sides in a diamond array. The strain rates for parallel sides were averaged to provide four values of£ corresponding to the directions 0°, 45°, 90° and 135°. Nye (1959; p.413) described a least squares method of obtaining Ex and £2 and also the shear strain rate, £zx. From these values, the magnitudes of the principal strain rates £1 and £3 in the zx plane were computed. In addition, the direction of the principal strain rate was calculated. The vertical component of strain, £2, was also found. The method described above required five stakes for each diamond array. The number of stakes available in the field was limited, however. Therefore, a second method of obtaining strain rates, which required fewer stakes, was used on the upper portion of Bjuvbreen. Two parallel lines of stakes, approximately 75 m apart, were placed running up the centre of the glacier. This line extended from the crest of the bulge to as far back as was safe to travel (Figure 4.5). The stakes in these lines were offset so that they formed a series of triangles. Using a Mohr circle construction (Ramsay, 1968), strain rates can be calculated from the measured deformation of the triangles. This method has been used glaciologically by Hambrey and Muller (1978), Raymond et al. (1987) and Williams and Knight (1987). Page 127 11 Chapter 4: Field area and methods 4. 5. 6 Calculation of survey errors The previous two sections described the methods employed in the reduction of the survey data to horizontal and vertical velocities and strain rates. Despite using highly accurate survey equipment, errors inevitably occurred during these attempts to measure small displacements over large distances. All the surveying of ice motion on Bjuvbreen was conducted from the survey station which was located near the glacier terminus (Figure 4.4). This was unsatisfactory, since it meant that the bulge and the upper glacier, the regions of greatest interest, were between 1 ·5 km and 2·5 km from the survey instruments. However, this problem was unavoidable. Unfortunately there were no suitable locations for survey stations up-glacier from the bulge. In this region, the valley slopes were too steep and susceptible to movement to guarantee the long-term stability of a survey station. A thorough analysis of errors in the measurement of glacier movement was presented by Hodge (1972). In other studies, less rigorous methods have been employed. Uncertainties in angle measurements are commonly determined using standard survey techniques. These usually involve taking repeat readings with the instrument in forward and reverse positions and using the average of the two angles obtained ( e.g. Meier, 1960; Iken and Bindschadler, 1986). A different approach to error quantification was used by Hooke et al. (1983a). They determined the uncertainties in their velocity measurements on Storglaciaren by assuming that the difference in strain rate with time between two adjacent strain diamonds on the glacier centreline remained approximately constant. The extent to which the strain rates were not constant was a measure of the uncertainty in the survey measurements. The calculated errors were then applied to velocity data. The following method was used to assess the accuracy of the survey measurements made on Bjuvbreen. For each target, the survey measurements consisted of horizontal and vertical angles measured using the theodolite and the distance measured using the EDM. Measurements were made only in the forward theodolite position. Reverse position measurements could not easily be made due to the configuration of the instruments. With the EDM attached above the theodolite it was not possible to invert the theodolite telescope through the zenith in order to bring the eyepiece to the observer's eye. Thus, the EDM would have to be removed each time a reverse position reading was required. This would have introduced an error, caused by moving the position of the survey instruments relative to the base reference point and also to the previous measurement. Although this effect was not quantified, noticeable movements of the theodolite were apparent when the EDM was removed in trials. It was considered that repeating this action for every target on the glacier would generate Page 128 Chapter 4: Field area and metlwds unnecessary errors and, therefore, reverse position measurements of horizontal and vertical angles were not made. Thus, another scheme was devised to assess the accuracy of the angular measurements made to markers on Bjuvbreen. At the start and finish of each complete survey of the targets on the glacier, a check on the consistency of the measurements was made by sighting and taking angles to a series of control points (Figure 4.4). These measurements were made with the theodolite in the forward and reverse positions, without the EDM in position. The standard error of the angles made to these points from the survey station woc.s calculated. The standard error for the horizontal angles made in 1989 from station 1 was Q0 00' 05". Using conventional trigonometry, this corresponds to an error of ± 44 mm for the closest target and ± 59 mm for the furthest target from station 1. The standard error of the vertical angles was 0°00'04". The corresponding target errors were ± 35 mm and ± 50 mm for the closest and furthest targets. In 1990, the same control points and error estimation procedures were used. The standard error of the horizontal angles was 0° 00' 05", corresponding to target errors of± 41 mm for the closest point and± 58 mm for the furthest point. The vertical angles to the control points had a standard error of 0° 00' 04". The target errors were thus, for the closest point, ± 32 mm, and for the furthest point, ± 46 mm. In addition to uncertainties in angle readings, there was also an error associated with the EDM distance measurements. These distance errors must be added to the angle uncertainties to obtain the total horizontal error. The Wild DI-3000 has a manufacturer's rated accuracy of 5 mm± 1 ppm. For distances measured from station 1 in 1989, the error for the nearest point was 5 mm± 1·8 mm, and 5 mm± 2·3 mm for the furthest point. Intolerances of similar measurements made from station 1 in 1990 were 5 mm± 1·7 mm and 5 mm± 2·4 mm. To obtain an estimate of the total errors resulting from the surveying, the worst case was calculated. Therefore, the maximum EDM error was added to the theodolite error to determine the total horizontal error. The resulting intolerances of measurements to the nearest and furthes~ points from both survey stations during the two field seasons are given in Table 4.2. The errors for all other targets on the glacier were calculated using similar procedures. The errors given in Table 4.2 are for successive measurements of the same stake. Thus, in order to apply these error estimates to the velocity data, the uncertainty was divided by the time interval between successive surveys. The error estimates were also applied to the strain calculations, by expressing the uncertainty as a fraction of the original side length. These estimates of the error in the velocity and strain measurements are considered to be reasonable assessments of the actual accuracy. A check on this Page 129 Chapter 4: Field area and methods assumption comes from a survey undertaken from station 1 to an NP trigonometric station, located on Crednermorenen. This was carried out by Fox (1989) to reference the topographic map of Bjuvbreen relative to the NP grid. The approximate distance between the two points was 10· 15 km. Ten separate measurements were made, with the instrument being resighted on each occasion. The standard errors of the horizontal and vertical angles and the distance measurements were calculated. These errors were Target Horizontal Error Vertical Error 1989:5 ± 50·8 mm ±35mm 1989: 44 ± 66.:3 mm ±50mm 1990: 1 ±47·7 mm ±32mm 1990: 13 ± 65·4 mm ±46mm Table 4.2 Total survey errors for the targets closest to and furthest from Station 1 during the 1989 and 1990 field seasons. The total error is the sum of angle errors from the theodolite and distance errors from the EDM. The actual error of a particular velocity measurement is obtained by dividing the total error for that target by the length of time between successive measurements. 0° 00' 20" for the ten horizontal angles and 0° 00' 16" for the vertical angles . The standard error of the ten EDM readings was 2 mm. When these estimates are proportionally reduced to distances typically measured on Bjuvbreen, there is a good agreement with error estimates prepared for the velocity and strain targets. Other sources of error which were considered were varying meteorological conditions, operator misreadings and curvature of the geoid. The prevailing weather conditions at the time of each survey can influence the accuracy of the measurements. Direct sunlight causes heat shimmer which distorts the electromagnetic beam and also makes sighting the target difficult (e.g. Andreason, 1985). This was not a problem as all surveying on Bjuvbreen was carried out during overcast conditions. However, strong winds were a problem on a number of occasions because they made the survey instruments vibrate strongly. This sometimes led to very large errors in the calculation of stake coordinates. Where these coordinates obviously did not represent the actual position of the stake concerned, they were discarded. Corrections were applied to the EDM at the time of surveying for the prevailing atmospheric pre~sure and temperature which were measured at the camp site. There were some occasions when it was apparent the operator had made an incorrect reading. These misreadings resulted in unrealistic stake coordinates. These faulty data points were not used in the velocity and strain calculations. A correction for variations in the curvature of the geoid was not applied to the data. Hodge (1972) and Bindschadler et al. (unpublished) reduced their Page 130 Chapter 4: Field area and metlwds glacier survey data to curvilinear coordinates to remove any uncertainty due to geoid curvature. However, on Bjuvbreen, the change in position of a stake between initial and final surveys was such a small proportion of the overall distance between station and target that a similar correction was not considered necessary. 4.6 GLACIER HYDROLOGICAL . INVESTIGATIONS 4. 6 .1 Introduction The links between the motion of a glacier and the behaviour of water within it have been demonstrated in a numbe_I of studies (e.g. Iken, 1973; Iken and Bindschadler, 1986; Kamb and Engelhardt, 1987). An increase in subglacial water pressure is believed to cause an increase in the surf ace velocity of normal glaciers and is also thought to be responsible for more dramatic events such as mini-surges and surges (section 1.3). Therefore, it is important to monitor the behaviour of water within a surge-type glacier to determine its influence on ice movement and surge initiation. A number of hydrological measurements and investigations were undertaken on B juvbreen during the two field seasons. The data collected will be discussed later in this thesis in comparison with the results of the motion studies. The timing of the field seasons was such that, cumulatively, they spanned almost the entire summer melt season. This provided an opportunity to analyse the dynamics of Bjuvbreen from the onset of melting during early spring to the closing down of the melt system as winter approached. The methods employed during these investigations are described in the following sections. 4. 6. 2 Proglacial stream discharge monitoring Fluctuations in sub glacial water pressure are an important cause of variations in the flow of a glacier. However, it is inherently difficult to measure, directly, water pressure beneath a glacier since it requires the drilling of access holes to the bed (section 4.6.4). Drilling equipment was not available during the first field season. Therefore, the discharge of the proglacial stream was measured to obtain an idea of the relative amount of water within the glacier (Tangborn et al., 1975) and, as such, a possible indirect indicator of subglacial water pressure (cf. Paterson, 1964; Hodge, 1972). For certain glaciers, the measurement of proglacial discharge has been complicated by the presence of more than one outflow stream. However, on Bjuvbreen this problem was not encountered. All proglacial meltwater was discharged from one large portal situated in the centre of the glacier terminus. The water flowed in a braided channel immediately in front of the snout before carving a channel through an area of Page 131 Chapter 4: Field area and metlwds Aufeis. The stream then broke through the large terminal moraine and flowed in an incised channel through glacigenic sediments to its confluence with Kjellstromelva, approximately 2·5 km from the glacier terminus (Figure 4.1). To obtain meaningful discharge results, selection of a site suitable for the gauging station was important. This selection inevitably involved some compromise. Pro glacial stream discharge should be measured · as close to the glacier terminus as possible to avoid 'contamination' of the record by water from other sources, such as snowmelt from the surrounding slopes. In order to select the best site possible, the character of the stream was observed for 2- 3 days at the start of each field season. A number of criteria was used in the selection of a gauging site. Firstly, a single, relatively stable channel was required. Any major changes in channel geometry would have necessitated a new rating curve to be determined. Secondly, the selected reach had to be sufficiently deep so that the pressure transducer would remain continuously submerged in the flow. Finally, the stretch of channel had to wadeable, since manual gauging was required to calibrate the automatically recorded measurements. The section of channel found to be most suitable was located on the distal side of the terminal moraine. In this area, the channel flowed in a trough incised into sediments, approximately 7 m wide and 3 m deep (Figure 4.6). At the base of the trough, the stream flowed over glaciofluvial sediments. At most locations, the flow was concentrated in one channel. Furthermore, significant lateral wandering of the channel was prevented by the high-sided trough. The gauging station was established in this trough approximately 350 m from the emergence of the steam at the glacier terminus. No tributaries entered the channel upstream from this point. The total catchment area of the stream above the gauging station was 4· 15 km2. Approximately 32 % ( 1 ·3 km2) of the catchment was glacierized. The aim was to collect a regular and continuous record of stream discharge, at a sampling interval of 1 hour, throughout both field seasons. To collect these data manually was impractical and, therefore, automatic methods were used. The basic components of the system were a pressure-sensitive transducer, a data logger and a power supply. For the 1989 field season, a pressure transducer was obtained on loan from the Department of Geography, University of Aberdeen. The transducer (Shape Instruments SH3100) was mounted in a 1 m length of dexion fr~me using jubilee clips. The frame was attached to a large boulder, situated in the chosen stretch of channel, with 6 rawbolts. Since there were no outcrops of bedrock on the stream bank, the transducer had to be attached to a boulder. This obviously increased the likelihood of it moving at some stage during the monitoring period. To minimise this possibility the largest boulder was used and then stabilised as best as possible in the stream channel. Page 132 Figure 4.6 Photograph of the proglacial stream on the distal side of Bjuvbreen's terminal moraine. The gauging site was located in the right middle ground. Page 133 Chapter 4: Field area and methods This proved successful, with only slight movement occurring during one period of particularly high discharge. The transducer was connected to a data logger (Grant Instruments SQ8) with a 15 m length of cable, allowing the logger to be placed at a safe distance from the stream. The data logger accepted a voltage signal as input, over a preset range 0-125 mV at intervals of 0-5 mV. The pressure transducer was configured to measure water depths between O and 1 ·5 m. This allowed the system to detect changes of 0-6 cm in stage. Before recordings were started, the transducer was calibrated in a known depth of water and allowed to stand for 24 hours in the stream. This period was required for the sensitive silicon diaphragm in the transducer to reach thermal equilibrium with the surrounding water. Readings were taken on the hour, every hour throughout the field season. An almost continuous record was obtained, with only short intervals where there were missing data. These data were missed due to the obstruction of the transducer by sediment following particularly high discharges. In order to convert the transducer readings recorded on the data logger to discharge values, the stream channel was gauged manually at a number of different flow levels to compile a rating curve. To ensure that the same cross-section of channel was used for gauging on each occasion, a length of rope was secured at each bank and stretched across the stream. Markers were placed on the rope at 50 cm intervals. At each of these markers, the depths from the rope to the water surface and the stream bed were recorded. The velocity of the stream was measured at 0·6 of the water depth using an Ott Current Meter (Type 02). Discharge was then calculated using a standard velocity-depth relationship (British Standards Institute, 1964). Similar measurements was planned for the 1990 field season using a Sandhurst Scientific Instruments pressure transducer (model LH-177). Discharge from the glacier began to emerge during the third week of the field season. Unfortunately, once the equipment was installed at the same gauging site as the previous season, the transducer developed a fault and was unable to take any measurements. As an alternative to a continuous record-of discharge, manual gauging of the channel was attempted. However, conditions in the stream were too hazardous to permit these measurements from being made. Consequently, no record of discharge is available for the 1990 season. 4. 6. 3 Proglacial stream suspended sediment concentrations. The sediment load of a proglacial stream can be divided into that in traction with the stream bed and that held in suspension in the flow (Fahnestock, 1963). It is the suspended sediment that gives glacial streams their characteristically turbid appearance. Page 134 Chapter 4: Field area and methods The temporal concentration of suspended solids does not remain constant, and has been shown to vary on both short (a few hours) and long (seasonal) time-scales (e.g. 0strem, 1975; Gumell, 1982). Suspended sediment properties of streams draining surge-type glaciers have also been investigated. Humphrey et al. (1986) observed pronounced peaks in stream turbidity following mini-surges on Variegated Glacier prior to its 1982-83 surge. They interpreted these results in terms of water behaviour beneath the glacier. Therefore, in the absence of being able to monitor hydrological conditions beneath the glacier directly, a useful body of information can be assembled from measurements made in the proglacial area. The location of the sampling site was an important consideration (e.g. Gurnell, 1982). Increasing distance from the glacier front might lead to the introduction of non- glacial sources of suspended sediment, confusing the record. In addition, the immediate site locality should be relatively stable. With these considerations in mind it was decided to site the sampling station at the same location as the discharge measuring equipment. No suitable site was found further upstream. Furthermore, by siting the two experiments at the same location, efficient use was made of personnel resources. An Epic Products 1101 automatic water sampler was used to collect samples of meltwater at hourly intervals through the field seasons. The sampling was controlled by programming a small microprocessor. Water was drawn into the sampler through a 7·5 m long hose following the release of a vacuum from a glass container. The hose was secured in such a position that the intake vent was wholly submerged in the flow. The time taken to collect a sample was approximately 30 seconds, which 0strem (1975) considered to be adequate for obtaining a representative sample in turbulent streams. Each sample of 500 ml was deposited sequentially in an individual plastic container by a rotating outlet pipe. The instrument had a capacity of 24 samples and, therefore, had to be emptied each day. It was considered impractical to return a large quantity of bulk meltwater samples to the laboratory for analysis. Therefore , the water samples were hand-filtered in the field and the suspended sediment residues and filter papers returned in petri dishes for subsequent concentration analysis. Prior to filtering, the contents of the sample bottles were shaken vigorously to ensure that material which had settled out since sampling was re-suspended. Samples of 250 ml were hand-pumped through a filter unit containing a preweighed Whatman 40 fil ter paper. These fil ter papers had a penetration pore size of 8 µm. A number of studies (Gumell, 1987) have demonstrated Whatrnan 40 to be a suitable choice of filter paper. Particles smaller than 8 µm in size have been found to constitute less than 5% of the volume of suspended sediments in glacial streams (Beecroft, 1981; Bezinge, 1987). The meltwater was expelled from the end of the filter unit and was not retained. The filter papers containing the sediment Page 135 Chapter 4: Field area and methods residue were placed in petri dishes. By incorporating a cling film seal, two samples could be accommodated in each dish without cross-contamination, therefore reducing the volume of material to be transported back to the laboratory. The samples returned to the laboratory were ashed to determine suspended sediment concentrations, following the procedure described by 0strem and Stanley (1969). The filter papers and their residue were placed individually in crucibles. Using an electronic fine balance, each crucible was weighed with and without their filter papers. The crucibles were then placed in a furnace at 800 ·c for 2 hours to burn off the filter papers. After the samples had been in the furnace for the required length of time, the crucibles were removed and placed immediately in a desiccator to prevent condensation forming while they cooled down. The crucible and remaining residue were then weighed, the residue brushed out and the crucible re-weighed to confinn accuracy. The amount of sediment on each paper was calculated by subtracting the weight of the crucible from the weight of the crucible and residue. A small amount of the residue remaining (0·01 %), representing the remnants of the filter paper, was also subtracted. The suspended sediment concentrations were determined by dividing the weight of residue by the filtrate volume. 4. 6. 4 Hot water drilling Investigations in boreholes which have penetrated to the bed of glaciers have yielded much information on, for example, the nature of the substrate, subglacial motion and basal hydrology (Engelhardt et al., 1978; Hodge, 1979; Clarke et al., 1984; Kamb et al., 1985; Ik:en and Bindschadler, 1986). The basic principle is simple. A vertical hole is melted with a hot fluid, usually water, pumped into the ice under pressure. Drilling rates of up to 5 m per minute may be be achieved in the upper layers of a glacier (Olesen, 1989) although as depth increases rates become progressively slower as the drilling fluid loses heat. The theoretical basis of the design of an efficient drilling system has been explored by Iken et al. (1989) and Humphrey and Echelmeyer (1990). However, in the present study, drilling to great depths was not envisaged and, as such, no special considerations for heat loss had to be made. Hot water drilling was intended to provide access holes down which pressure transducers could be lowered to record fluctuating basal water pressures. The Norsk Polarinstitutt hot water drill was used on Bjuvbreen during the 1990 field season. This system has a history of successful operation on sub-polar glaciers in Svalbard (A.C. S~trang, personal communication) and in colder ice in Antarctica (Orheim et al., 1991). At least 200 1 of water was required before downhole drilling could commence. During the operational season, there was no supraglacial melt available to supply the drill. Therefore, all water used for drilling had to be produced by melting ice. This was Page 136 Chapter 4: Field area and methods achieved by heating an original 25 1, needed to start the system, to melt more water from snow and ice. A reservoir tank was excavated into the glacier surface with a chainsaw and lined with a tarpaulin. Once enough liquid had been produced and collected, the water was then pumped from the reservoir through two heating units and into the ice through a high pressure hose. A small quantity of water was circulated back to the reservoir to melt further amounts of snow. In addition, a recirculation pump lowered into the borehole was able to pump water back to the tank to be used again. To gain experience with the system, a number of holes were drilled on the lower glacier, beneath the bulge. These holes were not intended to be instrumented although they did provide useful information on ice thicknesses. It was planned to deploy pressure transducers in three holes above the bulge. These were to be situated in a line up the long axis with the aim of detecting the spatial and temporal passage of water pressure peaks down the glacier. In addition, the boreholes werl;be used for injecting dye into the drainage system which would be detected as it emerged from the glacier. Such an experiment would provide information on water storage and throughflow rates. Water pressure variations can only be detected in boreholes which connect with the subglacial drainage system. Since the drainage system of each glacier has its own time-dependent and spatial characteristics, it is likely that not all holes drilled will achieve a connection. Hodge (1979) found, on South Cascade Glacier, that the chance of any hole connecting with the basal hydraulic network during drilling was about 30%, although this figure may increase, since holes may become connected some time after drilling has taken place. However, Hodge discovered that the success rate is not constant from year to year on the same glacier. In the course of further drilling on South Cascade Glacier during following seasons, the success rate decreased to 8% (Hodge, 1979). Ileen and Bindschadler (1986) reported that five holes, out of a total of eight drilled (63% ), successfully connected to the sub glacial drainage system on Findelengletscher in 1980. Two years later, on the same glacier, 11 holes from 25 drilled (44%) connected with the subglacial hydraulic network, although not all 11 holes were connected at the same time. Six holes were drilled through Bjuvbreen on the upper glacier, above the bulge (Figure 4.7). Unfortunately, none of the holes showed any evidence of connecting with the drainage system. A successfully connected hole is generally indicated by a marked drop in the level of water standing in the borehole (Hodge, 1976; Engelhardt et al., 1978; Kamb and Engelhardt, 1987; Hooke et al., 1988). Once a con.nection has been established, water levels in the hole will fluctuate in response to basal water pressure fluctuations. On B juvbreen, once a borehole had been drilled to the bed, the water level in the hole remained at the same level as during drilling. In previous studies, when this Page 137 \ \ \ \ \ \ \ \ \ Figure 4.7 / / / / / / ' 0 O sc 0 SA SB 0 04c 0 48 4A c.rest o - - - - - - - - brAlge. / / / / / / \ \ \ \ \ \ \ 0 100 200 m \ 01 \ L_ _______________ ...._ ____ ___. Location of hot water drilling points on Bjuvbreen. Holes 1, 2 and 3 are located on the stagnant lower portion of the glacier. The remaining holes are situated up-glacier from the bulge. Page 138 Chapter 4: Field area and methods behaviour has occurred, the detonation of a small amount of explosives (Rothlisberger et al., 1979), cable-tool drilling or sand-pump bailing (Engelhardt et al., 1978; Kamb and Engelhardt, 1987) at the base of the drill hole have been used to force a connection to the drainage system. However, these techniques were not available during fieldwork on Bjuvbreen. Further attempts would have been made to drill a connected hole but a fire in one of the heating units prevented the drilling system from working to the efficiency required in order to melt ice at depth. It is believed that all of the holes with the exception of lC (Figure 4.7) reached the bottom of the glacier. Reaching the base was suggested by the 'feel' of the drill tip, and also by the correspondence between the length of hose run out and the results from the radio echo sounding. 4. 6. S Meteorological observations During the periods spent in the field, a number of meteorological measurements were made. These were primarily to aid in the interpretation of weather-related hydrological events and to calibrate survey equipment. The amount of equipment available limited the parameters which could be measured to air temperature and precipitation. However, a complete record of meteorological observations was obtained from the weather station at Sveagruva (at sea level), 10 km to the west of Bjuvbreen (B. Aune, personal communication). There is a good agreement between measurements at Bjuvbreen and Sveagruva. Small amounts of precipitation occurred more frequently at Sveagruva because of its location on the coast. Due to the difference in altitude, temperatures at Bjuvbreen were generally 1 °C colder than at Sveagruva. This difference is in agreement with the reported temperature lapse rate of 0-009 °C m-1 for this part of Spitsbergen (Simoes, 1990). The meteorological measurements were made at the camp site situated on a terminal moraine, approximately 1 km from the upper glacier. The instruments were situated at an altitude of approximately 120 m above sea level. Air temperature was measured with a maximum- minimum thermometer. Precipitation was collected in a standard collecting jar and measured in the field. All readings were made at 0900 and 2100 hours daily. In addition, barometric pressure was recorded using an altimeter, to determine the correction factor to be applied to the survey measurements. Page 139 CHAPTER 5 THE GEOMETRY AND EVOLUTION OF BJUVBREEN 5.1 INTRODUCTION Surge-type glaciers are, by definition, not in steady-state. Instead, such glaciers undergo a progressive evolution as they pass through the surge cycle. Meier and Post ( 1969) described the characteristic changes associated with the build-up of glaciers towards a surge. The most noticeable changes are those associated with glacier geometry, namely the progressive thickening of the reservoir area coupled with a lowering of the ice surface in the stagnant receiving area. The quiescent phase evolution is particularly well known for three surge-type glaciers: Medvezhiy Glacier, Variegated Glacier and Trapridge Glacier. Medvezhiy Glacier has been studied in the build-up towards its two most recent surges in 1963 and 1973 (Dolgushin and Osipova, 1975). During the quiescent period prior to the 1973 surge, the reservoir zone increased in thickness by an average of 52 m while an average thinning of 60 m took place in the receiving area (Dolgushin et al., 1974). These geometrical changes were accompanied by an increase in ice velocity in the upper portion of the glacier. The velocity did not increase progressively on an annual basis but exhibited a strong seasonal component. Speed increased in a series of "wavy surges" (Dolgushin and Osipova, 1978). These events were accompanied by the advance of the boundary between active and stagnant ice, which Dolgushin and Osipova termed the dynamic balance line (DBL). During quiescence, the DBL advanced more than 4 km. Raymond and Harrison (1988) documented the progressive evolution of Variegated Glacier prior to its 1982-83 surge. Between 1973 and the start of the surge there were changes of up to 20% in ice thickness. Thicke~ing occurred in the upper 60% of the glacier and thinning in the lower 40%. The annual velocity of the active region of the glacier increased by 500%. The DBL, as defined by Dolgushin and Osipova (1978), advanced down-glacier by approximately 1 km (Raymond, 1987), which was considerably less than the advance observed on Medvezhiy Glacier. The quiescent phase evolution of Trapridge Glacier has also been documented, although this glacier has yet to be studied in detail during a surge. The most striking Chapter 5: Geometry and evolution of Bjuvbreen feature on Trapridge Glacier is a steeply-ramped bulge (Clarke et al., 1984), similar to that present on Bjuvbreen (Figure 4.2). This can be considered to represent the DBL. As the quiescent phase progresses, the bulge is propagating down-glacier at 30 m a-1 (Clarke and Blake, 1991). The base of the ramp has now completely overridden the area of stagnant ice which was a remnant from the last surge (personal communication from T. Murray, 1991). These changes have ·occurred as a gradual and continuous process (Clarke and Blake, 1991) and not as a series of pulses, such as those observed on Medvezhiy Glacier. During the period 1969-80 there was an almost doubling of the thickness of the glacier behind the bulge and a five-fold increase in flow velocity in this region (Clarke et al., 1984). Since 1980, however, ice velocities have remained fairly constant at 30 m a-1 (Clarke and Blake, 1991). Coupled with the geometric changes, there has been a thermal evolution of Trapridge Glacier. Warm-based ice behind the bulge is migrating down-glacier into a zone of colder ice, although the process is occurring at a slower rate than the movement of the ramp. 5 .1.1 Geometric evolution of Bjuvbreen: data sources The evolution of Bjuvbreen can be observed on a sequence of aerial photographs spanning the interval 1936-1990. The basic characteristics of the glacier in the various photographs were described in section 4.2.2, as were the reasons for classifying Bjuvbreen as of surge-type. These qualitative observations can be augmented by detailed analysis of the photographs using photogrammetry. Furthermore, with the aid of measurements made in the field, certain physical aspects of the glacier, such as ice volume, surface slope and driving stresses can be calculated from the photographic record. Stereo pairs of vertical aerial photographs of Bjuvbreen are available for 1970 and 1977, thus enabling them to be analysed photogrammetrically. These analyses were performed and described by Fox (1989). The primary aim of Fox's work was to assess the information sources and methods available to glaciologists who wish to study changes in the topography of ice masses through time (Fox, 1989; p. 2). However, he did not use the data: for describing the dynamic glaciology of Bjuvbreen and thus it is useful in the present chapter to incorporate some of the results of Fox's mapping work. The ground control used for the production of his maps was obtained during survey operations in the 1989 field season. Fox (1989) produced a topographic map of Bjuvbreen with a 10 m contour interval using the 1977 photographs. In mid-1991, NP released photographs taken over Svalbard during the summer of 1990. Images were obtained of Bjuvbreen at two scales, roughly corresponding to 1:50,000 and 1:15,000 at sea level. An attempt was made to produce a new topographic map of the glacier surface using the 1: 15,000 scale Page 141 Chapter 5: Geometry and evolution of Bjuvbreen photographs, in order to extend the sequence of started by Fox (1989). The Wild B-8 stereoplotter housed in Department of Geography, University of Cambridge, was used. The coordinates of the known control points in the Kjellstromdalen area were obtained from NP (B. Lytskjold, personal communication) so that the plotter could be calibrated. Since control points in the immediate vicinity of Bjuvbreen were not available, these points had to be taken from the 1 :50,000 photographs which covered a greater spatial area. The 1:15,000 scale photographs were then set up on the plotter using standard techniques (Wolf, 1983). However, in the latter stages of preparing the stereomodel for mapping the glacier, it was found that the exercise could not be completed. The difference between the highest and lowest elevations on the 1:15,000 scale photographs was roughly 900 m. The approximate flying height of the aircraft when taking the photographs was 2295 m a.s.l .. Thus, the elevation differences within the photograph were equivalent to about 40% of the total flying altitude. The Wild B-8 cannot cope with large terrain elevation differences within the photographic model. In general, the instrument cannot be used where these differences are greater than 15% of the flying height (Wild technical specifications, 1967). Fox (1989) did not encounter this problem because he used a Kern PG- 2 stereoplotter which can be adjusted to accommodate such extreme terrain elevation differences. Thus, the most recent topographic map of Bjuvbreen shows the geometry in 1977. Ground survey work undertaken mainly during the 1989 field season provided the coordinates of a number of points on and around the glacier which can be located on the aerial photographs. In addition, the elevation of the survey stations above sea level was established by surveying onto an NP trigonometry point, of known height, on Crednermorenen between Braganzav!l.gen and Van Mijenfjorden. This information enables estimates of the change in elevation of the surface of Bjuvbreen to be made. The depth of ice at a number of locations on the glacier was obtained from radio echo sounding and hot water drilling during the 1990 field season (section 4.3.4 and section 4.6.4). Therefore, it is possible to determine the volume of ice for the present glacier geometry which can be compared with estimates of ice volume derived from topographic maps of the 1977 geometry. Information is available concerning ice thicknesses and surface slopes. These data allow the calculation of basal shear stresses, both for the present glacier configuration and past geometries. 5 .1. 2 Aims of this chapter The aim of this chapter is to describe the geometry of Bjuvbreen. Using data from various sources, changes in the glacier geometry during quiescence will be quantified. In addition to describing the evolution of the glacier long profile and shear Page 142 Chapter 5: Geometry and evolution of Bjuvbreen stresses, a comparison of ice volumes in 1936 and 1990 will be used to predict the timing of the next surge. Finally, the thermal regime of Bjuvbreen will be estimated using a simple model. 5.2 SURFACE AND BED TOPOGRAPHY 5. 2 .1 Long profile changes The long profile of Bjuvbreen has been field surveyed on three occasions. The first survey was carried out by J.O. Hagen (NP) in 1986. Subsequent surveys were made during the 1989 and 1990 field-seasQns. The glacier long profile in 1977 was also obtained from a photogrammetrically-derived map (Fox, 1989). Data sources for the period prior to 1977 were not sufficiently detailed to obtain long profiles. The various long profiles are illustrated in Figure 5.1. 1936 profile The 1936 NP oblique aerial photographs are the earliest known source of pictorial data of Bjuvbreen. The appearance of the glacier in these photographs was described in section 4.2.2 and interpreted as illustrating the late stages of a surge or very early quiescence. The most noticeable difference between the 1936 and later photographs is that glacier occupies a much greater volume in 1936. Using the observation that the upper surface of the lower glacier is approximately level with two lateral moraines, a long profile of the Bjuvbreen in 1936 can be estimated based on a field survey of the position of lateral moraines relative to the 1989 surface. This profile shows that the glacier has a uniform surface slope along its entire length (-8°). Localised surface steepening or a bulge are not evident at this stage in the surge cycle. 1977 profile The longitudinal profile of Bjuvbreen in 1977, taken from a 1: 12,000 scale map produced by Fox (1989), represents the reference datum to which later long profiles can be compared. BJ this stage of the quiescent phase, the wave-like bulge was already well developed in the profile. The base of the bulge is located roughly 1860 m from survey station 1. The approximate gradient of the bulge front is 31 °. Up-glacier from the crest, the ice surf ace remains generally uniform with a slope of about 10°. 1986 profile The long profile of Bjuvbreen was field surveyed in August 1986 by J.O. Hagen. The base of the bulge was then located approximately 1825 m from survey station 1. The gradient of the steep front was approximately 34°. This was an increase Page 143 E ,-: C 0 ·- -ro CJ) Q) > 0 .D ro .c (J) Q) :r: 500 40 0 300 200 1936 · 1977 19 86 1990 1700 18 0 0 R7 R2 R1 19 0 0 2000 2100 22 00 2300 2400 Distance from Station 1, m Figure 5.1 Ice surface profiles for J3juvbreen in 1936, 1977, 1986 and 1990. For purposes of clarity, only the portion of the glacier upwards from near the bulge has been illustrated. Down-glacier from this region, the ice surface has downwasted vertically over the period 1977-90. The 1936 profile illustrates the glacier at the end of its last surge. This profile based on interpretation of oblique aerial photographs and field surveying of lateral moraines. The data for the 1977 longitudinal profile was obtained from a photogrammctric 111:1p of the glacier produced by Fox (1989). The 1986 and 1990 long profiles arc based on field data. A long profile w:1s also surveyed in 1989. That profile is not shown here because the diff erencc between 1989 and 1990 is too small to illustr:1te clearly. The bed profile was constructed from radio echo sou nding and hot water drilling depths. The locations or the echo sounding and drilling points arc shown. Note that there is no significant basal topography in the region or the bulge in the ice surface. Chapter 5: Geometry and evolution of Bjuvbreen of about 3° since 1977. The slope up-glacier of the bulge crest had a gradient of approximately 12°. This figure is only an approximation, since there were only four points surveyed on the upper glacier (personal communication from J.O. Hagen). Therefore, the detailed topography of this region is not known. The 1986 long profile illustrates a number of changes that have occurred since 1977. Near the terminus of Bjuvbreen, vertical downwasting of stagnant ice had taken place. In the lower 500 m of the glacier, thinning of up to 10 m had occurred during the previous nine years. Wasting in this part of the glacier results from a lack of ice being received from accumulation zone during quiescence. The location of the base of the bulge is approximately 35 m further-down_-glacier than in 1977. The base of the bulge can, therefore, be considered to represent the dynamic balance line (DBL) of Bjuvbreen (cf. Dolgushin and Osipova, 1978). There has been a steepening (-3°) of the bulge face. Most of this increase in slope was due to an increase in the thickness of ice on the bulge crest. There is also some evidence of a slight thickening of ice (-8 m) up-glacier from the bulge. 1989 profile The long profile of Bjuvbreen was surveyed on 20 August 1989. In situ stagnation and thinning of the lower glacier had continued since 1986. There has been a slight steepening (-1 °) of the upper reaches of the bulge. Up-glacier from the bulge, a spoon-shaped bowl began to become evident. The back of the bowl, approximately 250 m from the crest of the bulge, was marked by an increase in the ice surface gradient. 1990 profile The long profile of Bjuvbreen was surveyed on 3 July 1990. The position of the base of the bulge was 1812 m from survey station 1. This is roughly 50 m further down-glacier than the position of the bulge in the 1977 profile, which implies that the bulge is propagating at about 4 m a-1. The gradient of the ramp was 30°. This is a slight reduction in slope compared to the previous year's profile of the same area. The reason for this is an increased amount of snow present at the time of the 1990 survey. The base of the bulge acts as an accumulation hollow. Firn depths measured at this location during the 1990 season were in excess of 4.5 m. Earlier long profiles were based on data obtained in August when much of this firn had melted.' Therefore, the presence of extra snow during the 1990 survey had the effect of apparently lowering the gradient of the bulge. Page 145 - I Chapter 5 : Geometry and evolution of Bjuvbreen 5.2.2 Bed topography The configuration of the bed beneath Bjuvbreen was determined by radio echo sounding (section 4.3.4). Additional information was provided by the depths reached during hot water drilling (section 4.6.4). The radar and drilling work was generally done on the centreline of the glacier. However, two points on each side of the central axis were also sounded (Figure 4.3). Soundings were made with stationary equipment at each location. The bed topography was obtained by visual interpolation between the recorded values. Therefore, it does not represent a continuous echo sounding profile. Because of the danger posed by the presence of undetected crevasses, radio echo sounding was not carried out along tµe entire length of Bjuvbreen. Measurements began near the location of stake 13 (1990) and were made approximately every 50 m down the glacier centreline (Figure 4.3). The interpolated bed profile is shown in Figure 5.2. From the furthest point up-glacier, Rl, the bed dips down-glacier as far as R8 at an angle only slightly steeper than the ice surface ClXice = 13·6° compared to abed = 14·5°). At point R8 the bed topography slopes gently upward at an angle of 17 ·6° as far as RlO, marking the position of a small riegel (Figure 5.1). The crest of this riegel is approximately 250 m up-glacier from the bulge. The bed slope then dips in the opposite direction. Between RlO and the echo sounding point nearest the bulge, R12, abed = 4-4 °. This gradient is somewhat less than lXice = 9.4 ° between the same points. No radio echo sounding was carried out down-glacier of point R12. However, the elevation of the bed beneath the lower portion of Bjuvbreen can be reconstructed from the results of the hot water drilling. Three holes were drilled through the glacier on the centreline, down-glacier from the base of the bulge (Figure 4.7). Ice thickness decreased from 18 m at B3 to 8 mat Bl. The gradient between B3, at the base of the bulge, and the furthest down-glacier hole, B 1, is 7 ·5° . The ice surface over the same distance is 10·2°. As stated above, no measurements of the bed elevation were made between R12 and B3. Therefore, detailed knowledge of the basal topography in this region is not available. However, the inferred gradient of the bed can be computed between these two points. The result is .r:: (/J ro (/J ctl co Chapter 5 : Geometry and evolution of Bjuvbreen 200 Al 150 100 50 0 -+----.-.....----.-.....-----..--.------..--,------.--.,....-----.---, 1400 1600 1800 2000 2200 2400 2600 Distance from Station 1 (m) Figure 5.3 The variation of basal shear stress, 1°B, with horizontal distance on the centreline of Bjuvbreen. Shear stress was calculated with equation 5.1 using 1990 data. The glacier flows from right to left in the figure. In 1990, the base of the bulge was located approximately 1812 m from Station 1. results from another surge-type glacier in Svalbard. On Usherbreen, Hagen (1988) calculated the pre-surge shear stress to be about only 61 kPa. Problems have been pointed out for both the above examples, which may account for the differences in results. Paterson (1981) suggested that Robin and Weertman (1973) had failed to average the surface slope of Finsterwalderbreen over a distance long enough to eliminate the effects of longitudinal stress gradients. In the case of Usherbreen, Hagen (1988) recognised that he may have underestimated the ice thickness. S. 3. 4 Longitudinal averaging of basal shear stress The calculation of basal shear stress has been shown to be influenced by the distance used for the averaging of glacier surface slope (Budd, 1968; Raymond, 1980; Paterson, 1981) because of the effects of longitudinal stress gradients. Budd (1968) derived an expressiqn for shear stress which incorporated the effects of longitudinal stress gradients. However, Bindschadler (1982) considered Budd's equation too complicated for use on Variegated Glacier, since a number of its terms could not be determined from the glacier geometry alone. Instead, Bindschadler argued that reasonable estimates of basal shear stress could be obtained from equation 5.1, provided the surface slope component of this expression was averaged over 8-16 times the ice thickness. Both Budd and Bindschadler considered only the effect of changing surface slope on the pattern of shear stress. However, Meier et al. (1974) suggested that changes in ice thickness along the flowline were also important. Kamb and Echelmeyer (1986) developed a theory which accounts for the influence of longitudinal Page 151 Chapter 5: Geometry and evolution of Bjuvbreen stress gradients on basal shear stress, using both changes in surface slope and ice thickness. Kamb and Echelmeyer introduced a parameter, C, which is defined as the longitudinal coupling length. It represents the length scale over which the longitudinal averaging of changes in slope and thickness affect the local flow regime. This theory was used by Raymond and Harrison (1988) to calculate basal shear stresses during the quiescent phase of Variegated Glacier. They found that the values obtained using the Kamb-Echelmeyer method were able to explain many of the observed flow characteristics of this glacier. In order to incorporate the effects of longitudinal stress gradients in the analysis of shear stress on Bjuvbreen, Kamb and Echelmeyer's (1986) theory was used. Basal shear stress, TB, is expressed in terms of the local "slope" stress, Ts. The local slope stress was calculated from equation 5.1 (i.e. Ts = F Pig h sina). Basal shear stress was then found from the following equation: 'Z°B(X) = l f ~ Ts(x') h(x')l/n exp [+C'-~] dx' 2 Ch(x)lln (5.2) where C is the longitudinal coupling length, n is the exponent in Glen's flow law (here taken to be 3, as recommended by Paterson (1981)) and X and x'are longitudinal coordinates on the glacier surface. Kamb and Echelmeyer (1986) defined the quantity, exp [-Ix' - xl I C] , as an influence transfer function. This exponential function represents the longitudinal distance over which the effects of slope and thickness changes are averaged in shear stress calculations. Kamb and Echelmeyer found that the transfer function was approximated by a longitudinal averaging length of 4C in practical glacier flow calculations . The ratio C/h commonly falls between 1·5 and 2·5 for valley glaciers (Kamb and Echelmeyer, 1986; p. 280). For calculations on Bjuvbreen, a ratio of C/h "" 2 was assumed. Thus, the longitudinal coupling length, C, was kept constant at 0-25 km on the upper glacier and reduced to 0-036 for points below the bulge. Therefore, the longitudinal averaging lengths used in shear stress calculations on Bjuvbreen were 4C-= 1 ·00 km on the upper glacier, and 4C = 0· 14 km on the lower glacier. These averaging lengths are in excess of eight times the greatest ice thickness and are more than ten times the average depth. The results of the modified shear stress calculations are given in Table 5.2 and the variation of shear stress with distance along the long axis of Bjuvbreen is illustrated in Figure 5.4. The values of basal shear stress obtained using the Kamb-Echelmeyer method were not markedly different from the results of calculations performed using the conventional product of slope, thickness and shape factor. However, because the modified values account for longitudinal stress gradients, they are theoretically more Page 152 ro a.. 150 Chapter 5: Geometry and evolution of B juvbreen 6 100 U) U) ~ vi ro (I) ..c U) 0.06 ro --o-- 1990 C 0 en 0.04 Cll Q) Cf) 0.02 0.00 1600 1800 2000 2200 2400 2600 Horizontal distance from Station 1 (m) Figure 6.2 Longitudinal variation of seasonal velocities observed at the surface of Bjuvbreen. The base of the bulge is located at approximately 1820 m. Up-glacier from the bulge, seasonal velocities are noticeably higher than for points on the lower glacier. Seasonal velocities were generally greater during the 1989 field season. This difference is probably due to increased basal motion in 1989 (section 7.4.3 ). S ration 1 is located in Figure 4.4. Seasonal velocities 1990 The 1990 field measurements covered the period between late May and early July. The time span between the first and last velocity measurements was 22 days . Velocity measurements were made for 13 points on the glacier. The seasonal velocities measured in 1990 are shown in vector form in Figure 6. lb. This illustration clearly shows that surface movement down-glacier of the bulge is minimal. At stake 1 (90) the seasonal velocity was only 0-025 m d-1 . Mean velocities increase with distance up-glacier. The mean surface speed for the three markers at the base of the bulge was 0·049 m d-1. On the crest of the bulge the mean velocity of the three stakes was 0-095 m d-1. The highest mean velocity, 0· 106 m d-1, was calculated Page 172 Chapter 6: Flow and deformation of Bjuvbreen for the six stakes on the upper glacier. The variation of seasonal surface velocities with horizontal distance is illustrated in Figure 6.2. Comparison between field seasons The velocity measurements reported above were made at slightly different times of the year. In 1989, the measurements were made once the melt season was well established. The measurements in 1990 were made as the ablation season was beginning and continued into the melt season. Due to the differences in the timing of the observations, the velocity structure of Bjuvbreen may exhibit a slight variation between the field seasons. Different targets were used during each field season and, therefore, a direct comparison between identical points is not possible. However, an analysis can be made for markers that were located in approximately the same locations during both periods of measurement. The general trend is for seasonal velocities of roughly the same point on the glacier to be greater during the 1989 season than the 1990 season (Table 6.1 ). A possible explanation is that surf ace velocities are higher when more water is available during the summer (section 7.4). It is difficult to assess the magnitude of the seasonal difference in velocity because measurements made during the 1990 season included a period during which meltwater runoff occurred and water was most likely present at the glacier bed. Nine of twelve paired targets had higher seasonal velocities in 1989 (Table 6.1). From the available data, these nine points were moving, on average, 29% faster in 1989 than in 1990 (standard deviation ± 18%). Seasonal shifts in velocity of between 20-100% have been reported for a number of temperate glaciers, for example Nisqually Glacier (Hodge, 1974), Unteraargletscher (Iken, 1981), Storglaciaren (Hooke et al., 1983) and Variegated Glacier (Raymond and Harrison, 1988). However, not all temperate glaciers exhibit a seasonal acceleration. Echelmeyer (1983) found that Blue Glacier increased its velocity by only 5-10% at the start of the ablation season. There are two likely causes of variations in seasonal velocity. Changes in velocity can be generated by a change in glacier geometry as described by classical flow laws. Processes operating at the base of glaciers can also lead to differences in seasonal velocity. The evolution of the geometry of Bjuvbreen between 1989 and 1990 was discussed in section 5.2.1. A significant point to reiterate is that the geometric change occurring between the two field seasons was very small. Variation in basal shear stresses was accordingly small. The amount of geometry-driven flow of Bjuvbreen is discussed further in section 6.5. However, it is worth noting here that the observed changes in geometry cannot account for the differences in the seasonal velocity of Page 173 I ,, Chapter 6: Flow and deformation of Bjuvbreen Seasonal velocity m d-1 Mean short term deviation % Target pair 1989 1990 1989 1990 2-2 0·073 0·048 42 27 5-3 0·068 0·058 56 33 7-4 0·058 0·041 44 28 15-5 0·093 0·116 60 53 20-6 0·103 0·078 32 6 23- 7 0·105 0-093 38 20 31- 8 0-084 0•107 53 58 34-9 0·117 - 0·083 29 20 37-10 0·101 0·092 49 32 42-11 0·133 0-128 44 43 43- 12 0·121 0·141 38 55 44-13 0·122 0-083 36 12 Table 6.1 Comparison between seasonal and short-term velocities for similar points on the glacier surface, indicated by target pairs, during the 1989 and 1990 field seasons. In most cases, seasonal velocities were higher in 1989. The mean short term deviations are discussed in section 6.2.4. There is a trend for short term increases and decreases in velocity from the seasonal average to be less extreme in 1990 than in 1989. Stakes are located in Figure 4.5. Bjuvbreen between 1989 and 1990. Therefore, the velocity differences are most likely due to basal processes, probably a result of changes in the drainage system. 6. 2. 3 Short period velocity behaviour Fluctuations, at a variety of timescales, in the horizontal surface speed of glaciers are a well-known phenomenon. An original intention of the fieldwork on Bjuvbreen was to make velocity measurements of each stake at least once a day throughout both field seasons. However, weather conditions were frequently unsuitable for the a~curate surveying of small displacements. Low cloud often obscured many of the targets from the survey stations. An additional problem was strong winds buffeting the survey station causing the instruments to vibrate. Thus, in practice, the observational record of velocity does not have as fine a temporal resolution as was planned. On average, measurements of each stake were made every three or four days. The velocity of Bjuvbreen does not remain constant over short periods but is , instead, characterised by numerous fluctuations of varying amplitudes. The data are presented in terms of percentage deviations from the long-term (seasonal) average for each stake. Figure 6.3a illustrates the record obtained for a number of stakes lying close Page 174 .i- ·u 100 .Q 50 "' > "iii C 5l "' Q) V) E e C .Q iii ·;; Q) a -50 100 50 Marker 4 (seasonal velocity 0.070 m/d) 200 210 220 Julian day 230 Marker 19 (seasonal velocity 0. 111 m/d) 0 .................... ................... .. E _g C _Q iii ·;; Q) a -50 100 .i- ·u .Q 50 Q) > "iii C Sl "' 0 Q) V) E e C .Q iii ·;; "' a -50 200 21 0 220 Julian day 230 Marker 32 (seasonal velocity 0.125 m/d) 200 210 220 Julian day 230 240 240 240 Figure 6.3a Short term horizontal surface velocities for six targets located on the centreline of Bjuvbreen in the 1989 field season. Marker 4 is located at the base of the bulge. The furthest up-glacier target is marker 44. Marker locations are shown in the inset overleaf. Short term velocities are expressed as percentage deviations from the seasonal velocity for each marker (indicated by the dotted line) . Vertical lines represent error limits (section 4.5.6). Variations in velocity occur in phase over large parts of the glacier. [continued over. . . ] Page 175 Figure 6.3a Continued 100 ~ i- ·u o 50 w > "iii C ~ ro 0 ~ E _g C -50 .Q iii ·;; (l> 0 ~ i- u 100 .Q 50 (l> > Marker 36 (seasonal velocity 0 .. 123 mid) 200 210 220 Julian day 230 Marker 40 (seasonal veloci ty 0.106 m/d) 240 iii ·;; (l> 0 -1 00 +-r-~.--.-,--.~~~--r---r-~~-r-~-.-,--.-, ,o O' i- 'u 100 .2 50 (l> > "' C ~ ro 0 ~ E _g C ·SO .Q > (l> 0 200 210 220 230 240 Julian day Marker 44 (seasonal velocity 0.122 m/d) -100 -+-~~~~-r-~~------------ 200 210 220 230 240 Julian day n 100 200 m L_.__J_._j Chapter 6: Flow and deformation of Bjuvbreen to the longitudinal centreline of the glacier. Only these points are shown for the purpose of clarity, although the records of other targets are discussed below. Data on short period variations in velocity for all targets are given in Appendix A. The behaviour of points located close to the bulge is discussed first. At the beginning of the monitoring season in 1989, above average velocities were recorded for all markers immediately above and below the bulge (markers 1-8 and 15-23) between Julian days 201- 204 (20-23 July). The velocities recorded at all markers then decrease during the following period, although the magnitude of the velocity drop differs between markers. Speeds fall to below average for markers 1, 7 and 8 below the bulge, and all mark~rs above the bulge except 15 and 18 (Figure 4.5). A general trend of decreasing velocities continues until Julian day 214, although small increases were recorded for stakes 1, 4 and 7. The period spanning Julian day 214-223 (2-11 August) is characterised by asynchronous motion between different locations. Increases in velocity to approximately seasonal levels occurred at stakes 2, 16, 17, 19, 20, 21, 22 and 23. Other targets exhibited decreases in speed. Between Julian day 223- 225 there were increases in velocity to values close to or marginally above the seasonal average at most stakes. There is an interruption in the observational record of marker 5 so it is not known whether this location also experienced an increase in speed. The velocity increases below the bulge during this period were in contrast to the behaviour of the stakes above the bulge. At all markers except 15, there were slight decreases in speed between Julian day 223- 225. The following interval between Julian day 225- 231 is also characterised by asynchronous behaviour. All the markers below the bulge show a decrease in velocity, whereas some stakes on the crest of the bulge, namely stakes 15, 16, 17 and 18, displayed a speed increase. The last period of measurement during the 1989 field season, between Julian day 231- 234 (19-22 August), contained an almost unanimous increase in the velocity of stakes to values above the seasonal average. No increase was observed at stake 5 because the observational record was interrupted. The velocity record for the targets on the upper glacier, towards the corrie backwall from the crest of the bulge (markers 30-44), is now discussed. Observations were begun slightly later than those in the vicinity of the bulge. At the beginning of the survey period (Julian day 208-213, i.e. 27 July-1 August), all markers had measured velocities above the seasonal average, except for stakes 30 and 41 where measurements did not begin until later in the season. During the following two periods there is a common two-step decrease in velocities, when the speed of most stakes fell to below average. Marker 38 differs from all the other stakes in that the second decrease in velocity does not occur, and instead there is a slight increase. In the period Julian day 222-224 there were slight speed-ups at every location, except marker 38 which Page 176 Chapter 6: Flow and deformation of Bjuvbreen · decreased its velocity. At most of the stakes, the last two periods of measurement were characterised by a two-step decrease in velocity. Exceptions to this pattern were 34, 35, 38, 39 and 42. At these locations, there was either only one decrease or alternatively, velocity increased in the last observation interval of Julian day 230- 234 (18- 22 August). The pattern of velocity decrease on the stakes of the upper glacier contrasted with the marked peak in speed recorded for a number of the markers near the bulge. The 1990 velocity record also contains a good deal of variation on a short timescale (Figure 6.3b). The record of fluctuations for marker 1 is typified by smaller amplitudes than observed for other stakes. The measurements began with the period between Julian day 165- 172 (14-21 Jl!ne). Markers at most locations on the glacier at - this time had velocities close to or below the seasonal average. Both stakes 2 and 3 differed from the other targets in that they experienced above average velocities during this period. These targets then experienced a slowdown during the next interval between Julian day 17 2-17 4. This decrease is also found at markers 7 and 9. The remaining stakes all increased in speed at this time. At marker 5 this increase peaked during the subsequent interval between Julian day 174-176 (23-25 June). During this period, the remaining targets exhibited a decrease in velocity. Velocities were close to, or below average during the following 5 days, and generally remained below average until the end of the observational season. Stakes 2, 4, 5 and 9 all increased their speed during the final interval (30 June- 3 July), although only stake 2 reached a velocity well above the seasonal average. 6. 2. 4 Discussion of velocity behaviour of Bjuvbreen The above sections described the spatial pattern of seasonal velocities and the character of transient velocity variations recorded on Bjuvbreen. A number of features of these records are discussed in this section. Most of the discussion will centre on the seasonal velocities, with an analysis of the causes of the short-term fluctuations being presented in the subsequent chapter where they are linked with glacier hydrology. The lack of velocity measurements spanning a full year make it difficult to analyse the presence of yearly v~ations in the velocity of Bjuvbreen. A number of theoretical and observational studies have demonstrated that basal shear stress exerts a fundamental influence on the pattern of observed glacier velocity (e.g. Nye, 1952; Bindschadler, 1983; Raymond and Harrison, 1987). Exceptions to this relationship have been reported, however. Hodge (1972) found that surface velocity varied inversely with basal shear stress on Nisqually Glacier. The influence of shear stress on the spatial distribution of seasonal velocity on Bjuvbreen was examined by comparing shear stresses computed in section 5.3.4 with surface motion records. Figure 6.4 illustrates the relationship between seasonal surface velocities and basal Page 177 E _g C .Q ;;; ·;; a, a ~ .~ u 0 ai > '" C 0 Marker 1 (seasonal ve locity 0.025 m/d) 100 -50 160 170 180 Julian day Marker 3 (seasonal velocity 0.058 m/d) 100 50 0 -50 -100 160 170 180 100 0 Julian day Marker 6 (seasonal velocity 0. 078 m/d) I ..................... ...... ---nt 190 190 C -50 Q ;;; > a, 0 -1 00 +-,--~.--,r--,-...---,---,---,.--,---r---r----r---,--, 160 170 180 190 Julian day Figure 6.3b Short tem1 horizontal surface velocities for six targets located on the centreline of Bjuvbreen in the 1990 field season. Marker 3 is located at · the base of the bulge. The furthest up-glacier target is marker 13. Marker locations are shown in the inset overleaf. Short term velocities are expressed as percentage deviations from the seasonal velocity for each marker (indicated by the dotted line). Vertical lines represent error limits. The most noticeable motion event takes place between Julian days 174-176. The magnitudes of the short term deviations is slightly smaller in 1990 compared to 1989. [continued over ... ] Page 178 Figure 6.3b Continued Marker 9 (seasonal ve locity 0.083 m/d) 100 u .Q 50 o, > E e C -SQ .Q '" ·;: 0 -100 -+-~~-~~~~~-~~~~-~~~~ i- G 100 .Q 50 > -;;; C :; n, 0 :;: E _g C -50 .2 ;;; ·;: '" 0 160 170 180 190 Julian day Marker 11 (seasonal velocity 0. 128 m/d) -100 -~~-~~~~~-~~~~-~~~~ 100 i.} _(_) so .. > r.i C: ~.~ "' 0 :;: [: ~ ~ r- -50 g 160 170 180 Julian day Marker 13 (seasonal velocity 0.083 mid) 160 170 l_L-1._._J I , ,oo ,,, m Jul ian day 180 \ \ 190 190 Chapter 6: Flow and deformation of Bjuvbreen shear stress. The regression line through the data points has only a moderate degree of correlation. The r2 of 0·42 demonstrates that the relationship between surface velocity and shear stress on Bjuvbreen is not significant at the 0· 10 level. The large changes in surface slope near the bulge complicate the pattern of basal shear stresses in this region of the glacier. The difference between shear stresses calculated assuming "laminar" flow (equation 5.1) and the Kamb-Echelmeyer longitudinal averaging theory (equation 5.2) was most notable in the bulge region (Figures 5.5 and 5.6). However, Kamb and Echelmeyer (1986) noted that shear stresses computed using their theory were not able to account for observed flow patterns in the surge trigger zune on Variegated Glacier and near the icefall on Blue Glacier. Appreciable basal sliding was believed to be occurring in these areas. The region of Bjuvbreen where there is a poor relationship between surface velocity and shear stress is shown later in this chapter (section 6.5) to be where a significant amount of basal motion occurs. The relationship between velocity and shear stress is further complicated by the potential influence of local slope on the observed surface motion. Meier et al. (1974) found that there was a small but noticeable increase in the surface velocity of Blue 0.16 0.14 AR2 ~ 0.12 .s ;::.- 0.10 ·c::; _Q 0.08 a, > '° 0.06 C 0 (/) C1l 0.04 a, ([) 0.02 01 0.00 60 80 100 120 140 Shear stress (kPa) Figure 6.4 · Relationship between seasonal surface velocities and shear stress. The ·data used in this figure are 1990 values and the line was fitted using least squares regression. The velocity targets used in this analysis are numbered (Figure 4.5). The r2 is 0·42. The inability of shear stress to predict the seasonal velocity is significant at the 0· 10 level. Glacier which coincided with a peak in the local surf ace slope. The change in surf ace slope of Blue Glacier where the velocity increased was less than 3° occurring over a horizontal distance of about only one ice thickness. Kamb and Echelmeyer's (1986) theory was developed to take account of local slope variations by using an exponential weighting function, exp [-Ix' - xl IC], to set the longitudinal averaging length. The use Page 179 Chapter 6: Flow and deformation of Bjuvbreen of an exponential function ensures that the local glacier geometry will have an appreciable, although muted, influence on flow velocity (Kamb and Echelmeyer, 1986, p. 271). Applying their theory to field data from Blue Glacier, Kamb and Echelmeyer were able to reproduce a velocity peak in the area of the observed increase in flow. There are, however, substantial differences between slope changes on Bjuvbreen and those on Blue Glacier. In the area of the bulge on Bjuvbreen, over a longitudinal interval of about 100 m (slightly greater than one ice thickness in this region), surface slope increases from - 12° to 30° before decreasing to -10° down-glacier from the bulge (section 5.2.1). The mote pronounced changes in slope on Bjuvbreen compared to Blue Glacier probably explain part of the difficulty in accounting for flow velocity using basal shear stress. Seasonal velocities on the surface of Bjuvbreen were found to be different during the two field seasons (section 6.2.2) . The record of transient variations in velocity during the two seasons was compared to determine if the short-term velocity behaviour of the glacier also differs between seasons. Velocity variations were observed to occur during both periods of measurement at nearly every location on the glacier surface. 'Staircase' plots of the percentage deviation of short-term velocity from the seasonal average for comparable points on the glacier surf ace were examined (Figure 6.3). The deviations from average velocities were described in section 6.2.3. A common factor among all the targets surveyed was for the transient deviations never to exceed 100% of the long term average. There is a tendency for the deviations from the seasonal average to be generally smaller in 1990 than in 1989. The mean percentage deviation for each comparable pair of targets was computed. For all pairs, except 1989- 1990 pairs 31-8 and 43-12, the magnitudes of the deviations from the seasonal average velocities are smaller during 1990 (Table 6.1 ). The differences between the mean deviations for each target pair vary in size, but in general they are not large (usually <15%): The observations discussed above suggest that a number of differences exist in the dynamic behaviour of Bjuvbreen during the onset of the ablation season compared to the situation once the melt season has become established. One reason for these differences might be a change in the glacier ' s drainage system. The behaviour of other glaciers during the transition between winter and summer has been described by various authors. Data from Storglaciaren (Hooke et al., 1983a; Hooke et al., 1989a) appear to indicate that short-term motion events are more pronounced during the initial stages of the ablation season. Similar results were reported for the quiescent phase of Variegated Glacier (Kamb and Engelhardt, 1987). Mini-surge events usually occurred in late June and early July, although much smaller fluctuations in flow velocity continued to occur through the remainder of the ablation season . Data from White Page 180 Chapter 6: Flow and deformation of Bjuvbreen Glacier presented by Iken (1974) indicated that the greatest amplitude in the short-term velocity fluctuations occurred about one month after the beginning of the melt season. As the ablation season progressed the fluctuations became less pronounced. An analysis of the factors influencing the temporal variation in the velocity of Bjuvbreen, linked to observations on its hydrology, is presented in Chapter 7. 6.3 VERTICAL VELOCITIES AT THE GLACIER SURFACE 6. 3 .1 Seasonal vertical velocities The analysis of vertical velocity is _related to three components of motion in the vertical plane (Paterson, 1981). These are: (i) emergence, or submergence angle of horizontal flow, (ii) the effect of the vertically-averaged vertical strain rate, and (iii) uplift due to high water pressures or cavity formation and closure. Seasonal vertical velocities are a reflection of the first two processes, whereas short- term variations are due to transient events such as water pressure peaks and the opening and closing of cavities. The vertical inclinations of the seasonal horizontal velocities were calculated for all markers, from data averaged over time between the initial and final measurements during each field season. The angles of the flow vectors during the 1989 season are illustrated in Figure 6.5a. Flow is generally upwards relative to the glacier surface. Downward vertical velocities were observed only at markers 41-44. In an idealised glacier, flow is downward in the accumulation zone and upward in the ablation zone (Paterson, 1981; p.60). Markers 41-44 lie very close to the inferred position of the equilibrium line (section 4.2.2) and, thus, submerging flow in this region is not unusual. The angle of plunge steadily decreases down-glacier from marker 44 (Figure 6.5a). At marker 40 and points down-glacier, the angles of the flow vectors are upwards at the glacier surface. There is no obvious variation in the angle of emergence with distance towards the bulge. Markers at the base of the bulge have emerging flow vectors, although if anything, the emerging angle of flow is smaller here than at points up-glacier. This reduction reflects the pattern of ice flow in the region of the bulge. Ice at the base of the bulge is less than 20 m thick and, therefore, will not have a significant component of upwards motion. Instead, ice in this region is being pushed forward and downhill as the bulge migrates. The vertical angles of the horizontal flow vectors were also calculated for the 1990 field season (Figure 6.5b). Plunging flow vectors were observed at markers 12 and 13. At all other locations, the flow vectors were emerging at the glacier surface. The strongest emerging angles were observed for markers near the crest of the bulge Page 181 I I 1] I E C 0 500 ~ 400 (/) Q) > 0 .D "' i: 300 "' Q) I E ·- · C 0 50 0 ~ 400 (/) Q) > 0 .D "' i: 300 "' C, I A 1989 20 10 0 Cc:1/d • Scale for ve l ocilyvectors B 1990 EL A 20 . .... : : ·: ':: :. ·: .. : .. · . .. . Distance from Stat i on 1, m 6 . . . .. .... :_·: ·.:. ·:·_: .· . . . ·. . . : .... -~ Distance from Sta t ion 1, m Figure 6.5 Vertical components of the surface velocity vectors during the 1989 and 1990 field seasons. The targets shown are located on the centreline of Bjuvbreen (Figure 4.5). The flow vectors are generally emerging from the glacier surface, except at the furthest up-glacier targets where the flow is submerging. The presence of submerging flow in this part of the glacier was used to locate the ELA (see text) . Page 182 .. . . . · I I Chapter 6: Flow and deformation of Bjuvbreen (Figure 6.5b). A similar flow pattern was reported for Trapridge Glacier by Clarke et al. (1984, Figure 4). The vertical angles of the horizontal flow vectors can be used to provide a better estimate of the equilibrium line altitude than was given in section 4.2.2. Since downward flow should only occur in the accumulation zone (Paterson, 1981), the equilibrium line lies near markers 41 (in 1989) and 12 (in 1990). This corresponds to an altitude of 560--570 m a.s.1 .. Theoretically, the fastest horizontal surface velocities should be found at the equilibrium line (Paterson, 1981). This is indeed the case on Bjuvbreen (section 6.2.2). In 1989, the highest seasonal velocity was observed at marker 42 (0·133 m d·l). The highest seasonal velocity during 1990 was recorded at marker 12 (0·141 md-1). 6. 3. 2 Short period fluctuations in vertical velocity Vertical velocities exhibit transient changes in magnitude and direction in response to various events and processes taking place within the glacier over timescales measured in days. These changes may be occurring hourly, but because vertical velocities are much smaller than their horizontal counterparts, variations at such short timescales are very difficult to measure accurately by conventional surveying methods (Paterson, 1981; p. 62). During fieldwork on Bjuvbreen, vertical velocities were measured over intervals of several days. The record of vertical velocities for 1989 shows a certain amount of variation over the observation period (Figure 6.6a). The discussion will begin with the markers located below and on the crest of the bulge. The record for most of the stakes in this area starts with a period of uplift immediately followed by downward motion. A number of markers, namely stakes 20, 21, 22 and 23, began the period by having downward velocities which then increased in magnitude. On Julian day 211 (30 July) all the markers at the base of the bulge experience a marked increase in the magnitude of downward velocity which continued until Julian day 214 when velocities become upward. This event is not noticeable on the stakes along the crest of the bulge because these markers were_ not surveyed on Julian day 211 as a result of poor visibility. On Julian day 214 vertical velocities decreased in downward magnitude, although in some cases, notably stakes 3, 5 and 15, a peak in uplift for the whole field season was reached. For most of the remaining markers there was a 'two-step increase to their maximum rates of uplift which occurred during the interval Julian day 217-223 (5-11 August). The three stakes that had peaked earlier now experienced downward velocities. The remaining stakes ceased uplift in the interval Julian day 223-225. During the penultimate measurement interval in 1989, all markers had small downward velocities. These became slightly less or in some cases became upward during the final Page 183 Marker 4 0.20 0.15 ~ 0.10 .s 0.05 2:- ·u _Q -0.00 Q) > "iii -0.05 " t Q) -0.10 > -0.15 -0.20 200 2 10 220 230 240 Julian day Marke r 19 0 .20 0.15 ~ 0.10 .s 0 .05 2:- ·u 0 -0.00 ai > "iii -0.05 " 'i:' Q) -0. 10 > -0.15 -0.20 200 210 220 230 240 Julian day Marker 32 0.20 0.15 ~ 0.10 .§.. 0.05 ?:- ·u -0. 00 0 ai > "iii -0.05 " "i: Q) -0. 10 > -0.15 -0.20 200 210 220 230 240 Julian day Figure 6.6a Short term vertical velocities for six targets located on the centreline of Bjuvbreen in the 1989 field season. Marker 4 is located at the base of the bulge. The furthest up-glacier target is marker 44. Marker locations are shown in the inset overleaf. Vertical lines represent error limits (section 4.5.6). Markers near the bulge (4 and 19) show a different pattern of vertical movement to targets on the upper glacier. Targets on the upper glacier often display a similar pattern of motion although the magnitudes of the variations are not identical. [continued over. .. ] Page 184 11 Figure 6.6'a- Continued 0.20 0.15 ~ 0. 10 .s 0 .05 ?: u 0 -0.00 w > '" -0.05 u ·c (l> -0.10 > -0.15 -0.20 0.20 0.15 ~ 0.10 .s 0.05 -~ u _g -0.00 (l> > r., -0.05 u -~ -0.10 > -0.1 5 -0.20 0.20 0.15 0.10 O.C:S u .2 -0 00 C• :, " -0 05 'd > -0.10 -0.15 20 0 200 Marker 3G 210 220 230 Julia n day Mmker 40 I I ~ 210 220 Julian day Mc1rke r 44 230 240 2,0 -0.20 +~~-,--,--,-~~~~~~~~~~~~~--. 200 210 220 230 240 Juli.in day 11 Chapter 6: Flow and deformation of Bjuvbreen period between Julian day 231-234. Stakes 6 and 20, however, experienced an increase in downward velocities at this time. Short-term variations in vertical velocity were also recorded at the markers on the upper glacier. Downward velocities were recorded at all locations during the initial period of observation. In most cases the downward trend increased during the following interval of Julian day 213-216 (1-4 August). Slight increases were, however, noted for stakes 32, 34, 36 and 39. All stakes, except marker 38, maintained downward velocities until Julian day 222-224 (10-12 August), when there was a noticeable period of uplift. This event was least evident at stakes 35, 38, 39 and 41. In fact, marker 38 increased its downward velocity during this time. During the following period, Julian day 224-230, all stakes had downward velocities . Uplift was recorded during the final observation period of Julian day 230-234 (18-22 August) at markers 30, 33, 38, 39 and 41. Most remaining markers decreased their downward velocities during the same interval, although stakes 31, 35 and 37 increased the magnitude of their downward motion. Short-term variations in the pattern of vertical velocity were also recorded during the 1990 field season (Figure 6.6b). The record for marker 1, located down- glacier from the bulge is alone in not having any noticeable variations. At this point, vertical velocity remains less than 2 mm d-1 and is always downward. The pattern of velocity fluctuations at other markers exhibits a degree of variation for the first interval of observation (Julian days 165-172, i.e. 14-21 June). At most locations, vertical velocities are downward. Only stakes 5 and 7 experience uplift during this period. During the following interval between Julian day 172-17 4 most markers attain peak rates of uplift. Exceptions were stakes 2 and 7 which experienced downward motion. After the peaks of uplift, marked downward velocities were noted during the period Julian day 174-176. Phases of uplift were observed in the interval Julian day 176-181 at markers 3, 4, 5, 7, 8, 10, 11 and 12. At the remaining stakes, velocity decreased its downward component. The final period of observations was characterised by decreases in uplift velocity at markers 8 and 13, and downward motion at the remaining points. Marker 6 was slightly different because it experienced uplift during this final measurement interval. However, the observational record for this point only spans the last two surveyed periods. The interpretation of these results is presented in section 7.4.3. Page 185 Figure 6.6b Marker 1 0.20 0.15 ~ 0.10 .s 0.05 -~ u .Q -0.00 Q) > '° -0.05 u t (l) ·0.10 > -0.15 -0.20 160 170 180 190 Julian day Marke r 3 0.20 0.15 u 0 .1 0 ]. 0.05 -~ I u -0.00 [T+ .Q a, > '° -0.05 ,',! 't: Q) -0.10 > -0.15 -0.20 160 170 180 190 Julian day Ma1·ker 6 0.20 0.15 ~ 0.10 ! E _J 0.05 ,.. u 0 -0.00 '" I > -;:; -0 .05 ·" a, -o. rn > -0.15 -0.20 160 170 160 190 Julian day Short tem1 vertical velocities for six targets located on the centreline of Bjuvbreen in the 1990 field season. Marker 3 is located at the base of . the bulge. The furthest up-glacier target.is marker 13. Marker locations · are shown in the inset overleaf. Vertical lines represent error limits . The variation in vertical velocities on the lower glacier (markers 1 and 3) and the crest of the bulge (marker 6) is smaller in contrast to vertical variations of targets on the upper glacier. A significant uplift event occurs between Julian 172- 174. [continued over ... ] Page 186 Figure 6.6b Continued. -~ u 0 "' > 'iii u ·e n, > .?: u 0 '" > "' u ·1: a, > 0.20 0.15 170 100 190 Julian d~y Mc1 rkcr 11 100 J ulian day M;:irker 13 100 190 J ulian day Chapter 6: Flow and deformation of Bjuvbreen 6.4 SURFACE STRAIN RATES 6. 4 .1 Introduction Strain measurements were made on the surface of Bjuvbreen during the 1989 and 1990 field seasons using methods described in Section 4.5. The location of markers used in the analysis is shown in Figure 4.5. Measurements of strain rates were made, on average, every 15. days in order to reduce the proportion of errors. Strain rates can be measured over shorter periods using high-precision wire strain gauges (e.g. Evans et al., 1975; Raymond and Malone, 1986; Jansson and Hooke, 1989). 6. 4. 2 Strain rates in 1989 During the 1989 field season, strain markers (Figure 4.5) were surveyed on three occasions, allowing strain rates to be determined over two intervals equating to the middle and final stages of the ablation season. The two periods were slightly different for the strain diamonds in the region near the bulge and the strain triangles further up-glacier. The dates of the measurement intervals are listed in Table 6.2. The results are expressed as parts per year, although it is emphasized that the measurements were not made one year apart. During t1, the principal axis of strain at the base of the bulge is compressive for both diamonds 2 and '3 (roughly -0-030 a-1 ). At diamond 1, the principal strain axis is strongly compressive (-0-070 a-1 ). At diamond 4 the measured strain rate was very small and fell within the error limits, hence making it difficult to determine whether flow was extensional or compressive. Diamonds 5 and ~, however, experienced compressive strains of -0-035 a-1 and -0-043 a-1 respectively. Up-glacier from the bulge there were thirteen strain triangles arranged sequentially up the glacier centreline (Figure 4.5). No measurements are available for triangle 1 during t1. Thus, the description of the strain rate pattern begins with the results from triangle 2 and progresses towards the head of the glacier. The pattern of surface strain on the glacier cen'treline during t1 was generally compressive, although tensile in places (Figure 6.7a). Flow was compressive from triangle 2 until triangles 7 and 8, where extensional strains were recorded, before becoming compressive once again. Extending strain was measured at triangles 11 and 12 but was compressive at the furthest up-glacier triangle, 13. The spatial distribution of surface strain rate is illustrated in Figure 6.7a. During t2 in 1989, compressive strain rates were observed at all three diamonds at the base of the bulge. Values of the principal strain rate axis at these sites varied from -0-018 a-1 to-0-087 a-1. On the crest of the bulge, strain rates at all three diamonds are again compressive, ranging from -0-005 a-1 to - 0-045 a-1. Moving up the glacier Page 187 \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ .... \ .... \ \ .... .... .... .... o o. o5 o . ·t 0 .1 5 strain rate, a·' extending strain \ \ \ compressive st rain /' /' / - - - - ' L----------------.--__.__------::-----1 \ o 100 ioo rn , _ , ___ l _ ,_J Figure 6.7a . Surface strain rate on Bjuvbreen during t1 in the 1989 field season. Flow is compressive at most locations on the glacier. Extending flow was observed to occur only at triangles 7, 8 and 13 on the upper glacier (see inset). The pattern of flow in these areas is related to the sub glacial topography. Strain rates during t2 are illustrated overleaf. Page 188 If I \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ ........ - ..... 0 0 .05 0 .1 0 . 15 St r a in rate a_, E x tending s train Compr ess i ve st r a i n 0 100 200 "' '··- ·- --'- ·- · / / I Figure 6.7a . [continued] Surface strain rate on Bjuvbreen during t2 in the 1989 field season. The pattern of flow is similar to that observed during t1, with tensile and compressive strains occurring at the same locations as earlier. Chapter 6: Flow and deformation of Bjuvbreen Strain rate period Julian days Calendar dates 1989 t1 (bulge) 201-217 20 July-5 August 1989 t 1 (up-glacier) 208- 222 27 July-10 August 1989 t2 (bulge) 217- 234 5 August-22 August 1989 t2 (up-glacier) 222-233 10 August-21 August 1990 t1 165-176 14 June-25 June 1990 t2 176--184 25 June- 3 July Table 6.2 Intervals over which strain rates were measured on the surface of Bjuvbreen. In the 1989 field season, markers in the area of the bulge were surveyed at slightly different times than those on the centreline on the upper glacier. During the 1990 season, all markers were surveyed at the same times. Results were expressed in parts per year. centreline, strain rates are strongly compressive. Extending flow then occurs at triangles 7 and 8, before becoming compressive once again. This type of flow persists until triangle 13, where tensile strain was recorded. The large-scale longitudinal strain, cJu/dx, was calculated using velocity gradient data (Meier, 1960) from targets 30 and 44. This provides a measure of the gross dynamics of an area of Bjuvbreen up-glacier from the bulge, and serves as a useful comparison to the more spatially restricted strain diamonds. The distance over which the strain rate was calculated was approximately 500 m. Only the initial and final velocity measurements of the 1989 field season were used. Therefore, the strain rate, expressed in parts per year, represents the deformation taking place over a period of 25 days. A longitudinal strain rate of - 0·045 a-1 was calculated, indicating compressive flow in this region of the glacier. 6.4.3 Strain rates 1990 The location of the strain network during the 1990 field season is shown in Figure 4.5 . In 1990, the strain measurements were made over shorter time intervals. The strain diamonds above and below the bulge were surveyed once only, over an interval of 8 days. I{owever, the strain triangles located up the centreline from the bulge were surveyed over two intervals. The time periods used in the calculation of strain rates are given in Table 6.2. Strain measurements in the area of the bulge were made during the second interval. As with the 1989 data, the strain rates are expressed in parts per year (Figure 6.7b). Longitudinally compressive strain was recorded at diamond 1 located at the bottom of the bulge. Compression was equal to -0-015 a-1 which was less than values of -0·030 a- 1 and -0·018 a-1 recorded at the same site the previous summer. At diamond 2 located on the left margin above the bulge, the principal strain rate is weakly Page 189 Figure / / \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ / \ \ / / \ \ \ \ \ \ \ \ \ \ / \ \ \ / / / ' ' \ \ \ \ \ \ - - - 0 0 .05 0 .1 0 . 1 Strain rate a-1 ,, E x tending strain Com pressi ve st r a in ..__---------------.---.'-,,-------,,, i l1l~n1,1 I ~ •;' \ \ "/ \ \ \r J ' ' ,. ... :.~·:·::· ' .. '.t ,/ Dia m on d 1 n • ( <• ci~; . ' 6. 7b Surface strain rate on Bjuvbreen during the 1990 field season. Strain · arrays at the base and the crest of the bulge were only surveyed once. The strain triangles on the upper glacier were surveyed twice. Strains measured during t1 are indicated by the solid lines. The flow at the base of the bulge was less compressive than during the 1989 field season. A small amount of extending flow was measured at diamond 3 (see inset). On the upper glacier, extending flow occurs at the same locations as in 1989. Page 190 I i I Chapter 6: Flow and deformation of Bjuvbreen compressive (-0·005 a-1). However, on the opposite margin, at diamond 3, a small amount of extending flow (0·025 a-1) was measured. Four strain triangles were placed longitudinally up the glacier centreline with triangle 1 being located closest to the crest of the ramp. Over the first measurement interval, flow in triangle 1 was compressive, amounting to -0·055 a·1. In triangles 2 and 3, however, strongly extending flow (- 0-090 a-1) was measured. Compressive flow was recorded again in triangle 4. During t2, compressive flow persisted at triangle 1. The pattern observed during the previous interval was repeated with extending strains again recorded at triangles 2 and 3, although they were approximately only 66% as extensive compared with the initial observation period. Compressive strain of - 0-017 a-1 was measured at the furthest up-glacier triangle (triangle 4) (Figure 4.5). This alternating pattern of extensional and compressive strain rates is related to changes in the ice surface slope. Areas where strains were tensile coincided with increased gradients. Similarly, compressive flow was observed where the surface of the glacier reduced its gradient. The changes in surf ace slope result from variation in the subglacial topography. The decrease in surface gradient is caused by the glacier flowing over the riegel on the bed (section 5.2.3). The large-scale strain rate pattern, du/dx, was calculated in this region by computing the velocity gradients between targets 8 and 13. The distance between these points was roughly 300 m and the measurements were made 19 days apart. A strain rate of - 0·084 a-1 was calculated, indicating that strongly compressive flow was occurring in this region at that time. The flow is more compressive than was recorded in the same region during the 1989 field season. 6. 4. 4 Comparison of strain rates with crevasse distribution Crevasses are generated in response to the motion of ice. In the absence of direct measurements of strain rate, the pattern of crevasses can give an qualitative indication of the surface strain behaviour of a glacier. Crevasses form when the principal extensional strain rate exceeds a critical value. The critical strain rate differs according to the varying thermal regime of ice. Meier (1958) demonstrated that crevasses do not form on temperate glaciers until extending strain rates of +0·01 a-1 occur. On polar glaciers, however, much lower extensional strains are required for crevasses to form. Hambrey and Muller (1978) found on White Glacier that crevasses formed at strain rates of only +0·004 a-1. Similarly, Holdsworth (1969) reported crevasse formation on Meserve Glacier, Antarctica, at strain rates of +0·002 a-1. Crevasse orientation can indicate the nature of the flow of a glacier, following the observation that crevasses form perpendicularly to the principal axis of extending strain (Meier, 1960). Page 191 Chapter 6: Flow and deformation of Bjuvbreen Figure 6.8a illustrates the pattern of crevasses on the surface of Bjuvbreen in August 1990. Crevasses are limited to the upper region of the glacier, up-glacier from the bulge. The distribution of crevasses is fairly regular as would be expected on the surf ace of a glacier in quiescence. A noticeable feature of the crevasse pattern in 1990 is the absence of fractures in the area immediately behind the bulge. This area is a zone of compressive flow (Figure 6. 7), where ice being transported down from the upper glacier cannot move forward .at the same speed because of the presence of an obstruction at the bulge. The only crevasses near the bulge are a few small fractures on the crest near the margins. These fractures have formed at the break of slope, presumably as the result of locally extensile strains, caused by ice being pulled downhill by a rapid increase in local gradient. Field observations confirm that these cracks do not penetrate more than two metres into the glacier. Crevasses begin to appear on the ice surface approximately 310 m up-glacier from the crest of the bulge. The leading edge of the crevasse field has an arcuate pattern which is concave up-glacier. The limit of the crevasses traces roughly the topography the spoon-shaped depression described in Chapter 5. Nye (1952) illustrated various crevasse patterns on valley glaciers which result, theoretically, from different flow regimes. The crevasse pattern observed on Bjuvbreen corresponds to the arrangement Nye predicted would result from the combined effects of extending flow and the resistive drag of valley walls. Field measurements confirm that this area is experiencing extensional flow. Measured strain rates are in excess of those required for crevasse formation. As a comparison, the crevasse pattern in the same area of the glacier is shown for 1977 (Figure 6.8b). At this earlier stage of quiescence, the amount of crevassing present is considerably more limited. The most noticeable zones of fracturing are on the western margin and in the central eastern area. The snow cover on the glacier in the aerial photograph, from which the 1977 crevasse pattern was traced, was more extensive than on the 1990 image. Thus, it is conceivable that some crevasses were present beneath the snow but went undetected, although it is unlikely that a significant number of crevasses were missed. The crevasse pattern observed in 1977 suggests that although extensional flow was occurring in the uppermost part of the glacier, it was present over a more spatially limited area compared with 1990. Page 192 A 1977 0 500 m bulge B 1990 0 500 m Figure 6.8 Pattern of crevasses on the surface of Bjuvbreen in 1977 (A) and 1990 (B).The portion of the glacier illustrated lies,up-glacier from the base of the bulge. There were no crevasses down-glacier from this point in either year. The crevasses were traced from aerial photographs taken in August of the respective years. It is unlikely that a significant number of crevasses were concealed by late-lying snow. A considerable amount of · crevasse formation has occurred in the 13 years between photographs, indicating that tensile strains are becoming stronger and more widespread. Note the arcuate leading edge of the crevasse field in 1990, and the down-glacier crevasse-free zone. Compressive flow was measured in this location. Page 193 I I I I : I I : I I' Chapter 6: Flow and deformation of Bjuvbrecn 6.5 BASAL SLIDING AND INTERNAL DEFORMATION OF BJUVBREEN 6.5.1 Background This section investigates the separate components of the flow of Bjuvbreen. The relative proportion that basal motion and internal deformation contribute to the total flow of a glacier varies between ice masses. Paterson (1981) found that basal sliding accounted for roughly half the total movement of temperate glaciers, although this proportion varies considerably between individual glaciers (e.g. Table 5.2 in Paterson, 1981). Furthermore, the proportion_ of each component varies temporally and also spatially within a glacier. Raymond and Harrison (1988) reported that the change in velocity structure of Variegated Glacier as the surge cycle progressed was due to an increase in the amount of basal sliding taking place. This increase was especially marked in the five years immediately prior to the surge. Kamb et al. (1985) demonstrated that basal sliding was the most important component of motion during the surge of Variegated Glacier, accounting for 95% of the observed velocity. Therefore, motion at the base of glaciers is considered to be a crucial aspect of fast glacier flow. However, Echelmeyer and Harrison (1990) have suggested that the exceptionally high velocities of Jakobshavn Isbrre might not be related to basal sliding but instead could be caused by particularly rapid internal deformation associated with very large shear stresses. 6. 5. 2 Calculation of the basal sliding velocity The total surface velocity of a glacier, u, is the sum of two components, defined as: U=Uct+UB (6.1) where Uct is the velocity due to internal deformation and uB is the contribution arising from basal sliding. Basal sliding in this expression can include discrete movement along the ice-substrate contact or the deformation of subglacial sediments. The contribution of basal sliding can be calculated from Nye's (1952) theoretical work. The sliding velocity is found from the expression: UB = U - [__2A__J ~ h n+l (6.2) where n and A are components in the flow law. The above equation shows that the velocity resulting from internal deformation depends upon the quantity '(2A/n + l)rBnh. This quantity will not change greatly with time for a given location on a glacier, because both TB and hare geometry dependent and do not change rapidly. Thus, it is apparent Page 194 Chapter 6: Flow and deformation of Bjuvbreen that much of the variation observed in the velocity of glaciers cannot be explained by geometry-influenced deformation processes. The remaining velocity is usually assumed to be occurring at the base of a glacier. To examine the relative proportions of basal and deformational velocity present in the observed surface velocity records of Bjuvbreen, uB was computed using equation 6.2. Basal velocity was calculated using measured surface velocities and ice thicknesses for several points on the glacier centreline. Shear stresses and ice thicknesses were not reliably known for points lying away from the central axis, thus basal velocities were not calculated for these targets. Values of the flow law parameters were taken from data recommended for use in glaciological computations by Paterson (1981). The exponent, - n, was taken to be 3. A value of A was more difficult to select, since this parameter varies according to the temperature of the ice. For present purposes, the ice in Bjuvbreen was assumed to be close to O °C (Section 5.5). Therefore, A was taken to be 5.3 x lQ-1 5 s-1 kPa-3 (Paterson, 1981, p. 39), although the effect of varying this parameter is examined. In 1989, basal velocities were calculated for targets 4, 19, 30, 34, 38 and 42. In 1990, data from targets 1, 3, 6, 9, 11 and 13 were used to obtain basal velocities. Basal shear stresses used as input in equation 6.2 were obtained using the Kamb---Echelmeyer theory which incorporates longitudinal stress gradients (section 5.3.4) (Kamb and Echelmeyer, 1986). 6. 5. 3 The contribution of basal sliding to the total velocity of Bjuvbreen The basal sliding component was initially calculated for the seasonal velocities of the targets identified above. Basal velocity was expressed as a percentage of the total observed surface velocity. Figure 6.9a illustrates the proportion of basal motion at each of the selected targets during the 1989 season . This figure shows that there is a trend for the contribution of basal sliding to decrease with increasing distance up-glacier in response to increasing ice thicknesses. In 1989, 95% (0·067 m d-1) of target 4' s total velocity was accounted for by motion at the base. However, at the furthest up-glacier targets used in thi~ analysis, 38 and 42, only about 50% (0·045 and 0·067 m d-1 respectively) of the observed velocity was calculated to be occurring at the base. A similar pattern was found for the 1990 data (Figure 6.9b). The proportion of basal sliding was greatest for points near the bulge and decreased markedly up-glacier. At target 13, the furthest up-glacier point, only 20% (0·016 m d-1) of the observed surface motion can be explained by basal sliding. The low deformation velocities (0·001- 0·003 m d-1 ) calculated for the points down-glacier from the bulge are due to the small ice thicknesses present in this part of the glacier. According to equation 6.2, ice which is very thin, less than 20 m thick in Page 195 1989 ... ~ 1 , 111nu1, 1y \ 0 100 ai 4J •. _.42 (/) .'c./\., C1l ,)/\,. .0 Q) 80 \/\ ..c J7J-;:./·,:J. -cu \/\ CJ) 3 3 •-32 C 60 \~T•h n1.11111 ·t: JI •-....._.JO \~ A ::, ,I? ~ 11/tr:.1!' • (.) (.) •" .,[~~~'7'00 m l _ ,_ i__, _ J o I ~ 0 0 Q_ e Cl.. 3 6 9 11 13 Mc1rker Figure 6.9 Basal motion as a proportion of the total seasonal surface velocity at selected markers on the glacier centreline for 1989 and 1990. Stake locations are shown in the insets. Basal velocities were computed using equation 6.2. The proportion that basal motion contributes to the total · velocity decreases with increasing distance up-glacier. This decrease is balanced by an increase in the amount of internal deformation that can occur in regions with greater ice thicknesses. Although the proportion of basal motion is highest at the furthest down-glacier locations, total surface velocities in these areas are relatively small. Page 196 Chapter 6: Flow and deformation of Bjuvbreen this case, is unable to drive itself downslope. Therefore, much of the observed motion must be occurring at the base of the glacier. The high proportion of basal velocity to total velocity in this part of the glacier should be taken in context. Although basal velocity can account for up to 90% of the total surface velocity, the observed surface velocities are relatively low (about 0-05 m d-1 at the base of the bulge decreasing to 0·02 m d-1 further down-glacier). Actual basal velocities averaged over the season are greatest in a region 250-- 300 m up-glacier from the crest of the bulge. This pattern is repeated in both field seasons. Directly beneath this area of the glacier there is a noticeable hummock in the subglacial topography (section 5.2:2). The peak of the topographic high lies approximately beneath target 9 (target 34 in 1989). Basal shear stresses for this part of Bjuvbreen are relatively low compared to elsewhere on the upper glacier (section 5.3.4). Therefore, on Bjuvbreen, the highest basal sliding velocities are not associated with the greatest shear stresses. This observation contradicts Raymond and Harrison (1988), who found that regions of high basal sliding on the quiescent Variegated Glacier roughly coincided with regions with high driving stresses. The contrasting pattern observed on Bjuvbreen could result from a decrease in the effective normal stress caused by relatively higher subglacial water pressures under this part of the glacier. These higher pressures could be caused by a build-up of water behind the up- glacier side of the riegel. An increase in the production of water by pressure melting, resulting from glacier flow against the up-slope side of the riegel, might possibly contribute to increased amounts of water present in this area of the bed. The effect of varying the value of the flow law parameter The calculations of seasonal basal motion on Bjuvbreen, discussed above, assumed thatthe flow law parameter, A, was equal to 5·3 x lQ-15 s-1 kPa-3. This value was chosen on the basis that ice in the glacier was believed to be close to O °C (section 5.5). However, as was pointed out in section 5.5, the results of modelling the thermal regime of Bjuvbreen were sensitive to the boundary conditions used. Had the actual MAA T of -11 ·05 ° C at the glacier surf ace been used as the upper boundary condition instead of -2·8 °C (section 5.5.3), the resulting thermal characteristics of Bjuvbreen would have been markedly different. Therefore, because A is temperature-dependent, basal velocities were re-calculated by varying the ice temperature to-5 °C and -10 °C. The corresponding values of A were 1·7 x lQ-15 s-1 kPa-3 and 5·2 x lQ-16 s-1 kPa-3 respectively (Paterson, 1981 , p. 39). The results are given in Appendix B. The primary effect of reducing the ice temperature was to increase the calculated amount of basal motion that can occur at a given target. This is because ice at -10 °C is much stiffer than ice at O °C and, therefore, less able deform under its own weight Page 197 Chapter 6: Flow and deformation of Bjuvbreen (Paterson, 1981 ). The parameter A takes account of these differences in viscosity. The changes in basal velocity, when ice temperature is altered, are most noticeable (Appendix B) for the targets located furthest up-glacier (i.e. markers 38 and 42 in 1989 and markers 11 and 13 in 1990). These were the targets where the greatest proportion of the total surface motion was originally calculated to be occurring through internal deformation. At the other markers on the glacier surface, the proportion of their velocity contributed by creep was much smaller in the original calculations. Therefore, there was less scope for these targets to substantially increase their basal velocities when the ice temperature was reduced. Short tennfluctuations in basal velocity Because the amount of internal deformation that can occur is controlled by the flow law parameters and glacier geometry, short term variations in glacier velocity must be caused by another mechanism. This additional motion is assumed to occur at the base of an ice mass. The amount of basal motion occurring over short periods on Bjuvbreen was calculated for each of the selected centreline targets during both field seasons. The pattern of calculated basal velocities during the 1989 field season is roughly similar for each of the targets. Motion at the base is greatest (0·06 m d-1) at the start of the measurement period and then decreases in steps until the middle of the monitoring phase. Most targets experience minimum basal velocities at the same time. Near the base and crest of the bulge, minimum velocities (-0·02 m d-1) occurred between Julian days 217- 223. Minimum basal velocities (-0·01 m d-1) on the upper glacier were recorded between Julian days 216-222. The records of basal motion for the 1990 field season demonstrate a good deal of agreement between individual targets. Basal velocities at the start of the field season were low at all points (-0·05 m d-1 ), before increasing for the period between Julian days 174-176 (0·08 m d-1 at the base of the bulge and -0·11 m d-1 on the upper glacier). Between Julian days 176-181, subglacial motion decreased simultaneously at all stakes. Below the bulge, basal motion almost ceased (velocities were <0·01 m d-1 ). Sub glacial velocities were also much reduced on the upper portion of the glacier (-0·03 m d-1 ). There was then a small increase in basal velocity at all targets, except 6 and 13, during the final period of observation in 1990. The similarity in the record of basal motion at each target, observed during both field seasons, suggests that the cause of the variations is occurring repeatedly and almost simultaneously beneath large areas of the glacier base. The potential for these changes to be related to the hydrology of the glacier will be examined in the following chapter. Page 198 : II' Chapter 6: Flow and deformation of Bjuvbreen (Paterson, 1981). The parameter A takes account of these differences in viscosity. The changes in basal velocity, when ice temperature is altered, are most noticeable (Appendix B) for the targets located furthest up-glacier (i.e. markers 38 and 42 in 1989 and markers 11 and 13 in 1990). These were the targets where the greatest proportion of the total surface motion was originally calculated to be occurring through internal deformation. At the other markers on the glacier surface, the proportion of their velocity contributed by creep was much smaller in the original calculations. Therefore, there was less scope for these targets to substantially increase their basal velocities when the ice temperature was reduced. Short term fluctuations in basal velocity Because the amount of internal deformation that can occur is controlled by the flow law parameters and glacier geometry, short term variations in glacier velocity must be caused by another mechanism. This additional motion is assumed to occur at the base of an ice mass. The amount of basal motion occurring over short periods on Bjuvbreen was calculated for each of the selected centreline targets during both field seasons. The pattern of calculated basal velocities during the 1989 field season is roughly similar for each of the targets. Motion at the base is greatest (0·06 m d-1) at the start of the measurement period and then decreases in steps until the middle of the monitoring phase. Most targets experience minimum basal velocities at the same time. Near the base and crest of the bulge, minimum velocities (-0·02 m d-1) occurred between Julian days 217- 223. Minimum basal velocities (-0-01 m d-1) on the upper glacier were recorded between Julian days 216-222. The records of basal motion for the 1990 field season demonstrate a good deal of agreement between individual targets. Basal velocities at the start of the field season were low at all points (-0·05 m d-1 ), before increasing for the period between Julian days 174-176 (0·08 m d-1 at the base of the bulge and -0· 11 m d-1 on the upper glacier). Between Julian days 176-181, subglacial motion decreased simultaneously at all stakes. Below the bulge, basal motion almost ceased (velocities were <0·01 m d-1). Sub glacial velocities were also much reduced on the upper portion of the glacier (- 0·03 m d-1 ). There was then a small increase in basal velocity at all targets, except 6 and 13, during the final period of observation in 1990. The similarity in the record of basal motion at each target, observed during both field seasons, suggests that the cause of the variations is occurring repeatedly and almost simultaneously beneath large areas of the glacier base. The potential for these changes to be related to the hydrology of the glacier will be examined in the following chapter. Page 198 11 Chapter 6: Flow and deformation of Bjuvbreen 6.6 BALANCE VELOCITY AND FLUX 6. 6 .1 Introduction The glacier velocity that is required in order to maintain a steady profile can be calculated theoretically. This theoretical value is known as the balance velocity. Comparison of the balance velocity with the observed velocity provides a measure of the equilibrium state of a glacier. Surge-type glaciers are well known to have an imbalance between observed and balance velocities (Clarke, 1987a). During the quiescent phase, balance velocities exceed actual velocities, thus causing the glacier to build up mass. This situation is reversed during the active phase, when the glacier discharges ice at a rate which cannot be sustained. As a result the surge terminates and the glacier begins the process of building up towards another surge. 6. 6. 2 Calculation of balance velocity The theoretical velocity required to maintain an equilibrium profile on Bjuvbreen, given its present mass balance, was calculated using available field data. The first stage of this procedure was to compute the balance flux at the equilibrium line. The balance flux is equal to the areal addition of mass in a given period (usually a year) for the portion of glacier lying up-stream from the chosen point. The balance flux at point x, Qb(x), was found from the expression: Qb(x) - f w(xo) B(xo) dx (6.3) where w is glacier width as a function of distance (x = 0 at the head of the glacier) and B is the accumulation rate (Bindschadler, 1984; Clarke, 1987a). The flux was calculated at the equilibrium line (section 6.3.1). The surface area up-glacier from this point was digitized from the 1: 12,000 scale map of Fox (1989). Mass balance was not measured on B juvbreen, so a value was taken from Simoes' s (1990) work on nearby Skobreen. The variation of mass balance with altitude was not known. Therefore, the net accumulation was assumed to be constant at 0.36 m a-1 (the ice equivalent of Simoes's figure) for the area above the equilibrium line. The balance velocity, Vb, is the velocity that the glacier must move at in order to discharge its annual input of mass and maintain a steady profile. This value is an average of the velocity over the cross-sectional width at the equilibrium line. The balance velocity was found from: (6.4) Page 199 Chapter 6: Flow and deformation of Bjuvbreen where Ac is the area of the cross-section at the equilibrium line (section 5.4.2). These calculations provided a value of Vb of 6·5 m a-1. 6 . 6. 3 Comparison between balance velocity and actual velocity Field measurements from 1990 were used as the basis for estimating the actual velocity at the equilibrium line. It was explained in section 4.5.2 that annual velocities were not measured on Bjuvbreen. Therefore, only a seasonal velocity figure was available for comparison with the calculated annual balance velocity. Instead of using the measured seasonal velocity at marker 12, the theoretical amount of internal deformation occurring at that point in or~e year was calculated from the expression: Uct = (2A/n + l)rBnh. (6.5). The result of 16·9 m a-1 was taken to represent the actual surface velocity, Us, at the equilibrium line. The computed balance velocity, Vb, represents an average velocity over the channel cross-section. The actual velocity, however, indicates the speed on the surface at the glacier centreline. Therefore, in order to make the actual value strictly comparable to the balance value, the centreline velocity must be converted to a cross-sectional average. Had the surface velocity been measured at several points across the glacier width, numerical calculations performed by Nye (1965) showed that the average of these values is a good approximation of the average velocity through the cross-section. Raymond (1971), however, found from work on Athabasca Glacier, that if appreciable basal sliding occurs, average velocities predicted using Nye's method may be too low. In the absence of a transverse velocity profile at the equilibrium line, the average velocity of the cross-section had to be determined by another method. Nye (1965, p. 677) presented numerically derived solutions of the ratio of surface centreline velocity to average cross-sectional velocity. These solutions were only derived for parabolic glacier channels, although Bjuvbreen is assumed to flow in an approximately semi- elliptical channel (section 5.3.2). Furthermore, Nye's solutions assumed that all motion within a glacier occurs by internal deformation. The ratio of averaged cross-sectional velocity, u, to the observed centreline velocity, u5 , was estimated from values given in Nye (1965, Table IIIB, p. 677). This ratio, f*, is known as the flux shape factor (Bindschadler et al., 1977). The observed average cross-sectional velocity was then computed from the expression: U = f* Us (6 .6) . The result of these calculations is an actual cross-sectional velocity of 10-99 m a-1. Because the observed average velocity was calculated assuming no modon at the base, the realistic average velocity is likely to be somewhat higher due to the demonstrated occurrence of basal sliding (section 6.5.3) during at least part of the year. Page 200 [, I, r Chapter 6: Flow and deformation of Bjuvbreen where Ac is the area of the cross-section at the equilibrium line (section 5.4.2). These calculations provided a value of Vb of 6·5 m a-1. 6 . 6 . 3 Comparison between balance velocity and actual velocity Field measurements from 1990 were used as the basis for estimating the actual velocity at the equilibrium line. It was explained in. section 4.5.2 that annual velocities were not measured on Bjuvbreen; Therefore, only a seasonal velocity figure was available for comparison with the calculated annual balance velocity. Instead of using the measured seasonal velocity at marker 12, the theoretical amount of internal deformation occurring at that point in on~ year was calculated from the expression: Uct = (2A/n + l)rBnh. (6.5). The result of 16·9 m a-1 was taken to represent the actual surface velocity, Us, at the equilibrium line. The computed balance velocity, Vb, represents an average velocity over the channel cross-section. The actual velocity, however, indicates the speed on the surface at the glacier centreline. Therefore, in order to make the actual value strictly comparable to the balance value, the centreline velocity must be converted to a cross-sectional average. Had the surface velocity been measured at several points across the glacier width, numerical calculations performed by Nye (1965) showed that the average of these values is a good approximation of the average velocity through the cross-section. Raymond (1971), however, found from work on Athabasca Glacier, that if appreciable basal sliding occurs, average velocities predicted using Nye's method may be too low. In the absence of a transverse velocity profile at the equilibrium line, the average velocity of the cross-section had to be determined by another method. Nye (1965, p. 677) presented numerically derived solutions of the ratio of surface centreline velocity to average cross-sectional velocity. These solutions were only derived for parabolic glacier channels, although Bjuvbreen is assumed to flow in an approximately semi- elliptical channel (section 5.3.2). Furthermore, Nye's solutions assumed that all motion within a glacier occurs by internal deformation. The ratio of averaged cross-sectional velocity, u, to the observed centreline velocity, Us, was estimated from values given in Nye (1965, Table IIIB, p. 677). This ratio, f*, is known as the flux shape factor (Bindschadler et al., 1977). The observed average cross-sectional velocity was then computed from the expression: U = f* Us (6 .6) . The result of these calculations is an actual cross-sectional velocity of 10·99 m a-1 . Because the observed average velocity was calculated assuming no motion at the base, the realistic average velocity is likely to be somewhat higher due to the demonstrated occurrence of basal sliding (section 6.5.3) during at least part of the year. Page 200 Chapter 6: Flow and deformation of Bjuvbreen The comparison between predicted and observed velocities is illustrated in histogram form in Figure 6.10. This figure shows that the actual velocity at the equilibrium line of Bjuvbreen exceeds the balance velocity (u = 10·99 m a-1 and Vb = 6·5 m a-1 ). This situation is likely to be unrealistic, otherwise Bjuvbreen will be unable to build up mass before another surge. Evidence presented in section 5.2.1 demonstrated that ice thicknesses in the upper portion .of the glacier are increasing as the quiescent phase progresses, indicating that the actual velocity of Bjuvbreen is lower than its balance velocity. One potential explanation of why the actual velocity exceeds the balance velocity is that an incorrect value of A was used to calculate the amount to internal deformation (equation 6.5). In th~ above computations, A was assumed to be 5.3 x lQ-15 s-1 kPa-3 representing an ice temperature of O °C. Section 6.5.3 and Appendix B discussed the effect on the calculated internal deformation velocities if the ice temperature in Bjuvbreen is below O °C. When the actual velocity of marker 12 is re-calculated assuming an ice temperature of - 5 °C, the result is u = 3-6 m a-1. In this case, the balance velocity exceeds the actual velocity (Figure 6.10). It should be remembered that subglacial motion has not been considered in the above estimates of actual velocity. Therefore, the realistic actual velocity at the equilibrium line is probably slightly higher than 3·6 m a-1. The estimate of u ~ 3·6 m a-1 probably provides a reasonable value for comparison with the balance velocity, although it suggests that the ice in Bjuvbreen is stiffer than has been assumed elsewhere in this thesis. Bearing in mind that u might be slightly higher because of basal motion, this section has demonstrated that the dynamic imbalance present on Bjuvbreen is not very large. Bindschadler et al. (1977, unpublished) reported that Variegated Glacier was transporting much less ice than that required to maintain a steady-state configuration, ten years before its last surge (1982- 83). Furthermore; they found that changes in glacier geometry as a result of the mass imbalance were clearly observable after a one-year period. The same situation is not present on Bjuvbreen because the difference between the balance velocity and the actual velocity is quite small. In Chapter 5, the progressive evolution of the geometry of Bjuvbreen through quiese:ence was described and, in particular, the slow rate at which changes occur was emphasized. This slow build-up of Bjuvbreen, and the long duration of the quiescent phase on other surge-type glaciers in Svalbard in general, can be explained by the balance analysis presented in this section. Annual accumulation on Svalbard glaciers is considerably lower than on glaciers in non-polar regions (Hagen and LiestlZll , 1990) . The accumulation is low enough that many glaciers in the archipelago are able to discharge their mass input and remain close to steady-state. This situation is also approached by a number of surge-type glaciers. Even where all the Page 201 ,- 'C1l E 15 10 5 0 Chapter 6: Flow and deformation of Bjuvbreen El Balance velocity • Actual velocity Figure 6.10 Comparison between balance velocity and actual velocity at the equilibrium line. The balance velocity is the average speed through the cross-section that Bjuvbreen must move in order to discharge its annual mass input and maintain a constant profile. The actual velocity represents the cross-sectional average of the calculated creep velocity at marker 12 obtained using theoretical methods of Nye (1965) and Bindschadler et al . (1977) (see text). The actual velocity was calculated assuming ice temperatures of O °C (left) and-5 °C (right). motion within a glacier is by internal deformation, accumulation is small enough that slow deformational velocities are often sufficient to maintain a near-steady-state balance. In an environment such as south-east Alaska, the annual accumulation on glaciers is much greater than on glaciers in Svalbard (Harrison et al., 1983; Mayo, 1984). Consequently, annual velocities may not be sufficient to maintain a steady-state profile. Glaciers such as Variegated Glacier, therefore, build up towards surges at a faster rate than Bjuvbreen and other polar surge-type glaciers. 6.7 SUMMARY This chapter has discussed the flow behaviour observed on Bjuvbreen during two successive field seasons. The avai lable information was used to deduce the pattern of velocity and strain at the glacier surface. The study represents the first detailed examination in Svalbard of the motion of a surge-type glacier during quiescence. The main points of this chapter can be summarized as follows: • Seas·onal velocities are higher at points up-glacier from the bulge than those on the lower glacier. Horizontal velocities on the upper glacier, averaged over the duration of the field seasons, had a mean value of -0· 10 m d-1. Page 202 ii Chapter 6: Flow and deformation of Bjuvbreen • Seasonally-averaged velocities were slightly higher during the 1989 field season than in 1990. These differences could not be explained by changes in the glacier geometry and were probably due to a variation in seasonally- influenced basal processes. • Horizontal and vertical velocities were both observed to vary over short intervals of time. Transient fluctuations in horizontal velocity never exceeded 100% of the seasonal average. • Short-term increases in horizontal velocity were generally correlated with increases in vertical velocity. • At the base and the crest of the bulge, the flow of the glacier is compressive. Along the centreline on the upper glacier, the pattern of flow is also compressive although extending flow occurs in one small area. This flow pattern is related to local surface slopes which are influenced by the subglacial topography. • The calculated proportion that basal sliding and internal deformation contribute to the observed surface velocities varies spatially and temporally over Bjuvbreen. The greatest amount of seasonal basal sliding appears to be occurring in an area 250--300 m up-glacier from the bulge. Further up-glacier, the amount of calculated basal sliding decreases as greater thicknesses of ice become more effective at causing internal deformation. • The balance velocity exceeds the actual velocity at the equilibrium line if the ice temperature is assumed to be - 5 °C. However, the difference between the balance and actual velocities is not great which, therefore, explains the slow rate at which the glacier is building up towards its next surge. Page 203 CHAPTER 7 THE HYDROLOGY OF BJUVBREEN 7.1 INTRODUCTION 7.1.1 Background Glacier hydrology has received an-increasing amount of attention following the recognition that the presence of water has a profound effect on the motion of ice masses. The behaviour of water within a glacier system must be known in order to understand the major problems in glaciology, which Paterson (1981) identified as the mechanism of basal motion and the causes and mechanics of surging. The presence of water is an important influence on movement whether a glacier is resting on a rigid bedrock bed or on a layer of sedimentary material. On rigid bedrock, water acts as a lubricant between the glacier sole and the bed (Weertman, 1964). Where a glacier is underlain by a layer of sediments, a build-up of water can lead to a lowering of the effective strength of the material (Boulton and Jones, 1979). The effective strength may become sufficiently low to make the sediments much softer than the ice. Under these circumstances the sediment may begin to deform. If the glacier is coupled to this material, the ice will be carried along by the deforming sediments (Jones, 1979). Theories of the causes and mechanics of glacier surges have generally featured the presence of water playing a key role (Chapter 1). Weertman (1969) proposed that the build-up of a layer of subglacial water, to a thickness greater than the size of bed obstacles, would reduce basal friction enough to initiate a surge. The mechanism leading to the accumulation of a water was given further attention by Robin and Weertman (1973 ), who suggested that a strong gradient in basal shear stress would block the flow of water beneath a glacier. The two currently favoured theories of surging, again, both rely on the presence of water (sections 1.3.2 and 1.3.3). The linked cavity model advocated by Kamb (1987) involves the destruction of a tunnel drainage system beneath a glacier which causes water to accumulate at the base in a series of linked cavities. The cavity network is unable to discharge water as efficiently as the tunnel system, meaning that the water which cannot escape is confined under increasingly high pressures. High basal water pressures facilita te easier sliding of the glacier over its bed. The linked cavity model was developed for glaciers resting on rigid beds. A second model of surging that deals explicitly with unconsolidated glacier beds Chapter 7: Hydrology of Bju.vbreen was proposed by Clarke et al. (1984). Their theory also begins with the destruction of the basal hydrological system. However, this mechanism differs from the linked cavity model in that water then accumulates in subglacial sediments. This material is weakened and begins to deform. As the sediment is deforming, the glacier is carried along at rapid velocities. The occurrence of transient motion events, observed on surge-type and non- surge-type glaciers, is thought to be related to the mechanism responsible for triggering surges (Raymond, 1987). Numerous studies (e.g. !ken, 1974; Hodge, 1979; Ileen and Bindschadler, 1986) have correlated these events to the behaviour of the glacier hydrological system. Recent work on Jakobshavn Isbn:e, however, has implied that fast glacier motion may not be exclusively related to the presence of water (Echelmeyer and Harrison, 1990). Thus, there is a need to undertake integrated studies of glacier movement and hydrology on a greater number of glaciers to better understand the processes influencing the motion of ice masses. 7 .1. 2 Fieldwork programme The methodology used to study the hydrology of Bjuvbreen was described in section 4.6. A brief resume is presented here, prior to a discussion of the results. The obvious difficulties involved in accessing the glacier bed meant that proglacial stream discharge was used as the main indicator of water behaviour within Bjuvbreen. A discharge record is only available for the first field season due to equipment failure in 1990. Discharge measurements were made every hour for the duration of the fieldwork at a site approximately 350 m from the glacier terminus (Figure 4.1). In conjunction with the discharge measurements, samples of meltwater were collected at the same location and times for determination of suspended sediment concentrations. The record of sediment concentrations covers both field seasons. During the 1990 field season, the NP hot water drill was used to melt a number of access holes through the glacier to the bed with the intention of measuring water pressure in them. Water pressures were not actually measured because the holes did not connect with the basal drainage system. However, a number of useful results were obtained from this exercise. 7.2 MELTWATER DISCHARGE AND SUSPENDED SEDIMENT DYNAMICS 7. 2 .1 Discharge rating curves A record of the discharge of water passing through the proglacial stream of Bjuvbreen was obtained for the 1989 field season. The record contains hourly Page 205 i I Chapter 7: Hydrology of Bjuvbreen measurements of discharge made between 11 July-24 August 1989 (Julian days 192- 236). Occasional short gaps exist in the data, usually caused by the transducer being unable to take a reading because of a build-up of sediment around the instrument. A number of methods for filling in missing data points exist, such as cubic spline interpolation (Chatfield, 1984). However, interpolation synthesizes missing values on the basis of the observed trend. A synthesized record will not, therefore, indicate extreme events. In this study, extreme events were, arguably, of the greatest interest. The problems associated with data interpolation were summarized by Press et al. (1989), who stated (p. 77) that if the information required to apply an interpolation function accurately was available, !here would be no need for interpolation. Thus, in view of these shortcomings, gaps in the discharge record of Bjuvbreen were not interpolated. Measurements were made with a pressure transducer which recorded river stage. The stage values were converted to discharges using the results of manual stream gauging. Gauging was carried out at a variety of different flow levels, enabling stage- discharge relationships to be established. Three separate rating curves had to be constructed for the field season because of changes resulting from two unusually high discharge events. The first event substantially altered the channel pattern of the stream. During the second unusually high discharge, the large boulder, on which the transducer was fixed, was moved slightly. Although the boulder was displaced only a small distance, the movement resulted in the transducer sitting in a deeper pool of water. Therefore, the stage in relation to the discharge of the channel was different from the pre-flood conditions. The stage-discharge relationships are plotted in Figure 7 .1 . Regression coefficients for all three plots ranged from 0-69 to 0-91. This variation in values indicates the difficulties of measuring discharge in glacial streams. The current meter of the type used can be inaccurate when used in rocky, turbulent streams (Richards, 1982). Proglacial stream channels are susceptible to erosion. The range of r2 values for the Bjuvbreen proglacial rating curves reflects the relatively mobile nature of the channel. Despite frequent cross-section surveys, it is likely that changes occurred in the section between consecutive surveys as a result of bedload transport. The standard errors of the rating curves are, in chronological order, ±26%, ±23% and ±4%. 7. 2. 2 Characteristics of the 1989 discharge hydrograph The rating curves were used to transform the logged record of transducer values into the discharge hydrograph (Figure 7 .2) . The discharge record of the Bjuvbreen stream displays the characteristic diurnal cycle of variations, common to proglacial streams (Stenberg, 1970). For the first half of the field season, the base flow remains Page 206 1.4 1.2 Cl) 0 1.0 Q) E :::, 0.8 ~ Q) 0) 0.6 ~ ..c 0 0.4 Cl) 0 0.2 0.0 0 1.4 1.2 en 0 1.0 Q) E :::, 0.8 ~ Q) 0.6 2' ro ..c 0.4 0 -~ 0 0.2 0.0 0 1.0 0.8 en 0 Q) E 0.6 :::, ~ Q) 2' 0.4 ro ..c 0 -~ 0 0.2 0.0 23 14-30 July 1989 r"2 = 0.69 standard error ±26% • 2 3 4 Stage, cm 30 July-9 August 1989 r"2=0.71 standard error ±23% 5 10 10-24 August 1989 r"2 = 0.91 standard error ±4% 24 25 15 Stage, cm 26 27 Stage, cm • 5 6 7 8 • 20 25 30 28 29 Figure 7.1 Stage-discharge relationships for Bjuvbreen's proglacial stream. Stage was recorded automatically every hour by a pressure transducer. In order to convert the data to a runoff record, the stream discharge was gauged using the conventional mean section method. Three rating curves were constructed because of channel changes resulting from particularly high discharges. The range of correlation coefficients reflects the difficulties of gauging high energy gravel bed stream,sage 207 30 ,] Cl) 0 2 (1) E 1)1 ::, 0 . ,I (1) 1.5 OJ ~ ~ C1l _;::_ 0 Cl) 0 \ ~ '~\A) ~ ~ ~ \ \~ 0.5 ,} 0 ,--r-r,·-,- -,~ I I 19 5 200 205 210 215 220 225 230 235 Juli an day Figure 7.2 Hourly discharge hydrograph for the 1989 field season collected for the meltwater stream draining from Bjuvbreen. The stream runoff displays a characteristically diurnal pattern. In addition, there are several high discharge events occurring throughout the field season. These events were related to meteorological conditions (Figure 7.4), usually precipitation. The gaps in the data were caused by the transducer being buried by sediment. Chapter 7: Hydrology of Bjuvbrcen approximately constant at about 0·5 m3 s-1 . After Ju lian day 220 (8 August) , the minimum daily discharge drops to around 0·25 m3 s-1. There is a slight rising trend towards the end of the measurement period. During the initial period covered by the hydrograph , the regular diurnal cycles have approximately the same amplitude. The range between maximum and minimum daily flows is generally about 0·25 m3 s-1. In glacierized catchments, the maximum daily discharge usually occurs in the afternoon or evening following the peak of incoming solar radiation (Rothlisberger and Lang, 1987). The timing of the peak daily runoff from Bjuvbreen changed during the field season. During the first half of July, the maximum discharge was generally recorded in the early evening (1900-2_000 hours) . As the melt season progressed, the maximum discharge usually occurred in the late afternoon (1600-1800 hours) . From mid-August onwards, however, runoff was highest in the early evening. The phased shift in the timing of diurnal discharge peaks through the melt season is illustrated in Figure 7 .3. Average daily hydro graphs were prepared from the discharge time series for the three periods of the melt season identified above. This procedure is known as stacking (Humphrey et al. , 1986; Fountain, in press) and is expressed statistically as: (7 .1) where Qr is the mean discharge at time t , subscript d is the day and N is the total number of days during the period of interest. In calculating the stacked hydro graphs, days when rainfall occurred were omitted to eliminate the changing times of the peak discharge. The shifts in the timing of the meltwater peaks were related to changes in the snow cover on the surface of Bjuvbreen. Temporary water storage in snow covering most of the upper glacier in early July caused a delay between the maximum daily temperature and the runoff peaks during the early part of the season. As the snow cover receded up-glacier, the delay between maximum temperature and peak discharge was shortened. Meltwater was able to flow quicker across the bare ice surface to points where it drained englac;:ially. The shortened delay may also be related to the evolution of the glacier ' s drainage system. During the preceding winter, the discharge passageways created in the previous summer probably decreased in size considerably owing to the reduced flow of water. Early in the new melt season, the internal drainage network is likely to be composed of a large number of small cavities linked by conduits (Walder, 1986; Kamb, 1987; Rothlisberger and Larig, 1987). This linked cavity network is unable to discharge, efficiently, the increasing amount of meltwater that is produced as the melt season progresses. Thus, there is a delay between the temperature and discharge maxima. The diffuse network of conduits gradually evolves, by viscous Page 209 Chapter 7: Hydrology of Bjuvbreen dissipation, into a series of larger channels that discharge meltwater more efficiently as the ablation season progresses (cf.Burkimsher, 1983; Ileen and Bindschadler, 1986; Kamb, 1987; Seaberg et al., 1988). The delay of the discharge peaks that was observed towards the end of the season was most likely due to snow fall on the upper glacier. The increased distribution of snow had the effect of increasing the amount of water in temporary storage. Similar changes in the timing of the diurnal discharge peak have been observed on streams draining other glaciers, Q) 0) ~ .c u en i:S 0 - ......................... --.-........, ......... --..--. ........ ........,e--T-....-.--.--.-........ --,--, 0 12 24 36 48 Time (hours) Early season Mid season Late season Figure 7 .3 Stacked discharges for the 1989 field season. This figure was compiled from the average discharge for a given hour, calculated using equation 7 .1. Data from storm events were not included in this analysis because of the altered times of the peak discharges. The stacked curves have been normalized to enable comparison. A phase shift in the timing of the discharge peak occurs though the melt season. In the early and late parts of the season, peak runoff occurs between 1900--2000 hours. The timing of this peak is related to the amount of snow cover on the glacier surface. Temporary water storage occurs in the snowpack, thus delaying the time between maximum temperature and peak discharge. The peak discharge occurs earlier in the afternoon during the middle part of the melt season. This earlier runoff peak is due to the rapid passage of meltwater across the bare ice surface. including Gornergletscher, Swiss Alps (Elliston, 1973), Br~ggerbreen, Svalbard (Hagen et al., 1991) and South Cascade Glacier, Washington State (Fountain, in press). The regular diurnal hydrograph for the 1989 field .season is punctuated on occasions by extreme events. These peak discharges are usually related to meteorological conditions (Figure 7.4). The distinct peak on Julian day 198 followed a period of heavy rain. Other periods of high streamflow, on Julian days 209, 215 and 220, were also preceded by rainfall. Precipitation was recorded on two other occasions during the field season, on Julian days 227 and 232. The discharge hydrograph (Figure Page 210 10 15 9 Q) 8 '"O ~ 7 10 E .gi E C 6 Q) c (.) 0 2 5 :~ :::, ~ 4 D. ·u Q) 5 Q) D. 3 D.. E Q) 2 I- 0 185 195 205 215 225 235 Julian Day Figure 7.4 Mean daily temperature (curve) and precipitation (columns) recorded on the terminal moraine of Bjuvbreen (-100 m a.s.1.) during the 1989 field season. Precipitation usually fell as rain, except on Julian day 227 when . snow fell. Page 211 Chapter 7: Hydrology of Bjuvbreen 7 .2) following these events is not as dramatic as following rainfall earlier in the field season. On Julian day 227, the precipitation fell as snow, resulting in a muted runoff response. Late on Julian day 228 the discharge began to increase as melting snow started to pass through the drainage system. Streamflow remained relatively high and did not display a strong diurnal component during the following 24 hours. This was partly due to a general rise in temperatures (Figure 7.4). Approximately 1 mm of rain fell on Julian day 232, although the runoff response to this event is not clear because of an interruption in measurements at this time. The stream runoff record ended on Julian day 236 (24 August). Discharges were noticeably lower towards the end of the monitoring period than they were at the beginning of the field season. Air temperature records from both Bjuvbreen and Sveagruva indicated that temperatures began to drop at the end of August. In the week following the termination of discharge monitoring, temperatures at Svea remained close to O ·c and there were several snowfalls. Assuming a lapse rate of 0·009 ·c m-1 (Simoes, 1990), it is likely that sub-zero temperatures prevailed on the upper region of Bjuvbreen at this time. The low temperatures would have caused a significant reduction in the stream discharge. It is not known when the meltwater ceased to be discharged from Bjuvbreen. Hydrographs from other proglacial streams in Svalbard (Sollid et al., 1991; Pettersen, 1991), however, demonstrate markedly reduced flows from mid- to late-August onwards. Thus, it is assumed that the 1989 discharge time series from Bjuvbreen covers the middle and late stages of the melt season. Runoff first appeared in the proglacial stream on 5 June during the 1990 season. The possibility that extreme events occurred after the termination of the monitoring programme cannot be excluded. Pettersen ( 1991) reported that the peak discharge of the Br0ggerbreen pro glacial stream during 1989 ablation season occurred on 26 September, following heavy rain. 7. 2. 3 Suspended sediment variations 1989 Streams draining glaciers possess a characteristically milky appearance, caused by the transport in suspension of varying quantities of sediment. Suspended sediment is the product of glacial erosion and rock weathering. The turbulent action of water flowing at the base of a glacier releases sediment from storage. Different proglacial streams carry varying amounts of sediment and, within a particular stream, the concentration of suspended sediments fluctuates with time. During the 1989 field season, samples of meltwater were collected hourly between Julian days 193 and 221 at the same site as discharge measurements were made (Figure 4.1 ). Samples were filtered in the field and the residue returned to the laboratory for determination of suspended sediment concentrations (SSC) (section 4.6.3). The record of sediment concentration for the 1989 field season is shown in Page 212 600 Ol 500 -- E C 0 -ro ~ -C Q) (.) C 0 (.) C Q) E -0 Q) U) --- 400 300 · \ 200 -· \1 ) 1~~,~ 100 - 0 195 200 205 210 215 Julian day Figure 7.5a Record of proglacial suspended sediment concentrations for the 1989 field season. Samples were collected hourly at a site on the meltwater stream approximately 350 m from the glacier terminus. Gaps in the data are due to equipment failure. Note the highly variable nature of the sediment concentrations. 600 Ol 500 · E C 0 -ro ~ C Q) (.) C 0 (.) C C 0 <.> c Q) E "O Q) (f) 0 12 24 Time (hours) 36 48 Figure 7.6 Stacked record of suspended sediment concentrations during the 1989 field season. In compiling this figure, the average sediment conc-entration was calculated for each hour. Note that, although the highest sediment concentrations generally occur in the late afternoon, there is often more than one sediment concentration peak in any one day. 7. 2. 4 Suspended sediment variations 1990 Suspended sediments were collected during the 1990 field season at the same site as the previous year. The sampling began on Julian day 164, two days after open channel flow began at the gauging site. Water first emerged from the glacier terminus on Julian day 156 although the proglacial flow was initially beneath the snow surface. Page 214 Chapter 7: Hydrology of Bjuvbreen Sampling ended on Julian day 184, although there was a hiatus between days 175 and 181 because of equipment failure. The hourly time series of suspended sediment concentrations is shown in Figure 7.5b. The pattern of sediment concentrations for the 1990 field season is similar to that recorded in 1989. Apart from the spiked signal, it is noticeable that most concentrations were below 250 mg 1-1, which is similar to the previous season. In addition, there were a number of peaks in the record. The greatest concentration of sediment was 550 mg I-1, recorded on Julian day 168. The variation in daily maximum and minimum concentrations became much larger towards the end of the field season, i.e. between Julian days 181-184. Due to the absence of discharge measurements for the 1990 field season it is not possible to compare the sediment dynamics to the flow regime of the stream. Furthermore, it is not possible to calculate cumulative daily totals of sediment transport. 7. 2. 5 Relationship between discharge and water quality The amount of sediment carried in suspension varies markedly between different proglacial streams. Church and Gilbert (1975; Table 3) sununarised data from a number of studies, demonstrating the wide variety of different concentrations and suggesting an apparent dependence on local geology. Several studies (e.g. Mathews, 1964; 0strem, 1975; Bogen, 1980) have attempted to link the amount of suspended sediment in transport with the discharge at the same time. The relationship between SSC and discharge is usually expressed in terms of a suspended sediment rating curve based on regression of the two variables. It has sometimes been difficult, however, to obtain reliable curves for individual proglacial streams (Gurnell, 1987). These difficulties arise because of the variability in the supply of sediment for water transport. Discharge and SSC can be compared only for the 1989 field season, since no discharge record was obtained in 1990. A power law relationship of the form SSC= aQb, where a and b are best fit parameters (Gurnell, 1987), was used to assess the level of dependence of sediment concentration on discharge. The relationship was applied, using least squares, .to the 1989 SSC and streamflow data, giving the expression: SSC = 84-186Q0·796 (7 .2). There is only a poor degree of association between the discharge of the stream and the amount of sediment that it carries in suspension (Figure. 7.7). The coefficient of correlation (r2) was low, at O· 29, reflected by the wide scatter of residuals from the best fit line. Studies undertaken elsewhere have had varying amounts of success at finding a relationship between discharge and suspended sediment (Gurnell, 1987). Both Mathews (1964) and 0strem (1975) were unable to identify any clear relationship between SSC and proglacial stream discharge, although 0strem Page 215 Chapter 7: Hydrology of Bjuvbreen 600 SSC = 84·1860°.796 ' 0) E X r:: 500 .Q X cii X '- X . X E 400 (1) C) X C: X 0 X C) 300 E X (1) E '6 200 (1) en "O (1) "O 100 C: (1) a. rJJ ::l 0 (/") 0 0.5 1.5 2 2.5 3 Discharge, m3 s-1 Figure 7.7 Power law relationship between discharge and suspended sediment in the proglacial stream during the 1989 field season. There is a considerable scatter of the residuals from the best fit line, indicating that the suspended sediment concentration of the proglacial stream is not closely related to the water discharge. Logarithmic transformation of the data did not improve the relationship. concluded that, in general, years of low total discharge gave less sediment transport than years of high total discharge. Logarithmic transformation of the data did not significantly improve the regression relationship. The correlation coefficient remained moderately low at 0·27. The above regression tests suggest that discharge of Bjuvbreen's terminal stream is not closely related to the amount of sediment that it carries in suspension. The weakness of this relationship is further demonstrated when the hourly records of strearnflow and sediment concentration are plotted together (Figure 7 .8). There appears to be some association between periods of high discharge and increases in suspended sediment concentration. However, the highest discharges did not always result in the greatest sediment concentrations. For this reason, the regression coefficients between discharge and SSC were understandably poor. Nevertheless, when the discharge was lagged behind the suspended sediment record by one hour, and vice versa, there was still no improvement in the regression relationship. During the first four days of the 1989 time series, there were several distinct peaks of sediment concentration, where values exceed ' background' levels. The hydrograph for this same period, however, did not display any unusual events. The most extreme storm event recorded during the 1989 field season took place on Julian Page 216 11 (/) 0 (l) E ::i 0 - (l) 0) '-- CU ..c 0 (/) 0 3 2 .5 2- 1.5- 0. 5 2 0 J u li a n day Figure 7 .8 Combined hourly record of pro glacial stream discharge (bottom) and suspended sediment concentrations (top) for the 1989 field season. Major peaks in suspended sediment concentration are related to increases in discharge, for example on Julian day 197, 202 and 209. However, the largest runoff events do not neccessarily produce the highest sediment concentrations. Very high sediment concentrations are recorded on Julian day 217 although the discharge at the same time is not particularly high. Page 217 600 325 ::::::: 0) E c- 1 1 0 5 o --;;; 225 '-- -C (l) 0 C 0 0 -C (l) E u Q) (.{) Chapter 7: Hydrology of Bjuvbreen day 198. The corresponding SSC record for that time did show an increased level of suspended sediment in the stream (Figure 7 .8). Nevertheless, the increase in sediment quantity was not as impressive as events later in the season. A smaller discharge event occurring on Julian day 202, in contrast, was accompanied by a dramatic increase in stream turbidity. Over the course of the next day, the discharge decreased in a series of progressively smaller peaks. However, during the evening of the day following the storm discharge, there was a second, although slightly smaller, peak in sediment concentration (Figure 7.8). The sediment concentration record was interrupted between Julian days 204- 208 (Figure 7 .5a). When the data resumed, the amount of sediment in suspension showed a rapid increase which can be linked to a small increase in stream discharge. A larger increase in discharge the following day (Julian day 210) was also associated with a sediment pulse. Immediately after this event, the stream discharge quickly returned to 'background' levels although there were two subsequent, but smaller, peaks in suspended sediment concentration on the following two evenings (Figure 7.5a). The sediment record was interrupted again between Julian days 214-218. During this period, there was a noticeable increase in stream runoff, beginning on Julian day 215 and gradually declining thereafter. Discharge was still decreasing when the sediment record resumed, with the highest concentrations of sediments observed during the 1989 field season (Figure 7.8). This exceptional event was followed by a series of lesser peaks until the data stopped being collected. Discharge data are missing for Julian days 219 and 220. Observations made at the gauging site during that period indicated, however, that discharges were not exceptional. The above results indicate that if suspended sediment concentrations are related to stream discharge, the relationship is difficult to define statistically over the course of a field season. This observation lends support to 0strem's (1975) suggestion, that sediment rating curves can be constructed for short time intervals within a field season. Collins (1979) constructed individual rating curves for the rising and falling limbs of the diurnal hydrograph recorded at Gomergletscher. Similar curves were presented for Decade Glacier river:, Baffin Island, by Church and Gilbert (1975). Hammer and Smith (1983) estimated suspended sediment rating curves for the early and late ablation season hydrographs of the Hilda Glacier proglacial stream, Alberta. Using data recorded at the terminus of Bjuvbreen (Figure 7 .8), one sediment-discharge curve could be established for the period between Julian days 192-200, another for the interval covering Julian days 201-204, and so on. In effect, a separate rating curve has to be established each time there is a significant peak in sediment concentration. However, individual short-period rating curves would be of little value. 0strem (1975) Page 218 Chapter 7: Hydrology of Bjuvbreen pointed out that a rating curve constructed using a short interval of data during one year would not be applicable for the same site at the same time the following year. Difficulties in establishing a reliable rating curve for the Bjuvbreen proglacial time series could be caused by the presence of a more subtle relationship between discharge and sediment concentration than is observed on the hourly record. An alternative potential relationship could exist with high streamflow events being associated indirectly with increased sediment concentrations. If this were the case, increased runoff may be an indication of greater flushing potential at the glacier bed. Sediment concentrations do not reach their maximum at the time of the highest discharge because of dilution effects. On the falling limb of the runoff, however, suspended sediments form an increasing part of the load, probably not because lower discharges are more effective at transporting sediment but because dilution effects operate during periods of high streamflow (Collins, 1979). Hysteresis effects The competition between the supply of sediment and the flushing potential of subglacial waters means that the concentration of suspended sediment in a proglacial stream does not remain constant. The changing pattern of the glacier drainage system will exploit and then exhaust new sediment supplies. These changes will manifest themselves in the form of hysteresis effects in the terminal discharge. Two examples of hysteresis in the SSC record from Bjuvbreen are discussed below. The first case represents 'normal' flow conditions, and the second example indicates processes occurring during flood events. The hysteresis rating loops are illustrated in Figure 7 .9. Figure 7 .9a illustrates the hysteresis dynamics occurring during a period of 'normal' flow conditions, in this case on Julian day 197 in 1989. The main trend in this figure is for the loop to follow an anti-clockwise direction, although there is a small involution leading to a brief clockwise rotation. Figure 7.9b shows the hysteresis loop occurring during a period of storm discharge (Julian day 209 in 1989). The most noticeable difference between storm and normal f19w conditions is the greater sediment variability during the storm event. In Figure 7.9b, the rating loop begins in a clockwise direction, but switches to an anti-clockwise trend following the first extremely high concentration of suspended sediments. Proglacial streams normally exhibit clockwise hysteresis effects (Rainwater and Guy, 1961; Collins, 1979; Richards, 1982; Humphrey et al., 1986). Clockwise hysteresis indicates that discharges on the rising limb of a hydrograph carry a greater amount of sediment in suspension than equivalent discharges on the falling limb (Richards, 1982). This situation arises as increasing discharges gain access to new sediment supplies, which become exhausted before the discharge begins to decline. Page 219 300 A Ju lian day 197 (/) C .Q ~ c 0 200 C 0 0 c E -0 100 (/) -0 -0 C a. (/) :l Cf) 0 0.5 0.6 0.7 0.8 0.9 Discharge 400 B Julian day 209 C .Q cii ..... c Q) 300 0 C 0 0 c Q) 200 E -0 Q) (/) -0 Q) 100 -0 C Q) a. (/) :l Cf) 0 0 2 3 Discharge Figure 7 .9 Hysteresis rating loops for a period of normal discharge (A) and storm discharge (B). Both rating loops illustrate the interaction between discharge and suspended sediment concentrations. The principal difference between normal and storm conditions is the greater variability in the sediment concentrations during the flood event. Both hysteresis loops display a dominant anti-clockwise trend, indicating that the highest sediment concentrations lag the peak discharge. Page 220 Chapter 7: Hydrology of Bjuvbreen The situation observed on Bjuvbreen, where peak diurnal discharges precede maximum daily sediment concentrations, is not unique to proglacial streams. Fountain (in press) reported the occurrence of anti-clockwise hysteresis loops in the terminal stream draining South Cascade Glacier. Fountain suggested that the peak concentration would lag behind the peak discharge if there was a large number of pools along the channel. The pools trap suspended sediment thereby delaying the arrival of the concentration peak at the gauging site. A large stream runs supraglacially down the stagnant portion of B juvbreen from a portal at the base of the bulge. This stream is incised several metres into the ice and follows a tortuous course. Sloughs and slackwater pools were observed on many of the meander bends. Suspended sediments entrained during high discharges could become trapped in these locations and released gradually after the peak runoff recedes. High SS Cs would, thus, be recorded several hours after high discharges. Sediment exhaustion may occur over the course of a melt season. To determine if exhaustion took place in Bjuvbreen's proglacial stream, the running totals of discharge and suspended sediment transport were compared. This comparison was made only for the 1989 data series. Figure 7 .10 shows the running totals of discharge and suspended sediment transport. The period for which data were available covered the interval between Julian days 193-220 (12 July-8 August); after Julian day 220, the record of suspended sediment concentrations was interrupted too frequently to be of any use. The total amount of water discharged in this period was -1 ·707 x 1Q9 1. During the same interval, -101 ·435 t of sediment was transported in suspension. This value was obtained by multiplying the measured SSCs by the total hourly discharge. There were two significant gaps in the suspended sediment record, between Julian days 204-208 and 214-217. The amount of sediment transported during these two intervals was interpolated using a cubic spline function . A curve was fitted between adjacent points on either side of the gaps by using a spline function to synthesize data based on the before and after trend. An inherent assumption of spline fitting is that no extreme events took place during the synthesized interval. Since the actual dynamics of suspended sediment .concentrations during the two unsampled intervals is not known, the absence of extreme events cannot be ruled out. Figure 7 .10 demonstrates that the pattern of discharge and suspended sediment transport are not identical over the course of the period ofinvestigation. During the initial stages, the proportion of total streamflow discharged exceeded the proportion of suspended sediment that had been transported. By midnight on Julian day 197, 10·7% of the total discharge had transported 8-5% of the total suspended sediments. As the Page 221 I , 100 0 ..0 ro ~ 3 00 (J) Q) I 2 00 Ice surfac e Hydraulic g ra de lin e :- : . . .. · . · .. · Glacier bed ·.· .·:.·.· .. -·.· ------ _,..,,....,,....,,... ------- ,..- --- ----- ------- --- -- ------ __ ,.. ,..-- ,.,,,. ------ ·. · · . . ... · .·· . . . . _.,. .,.---- --- . .... . .__ ______ 1 ~ 7 - 0 - 0 ----1-s~o~o----19~0-0 ____ 2 ~o ~0-0 ____ 2_1 ~0_0 ____ 2,...2~0,......o-,-----2-3 ...... o-o----2-4L..o-o-- Distance from Station 1, m Figure 7.12 Theoretical hydraulic grade line (dashed line) with respect to the surface of Bjuvbreen and the glacier bed, based on a numerical evaluation of equation 7.3 (Rothlisberger, 1972). The hydraulic grade line represents the height a column of water would reach in a piezometric tube. It should be noted that the longitudinal distribution of calculated water pressures represents the steady-state situation of a single subglacial conduit carrying all the water. The hydraulic grade li ne does not, therefore, indicate actual water pressures beneath Bjuvbreen at any particular time. E - T""" C 0 - 500 ro 40 0 (/') (l) > 0 .D ct1 ~ 30 0 0) (l) I 20 0 Ice surface Hydraulic grade line Glacier bed --- -- .,.,... ,....,. ------ ----______ , .,,....,,,,.. ---- - -- --- --- ---- .... • ... . ... : . · .. ·.· .· .... : . '---------=- 1 ...... 7 - 0 - 0 ----1-s_,o'-o----1_,9'-o-0 ____ 2_0__._0_0 ___ 2_1..__o_o ___ ----=-2~2-=-o-=-o----2-3..1...0_0 ____ 2_4..__o_o __ D ista nce from Station 1, m Figure 7.12 Theoretical hydraulic grade line (dashed line) with respect to the surface of Bjuvbreen and the glacier bed, based on a numerical evaluation of equation 7.3 (Rothlisberger, 1972). The hydraulic grade line represents the height a column of water would reach in a piezometric tube. It should be noted that the longitudinal distribution of calculated water pressures represents the steady-state situation of a single subglacial conduit carrying all the water. The hydraulic grade line does not, therefore, indicate actual water pressures beneath Bjuvbreen at any particular time. Chapter 7: Hydrology of Bjuvbreen insufficient time to adjust to the new volume of liquid. If the increased volume of water is sustained, the drainage channel will eventually adjust to a new steady-state and water pressures will drop. Despite the limiting assumptions involved, this analysis provides an approximation of the longitudinal variation of water pressure beneath Bjuvbreen as it LS influenced by glacier geometry and water discharge. The results are used in the following section in a discussion of the influence of basal hydrology on the observed long-term (seasonal) velocities of the glacier. 7.4 RELATIONSHIP BETWEEN GLACIER MOTION AND HYDROLOGY 7. 4 .1 Introduction The data collected on Bjuvbreen and discussed in the preceding sections of this thesis allow a comparison to be made between the motion of the glacier and changes in the hydrological regime of the glacier. The aim of this comparison is to determine the influence of water on the dynamics of Bjuvbreen. The surface velocities of temperate surge-type, and non-surge-type, glaciers have been demonstrated to respond to fluctuations in the behaviour of their glacier's drainage system (e.g. Variegated Glacier: Kamb and Engelhardt, 1987; Findelengletscher: Ileen and Bindschadler, 1986). Similar correlations have been observed on a sub-polar surge-type glacier (Trapridge Glacier: Clarke and Gerin, 1989). The behaviour of Bjuvbreen is examined at two scales. The first part of this section investigates the relationship between seasonal velocity and calculated water pressures. This is followed by an analysis of the short-term variations in surface velocity and how well they correlate with the available hydrological data. 7. 4. 2 Steady-state water pressures and seasonal velocities Seasonal surface velocities represent the horizontal motion of a stake averaged over the course of a field season (section 6.2.2). The variation of seasonal velocities with longitudinal disfance along the centreline of Bjuvbreen is shown in Figure 6.2. The influence of theoretical steady-state water pressures on the observed surf ace velocities is discussed in this section. Absolute magnitudes of water pressure are not the main focus of interest in this analysis. Instead, the excess pressure remaining from the subtraction of water pressure from ice overburden pressure is more important in influencing glacier flow. This remainder is known as the effective normal stress, Neff, i.e. (7.4). Page 234 I 1 Chapter 7: Hydrology of Bjuvbreen Effective normal stresses were calculated for points on the long profile of Bjuvbreen using basal normal stress data from section 5.3.5. The profile of Neff is illustrated in Figure 7.13 . There is a considerable amount of spatial variation in the values of Neff· The most significant features of the distribution of Neff are the minimum values recorded approximately 350 m up-glacier from the crest of the bulge followed by a large increase to peak values another -175 m further up-glacier. Neff then decreases at the furthest up-glacier points. Included in Figure 7.13 is the profile of seasonal basal sliding velocities compiled using centreline targets in 1989. It can be seen that there is not a simple one- to-one relationship between sliding v~locity and effective normal stress on Bjuvbreen. The highest sliding velocities should occur where Neff is at a minimum. This theoretical prediction is not supported by data from B juvbreen. At the point of lowest effective normal stress, the calculated sliding velocity does not increase. Instead, the greatest amount of basal motion is found to occur slightly up-glacier where there is an accompanying increase in Neff· However, the lowest sliding velocity calculated during ai z 0.25 0.2 0.15 0.1 0.05 --Neff · · · · · Basal velocity 2000 2100 2200 2300 2400 2500 Distance from Station 1, m 0.12 0.1 u 0.08 E >, ...... 0.06 ·c3 0 Q) > 0.04 C'Cl C.{) C'Cl O:l 0.02 0 Figure 7 .13 Profile of effective normal stress with longitudinal distance on Bjuvbreen. Effective normal stresses were calculated using basal normal stress and calculated water pressure data. Seasonal basal velocities for 1989 are also shown. Theoretically, basal motion should be greatest where the effective normal stress is at a minimum. On Bjuvbreen, the highest basal velocity occurs slightly up-glacier from the effective stress · minimum. However, where the effective normal stress is highest, minimum basal basal motion rates were recorded. Page 235 Chapter 7: Hydrology of Bjuvbreen the 1989 season was observed to coincide with the region of highest Neff· The up- glacier decrease in Neff from its maximum is accompanied by a slight increase in basal sliding, which then drops off again as Neff increases at the furthest up-glacier location. These data indicate that there is a crude relationship between the amount of basal motion and the effective normal stress at a given point on Bjuvbreen but that the correlation between the two quantities is not clearly defined'. A discussion of possible factors affecting this relationship is provided later. The growth of basal cavities has been considered to be important to the mechanism of sliding (Lliboutry, 1968; Iken, 1981). A theoretical ratio, known as the bed separation index (S), was defined by Bindschadler (1983) who found that there was a good agreement between the ratio and the sliding velocity of three glaciers. The bed separation index is given by the expression: S = 'CB/Neff (7 .5). Although calculated values of S are not proportional to a specified amount of bed separation, as the ratio increases so does the probability of cavity growth on the down- glacier side of subglacial roughness features. The distribution of S along the central 0.8 140 0.7 120 X Q.) 0.6 "O 100 (/) C (/) Q.) 0.5 ,._ C ..... 0 80 (/) ro ,._ 0.4 CU ,._ Q.) CU 60 ..c n. (/) Q.) 0.3 CU (/) - - - - - Bed separation index "O (/) 40 CU Q.) 0.2 a:i a:i --Basal shear stress 0.1 20 0 0 2000 2100 2200 2300 2400 2500 Distance from Station 1, m Figure 7.14 Profile of bed separation index with distance ori Bjuvbreen. The bed separation index was calculated with equation 7 .3, according to Bindschadler (1983). The profile of basal shear stress (section 5.3.4) is also shown. The pattern of bed separation compares better with basal . shear stress than effective normal stress (Figure 7 .13). The only noticeable difference between shear stress and bed separation is that the minimum shear stress occurs - 50 m up-glacier from the bed separation minimum. Page 236 11 Chapter 7: Hydrology of Bjuvbreen axis of Bjuvbreen was computed using data on Neff derived above and shear stresses calculated in section 5.3.4 with the Kamb--Echelmeyer method. Figure 7.14 illustrates this distribution. There is a considerable spatial variation in the computed values of the bed separation index, although this variation does not match the variation of effective normal stresses (Figure 7 .13). The similarity is better between S and basal shear stress in 1990, which is included in Figure 7.14 for comparison. The only major difference between the two profiles is that the minimum value of shear stress occurs roughly 50 m down-glacier from the minimum value of bed separation. The low point in the profile of Sis displaced slightly up-glacier due to the larger gradient in Neff than 'X'B over the same region. The second decrease in S, near _2330 m, is more pronounced than the slight decrease in basal shear stress at this point. This difference is also due to stronger gradients in effective normal stresses. Bindschadler (1983) computed bed separation ratios for Variegated and Columbia glaciers and Ice Stream B, and compared calculated values with basal sliding rates beneath each glacier. This analysis demonstrated that, within each glacier, the spatial differences in the amount of basal motion could be explained by variations in S. Differences in the values of S were, in tum, controlled by the distribution of basal water pressures. Raymond and Harrison (1988) extended Bindschadler's analysis of Variegated Glacier data by computing profiles of the bed separation index for the ten years prior to the 1982- 83 surge. They discovered that the region of glacier which had the greatest basal velocity increase through this period was also the area with the largest rise in S. This was the region of the glacier where the surge was initiated (Bindschadler, 1983; Raymond and Harrison, 1988). In contrast to the work elsewhere, the profile of bed separation ratios does not closely match the spatial variation of basal velocities beneath Bjuvbreen (Figure 7.14). Theoretically, the highest basal velocities should occur in the areas with the greatest bed separation ratios (Bindschadler, 1983). However, on Bjuvbreen the opposite situation is found. The greatest amount of seasonal basal motion was calculated to occur at approximately 2040 m. At this same point, the minimum value of S was computed. When the bed separation index increases at -2275 m, the calculated basal velocity decreased. The marked rise in Sat roughly 2375 m is accompanied by a small increase in the amount of basal motion, but the continuing trend of increasing Sup-glacier is not followed by the record of basal velocity. The above discussion reveals that the basal motion of Bjuvbreen is not clearly related to the amount of bed separation or effective normal stress. There are a number of reasons which might account for this poor correlation. Firstly, seasonal basal velocities occurring beneath Bjuvbreen were not actually measured, but were calculated from observed surface velocities using glaciological theory (section 6.5.3). It is Page 237 Chapter 7: Hydrology of Bjuvbreen conceivable that these calculated values do not accurately reflect the true amount of sliding at the base. These inaccuracies could result, for example, if the flow law constants used in the calculation of sliding velocities were not suitable for Bjuvbreen. However, recalculating the basal velocities using A = 1 ·7 x lQ-15 s-1 kPa-3 and 5·2 x lQ-16 s-1 kPa-3, equivalent to ice temperatures of - 5 °C and - 10 °C respectively (Paterson, 1981) (Appendix B), does not substantially alter the relationship with S. The ideal method of determining basal sliding rates is to compare summer surface velocities with winter velocities (e.g. Raymond and Harrison, 1988). Motion during winter can be assumed to be occurring by internal deformation alone. The excess velocities observed during summer can, therefore, be taken to represent the amount of basal motion. Such a comparison could not be made for the Bjuvbreen data and, thus, the potential remains that the calculated basal velocities are not an accurate representation of actual rates. A comparison was made between seasonal surface velocities and S to test whether the relationship between motion and bed separation was improved. A better relationship might have confirmed that calculated basal velocities were inaccurate. However, no substantial improvement in the correlation between velocity and bed separation was obtained. A second difficulty which could obscure the relationship between basal sliding and effective normal stress and the ratio of bed separation, concerns the calculation of sub glacial water pressures. A number of simplifications were made in order to perform the calculations (Rothlisberger, 1972). Firstly, it was assumed that discharge remained constant with time. This assumption is clearly violated, as the discharge record (Figure 7.2) indicates. Secondly, it was assumed that all the subglacial drainage was concentrated in one, straight, cylindrical conduit. In the absence of flow tracing studies, it is not possible to confirm this assumption for Bjuvbreen. Nevertheless, there is some evidence to suggest that the assumption is not unreasonable. The rapid response of the discharge hydrograph to increased water input to the glacier, for example, following periods of rainfall, is consistent wi th the flow of water through a glacier with a well developed drainage system. In addi tion, the inability to penetrate part of the active drainage system during drilling, suggests that the flow of water is concentrated in one, or at most a few, subglacial conduits. Rothlisberger's assumption has been challenged by other studies. For example, actual water pressures will be higher than calculated values if the,drainage takes place in a number of smaller conduits (Bindschadler, 1983), or in a single conduit with a meandering path (I.ken and Bindschadler, 1986) or a broad, low cross-section (Hooke et al., 1990). Given these difficulties, it would be unwise to place too much emphasis on the lack of a relationship between the bed separation index and basal velocities (cf. Raymond and Harrison, 1988). Paterson (1986) has questioned the wisdom of Page 238 1 JJ Chapter 7: Hydrology of Bjuvbreen applying Rothlisberger's (1972) model to an analysis of glacier motion. The model calculates steady-state water pressures and, therefore, predicts that basal motion will be greater in winter than in summer. However, the poor correlation between S and calculated basal velocities could be real and not simply the product of difficulties with the calculation of each parameter. The absence of a relationship could indicate that subglacial cavitation is not an important process in influencing the basal velocity of Bjuvbreen. If this is the case, then some other process may be responsible for motion at the glacier bed. Alternatively, the poor comparison between the two values might suggest that basal water pressures play a minor role in the observed velocity of Bjuvbreen during the melt season. 7. 4. 3 The influence of glacier hydrology on the short-term velocity behaviour of Bjuvbreen Judging from the poor correlation between the amount of seasonal basal motion and the water pressure-derived bed separation index, sub glacial water pressures would appear to have only a small influence on the velocity of Bjuvbreen. This influence is examined further by looking at the relationship between short-term variations in surface velocities observed during the two field seasons and the hydrological behaviour monitored at the same time. A discharge hydrograph is available only for the 1989 season. Therefore, velocity data from 1990 have be to be related to other variables, such as suspended sediment concentration and meteorological conditions. The potential for such a comparison was demonstrated by Humphrey et al. (1986), who found that the record of turbidity for the stream draining the major part of Variegated Glacier was able to provide a useful insight into the subglacial hydraulic processes operating during mini- surge events. Short-term variations in the horizontal and vertical velocities observed at the surface of Bjuvbreen were described in sections 6.2.3 and 6.3.2, respectively. Figure 7 .15 illustrates a time series of data depicting the horizontal and vertical velocities, proglacial stream characteristics and meteorological conditions recorded during the 1989 field season. At the beginning of the velocity observations in early to mid-July (Julian days 205-212), the motion of nearly all the targets is above their seasonal average. Weather conditions at this time consisted of rising mean daily temperatures and -1 mm of rainfall on Julian day 208. The discharge hydrograph exhibited a peak that day which probably represented the through passage of the storm runoff. Discharge remained high on Julian days 209 and 210 for several reasons. At this stage of the ablation season much of the glacier still maintained an extensive snow cover. Therefore, some of the rainfall was stored in the snowpack and released slowly over Page 239 C 0 +50 -~ ~ +25 "O 0 Q) > -25 r-i o -50 I 10 Q) 8 "O "' -~ C Q) 0 Q) ::, "' Q) Cl. E Q) I- " u Q) 6 4 2 0 2.5 2 E 1.5 ::, 0 Q) Ol "' .c 0 " 0 0.5 1 0 _ Horiz. ve l ----- Vert ical vel. 2 0 2 0 Julian day 2 0 2 0 +10 +5 0 -5 E E Q) > "' 0 Q) -10 > 14 12 10 8 6 4 2 E E c 0 "' Cl. 0 Q) n. 0 Ol 600 E c 0 400 ~ C Q) 0 200 g 0 2 0 0 C Q) E "O Q) u, Figure 7 .15 Combined record of short term horizontal and vertical velocity, stream discharge and suspended sediment concentration, and meteorological conditions for the 1989 field season. The average velocities of fourteen targets near the centreline on the upper glacier are shown. There is reasonable degree of association between . velocities and the other variables, with colder weather leading to a reduction in motion. Periods of rainfall usually cause an increase in glacier movement, although this is not always the case (e.g -Julian day 216). Page 240 Chapter 7: Hydrology of Bjuvbreen the days following the storm. In addition to the releasing liquid water, the rain probably enhanced the melting of the snow contributing to a larger runoff. Furthermore, the weather remained relatively warm, promoting ablation. Several pulses of turbid water were also recorded around this time. These pulses were likely to be the result of the increased sediment entrainment and transport capabilities of the swollen stream. However, extra sediment could have been made available from previously inaccessible supplies. Humphrey et al. (1986) found that following mini-surges of Variegated Glacier, the proglacial discharge became heavily charged with sediment. This change in water quality was suggested to be caused by the flushing of debris from newly opened cavities at the glacier bed due to high water pressures. If this was the case for the event in discussion on Bjuvbreen, then tne prolonged period of increased suspended sediment concentrations suggests that subglacial water had continued access to large sections of the bed for several days. Temporary trapping of sediment in pools and slackwater regions may have had some effect on the delayed passage of increased SSCs. The temporal resolution of the velocity measurements is such that the exact timing of the period of increased surface motion cannot be determined to an accuracy of more than a few days. Thus, although the velocity appears to remain high for several days, much of the motion may have occurred over a much shorter period but was significant enough to increase the average velocity during the measurement interval. Discharge quickly decreased to background levels on Julian day 210, although surface motion was above average and stream turbidity was increased. Thus, the data do not conclusively show that the cause of the increased glacier motion was greater water pressures affecting much of the bed. On Julian day 216, observed velocities fell below the seasonal average at nearly all locations on the glacier. This decrease is unusual because there had been -7 mm of rainfall during the previous two days and temperatures were relatively warm (Figure 7 .15). The stream discharge had increased in response to the input from rainfall. During the following day the hydrograph remained quite high and only returned slowly to background conditions. The possible reason for the decrease in velocity was a reduction in subglacial water pressures. Although as discharge increases and steady- state water pressures decrease (Rothlisberger, 1972), there will be an initial increase in water pressure due to the greater amount of liquid entering the system. Increased water pressures are caused by a larger volume of water flowing in a 9Tainage network which is adjusted for a smaller steady-state discharge. As time progresses, a new steady-state evolves and water pressures decrease (Bindschadler, 1983). Such a situation was probably in operation on Bjuvbreen at this time. The rainfall of Julian days 214 and 215 caused an increase in subglacial water pressures which contributed to the enhanced surface velocities. However, as the input of water remained high, melting of the Page 241 I' I I I · 111 I Chapter 7: Hydrology of Bjuvbreen drainage network walls by viscous dissipation enlarged the passageways and led to a decrease in pressure. The resulting increase in the effective normal stress would have caused the glacier to slow down. There was a further speed-up in surface velocities between days 222-225. The cause of this increase is unclear. The discharge at this time was relatively small, although there had been 3 mm of rain on Julian day 220. Temperatures had also dropped and ice formation was observed on surface water pools, up -glacier from the bulge. The amount of water entering the glacier's drainage system was, thus, likely to have been diminished. Storage of water at the bed may have led to a decrease in the effective normal stress, but this does not seem likely since no large anomalous discharges were observed during the foflowing ten days. If water was stored for the whole of that period, the surface velocities should also have remained high. As Figure 7.15 illustrates, velocities decreased noticeably on Julian day 225. Unfortunately the water sampler had ceased functioning by this stage and therefore the added information which could be obtained from the turbidity record is not available. The next interval of velocity measurements was characterised by a marked reduction in the amount of observed surface motion. During this interval there was a fall of snow at all altitudes on the glacier. Temperatures of -3 °C and -4 °C were recorded at the moraine (110 m a.s.l.) and Sveagruva (sea level), respectively . Assuming a lapse rate of 0·009 °C m-1 (Simoes, 1990), freezing temperatures were likely to have prevailed on the upper glacier at this time. Thus, the amount of water produced by ablation would have been much reduced. During the final velocity measurement interval (Julian days 231-234) the hydrograph displayed a gradual trend towards increasing discharges. This was partly the result of 1 mm ofrain on Julian day 232 and the melting of snow which had fallen during the previous week. This increase was enhanced by a slight amelioration in temperatures. Despite the potential increase in the amount of water reaching the glacier bed, however, there was no observed speed- up of the ice surf ace velocities over this period. Below average velocities were recorded at nearly all targets at this time. No discharge hydrograph is available for the 1990 field season. Nevertheless, a rough estimate of the quanti ty of liquid likely to be in the glacier drainage system can be obtained from the meteorological records. The history of suspended sediment variations is illustrated in Figure 7.16 for comparison with weather conditi~ms and fluctuations in glacier surface velocity. Between Julian days 165-172, surface motions were below average at all targets on the glacier. During the first half of this interval, suspended sediment concentrations were at normal levels. Based on temperature measurements made on the moraine and in Sveagruva, the temperature on the upper part of the glacier were likely to have been close to freezing. There was light drizzle on Julian day 165. Page 242 +5 0 +10 C _Horiz vel . 0 "' +2 5 _____ _ Verti ca l ve l . +5 E > E Q) u 0 -----------, 0 .; ' ' > 'If. ' L- -- -- ---- ---------- ----- --- __ _ J '" .; -25 -5 u > (I) N -50 -10 > 0 6 I 8 5 (I) u "' 6 g , 4 C: (I) E u 4 E (I) 3 ::, r: 0 "' "' (I) 0. 2 ~ 2 '?-E (I) u ..... (I) ;;_ 0 - 2 0 60 0 500 0) E r: 0 400 "' C (I) u 300 C II 0 u C: j "' 20 0 I E u f (I) !1t.l 1 ~J I Cl) 100 I~~ I~ i,' 1, ~·~ I ' I r1 I r 1 5 1 0 1 5 180 1 5 Ju l ia n day Figure 7.16 Combined record of short term horizontal and vertical velocity, suspended sediment concentration, and meteorological conditions for _the 1990 field season. No discharge record was obtained in 1990. The average velocities of six targets near the centreline on the upper glacier are shown. The major motion event of the field season took place between Julian days 172-17 6. This increase in velocities was related to a small amount of rainfall and rising temperatures. The weather conditions probably enhanced melting of the snowpack and led to an increase in the amount of liquid water reaching the glacier bed. Page243 11 I I I Chapter 7: Hydrology of Bjuvbreen On day 168 there was a dramatic increase in the turbidity of the proglacial stream. The highest sediment concentrations (up to 550 mg J-1) of the field season were recorded at this time. This increase in turbidity is not related to an increase in surface motion of Bjuvbreen, in contrast to observations of Humphrey et al. (1986) on Variegated Glacier. This suggests that the cause of the increased suspended sediment concentrations was not a change in the structure of the drainage system exposing new areas of the glacier bed to the action of flowing water. Had this been the case, the increased lubrication and potentially higher water pressures would have led to a speed- up of the glacier. Vertical velocities did not record a substantial amount of uplift during this period. The more likely explanation_ for the increased sediment concentrations was a rise in stream discharge, and hence competence, brought about by enhanced melting of snow cover in milder temperatures. At all targets there was a dramatic increase horizontal velocity, accompanied by vertical uplift, between Julian days 172-176. This was the major motion event of the field season. The uplift observed during this period (-0-03 m) was much larger than that recorded during other intervals on Bjuvbreen, although it did not match the uplift of 0·5 m reported by Iken et al. (1983) to occur on Unteraargletscher at the start of the melt season. The fastest velocities were recorded during the interval from Julian days 172- 174, although horizontal motion was generally above average, also, during the following two days . There was light rainfall (1 mm) on Julian day 172 and temperatures were above freezing. The rain probably enhanced the melting of the snowpack, increasing the input of water to the glacier bed. The water sampler was not in operation at this time, thus, it is not known whether the turbidity increased. In all cases, vertical velocities were upward during the period between Julian days 172-17 4 and downward during the following two days . The uplift is considered to be related to high subglacial water pressures which could have enhanced basal cavitation or dilation of sub glacial sediments. During the remainder of the survey period, below average horizontal velocities were recorded at all targets. However, during the interval of Julian days 176-181, uplift was observed at nearly all the markers . The weather conditions at this time had deteriorated with heavy snow on days 177 and 178 and sub-freezing temperatures (Figure 7 .16). Discharge in the pro glacial stream was drastically reduced and only a very small runoff was observed to emerge from the glacier terminus. Normal flows were resumed in the afternoon of Julian day 180 shortly after temperatures rose above freezing. The diminution of the streamflow was caused by the temporary cessation of melting on the glacier. A potential cause of the observed uplift could have been an increase in water storage at the glacier bed, resulting in higher water pressures. This, however, is considered unlikely, given the fact that horizontal velocities decreased Page 244 Chapter 7: Hydrology of Bjuvbreen during this period. Furthermore, the amount of free liquid in storage at the bed would not have been substantial. Much of the melted snow would probably have refrozen either in the snowpack or in englacial veins and capillaries, especially near the cold surface layer of the glacier. An alternative explanation for the anomalous amount of uplift is vertical straining of the ice column. An attempt was made to use Hooke et al.' s (1983a, 1989a) method of calculating uplift due t.o the formation of basal cavities to determine whether cavitation or vertical straining was responsible for the anomalous uplift during the Julian day period 17 6-181. Their method requires the vertically averaged strain rate to be known. This was not measured in the field, although it can be derived, theoretically, from the surface s_train rate data assuming ice incompressibility. However, the period of interest did not coincide with the interval over which the strain rates were originally calculated (Table 6.2). Data were available with which strain rates could be calculated for the interval between Julian days 176-181, although when this was done the errors (section 4.5.6) were found to constitute a large proportion (>50%) of the final results. Since the analysis was attempting to explain a small amount of uplift (-0·02 m), the uncertainties involved in computing the vertical strain rate would have produced an inconclusive result. Therefore, it is not possible to provide an explanation for this anomalous uplift. 7. 4. 4 Discussion The above discussion demonstrated that the short-term velocity behaviour observed on Bjuvbreen shows some dependence on the nature of the drainage system. In particular, speed-up events do appear to be correlated with relatively extreme changes in glacier hydrology, although the temporal resolution of the velocity observations is not fine enough to resolve the relative timing of increases in speed and discharge. The usual situation is for glacier motion to increase first, followed by increased discharges of water and increased turbidity (e.g. Humphrey et al., 1986). This is because water is stored or trapped at the bed, promoting higher water pressures and leading to faster sliding or sediment deformation. Increases in the amount of water discharged at the terminus, and changes in its quality, signal the reduction of water pressures and a slow down in glacier velocity. The presence of this pattern cannot be confirmed from the Bjuvbreen data. The character of the transient velocity pattern on Bjuvbreen is relatively undramatic, in that motion variations of the stakes never exceeded 100% of their average seasonal values. This observation is despite the fact that a number of reasonably significant hydrological events occurred during the periods of measurement. The most suitable explanation for this weak dependence of the glacier motion on hydrological behaviour is that the drainage system has a limited spatial extent beneath Page 245 11 11 Chapter 7: Hydrology of Bjuvbreen Bjuvbreen (section 7 .3.2) . A large proportion of the discharge measured in the proglacial stream, flows supraglacially and only enters the englacial or basal drainage network very close to the bulge. Given that the total discharge of water from Bjuvbreen is relatively low (-1 m3 s-1 ), the amount of water present sub glacially is likely to be very small. Therefore, unless increases in water reaching the bed are very substantial indeed, or significant storage occurs, then basa1 hydrology will exert little influence on the observed surface velocity of the glacier. 7.5 DISCUSSION AND SUMMARY - 7. 5 .1 The hydrology of Bjuvbreen compared with other glaciers The hydrology of Bjuvbreen and its influence on the movement of the glacier displays a number of the characteristics commonly observed on other glaciers. There are also several aspects of its hydrology which differ from other examples. The differences in the structure of the drainage network impact upon the future dynamic behaviour of the glacier, especially with regard to surging. The differences observed on Bjuvbreen are probably common to other high Arctic glaciers and could imply that the triggering of surges on glaciers in these environments is slightly different from that envisaged to occur on temperate glaciers. Certain features of Bjuvbreen's hydrology are observed on most other glaciers, irrespective of their geographical location and environment. Discharge of water from the glacier follows a clearly defined seasonal and diurnal rhythm which has been described by many authors (e.g. Fahnestock, 1963; Stenborg, 1970; Iken, 1974). The response of the proglacial stream to meteorological conditions was also observed. Changes in the inferred amount of water within the glacier were correlated with increases in the glacier motion. This type of behaviour has been observed on temperate (e.g. Iken and Bindschadler, 1986; Kamb and Engelhardt, 1987) and sub-polar glaciers (e.g. Iken, 1974). However, variations in the velocity of Bjuvbreen were not great. The pattern of drainage within the glacier was inferred from hot water drilling experiments. The p_rincipal finding to arise from this work was that the flow of water beneath the glacier is limited in its spatial extent. The drilling was undertaken during the early part of the melt season. Several authors have suggested, experimentally and theoretically (e.g. Hodge, 1972; Walder, 1986; Kamb; 1987), that at this time, subglacial drainage networks are characterised by a distributed series of cavities and small conduits. Kamb (1987) theoretically predicted that this spatially extensive network evolves into a system of one, or a few, large tunnels as the melt season progresses. Given that drilling on Bjuvbreen was carried out during the early part of the melt season, it is perhaps surprising that none of the holes drilled connected with a Page 246 II Chapter 7: Hydrology of Bjuvbreen hydraulically-linked cavity or conduit. Based on work by Engelhardt et al. (1978), an estimate was made of the area of hydraulically-inactive bed in the vicinity of drilling on Bjuvbreen. Further evidence for the limited subglacial drainage network was the observation that a large supraglacial stream flowed down Bjuvbreen from the equilibrium line but did not drain into the glacier until within -75 m of the crest of the bulge. The discharge of this supraglacial stream was estimated to carry slightly less than half the runoff measured in the pro glacial stream. Large supraglacial streams have been observed on other quiescent surge-type glaciers in Svalbard (e.g. Liest01 et al., 1980; Hagen et al., 1991). These streams form because of the cold surface layers and the absence of crevasses on some glaciers in the archipelago. The most likely factor which would account for a spatially-limited internal drainage network on Bjuvbreen is the almost complete absence of crevasses in the ablation area of the glacier. Crevasses are only present in large numbers in the accumulation zone, where surface melting is probably restricted. In the ablation zone, there is unlikely to be a significant amount of meltwater percolation into the glacier (e.g. Raymond and Harrison, 197 5) because the upper layers are believed to cold (section 5.5.3) and, therefore, impermeable (Hooke, 1988). The current paradigm for theories of the surge mechanism involves water being trapped at the glacier bed in sufficient quantities to initiate rapid motion, either by sliding (Kamb, 1987) or bed deformation (Clarke et al. , 1984). Evidence gathered on Bjuvbreen suggests that its internal drainage system does not extend over a very large area of the bed and, also, that hydrological changes which occur do not have a particularly dramatic effect on the motion of the glacier. If this situation continues, it is difficult to imagine how a surge on Bjuvbreen could be initiated by either Clarke et al. 's or Kamb's mechanism. This is because it does not seem likely that enough water will reach the bed for either mechanism to come into operation. If this were the case, another mechanism would have to be sought in order to explain a future surge. However, as the quiescent phase progresses, Bjuvbreen will probably become increasingly active an~ a form of feedback reaction could occur. Changes in the geometry of the glacier will increase the amount of internal deformation that can occur. As the glacier becomes more active, crevasses will form as a result of the increasing strains taking place. Once a large number of crevasses have opened below the equilibrium line, more meltwater will be able to penetrate to the base of the glacier. This should then present more potential for the development of distributed hydrological network. This type of drainage system is better able to transmit the effects of transient increases in subglacial water pressure to larger areas of the glacier than the presently restricted drainage network. One of the potential effects of the increased coupling between glacier dynamics and hydrology would be an increase in ice velocity. The Page 247 Chapter 7: Hydrology of Bjuvbreen enhanced velocities would promote further crevasse formation and, hence, improve meltwater access to the bed. Echelmeyer and Harrison (1990) advanced this type of feedback reaction further, by postulating that the heat released by the enhanced motion of the glacier produces an increased amount of water. This water would contribute to enhancing the accelerating ice velocities. At this stage of the surge cycle on Bjuvbreen, however, it is too early to predict if the above mechanism will be appropriate to explain an hydrological surge trigger. It is interesting to hypothesize that the hydrological processes occurring on Bjuvbreen during quiescence are common to other high Arctic surge-type glaciers. Sub- polar glaciers are relatively inactive compared to temperate glaciers. This reduced dynamic behaviour, combined with the impermeability to water caused by their thermal regime, serves to restrict the amount of liquid that can reach their beds. Therefore, surge-type glaciers in these environments will remain relatively inactive during the majority of their quiescent phase. The feedback reaction described above may be a common mechanism occurring during the quiescent phase on high Arctic surge-type glaciers. 7 .5.2 Summary The main points to emerge from this chapter are identified below: • The discharge of Bjuvbreen's terminal stream displays a characteristic diurnal rhythm and also responds to transient increases in water input to the glacier. • Meltwater leaving the glacier is charged with varying amounts of suspended sediment, although concentrations appear to be lower than these observed in other proglacial streams. The reduced sediment concentrations are probably a reflection of a spatially-limited drainage network and the small amount of subglacial erosion currently occurring. The amount of sediment carried in suspension by the stream is not directly related to the simultaneous water discharge. Hysteresis i_s observed in the sediment concentration record, with the peak discharge often preceding the peak turbidity. This lag is probably caused by the temporary trapping of sediments in pools and slackwater regions of the drainage network. • Hot water drilling indicates that a sizeable area of the centre of Bjuvbreen has a hydraulically-inactive drainage network. • The absence of a widespread internal drainage system is a reflection of a cold surface layer and the lack of crevasses in the ablation zone which prevent the penetration of water to the glacier bed. Page 248 Chapter 7: Hydrology of Bjuvbreen • Seasonal basal velocities did not show a clear dependence on effective normal stress or an index of bed separation. Any relationship could have been obscured by inaccuracies in the calculation of basal velocities or water pressures. • Transient increases in the velocity of Bjuvbreen are correlated with inferred changes in the hydrology of the glacier. However, not every dramatic hydrological event is accompanied by an increase in ice motion. • The reduced influence of water on the seasonal and short-term velocities of Bjuvbreen is thought to be related to the restricted spatial extent of the glacier drainage system. Page 249 CHAPTER 8 SUMMARY AND FURTHER WORK 8.1 INTRODUCTION The foregoing chapters of this thesis have presented the results of investigations on surge-type glaciers in Svalbard. These investigations have taken two approaches to study the phenomenon of surging in the archipelago. One approach involved the statistical analysis of a large sample of normal and surge-type glaciers which identified some of the potential factors controlling the spatial distribution of surge behaviour in Svalbard. The second approach analysed field data to describe the dynamics and hydrology of a small, quiescent surge-type glacier in central Spitsbergen. Both of these approaches are innovative in terms of research in the archipelago. Several of the findings may have implications for other glacierized regions of the world. In this final chapter, the significant findings of the thesis are summarized. The results of the statistical analysis are presented first, followed by a summary of the work undertaken on Bjuvbreen. The work presented in this thesis is by no means a complete analysis of surge-type glaciers in Svalbard. Therefore, the final section of this chapter outlines some potential areas for future research on the topic and suggests some methods which this research might use. 8.2 SUMMARY OF SIGNIFICANT RESULTS 8. 2 .1 The significance of surge-type behaviour on Svalbard glaciers The widespread occurrence of surge-type glaciers in Svalbard has long been recognized by glaciologists. In addition to the 69 glaciers in the archipelago that have been observed to surge this century Liest!Zll (in press) there is strong morphological evidence that a substantial number of other glaciers are surge-type. However, until the work presented in Chapter 3 was undertaken, there was no accurate estimate of the number or percentage of glaciers in the archipelago that are of surge-type. Figures that have been quoted in the literature probably over-estimate the proportion of surge-type glaciers in Svalbard. For example, Hagen and Liest!Zll (1990) suggested that up to 90% of glaciers in islands were surge-type, although the method they used to compile this estimate was not made clear. Chapter 8: Summary and further work As part of the investigation into the characteristics of surge-type glaciers and the potential controls on the geographical distribution of this behaviour (Chapter 3), a systematic analysis of every glacier in selected areas of Spitsbergen was undertaken. On the basis of evidence derived mainly from aerial photographs, the dynamic behaviour of each glacier in the sample was classified. The sample contained 615 glaciers, of which 224 were classified to be of surge-type. If this sample is representative of conditions elsewhere in the archipelago, then approximately 35% of glaciers in Svalbard are surge-type. This figure is considerably lower than Hagen and Liest0l's estimate, but is likely to be more accurate given that it was based on a systematic analysis of a large sample of glaciers. Despite the substantial difference between the estimate of Hagen and Liest01 (1990) and that presented in Chapter 3 of this thesis, surge-type behaviour is still considered to be very significant in Svalbard. Surge-type glaciers are known to be clustered in certain glacierized areas of the world (Raymond, 1987). The only known estimate of the proportion of surge-type glaciers in a particular geographic region was prepared by Clarke et al. (1986) for the Yukon Territory. Their estimate of 6-4% was substantially lower than that obtained for the Svalbard data set. This difference is likely to imply that the factors controlling the surge process occur more frequently in Svalbard than in the Yukon. Within Svalbard there is a geographical variation in the concentration of surge-type glaciers, indicating that the processes controlling the distribution on the global scale also operate on a regional scale. 8. 2. 2 Factors influencing the occurrence of surge-type glaciers in Svalbard · The distinct distribution of surge-type glaciers on both global and regional scales indicates that special factors are required for surges to occur. Chapter 3 presented a statistical analysis of the Svalbard glacier data set with the aim of identifying factors common to surge-type glaciers in Svalbard. Identifying such factors is potentially useful for clarifying the mechanism of glacier surges. Several glacier attributes were compared with the probability that a glacier is surge-type and were found to be unrelated to surging. These characteristics included glacier slope, elevation, orientation, curvature and the presence of tributaries. The observation that no particular slope was related to surge-type gl~ciers does not support the suggestion by Kamb (1987) that low slopes favour surging. Furthermore, the width-slope product defined by Fowler (1989) did not distinguish between normal and surge-type gl_aciers in Svalbard. Glacier length was related to surging, with long glaciers having a higher probability of being surge-type than short glaciers. A similar relationship was found for glaciers in the Yukon analyzed by Clarke et al. (1986) and Page 251 Chapter 8: Summary and further work Clarke (1991). This observation is difficult to explain, both in the Yukon and Svalbard sample populations, although the increased likelihood of identifying a surge-type glacier that is long probably had some influence on the result. Two characteristics in particular demonstrated strong relationships with the probability of surge-type behaviour. These characteristics were concerned with the lithology of the sub glacial substrate and the bulk thermal regime of the glacier. Glaciers which were underlain by sedimentary rocks had a much increased probability of being surge-type compared with those resting on metamorphic and igneous rocks. However, the surge probabilities associated with these three lithologies were unable to predict the observed geographical distribution of surge-type glaciers. This inability implied that either geology was not the only factor influencing surging or that a three-fold classification of lithologies was not detailed enough. Sample sizes for populations of glaciers resting on igneous and metamorphic rocks were too small to be statistically analyzed further. When the glaciers underlain by sedimentary rocks were re-classified according to the individual rock type, limestone and Old Red Sandstone showed the best correlations with the probability of surging. However, the more detailed classification of geology was still unable to explain the spatial distribution of surge-type glaciers fully. One possible reason for the connection between surging and sedimentary rocks is that these lithologies are better suited to the formation of sedimentary, and potentially deformable, glacier beds. Clarke et al. (1984) proposed a surge mechanism based on the deformation of unlithified subglacial sediments. Glaciers with a two-layered thermal structure had a much greater probability of being surge-type compared to glaciers with entirely cold or temperate thermal regimes. The sample on which this statistical analysis is based, however, was smaller than the original dataset. The meaning of this result was not clear. The boundary between the cold upper layers and the warmer bottom layers of a two-layered glacier is believed to be marked by an englacial water channel (Bamber, 1987a). One possible reason that two-layered glaciers have a high probability of being surge-type is that as the quiescent phase progresses, ice overburden pressures will increase and may drive water out of the englacial conduit. Water that is driven out will be forced through the permeable lower layers of the glacier and will accumulate at the bed. These glaciers may not possess drainage systems which are efficient enough to discharge this increase in basal water. Therefore, subglacial water pressures could increase, enhancing basal sliding or sediment deformation. One result that was consistently observed in the statistical analysis was that no single charaGteristic or factor was able to adequately explain the geographical distribution of surge-type glaciers in Svalbard. This observation implies that a combination of suitable conditions is required for surges to occur. Page 252 Chapter 8: Summary and further work 8. 2. 3 The dynamics of Bjuvbreen Two field seasons were spent studying the geometry, motion and hydrology of Bjuvbreen in central Spitsbergen. This glacier was classified as being of surge-type, based on the analysis of a series of aerial photographs. The principle features leading to this classification were the increased ice thicknesses and presence of crevasses near the terminus of the glacier in 1936, and the subsequent stagnation of much of the lower glacier and evolution of a large wave-like bulge, clearly seen in 1977. Radio echo sounding and drilling have confirmed that the formation of the bulge is not influenced by the subglacial topography. Changes in the geometry of Bjuvbreen have been determined mainly through the study of the glacier's long profile. Profiles surveyed in the field in 1989 and 1990 were supplemented by an additional field-surveyed profile from 1986 (Hagen, personal communication) and a profile derived from a 1977 photogrammetric map produced by Fox (1989). These profiles show a consistent decrease in ice thickness down-glacier from the bulge and an increase in thickness up-glacier. The bulge, therefore, marks the boundary between active and stagnant regions of the glacier, and can be considered to represent the dynamic boundary line (DBL). Since 1977, the base of the bulge has propagated down-glacier at an average rate of 4 m a-1 . Increases in ice thicknesses on the up-glacier portion of Bjuvbreen in the 13 years since 1977 have been relatively small, compared with thickness changes observed on surge-type glaciers in other environments. The main cause of this difference is the slow rate at which mass is added to glaciers in Svalbard. This low accumulation rate, furthermore, means that the quiescent phase of Bjuvbreen is much longer than for surge-type glaciers in areas of higher accumulation. A simple series of calculations were made to estimate the period required for the next surge of Bjuvbreen to be initiated. The results of the these calculations, which assumed that the glacier had surged sometime around 1936, indicated that another 32- 55 years from 1990 are required for Bjuvbreen to reach a geometry at which it can surge. Thus, the return period of surges of this glacier is between about 90 and 130 years. Data from the 1ong profile surveys and radio echo sounding of ice thicknesses allowed the computation of basal shear stresses using the longitudinal averaging theory of Kamb and Echelmeyer (1986). Shear stresses decrease rapidly down-glacier from the bulge. The highest shear stresses were calculated for points located -400 m up- glacier from the bulge as a result of the combination of greatest thicknesses of ice and relatively steep surface slopes. In a zone midway between these furthest up-glacier points and the bulge, there is an area of reduced basal shear stresses. These lower Page 253 I' I JI Chapter 8: Summary and further work values are a response to decreased ice thicknesses and a reduced surface slope. Shear stresses increase again towards the bulge as a result of increasing surface slopes. The history of shear stress changes was determined using the series of long profile data. Shear stresses in a zone -400 m up-glacier from the bulge have shown a steady increase since 1977. The pattern of shear stress change near the bulge is more complex. The points at which shear stresses are calcµlated are fixed in space on each long profile. Therefore, the changes in shear stress through time of points near the bulge are due to the down-glacier propagation of the bulge. Measurements of the distribution of temperature within Bjuvbreen were not undertaken in the field. The thermal regirpe of the glacier was, therefore, estimated using a moving column model. The model could only be applied to the accumulation area of the glacier. The results suggested that the base of Bjuvbreen up-glacier from the bulge is at the pressure melting point. However, this result was found to be very sensitive to the surface boundary condition, so that the prescribed surface temperature had to be only 0·02 °C colder than the assumed value for the base of glacier to become frozen. Ice down-glacier from the bulge is so thin that it is most likely to be frozen to its bed. The sub-polar thermal regime of Bjuvbreen is similar to the thermal regime of the surge-type Trapridge Glacier (Jarvis and Clarke, 1975) and is also consistent with the thermal regimes of other glaciers in Svalbard. The velocity of Bjuvbreen varies both spatially and temporally. The highest seasonally-averaged velocities (-0· 12 m d-1) occur in an area 250-350 m up-glacier from the crest of the bulge. At the base of the bulge, the seasonal velocities are much lower, and further down-glacier there is barely any movement. Seasonal velocities for comparable targets on the glacier surface were, generally, slightly higher during the 1989 field season than in 1990. This difference was small, however, and most likely caused by an increased amount of basal motion during the 1989 season. Fieldwork in 1989 took place during the ablation season whereas work in 1990 was undertaken before and during the melt season. Transient fluctuations in the surface velocity of Bjuvbreen were recorded over short intervals during both field seasons. Increases in surface velocity usually occurred simultaneously at points on the upper portion of the glacier, and were often accompanied by vertical uplift. A general feature of the record of short-term horizontal velocities was that the transient variations at any marker never exceeded 100% of the marker's seasonally-averaged velocity. Dramatic accelerations and mini-surges, which have been observed on other glaciers, were not recorded on Bjuvbreen. Strain rates at the surface of Bjuvbreen were compressive at nearly all measured locations on the glacier. This is because the measurements were made within 500 m of the crest of the bulge. Flow in this part of the glacier is compressive because ice is Page254 Chapter 8: Summary and further work moving towards the bulge faster than the bulge is propagating down-glacier. The pattern of measured strain rates corresponds with the distribution of crevasses on the glacier. A comparison of the distribution of crevasses in 1977 and 1990 showed that crevasses were more widespread in 1990. This change in distribution indicated that extending flow was occurring over a greater area of the glacier. This evolution of strain rates suggests that the glacier is becoming more active in its upper reaches as the quiescent phase progresses. The hydrology of Bjuvbreen was investigated during the two field seasons although a discharge record was obtained only during the 1989 season. The record of runoff in the proglacial stream displayed a diurnal cycle characteristic of proglacial streams. This pattern was supplemented at irregular intervals by extreme discharge events. These increases in streamflow were related to weather conditions. An hourly record of suspended sediment concentrations was obtained for the proglacial stream during both field seasons. The amount of sediment carried in suspension by the stream displayed variations typical of proglacial environments. Throughout both seasons, there were several peaks in stream turbidity. However, the relationship between discharge and suspended sediment concentrations was not particularly good. Peaks in turbidity were usually related to high runoff events although the highest discharges did not always produce the highest sediment concentrations. This relationship was probably the result of the longer term influence of high discharges in establishing new subglacial sediment sources. The concentrations of suspended sediment in Bjuvbreen's proglacial stream were substantially lower than concentrations reported for other proglacial streams. There are two likely reasons for this difference. The first is that much of the sediment eroded during the last surge of the glacier has now been depleted. The present comparative inactivity of the glacier and its low velocities probably mean that the production of new sediment is occurring at a slower rate than sediment is being removed from the system in suspension. The second potential reason for the low sediment concentrations is that much of the drainage system on Bjuvbreen does not flow sub glacially. If the amount of drainage occurring at the bed is limited, then there will be less opportunities for sediment to be entrained by water. An attempt was made to link the motion of Bjuvbreen with the geometrical and hydrological characteristics of the glacier. Seasonal velocities along the central flow line of the glacier were compared with profiles of basal shear stress, effective normal stress and Bindschadler's (1983) index of bed separation. The latter two variables required the computation of the longitudinal variation of water pressures beneath the glacier. Seasonal velocities showed only a moderate degree of correlation with basal shear Page255 Chapter 8: Summary and further work stress. One possible reason for this was that the large changes in surface slope led to inaccuracies in the calculation of shear stresses. Seasonal basal velocities were also compared with effective normal pressure, Neff, and a ratio of bed separation, S, both of which depend upon subglacial water pressures. There is a crude correlation between basal velocity and Neff, but an unconvincing relationship between velocity and S. i:>otential explanations for these poor correlations are difficulties in the calculation of shear stresses and water pressures. In addition, the calculated basal velocities may not accurately represent the actual amount of motion at the base. However, if the relationships between velocity and effective stress and bed separation are accurate2 the implication is that steady-state water pressures do not exert a significant amount of influence on the seasonal motion of Bjuvbreen. The short-term surface velocities observed on Bjuvbreen were compared to the characteristics of the hydrological system. Increases in glacier velocity occurred simultaneously with inferred increases in the amount of water within the internal drainage system. However, several significant events were observed in the stream discharge and sediment concentration records. These events did not have a noticeable effect on the motion of the glacier. It was suggested that the apparently limited influence of hydrology on the dynamics of Bjuvbreen was related to a restricted spatial distribution of the active internal drainage network. If the relative amount of water at the glacier bed was small, and the water was concentrated in one, or a few, reasonably large channels the drainage system would be unable to transmit the effects of water pressure increases to large areas of the glacier. Some evidence exists to support the hypothesis of a limited basal drainage network. The relatively small concentrations of sediment carried in suspension by the proglacial stream can be taken as an indicator of a short transport path of water at the glacier bed. Six holes were drilled to the base in the central area of the upper glacier, although no hydraulically-active connections were achieved. A comparison of this observation with the results of previous drilling experiments (Engelhardt et al., 1978; Hodge, 1979) suggested that the active drainage channels within the glacier were a considerable distance apart. Two probable factors explained the absence of a well developed network of drainage channels within Bjuvbreen. Firstly, the glacier is moving at relatively slow velocities. In addition, flow is generally compressive in the area on the upper glacier which lies below the equilibrium line. This dynamic condition has prevented the formation of large number of crevasses in the ablation zone. Secondly, the temperature of the surface layer of Bjuvbreen is believed to be several degrees below the melting point. The combination of the above factors prevents the penetration of water to depth Page 256 i!il Chapter 8: Summary and further work over large portions of the glacier. Since water cannot readily drain into the glacier, a substantial proportion of surface meltwater drains supraglacially. A large supraglacial meltwater stream is believed to form in the same location each melt season. This stream drains englaciaUy only a short distance from the crest of the bulge. These factors are intimately related to the glacier thermal regime. The presence of similar hydrological systems has been reported for other glaciers in Svalbard (Liest01 et al., 1980; Hagen et al., 1991). It was suggested that the drainage system of Bjuvbreen will evolve as the quiescent phase progresses. This evolution would occur as a form of feedback reaction. Changes in the glacier geometry will cause the ice to become more active. 1be increase in dynamic activity will promote the formation nf crevasses in the ablation zone, thus allowing a greater amount of water to reach wider areas of the glacier bed. Increased glacier motion will be further enhanced by the development of the subglacial drainage system, leading to the formation of more crevasses. This type of feedback mechanism is probably common on other sub-polar surge-type glaciers in Svalbard and similar high Arctic regions. 8.3 SUGGESTIONS FOR FURTHER WORK The results of investigations presented in this thesis have provided new information concerning surge-type glaciers in Svalbard. However, the interpretations of the data discussed in the foregoing chapters are considered as hypotheses which are open to falsification. These hypotheses provide a framework for suggesting future directions of research on surge-type glaciers in Svalbard. Several of these suggestions are outlined briefly below. The statistical analyses discussed in Chapter 3 identified several factors which appeared to be associated with surge-type glaciers. The analysis was carried out using approximately 30% of the glaciers in the archipelago. There is scope to expand the study to include all the ice masses in Svalbard and also to test some of the surging- related factors in sample populations taken from other geographical regions. The Svalbard database can also be used to test the significance of additional factors which might be related to surging. For example, Wilbur (1988) found that surge-type glaciers in northwestern North America were associated with certain hypsometric profiles. Several factors stood out as being particularly well related to the occurrence of surging in the Svalbard sample population. However, no single characteristic adequately explained the_ observed geographical distribution of surge-type glaciers in the sample region. This result suggested that surging is controlled either by an unidentified factor, or a combination of factors. Factor analysis, or similar statistical techniques, can be Page 257 Chapter 8: Summary and further work used to determine which combination of characteristics is most closely related with surging. For example, Clarke et al. (1986) speculated that glacier length and substrate lithology, when analysed in combination, might explain the geographical distribution of surge-type glaciers. This suggestion was made on the basis that glacier length may be a proxy for ice thickness and, hence, overburden pressure. Ice overburden pressure will influence the mechanical behaviour of subglacial sediments and will, therefore, be an important factor if deforming bed theories of glacier surges are accurate. Several aspects of the glaciology of Bjuvbreen deserve further attention. Continued monitoring of the glacier as its quiescent phase progresses would be particularly useful. A long-term study of a high Arctic surge-type glacier would provide a useful counterpoint to existing studies carried out at lower latitudes. Future studies can build on the baseline data collected for this study, concerning for example geometry, velocity, strain, stress and hydrology. Particular aspects of the glaciology of Bjuvbreen which should be addressed are the causes of the variations in velocity and the nature of the hydrological system. The amount of basal motion which occurs has not been measured. Furthermore, it is not known if basal motion is caused by sliding over a rigid bed or the deformation of sub glacial sediments. The techniques required to answer these problems have been used elsewhere, e.g. Kamb and Engelhardt (1987), Hooke et al. (1987), Blake and Clarke (1991) and Blake et al. (in press). Additional work on the hydrology of Bjuvbreen is warranted in order to decipher the relationship between water, ice motion and the thermal regime of the glacier. Further hot water drilling may result in a connected hole being obtained. Direct access to the internal drainage system would allow the pattern of water storage within the glacier to be investigated. Dye tracing would be an appropriate method to employ in such a study. Clearly, the observational database of surge-type glaciers needs to be expanded. In comparison to glaciers in northwestern North America and the Pamirs, surge-type glaciers in high Arctic regions have received little detailed field attention. Field observations from a variety of environments will enable similarities and differences in the dynamics of surge-type glaciers to be identified. Until these observations are made, it will not be possible to unequivocally accept Paterson 's (1986) assertion that a single mechanism is capable-of explaining all surges. Page 258 REFERENCES Abramowitz, M. and Stegun, I.A., 1965. Handbook of mathematical functions. Dover, New York. Aellen, M., 1985. Evidence for uplift due to water pressure at the glacier bed at Grosser Aletschgletscher (abstract). International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September 1985, foterlaken, Switzerland. Mitteilungen der Versuchsanstaltfur Wasserbau, Hydrologie und Glaziologie, Nr. 90, p. 23. Ahlmann, H.W:son, 1933. Scientific results of the Swedish-Norwegian Arctic Expedition in the summer of 1931. Part VIII. Glaciology. Geografiska Annaler, v. 15, p. 161-216 and 261-295. Ahlmann, H.W:son, 1935. The stratification of the snow and firn on Isachsen's Plateau. Geografiska Annaler, v. 17, p. 29--42. Alley, R.B., 1989. Water-pressure coupling of sliding and bed deformation: I. Water system. Journal of Glaciology, v. 35, p. 108-118. Alley, R.B., Blankenship, D.D., Bentley, C.R. and Rooney, S.T., 1987a. Till beneath Ice Stream B. 3. Till deformation: evidence and implications. Journal of Geophysical Research, v.92, p. 8921- 8929. Alley, R.B., Blankenship, D.D., Rooney, S.T. and Bentley, C.R., 1987b. Till beneath Ice Stream B. 4. A coupled ice-till flow model. Journal of Geophysical Research, v. 92, p. 8931-8940. Anandakrishnan, S. and Bentley, C.R., 1989. Seismic activity beneath Ice Stream C and a comparison with Ice Stream B (abstract) . EOS, v. 70, p. 1081. Andreason, J.-0., 1985. Apparent short-term glacier velocity variations. Journal of Glaciology, v. 31, p. 49-53. Bamber, J.L., 1987a. Internal reflecting horizons in Spitsbergen glaciers. Annals of Glaciology, v. 9, p. 5-10. Bamber, J.L., 1987b. Radio echo sounding studies of Svalbard glaciers. Unpublished PhD Thesis , University of Cambridge, Cambridge. Bamber, J.L and Dowdeswell, J.A., 1990. Remote sensing studies of Kvit0yj0kulen, an ice-cap on Kvit0ya, north-east Svalbard. Journal of Glaciology, v. 36, p. 75-82. Baranowski, S., 1977. The subpolar glaciers of Spitsbergen seen against the climate of this region. Act a Universitatis Wratislaviensis, No. 410, 111 pp. Baranowski, S. and Karlen, W., 1976. Remnants of Viking age tundra in Spitsbergen and Northern Scandinavia. Geografiska Anna/er, v. 58A, p. 10-20. Beecroft, I., 1981. Variations, over a 24 h period, in suspended sediment concentration and size distribution in meltwater from the Tsidjiore Nouve glacier, Arolla, Valais, Switzerland. Wessex Geographer,v. 16,p. 12-18. Bentley, C.R., 1987. Antarctic ice streams: a review. Journal of Geophysical Research, v. 92, p. 8843-8859. Bezinge, A., 1987. Glacial meltwater streams, hydrology and sediment transport: the case of the Grande Dixence hydroelectricity scheme. In: A.G. Gumell and M.J. Clark (editors) Glacio-fluvial sediment transport: an Alpine perspective, p. 473--498. John Wiley and Sons, Chichester. Bindschadler, R., 1982. A numerical model of temperate glacier flow applied to the quiescent phase of a surge-type glacier. Journal of Glaciology, v. 28, p. 239-265 References Bindschadler, R., 1983. The importance of pressurised subglacial water in separation and sliding at the glacier bed. Journal of Glaciology, v. 29, p. 3-19. Bindschadler, R., 1984. Jakobshavn Glacier drainage basin: a balance assessment. Journal of Geophysical Research, v. 89, p. 2066-2072. Bindschadler, R. and Rasmussen, L.A., 1982. Finite difference model predictions of the drastic retreat of Columbia Glacier, Alaska. United States Geological Survey Professional Paper, No. 1258-D. Bindschadler, R., Harrison, W.D., Raymond, C.F. and Crosson, R., 1977. Geometry and dynamics of a surge-type glacier. Journal of Glaciology, v. 18, p. 181- 194. Bindschadler, R., Harrison, W.D. and Raymond, C.F., unpublished. Variegated Glacier studies. Unpublished Report Series, Geophysics Program, University of Washington, Seattle. Birkenmajer, K., 1981. The geology of Svalbard, the Western Part of the Barents Sea, and the Continental Margin of Scandinavia. In: A.E.M. Nairn, M. Churkin and F.G. Stehli (editors) The Ocean Basins and Margins, v. 5: the Arctic Ocean, p. 265- 329. Plenum Press, New York. Blake, E.W. and Clarke, G.K.C., 1991. Subglacial water and sediment samplers. Journal of Glaciology, v. 37, p. 188-190. Blake, E.W., Clarke, G.K.C. and Gerin, M.C., in press . Tools for examining subglacial bed deformation. Journal of Glaciology. Blake, W., Jr., 1962. Geomorphology and glacial geology in Nordaustlandet, Spitsbergen. Unpublished PhD Thesis, The Ohio State University, Columbus. Blankenship, D.D., Bentley, C.R., Rooney, S.T. and Alley, R.B., 1986. Seismic measurements reveal a saturated porous layer beneath an active Antarctic ice stream. Nature, v. 332, p. 54-57. Blankenship, D.D., Bentley, C.C., Rooney, S.T. and Alley, R.B., 1987. Till beneath Ice Stream B. 1. Properties derived from seismic travel times. Journal of Geophysical Research, v. 92, p. 8903- 8912. Bogen, J., 1980. The hysteresis effect of sediment transport systems. Norsk Geografisk Tidsskrift, v. 34, p. 45- 54. Bogen, J., 1991. Erosion and sediment transport in Svalbard. In: Y. Gjessing, J.O. Hagen, K.A . Hassel and B. Wold (editors) Arctic hydrology: present and future tasks. Hydrology of Svalbard, p. 147-158. Norwegian National Committee for Hydrology Report No. 23. Bogorodsky, V.V., Bentley, C.R. and Gudmandsen, P.E., 1985. Radioglaciology. D. Reidel, Dortrecht. 254 pp. Boulton, G.S. and Jones, A.S., 1979. Stability of temperate ice caps and ice sheets resting on beds of deformable sediments. Journal of Glaciology, v. 24, p. 29-43. British Standards Institution, 1964. Measurement of liquid flow in open channels: velocity-area methods. BS 3680, part 3. Brown, R.N.R., 1911. British work in Spitsbergen: some historical notes. Scottish Geographical Magazine, v. 27, p. 180-187. Brown, N.E., Hallet, B. and Booth, D.B. , 1987. Rapid soft bed sliding of the Puget Glacial Lobe. Journal of Geophysical Research, v. 92, p. 8985-8998. Brugman, M.M., 1986. Water flow at the base of surging glacier. Unpublished PhD Thesis (Part 2), California Institute of Technology, Pasadena. Budd, W.F., 1968. The longitudinal velocity profile of large ice masses. International Association of Scientific Hydrology Publication, No. 79, p. 58-75. Page 260 [I'' References Budd, W.F., 1975 . A first simple model of periodically self-surging glaciers. Journal of Glaciology, v. 14, p. 3-21. Budd, W.F. and Radok, U., 1971. Glaciers and other large ice masses. Reports on Progress in Physics, V. 34, p. 1- 70. Burkimsher, M., 1983. Investigations of glacier hydrological systems using dye tracer techniques: observations at Pasterzengletscher, Austria. Journal of Glaciology, v. 29, p. 403-416. Chatfield, C., 1984. The analysis of time series: an introduction. 3rd edition. Chapman and Hall, London. 286pp. Church, M. and Gilbert, R., 1975. Proglacial fluvial and lacustrine environments. In: A.V. Jopling and B.C. McDonald (editors) Glaciofluvial and glaciolacustrine sedimentation, p. 22- 100. Society of Economic Palaeontologists and Mineralogists, Tulsa, Oklahoma, Special Publication 23. Clarke, G.K.C., 1976. Thermal regulation of glacier surging. Journal of Glaciology, v. 16, p. 231- 250. Clarke, G.K.C., 1986. Trapridge Glacier studies 1969- 1985; inferences concerning the surge mechanism (abstract). International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September, 1985, Interlaken, Switzerland. Mitteilungen der Versuchsanstaltfur Wasserbau, Hydrologie und Glaziologie, Nr. 90, p. 33- 34. Clarke, G.K.C., 1987a. Fast glacier flow: ice streams, surging and tidewater glaciers. Journal of Geophysical Research, v. 92, p. 8853- 8842. Clarke, G.K.C., 1987b. Subglacial till: a physical framework for its properties and processes. Journa l of Geophysical Research, v. 92, p. 9023-9036. Clarke, G.K.C., 1991. Length, width and slope influences on glacier surging. Journal of Glaciology, v. 37, p. 236-246. Clarke, G.K.C., unpublished. Trapridge Glacier studies: an overview. Paper presented at the Northwest Glaciologists Meeting, Vancouver, 1-2 December 1989. Clarke, G.K.C. and Goodman, R.H., 1975. Radio echo soundings and ice temperature measurements in a surge-type glacier. Journal of Glaciology, v. 14, p. 71-78. Clarke, G.K.C. and Gerin, M., 1989. Presurge fluctuations in water pressure and velocity, Trapridge Glacier, Yukon Territory. EOS, v. 70, p. 1084. Clarke, G.K.C. and Blake, E.W., 1991. Geometric and thermal evolution of a surge-type glacier in its quiescent state: Trapridge Glacier, Yukon Territory, Canada, 1969-89. Journal of Glaciology, v. 37, p. 158- 169. Clarke, G.K.C., Nitsan, U. and Paterson, W.S.B., 1977. Strain heating and creep instability in glaciers and ice sheets. Reviews of Geophysics and Space Physics, v. 15, p. 235-247. Clarke, G.K.C., Collins, S.G. and Thompson, D.E., 1984. Flow, thermal structure and subglacial conditions of a surge-type glacier. Canadian Journal of Earth Sciences, v. 21, p. 232-240. Clarke, G.K.C., Schmok, J.P., Ommanney, C.S .L. and Collins, S.G., 1986. Characteristics of surge- type glaciers. Journal of Geophysical Research, v. 91, p. 7165-7180. Clarke, T.S., 1991. Glacier dynamics in the Susitna River basin, Alaska, U.S.A .. Journal of Glaciology, v. 37, p. 97- 106. Clayton, L., Teller, J.T. and Attig, J.W., 1985. Surging of the southwestem part of the Laurentide Ice Sheet. Boreas, v. 14, p. 235-241. Page 261 I 111 References Collins, D.N., 1979. Sediment concentration in meltwaters as an indicator of erosion processes. Journal of Glaciology, v. 23, p. 247-252. Collins, S.G.; 1972. Survey of the Rusty Glacier area, Yukon Territory, Canada, 1967-70. Journal of Glaciology, v. 11, p. 235- 253. Costner, F., 1925. Results of the Swedish Expedition to Spitzbergen in 1924. 1. Quaternary geology of the region around the Kjellstrom valley. Geografiska Annaler, v. 7, p. 104-121. Cromack, M., 1991. A glaciolacustrine sedimentary system, north-west Spitsbergen. Unpublished PhD Thesis, University of Cambridge, Cambridge. Croot, D.G., 1988. Glaciotectonics and surging glaciers: a correlation based on Vestspitsbergen, Svalbard, Norway. In: D.G. Croot (editor) Glaciotectonics: Forms and Processes, p. 49-62. A.A. Balkema, Rotterdam. Cruikshank, J., 1990. Life lived like a story. Uaiversity of Nebraska Press, Lincoln. 404pp. Dege, W., 1948. Das Nordostland von Spitzbergen. Polarforschung, v. 2, p. 72-83 and 152-163. Dege, W., 1949. Meine umsegelung des Nordostlandes von Spitzbergen. Fetschrift zum 70 Geburtstag dae Ord. Professors der Geographie. Dr. L. Mecking. Bremer-Hom, p. 79- 96. Desio, A., 1954. An exceptional glacier advance in the Karakoram-Ladakh. Journal of Glaciology, v. 2, p. 383- 385. Dolgushin, L.D. and Osipova, G.B., 1973. Regime of a surging glacier between advances. International Association of Hydrological Sciences Publication, No. 107, p. 1150-1159. Dolgushin, L.D. and Osipova, G.B., 1975. Glacier surges and the problem of their forecasting. International Association of Hydrological Sciences Publication No. 104, p. 292-304. Dolgushin, L.D., Yevteyev, S.A. , Krenke, A.N., Rototayev, K.G. and Svatkov, N.M., 1963. The recent advance of the Medvezhi Glacier. Priroda II, p. 85- 92. Dowdeswell, J.A., 1984. Remote sensing studies of Svalbard glaciers. Unpublished PhD Thesis, University of Cambridge, Cambridge. Dowdeswell, J.A., 1989. On the nature of Svalbard icebergs . Journal of Glaciology, v. 35, p. 224- 234. Dowdeswell , J.A. and Collin, R.L., 1990. Fast-flowing outlet glaciers on Svalbard ice caps. Geology , V. 18, p. 778-781. Dowdeswell, J.A. and Drewry, D.J., 1989. The dynamics of Austfonna, Nordaustlandet, Svalbard: surface velocities, mass balance, and subglacial melt water. Annals of Glaciology, v. 12, p. 37- 45. Dowdeswell , J.A., Drewry, D.J., Liest01, 0. and Orheim, 0., 1984. Airborne radio echo sounding of sub-polar glaciers in-Spitsbergen. Norsk Polarinstitutt Skrifter Nr. 182. Dowdeswell, J.A., Drewry, DJ., Cooper, A.P.R., Gorman, M.R., Liest01, 0. and Orheim, 0., 1986. Digital mapping of the Nordaustlandet ice caps from airborne geophysical investigations. Annals of Glaciology, v. 8, p. 51-58. Dowdeswell, J.A., Drewry, DJ. and Simoes, J.C., 1990. Comments on: "6000-year climate records in an ice core from the H0ghetta ice dome in northern Spitsbergen". Journal of Glaciology, v. 36, p. 353-356. Dowdeswell, J.A., Hamilton, G.S . and Hagen, J.O., 1991. The duration of the active phase on surge- type glaciers: contrasts between Svalbard and other regions. Journal of Glaciology, v. 37, p. 388-400. Page 262 I I References Drewry, D.J., 1987. Radar and seismic ice-thickness measurements compared on sub-polar glaciers in Svalbard. Annals of Glaciology, v. 9, p. 246. Drewry, DJ. and Liest01, 0., 1985. Glaciological investigations of surging ice caps in Nordaustlandet, Svalbard, 1983. Polar Record, v. 22, p. 357-378. Drewry, DJ., Dowdeswell, J.A. and Riley, N.W., unpublished. Surge activity on Bakaninbreen, Spitsbergen, Svalbard. Unpublished manuscript, Scott Polar Research Institute, Cambridge. Echelmeyer, K., 1983. Response of Blue Glacier to a perturbation in ice thickness-theory and observation. Unpublished PhD Thesis, California Institute of Technology, Pasadena. Echelmeyer, K. and Harrison, W.D., 1989. Surge of West Fork Glacier, Alaska, U.S.A. (abstract). Annals of Glaciology, v. 12, p. 212. Echelmeyer, K. and Harrison, W.D., 1990. Jakobshavn Isbrre, West Greenland: seasonal variations in velocity-or lack thereof. Journal of Gluciology, v. 36, p. 82-88. Echelmeyer, K. and Wang Zhongxiang, 1987. Direct observation of basal sliding and deformation of basal drift at sub-freezing temperatures. Journal of Glaciology, v. 113, p. 83-98. Echelmeyer, K., Butterfield, R. and Cuillard, D., 1987. Some observations on the recent surge of Peters Glacier, Alaska, U.S .A .. Journal of Glaciology, v. 33, p. 341-345. Echelmeyer, K., Clarke, T.S. and Harrison, W.D., 1991. Surficial glaciology of Jakobshavn Isbrre, West Greenland: Part I. Surface morphology. Journal of Glaciology, v. 37, p. 368-382. Elliston, G.R., 1973. Water movement through the Gornergletscher. International Association of Scientific Hydrology Publication, No. 95, p. 79-84. Elverh0i, A. and Solheim, A., 1983. The Barents Sea Ice Sheet a sedimentological discussion. Polar Research, v. 1, p. 23--42. Elverh0i, A. and Solheim, A., 1987. Late Weichselian glaciation of the northern Barents Sea: a discussion. Polar Research, v. 5, p. 285-288. Engelhardt, H.F., Harrison, W.D. and Kamb, W.B., 1978. Basal sliding and conditions at the glacier bed as revealed by borehole photography. Journal of Glaciology, v. 20, p. 469-508. Engelhardt, H.F., Kamb, B., Echelmeyer, K. and Harrison, W.D., 1986. Basal water pressure in a surge-type glacier: Variegated Glacier, Alaska (abstract). International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September, 1985, Interlaken, Switzerland. Mitteilungen der Versuchsansta ltfur Wasserbau, Hydrologie und Glaziologie, Nr. 90; p. 43-45. Engelhardt, H.F., Humphrey, N., Kamb, B. and Fahnestock, M., 1990. Physical conditions at the base of a fast moving Antarctic ice stream. Science, v. 248, p. 57-59. Evans, K., Goodman, DJ. and Holdsworth, G., 1975. Recording wire strain meters on the Barnes Ice Cap, Baffin Island, Canada. Journal of Glaciology, v. 20, p. 409-423. Fahnestock, R.K., 1963 . Morphology and hydrology of a glacial stream, White River, Mount Rainier, Washington. United States Geological Survey Professional Paper No. 422-A. Farmer, I.W., 1983. Engineering behaviour of rocks. Chapman and Hall, London. Fastook, J.L., 1987. Use of a finite element continui ty model to study the transient behaviour of Ice Stream C and the causes of its present low velocity. Journal of Geophysical Resfarch, v. 92, p. 8941- 8950. Flood, B., Nagy, J. and Winsnes, T.S., 1971. Geological map of Svalbard, 1:500,000. Sheet lG: Spitsbergen, southern part. Norsk Polarinstitutt Skrifter Nr. 154-A. Page 263 References Flotron, A., 1973. Photogrammetrische Messungen von Gletscherbewegungen mit automatischer Kamera. Photogrammetrie und Kulturtechnik, v. 71, p. 15-17. Forman, S.L. , 1988. The solar resetting of thermoluminescence of sediments in a glacier-dominated fjord environment in Spitsbergen: geochronologic implications. Arctic and Alpine Research, v. 20, p. 243-253. Fountain, A.G., in press. Subglacial water flow inferred from stream measurements at South Cascade Glacier, Washington, U.S.A .. Journal of Glaciology. Fowler, A.C,, 1987. A theory of glacier surges. Journal of Geophysical Research, v. 92, p. 9111- 9120. Fowler, A.C., 1989. A mathematical analysis of glacier surges. Society of Industrial and Applied Mathematics Journal of Applied Mathematics, v. 49, p. 246-263. Fowler, A.C. and Larson, D.A., 1980. Thermal stability properties of a model of glacier flow . Geophysical Journal of the Royal Astronomical Society, v. 63, p. 347-359. Fox, A.J., 1989. A photogrammetric and image-processing approach to the quantification of ice volume change over time in two surge-type glaciers in west Spitsbergen. Unpublished MSc Dissertation, University of Aberdeen, Aberdeen. Fujii, Y., Kamiyama, K., Kawamura, T., Kameda, T., Izumi, K., Satow, K., Enomoto, H., Nakamura, T., Hagen, J.O., Gjessing, Y. and Watanabe, 0., 1990. 6000-year climate records in an ice core from the H0ghetta Ice Dome in northern Spitsbergen. Annals of Glaciology, v. 14, p. 85- 89. Gillet, F., 1975. Steam, hot-water and electrical thermal drills for temperate glaciers. Journal of Glaciology, v. 14, p. 171-179. Glazyrin, G.E. , 1978. Identification of surging glaciers by morphometric characteristics. Materialy Glyatsiologicheskikh Isseldovaniy. Khronika. Obsuzhdeniya, v. 33, p. 136-138. Glen , A.R ., 1937. The Oxford University Arctic Expedition, North East Land, 1935-1936. Geographical Journal, v. 90, p. 193-222 and 289- 314. Glen, A.R., 1941 . A sub-arctic glacier cap: the West Ice of North East Land. Geographical Journal, v. 98, p. 65-76 and 135-146. Glen, J.W., 1955. The creep of polycrystalline ice. Proceedings of the Royal Society of London Series A, V. 228, p. 519-538. Gordiyenko, F.G., Kotlyakov, V.M., Punning, Y.K.M. and Vaikmae, R. , 1981. Study of a 200 mice core from the Lomonosov Plateau on Spitsbergen and the palaeoclimatic implications. Polar Geography and Geology, v. 5, p. 242- 251. Gurnell, A.M., 1982. The dynamics of suspended sediment concentration in a proglacial stream. International Association of Hydrological Sciences Publication, No. 138, p. 319-330. Gurnell, A.M., 1987. Suspended sediment. In: A.M. Gurnell and M.J. Clark (editors) Glacio-fluvial sediment transport: an Alpine perspective. p. 305-354. John Wiley and Sons, Chichester. Gus'kov, A.S., 1981. Vodno-ledovy balans lednikov Shpitsbergena v 1979/80 balansovom goda [Water-ice balance of glaciers in Spitsbergen in the 1979/80 balance year]. Materialy Glyatsiologicheskikh Isseldovaniy. Khronika. Obsuzhdeniya, v. 46, p. 136-139. Haeberli , W., 1986. Factors affecting the distribution of rocky and sedimentary glacier beds (abstract) . Intematio_nal Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September, 1985, Interlaken, Switzerland. Mitteilungen der Versuchsanstalt fur Wasserbau, Hydrologie und Glaziologie, Nr. 90, p. 48-49. Page 264 References Hagen, J.O., 1987a. Glacier surge in Svalbard with examples from Usherbreen. Norsk Geografiska Tidsskrift, v. 42, p. 203- 213. Hagen, J.O., 1987b. Glacier surge at Usherbreen, Svalbard. Polar Research, v. 5, p. 239-252. Hagen, J.O., 1988. Glacier mass balance investigations in the balance year 1986-87. Polar Research, V. 6, p. 205-209. Hagen, J.O. and Liest01, 0., 1990. Long-term glacier mass-balance investigations in Svalbard, 1950- 88. Annals of Glaciology , v. 14, p. 102-106. Hagen, J.O. and Sretrang, A.C., 1991. Radio echo soundings of sub-polar glaciers with low-frequency radar. Polar Research, v. 9, p. 99-107. Hagen, J.O., Korsen, O.M. and Vatne, G., 1991. Drainage pattern in a sub-polar glacier: Br0ggerbreen, Svalbard. In: Y. Gjessing, J .O. Hagen, K.A. Hassel and B. Wold (editors) Arctic hydrology: present and future tasks. Hydrology of Svalbard, p. 121- 131. Norwegian National Committee for Hydrology Report No. 23. Hallet, B., Gregory, C.E., Stubbs, C.W. and Anderson, R.S., 1986. Measurements of ice motion over bedrock at subfreezing temperatures (abstract). International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September, 1985, Interlaken, Switzerland. Mitteilungen der Versuchsanstaltfiir Wasserbau, Hydrologie und Glaziologie, Nr. 90, p. 53-54. Hambrey, M.J., 1984. Sedimentary processes and buried ice phenomena in the pro-glacial areas of Spitsbergen glaciers. Journal of Glaciology, v. 30, p. 116-119. Hambrey, M.J. and Muller, F., 1978. Structures and ice deformation in the White Glacier, Axel Heiberg Island, Northwest Territories, Canada. Journal of Glaciology, v. 20, p. 41-66. Hammer, K.M. and Smith N.D., 1983. Sediment production and transport in a proglacial stream: Hilda Glacier, Alberta, Canada. Boreas, v. 12, p. 91-106. Hanssen-Bauer, I., Kristensen So!As, M. and Steffensen. E.L., 1990. The climate of Spitsbergen. Klima. Det Norske Meteorologiske Institutt Rapport Nr. 39/90. Harrison, W.D., unpublished. Russell Fjord and its damming. Unpublished manuscript, Geophysical Institute, University of Alaska, Fairbanks . Harrison, W.D., Drage, B.T, Bredthauer, S., Johnson, D., Schoch, C. and Follett, A.B., 1983. Reconnaissance survey of the glaciers of the Susitna River basin in connection with proposed hydroelectric development. Annals of Glaciology, v. 4, p. 99-104. Harrison, W.D., Kamb, B. and Engelhardt, H., 1986a. Morphology and motion at the base of a surge- type glacier (abstract). International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September, 1985, Interlaken, Switzerland. Mit1eilungen der Versuchsanstaltfur Wasserbau, Hydrologie und Glaziologie, Nr. 90, p. 55-56. Harrison, W.D., Raymond, C.F. and MacKeith, P., 1986b. Short period motion events on Variegated Glacier as observed by automatic photography and seismic methods. Annals of Glaciology, v. 8, p. 82- 89. Hartog, J.M. and Thompson, H.R., 1950. Oxford University Expedition to North East Land, 1949. Oxford University Exploration Club Bulletin, v. 3, p. 5- 10. , Hattersley-Smith, G., 1964. Rapid advance of glacier in northern Ellesmere Island. Nature, v. 20, p. 176. Heinrichs, T.A., Mayo, L.R., Echelmeyer, K. and Harrison, W.D., 1991. Black Rapids Glacier, Alaska-unexpected behaviour during the quiescent phase of a surge-type glacier. EOS, v. 72, p. 158. Page 265 References Hisdal, V., 1985. The geography of Svalbard. 2nd edition. Norsk Polarinstitutt, Polarhfmdbok, No. 2, 84pp. Hjelle, A. and Lauritzen, 0., 1982. Geological map of Svalbard, 1 :500,000. Sheet 3G: Spitsbergen, northern part. Norsk Polarinstitutt Skrifter Nr. 154-C Hodge, S., 1972. The movement and basal sliding of the Nisqually Glacier, Mount Rainier. Unpublished PhD Thesis, University of Washington, Seattle. Hodge, S., 1974. Variations in the sliding of a temperate glacier. Journal of Glaciology, v. 13, p. 349- 369. Hodge, S., 1976. Direct measurement of basal water pressures: a pilot study. Journal of Glaciology, v. 16, p. 205- 218. Hodge, S., 1979. Direct measurement of basal water pressures: progress and problems. Journal of Glaciology, v. 16, p. 205- 217. - Hoel, A., 1919. A brief account of Norwegian scientific exploration in Spitsbergen, 1827-1918. S. and Jui S0rensen's Boktrykkeri A/S, Kristiana. 38pp. Holdsworth, G., 1969. Primary transverse crevasses. Journal of Glaciology, v. 8, p. 107- 129. Holdsworth, G., 1977. Surge activity on the Barnes Ice Cap. Nature, v. 269, p. 588- 590. Holdsworth, G., 1984. Glaciological reconnaissance of an ice-core drilling site, Penny Ice Cap, Baffin Island. Journal of Glaciology, v. 30, p. 3- 15. Hollin, J.T., 1956. Oxford University expedition to Nordaustlandet, 1955. Polar Record, v. 8, p. 28. Hooke, R. leB., 1981. Flow law for polycrystalline ice in glaciers; comparison of theoretical predictions, laboratory data and field measurements. Reviews of Geophysics and Space Physics, V. 19, p. 664-672. Hooke, R. leB., 1984. On the role of mechanical energy in maintaining subglacial water conduits at atmospheric pressure. Journal of Glaciology, v. 30, p. 180-187. Hooke, R. leB., 1988. Englacial and subglacial hydrology: a qualitative review. Arctic and Alpine Research, v. 21, p. 221-233 . Hooke, R. LeB., Brzozowoski, J. and Bronge, C., 1983a. Seasonal variations in surface velocity, Storglaciaren, Sweden. Geografiska Annaler, v. 65A, p. 263- 277. Hooke, R. leB:, Gould, J.E. and Brzozowoski, J., 1983b. Near-surface temperatures near and below the equilibrium line on polar and subpolar glaciers. Zeitschrift fur Gletscherkunde und Glazialgeologie, v. 19, p. 1-25. Hooke, R. JeB., Wold, B. and J.O. Hagen, 1985. Subglacial hydrology and sediment transport at Bondhusbreen, south west Norway. Geological Society of America Bulletin, v. 96, p. 388- 397. Hooke, R. leB., Holmlund, P. and Iverson, N.R., 1987. Extrusion flow demonstrated by borehole deformation measurements over a riegel, Storglaciaren, Sweden. Journal of Glaciology, v. 33, p. 72-78. Hooke, R. leB ., Miller, S.B. and Kohler, J., 1988. Character of the englacial and subglacial drainage system in the upper part of the ablation area of Storglaciaren, Sweden. Journal of Glaciology, v. 34, p. 228-231. Hooke, R. leB., Calla, P., Holmlund, P., Nilsson, M. and Stroeven, A., 1989a. A 3-year record of seasonal variations in surface velocity, Storglaciaren, Sweden. Journal of Glaciology, v. 35, p. 235-247. Page 266 References Hooke, R. leB., Laumann, T. and Kennett, M.I., 1989b. Austdalsbreen, Norway: expected reaction to a 40 m increase in water level in the lake into which the glacier calves. Cold Regions Science and Technology, v. 17, p. 113-126. Hooke, R. leB., Laumann, T. and Kohler, J., 1990. Subglacial water pressures and the shape of subglacial conduits. Journal of Glaciology, v. 36, p. 67- 71. Horvath, E.V. and Field, W.O., 1969. References to glacier surges in North America. Canadian Journal of Earth Sciences, v. 6, p. 845-851. Humphrey, N. and Echelmeyer, K., 1990. Hot-water drilling and bore-hole closure in cold ice. Journal of Glaciology, v. 36, p. 287-298. Humphrey, N., Raymond, C.F. and Harrison, W.D., 1986. Discharges of turbid water during mini- surges of Variegated Glacier, Alaska. Journal of Glaciology, v. 32, p. 195-207. Hutchins, P.F., 1952. British scientific work in Spitsbergen. Endeavour, v. 11, p. 17- 21. Hutter, K., 1983. Theoretical glaciology: material science of ice and the mechanics of glaciers and ice sheets. D. Reidel, Dortrecht 510pp. Iken, A., 1973. Schwankungen der Oberflachengeschwindigkeit des White Glacier, Axel Heiberg Island im Zusammenhang mit Schwankungen der Wal3erfiihrung von Gletscherbachen und des Wal3erdruckes in Gletschennilhlen. Zeitschriftfii.r Gletscherkunde und Glazialgeologie, v. 9, p. 207- 219. Iken, A., 1974. Velocity fluctuations of an Arctic valley glacier. A study of White Glacier, Axel Heiberg Island, Canadian Arctic Archipelago. Axel Heiberg Island Research Report, McGill University, Glaciology, No.5. Iken, A., 1978. Variations in the surface velocities of some Alpine glaciers measured at intervals of a few hours: comparison with Arctic glaciers. Zeitschriftfii.r Gletscherkunde und Glazialgeologie, v. 13, p. 23- 35. Iken, A., 1981. The effect of the subglacial water pressure on the sliding velocity of a glacier in an idealized numerical model. Journal of Glaciology, v. 27, p. 407-422 Iken, A. and Bindschadler, R., 1986. Combined measurements of subglacial water pressure and surface velocity of Findelengletscher, Switzerland. Conclusions about the drainage system and sliding mechanism. Journal of Glaciology, v. 32, p. 101-119. !ken, A., Rothlisberger, H., Flotron, A. and Haeberli, W., 1983. The uplift of Unteraagletscher at the beginning of the melt season-a consequence of water storage at the bed? Journal of Glaciology, V. 29, p. 28-47. Iken, A., Echelmeyer, K. and Harrison, W.D., 1989. A light-weight hot-water drill for large depth: experiences with drilling on Jakobshavn Glacier, Greenland. In: C. Rado and D. Beaudoing (editors) Ice core drilling. Proceedings of the Third International Workshop on lee Drilling Technology, Grenoble, 10-14 October, 1988. Laboratoire de Glaciologie et Geophysique de !'Environnement, S_t. Martin d'Hcres. p. 123- 136. Jania, J., 1986. Calving processes of South Spitsbergen tidewater glaciers (abstract). Chapman Conference on Fast Glacier Flow, 4-8 May 1986, Whistler, British Columbia, American Geophysical Union. p. 22. Jansson, P. and Hooke, R. leB., 1989. Short-term variations in strain and surface tilt on Storglaciaren, Kebnekaise, northern Sweden. Journal of Glaciology, v. 35, p. 201-208. Jarvis, J.T. and Clarke, G.K.C., 1975. Thennal regime of Trapridge Glacier and its relevance to glacier surging. Journal of Glaciology, v. 14, p. 235-250. Jones, A.S., 1979. The flow of ice over a till bed. Journal of Glaciology, v. 87, p. 393-395. Page 267 I I References Jonsson, S., 1982. On the present glaciation of Ston1ya, Svalbard. Geografiska Anna/er, v. 64A, p. 53-79. Kamb, B., 1986. An observationally based mechanism of glacier surging (abstract). International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September 1985, Interlaken, Switzerland. Mitteilungen der Versuchsanstaltfur Wasserbau, Hydrologie und Glaziologie, Nr. 90, p. 66-67. Kamb, B., 1987. Glacier surge mechanism based on linked cavity configuration of the basal water conduit system. Journal of Geophysical Research, v .. 92, p. 9083-9100. Kamb, B., 1991. Rheological nonlinearity and flow instability in the deforming bed mechanism of ice stream motion. Journal of Geophysical Research, v. 96, p. 16585- 16595. Kamb, B. and La Chapelle, E.R., 1964. Direct observation of the mechanism of glacier sliding over bedrock. Journal of Glaciology, v. 5, p. 159-172. Kamb, B. and Echelmeyer, K., 1986. Stress-gradient coupling in glacier flow. I. Longitudinal averaging of the influence of ice thickness and surface slope. Journal of Glaciology, v. 32, p. 267-284. Kamb, B. and Engelhardt, H.F., 1987. Waves of accelerated motion in a glacier approaching surge: the mini-surges of Variegated Glacier, Alaska, U.S .A .. Journal of Glaciology, v. 33, p. 27-46. Kamb, B., Raymond, C.F., Harrison, W.D., Engelhardt, H.F., Echelmeyer, K., Humphrey, N., Brugman, M. and Pfeffer, T., 1985. Glacier surge mechanism: the 1982-1983 surge of Variegated Glacier, Alaska. Science, v. 227, p. 469-479. Kotlayakov, V.M., Macheret, Y.Y., Gordiyenko, F.G. and Zhurav!ev, A.B., 1980. Geofizicheskiye i isotopnyye issledovaniya lednikov Shpitsbergena [Geophysical and isotopic investigations of Spitsbergen glaciers] . Vestik Akademii Nauk SSSR, 1980, p. 132-138. Krimmel, R.M., 1987. Columbia Glacier in 1986: 800 meter retreat. United States Geological Survey Open File Report, 87-207. Krimmel, R.M. and Vaughn, B.H., 1987. Columbia Glacier, Alaska: changes in velocity 1977-86. Journal of Geophysical Research, v. 92, p. 8961-8968. Kristiansen, K.J. and Sollid, J.L., 1987. Svalbard, Jordartskart, 1:1,000,000. Nasjonalatlasfor Norge, Geografisk Institutt, Universitetet i Oslo. Lamplugh, G.W., 1911. On the shelly moraine of the Sefstrtim Glacier and other Spitsbergen phenomena illustrative of British glacial conditions: with a note by A. Strahan, and a list of shells by Prof. Gerard de Geer. Proceedings of the Yorkshire Geological Society, v. 17, p. 216- 241. Landvik, J.Y., Mangerud, J. and Salvigsen, 0., 1988. Glacial history and permafrost in the Svalbard area. In: K. Senneset (editor) Permafrost. Fifth International Conference, 2-5 August 1988. Proceedings. Volume 1, p. 194-198. Lefauconnier, B. and Hagen, J.O., 1990. Glaciers and climate in Svalbard: statistical analysis and reconstruction of the Br0ggerbreen mass balance for the last 77 years. Annals of Glaciology, v. 14, p. 148-152. Lefauconnier, B. and Hagen, J.O., 1991. Analysis of NP photography: surging and calving glaciers. Report to Operat0rkomite Nord. Liest0l, 0., 1969. Glacier surges in West Spitsbergen. Canadian Journal of Earth Sciences, v. 6, p. 895-898. Liest0l, 0., 1975. Glaciological work in 1973. Norsk Polarinstitutt Arbok 1973, p. 181-192. Liest0l, 0., 1976. Glaciological work in 1974. Norsk Polarinstitutt Arbok 1974, p. 183-194. Page 268 References Liest0l, 0., 1977. Pingos, springs and permafrost in Spitsbergen. Norsk Polarinstitutt Arbok 1975, p. 7- 29. Liest01, 0., 1980. Glaciological work in 1979. Norsk Polarinstitu/1 Arbok 1979, p. 43-51. Liest01, 0., 1990. Glaciers in the Kongsfjorden area. Norsk Polarinstitu/1 Arbok 1989, p. 51-61. Liest01, 0 ., in press. Glaciers of Svalbard, Norway. In: R.S. Williams and J.G. Ferrigno (editors) Satellite Image Atlas of Glaciers of the World, United States Geological Survey Professional Paper, 1386. Liest01, 0., Repp, K. and Wold, B., 1980. Surpa-glacial lakes in Spitsbergen. Norsk Geografisk Tiddsskrift, v. 34, p. 89-92. Lingle, C.S., Hughes, T.J. and Kollmeyer, R.C., 1981. Tidal flexure of Jakobshavn Glacier, West Greenland. Journal of Geophysical Research; v. 86, p. 3960-3968. Lliboutry, L., 1958. Studies of the shrinkage after a sudden advance, blue bands and wave ogives on Glaciar Univesidad (Central Chilean Andes). Journal of Glaciology, v. 3, p. 261-272. Lliboutry, L., 1968. General theory of subglacial cavitation and sliding of temperate glaciers . Journal of Glaciology, v. 7, p. 21- 58. L0ken, O.H., 1969. Evidence of surges on the Barnes Ice Cap, Baffin Island. Canadian Journal of Earth Sciences, v. 6, p. 899-901. McMeeking, R.M. and Johnson, R.E., 1986. On the mechanics of surging glaciers. Journal of Glaciology, v. 32, p. 120-132. Macheret, Y.Y., 1981. Forms of glacial relief on Spitsbergen glaciers. Annals of Glaciology, v. 2, p. 45-51. Macheret, Y.Y. and Zhuravlev, A.B., 1982. Radio echo sounding of Svalbard glaciers. Journal of Glaciology, v. 28, p. 295-314. Macheret, Y.Y., Zhuravlev, A.B. and Bobrova, L.I., 1984. Thickness, subglacial relief and volume of Svalbard glaciers based on radio echo sounding data. Materialy Glyatsiologicheskikh !sseldovaniy. Khronika. Obsuzhdeniya, v. 51, p. 49-63. Machcrel, Y.Y ., Bobrova, L.I. and Sankina, L.V., in press. Volume hydrothermal state and regime of Spitsbergen glaciers deduced from airborne radio echo sounding data. Materialy Glyatsiologicheskikh lsseldovaniy. Khronika. Obsuzhdeniya. Mathews, W.H., 1964. Water pressure under a glacier. Journal of Glaciology, v. 5, p. 235-240. Mathews, W.H. and McKay, J.R., 1960. Deformation of soils by glacial ice and the influence of pore pressure and permafrost. Royal Society of Canada Transactions, v. 54, p. 3 7-43 . Mayo, L.R., 1984. Glacier mass balance and runoff research in the U.S.A .. Geografiska Anna/er, v. 66A, p. 215- 228. Mayo, L.R., 1989. Advance of Hubbard Glacier and the 1986 outburst of Russell Fjord, Alaska, U.S.A .. Annals of Glaciology, v. 13, p. 189-194. Meier, M.F., 1958. The mechanics of crevasse formation. International Association of Scientific Hydrology Publication, No. 4, p. 500-509. Meier, M.F., 1960. Mode of flow of Saskatchewan Glacier, Alberta, Canada. United States Geological Survey Professional Paper, 351. Meier, M.F., 1968. Calculations of slip of Nisqually Glacier on its bed. International Association of Hydrological Sciences Publication, No. 79, p. 49- 57. Page 269 r References Meier, M.F., 1969. Seminar on the causes and mechanics of glacier surges, St. Hilaire, Canada, September 10-11, 1968: a summary. Canadian Journal of Earth Sciences, v. 6, p. 987-989. Meier, M.F. and Post, A.S., 1969. What are glacier surges? Canadian Journal of Earth Sciences, v. 6, p. 807-819. Meier, M.F. and Post, A.S., 1987. Fast tidewater glaciers. Journal of Geophysical Research, v. 92, p. 9051-9058. Meier, M.F., Kamb, B., Allen, C.R. and Sharp, R.F. , 1974. Flow of Blue Glacier, Olympic Mountains, Washington, U.S.A.. Journal of Glaciology, v. 13, p. 187-212. Meier, M.F., Rasmussen, L.A., Post, A.S., Brown, C.S., Sikonia, W.G., Bindschadler, R.A., Mayo, L.R. and Trabant, D.C., 1980. Predicted timing of the disintegration of the lower reach of Columbia Glacier, Alaska. United States Geological Survey Open File Report, 80-582. Menzies, J., 1989. Subglacial hydraulic conditions and their possible impact upon subglacial bed formation . Sedimentary Geology, v. 62, p. 125-150. Moffit, F.H., 1942. Black Rapids Glacier, Alaska. United States Geological Survey Bulletin, 926-B, p. 146-160. Moran, S.R., 1971. Glaciotectonic structures in drift. In: R.P. Goldthwaite (editor) Till, a symposium. p. 127-148. Ohio State University Press, Columbus, Ohio. Moss, R., 1938. The physics of an ice cap. Geographical Journal, v. 92, p. 211-231. Moss, R. and Glen, A.R., 1939. The retreat of Franklin Glacier, North East Land. Geographical Journal, v. 93, p. 228- 229. Murray, T., 1990. Deformable glacier beds: measurement and modelling. Unpublished PhD Thesis, University of Wales, Aberystwyth. Napoleoni, J.-G.P. and Clarke, G.K.C., 1978. Hot water drilling in a cold glacier. Canadian Journal of Earth Sciences, v. 15, p. 316-321. Nathorst, A.G., 1909. The historical sketch of Swedish exploration in Spitsbergen, 1758-1908. Ymer, V. 29, p. 4-22. National Research Council, 1985. Glaciers, ice sheets, and sea level: effects of a C02-induced climatic change. Ad Hoe Committee on the Relationship between Land Ice and Sea Level, Polar Research Board, National Research Council. National Academy Press, Washington D.C., 330pp. Nixon, W.A., Dowdeswell, J.A., Cooper, A.P.R., Drewry, DJ., Watts, L.G., Liest01, 0. and Smith, R.A., 1985. Applications and limitations of finite element modeling to glaciers: a case study. Journal of Geophysical Research, v. 90, p. 11303- 11311. Nordenskiold, N.A.E., 1873. D"ie schilttenfahrt expedition im nord-ostlichen Theile von Spitzbergen, 24 April-15 Juni, 1873. Petermann' s Geographische Mittheilungen, v. 12, p. 444-453. Nye, J.F., 1952. The mechanics of glacier flow. Journal of Glaciology, v. 2, p. 82-93 Nye, J.F., 1957. The distribution of stress and velocity in glaciers and ice sheets. Proceedings of the Royal Society, Series A, v. 239, p. 113-133. Nye, J.F., 1965. The flow of a glacier in a channel of rectangular, elliptic or parabolic cross-section. Journal of Glaciology, v. 5, p. 661-690. Nye, J.F., 1959. A method of determining the strain-rate tensor at the surface of a glacier. Journal of Glaciology, v. 3, p. 409-419. Page 270 References Nye, J .F., 1973. Water at the bed of a glacier. International Association of Hydrological Sciences Publication, No. 95, p. 189-194. Olesen, O.B., 1989. A Danish contribution to the family of hot-water glacier drills. In: C. Rado and D. Beaudoing (editors) Ice core drilling. Proceedings of the Third International Workshop on Ice Drilling Technology, Grenoble, 10-14 October, 1988. Laboratoire de Glaciologie et Geophysique de !'Environnement, St. Martin d'Heres. p. 140-148. Ommanney, C.S.L., 1970. The ice masses of Axel Heiberg Island, Canadian Arctic Archipelago: a study in glacier inventory. Axel Heiberg Island Research Reports, Glaciology 3, McGill University, Montreal. Orheim, 0., Hagen, J.O., 0sterhus, S. and Sretrang, A.C., 1991. Studies on, and underneath, the ice shelf Fimbulisen. In: 0. Orheim (editor) Report of the Norwegian Antarctic Research Expedition 1989/90. Norsk Polarinstitutt Meddelelser Nr. 113, p. 59-73. Parry, W.E., 1828. Narrative of an attempt to reach the North Pole. J. Murray, London, 148pp. Paterson, W.S .B., 1961. Movement of Sefstr0ms Gletscher, north-east Greenland. Journal of Glaciology, v. 3, p. 844-849. Paterson, W.S.B., 1964. Variations in the velocity of Athabasca Glacier with time. Journal of Glaciology, v. 5, p. 277-285. Paterson, W.S.B., 1972. Laurentide Ice Sheet: estimated volumes during Late Wisconsin. Reviews of Geophysics and Space Physics, v. 10, p. 885-917. Paterson, W.S.B., 1981. The physics of glaciers. Pergamon Press, Oxford. 380pp. Paterson, W.S.B., 1986. The current state of research on hydraulic effects at the glacier bed: an introduction to the Workshop. International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September 1985, Interlaken, Switzerland. Mitteilungen der Versuchsanstaltfil.r Wasserbau, Hydrologie und Glaziologie, Nr. 90, p. 7-14. Paterson, W.S.B., Nitsan, U. and Clarke, G.K.C., 1978. An investigation of creep instability as a mechanism of glacier surges. Materialy Glyatsiologicheskikh Isseldovaniy. Khronika. Obsuzhdeniya, v. 32, p. 201-209. Pettersen, L.-E., 1991. Hydrometric investigations in Svalbard, 1989- 1990. In: Y. Gjessing, J.O. Hagen, K.A. Hassel and B. Wold (editors) Arctic hydrology: present and future tasks. Hydrology of Svalbard, p. 133-138. Norwegian National Committee for Hydrology Report No. 23. Pfeffer, W.T., 1988. Structure and deformation in a propagating surge front. Unpublished PhD Thesis, University of Washington, Seattle. Pfirman, S.L. and Solheim, A., 1989. Subglacial meltwater discharge in the open marine tidewater glacier environment: observations from Nordaustlandet, Svalbard archipelago. Marine Geology, V. 86, p. 265-381. Pillewizer, W., 1965. Bewegungsstudien en einem arktische Gletscher. Polarforschung, v. 34, p. 247- 253. Post, A.S., 1960. The exceptional advances of the Muldrow, Black Rapids, and Susitna Glaciers. Journal of Geophysical Research, v. 65, p. 3703-3712. Post, A.S., 1965. Alaskan glaciers: recent observations in respect to the earthquake-advance theory. Science , v. 148, p. 366-368. Post, A.S ., 1969. Distribution of glacier surges in western North America. Journal of Glaciology , v. 8, p. 229-240. Post, A.S. , 1975. Preliminary hydrography and historic terminal changes of Columbia Glacier, Alaska. United States Geological Survey Hydro/ogic Investigations Atlas, No. 599, 3 sheets. Page 271 1 I II I References Press, W.H., Flannery, B.P. , Teukolsky, S.A. and Vetterling, W .T., 1989. Numerical recipes: the art of scientific computing. Cambridge University Press. Punning, A., Troitskiy, L.G. and Rajamae, E., 1976. The genesis and age of Quaternary deposits in the eastern part of Van Mijenfjorden, west Spitsbergen. Geologiska Foreningene i Stockholm Forhandlingar, v. 98, p. 343-347. Radok, U., Jenssen, D. and Budd, W., 1970. Steady-state temperature profiles in ice sheets. International Association of Scientific Hydrology Publication 86, p. 151-165. Radok, U., Jenssen, D. and Mclnnes, B., 1987. On the surging potential of polar ice streams. United States Department of Energy Contract Report DE-AC02-84ER60197. Rainwater, F.H. and Guy, H.P., 1961. Some observations on the hydrochemistry and sedimentation of the Chamberlain Glacier area, Alaska. United States Geological Survey Professional Paper 414- C. Ramsay, J.G., 1968. Folding and fracturing of rocks. McGraw Hill, New York. Raymond, C.F., 1971. Flow in a transverse of Athabasca Glacier, Alberta, Canada. Journal of Glaciology, v. 10, p. 55-84. Raymond, C.F., 1980. Temperate valley glaciers. In: S.C. Colbeck (editor) Dynamics of snow and ice masses, p. 79-139. Academic Press, London. Raymond, C.F., 1987. How do glaciers surge? A review. Journal of Geophysical Research, v. 92, p. 9121-9134. Raymond, C.F. and Harrison, W.D ., 1975. Some observations on the behaviour of the liquid and gas phases in temperate glacier ice. Journal of Glaciology, v. 14, p. 213- 233. Raymond, C.F. and Harrison, W.D., 1986. Winter initiation of surges (abstract). International Workshop on Hydraulic Effects at the Glacier Bed and Related Phenomena, 16-19 September 1985, Interlaken, Switzerland. Mitteilungen der Versuchsanstaltfur Wasserbau, Hydrologie und Glaziologie, Nr. 90, p. 85-86. Raymond, C.F. and Harrison, W.D., 1987. Fit of ice motion models to observations from Variegated Glacier, Alaska. International Association of Hydrological Sciences Publication 170, p. 153- 166. Raymond, C.F. and Harrison, W.D ., 1988. Evolution of Variegated Glacier, Alaska, U.S.A., prior to its surge. Journal of Glaciology, v. 34 , p. 1-15. Raymond, C.F. and Malone, S.D., 1986. Propagating strain anomalies on Variegated Glacier, Alaska , U.S.A .. Journal of Glaciology , v. 32, p. 178-19 1. Raymond, C.F., Johannesson, T., Pfeffer, T. and Sharp, M.J ., 1987. Propagation of a glacier surge into stagnant ice. Journal of Geophysical Research, v. 92, p. 9037-9050. Richards, K.S., 1982. Rivers:form and process in alluvial channels. Methuen, Land. Riley, K.F., 1974. Mathematical methods for the physical sciences. Cambridge University Press. Robin , G. deQ., 1955. Ice movement and temperature distribution in glaciers and ice sheets . Journal of Glacio logy, v. 2, p. 523-532. Robin, G. deQ., 1975. Velocity of radio waves in ice by means of a borehole interferometric technique. Journal of Glaciology, v. 15, p. 151-159. Robin, G. deQ. and Weertman, J. , 1973. Cyclic surging of glaciers. Journal of Glaciology, v. 12, p. 3-18. Page 272 References Rose, K.E., 1979. Characteristics of ice flow in Marie Byrd Land, Antarctica. Journal of Glaciology, v. 24, p. 63-75. Rothlisberger, H., 1972. Water pressure in intra- and subglacial channels. Journal of Glaciology, v. 11, p. 177-203. Rothlisberger, H. and Lang, H., 1987. Glacial hydrology. In: A.M. Gurnell and M.J. Clark (editors) Glacio-fluvial sediment transport: an Alpine perspective. p. 207-284. John Wiley and Sons , Chichester. Rothlisberger, H. , Iken, A. and Spring, U., 1979. Piezometric observations of water pressure at the bed of Swiss glaciers. Journal of Glaciology, v. 23, p. 429. Rowan, D.E., Pewe, T.L. , Pewe, R.H. and Stuckenrath, R., 1982. Holocene glacial geology of the Svea lowland, Spitsbergen, Svalbard. Geografiska Annaler, v. 64A, p. 35- 51. Rutishauser, H., 1971. Observations on a surging glacier in East Greenland. Journal of Glaciology, v. 10, p. 227-236. Schytt, V., 1964. Scientific results of the Swedish Glaciological Expedition to Nordaustlandet, Spitsbergen, 1957 and 1958. Parts I and II. Geografiska Anna/er, v. 46, p. 243- 281. Schytt, V., 1969. Some comments on glacier surges in eastern Svalbard. Canadian Journal of Earth Sciences, v. 6, p. 867-873. Seaberg, S.Z., Seaberg, J.Z., Hooke, R. leB. and Wiberg, D.W., 1988. Character of the englacial and subglacial drainage system in the lower part of the ablation area of Storglaciaren, Sweden, as revealed by dye-trace studies. Journal of Glaciology, v. 34, p. 217- 227. Sexton, D.J., Dowdeswell, J.A., Solheim, A. and Elverh0i, A., 1992. Seismic architecture and sedimentation in northwest Spitsbergen fjords. Marine Geology, v. 103, p. 53-68. Shabtaie, S. and Bentley, C.R., 1987. West Antarctic ice streams draining into the Ross Ice Shelf: configuration and mass balance. Journal of Geophysical Research, v. 92, p. 1311-1336. Shabtaie, S., Whillans, I.M. and Bentley, C.R., 1987. The morphology of Ice Streams A, B and C, West Antarctica, and their environs. Journal of Geophysical Research, v. 92, p. 8865-8884. Shabtaie, S., Bentley, C.R., Bindschadler, R. and MacAyeal, D.R., 1988. Mass balance studies of ice streams A, B and C, West Antarctica, and possible surging behaviour of Ice Stream B. Annals of Glaciology, v. 11, p. 137-149. Sharp, M.J., 1985. Crevasse-fill ridges: a landform type characteristic of surging glaciers? Geografiska Anna /er , v. 67A, p. 213-220. Sharp, M.J., 1988. Surging glaciers: behaviour and mechanisms. Progress in Physical Geography, v. 12, p. 349-370. Sharp, R.P., 1947. The Wolf Creek glaciers, St. Elias Range, Yukon Territory. Geographical Review, V. 37, p. 26-52. - Shreve, R.L., 1984. Glacier sliding at subfreezing temperature. Journa l of Glaciology, v. 30, p. 341- 347. Sikonia, W.G., 1982. Finite element glacier dynamics model applied to Columbia Glacier, Alaska. United States Geological Survey Professional Paper, 1258-A. Simoes, J.C., 1990. Environmental interpretation from Svalbard ice cores. Unpublished PhD thesis, University of Cambridge, Cambridge. Smith, B.M.E. and Evans, S., 1972. Radio echo sounding: absorption and scattering by water inclusions and ice lenses. Journal of Glaciology , v. 11, p. 133-146. Page 273 References Solheim, A., 1991. The depositional environment of surging subpolar tidewater glaciers. Norsk Polarinstitutt Skrifter Nr. 194. Solheim, A. and Pfirman, S.L., 1985. Sea floor morphology outside a grounded, surging glacier; Bnlsvellbreen, Svalbard. Marine Geology , v. 65, p. 127-143. Sollid, J.L. , B0 , P.H., Etzelmiiller, B., Vatne, G. and 0deg:rrd, R.S., 1991. Glacial and glaciofluvial material transport in subpolar glaciers; examples from Svalbard. In: Y. Gjessing, J.O. Hagen , K.A. Hassel and B. Wold (editors) Arctic hydrology: present and futu re tasks. Hydrology of Svalbard, p. 159-165. Norwegian National Committee for Hydrology Report No. 23. Steffensen, E.L. , 1969. The climate and its recent variations at the Norwegian Arctic stations . Meteorologiske Annaler, v. 5, no. 8. Steffensen, E.L., 1982. The climate at Norwegian Arctic stations. Klima, no. 5. Stenborg, T., 1970. Delay ofrunoff from a glacier basin. Geografiska Annaler, v. 52A, p. 1-30. Stuart, A. and Ord, J.K., 1987. Advanced theory of statistics. 5th edition. Edward Arnold, London. Sturm, M., 1987. Observations on the distribution and characteristics of potholes on surging glaciers. Journal of Geophysical Research, v. 92, p. 9015-9022. Sturm, M. and Cosgrove, D.M., 1990. An unusual jokulhlaup involving potholes on Black Rapids Glacier, Alaska Range. Journal of Glaciology, v. 36, p. 125. Sverrison, M., Johannesson, A. and Bjpmsson, H., 1980. Radio echo equipment for depth sounding of temperate glaciers. Journal of Glaciology, v. 25, p. 477-486. Swithinbank, C.M.W. , 1954. Ice streams. Polar Record, v. 7, p. 185-186. Tangborn, W.V., Krimmel , R.M. and Meier, M.F., 1975. A comparison of glacier mass balance by glaciological, hydrological and mapping methods, South Cascade Glacier Washington . International Association of Hydrological Sciences Publication, No. 104, p. 185-196. Tarr, R.S . amd Martin, L., 1914. Alaska Glacier Studies. National Geographic Society , Washington D.C. Thompson, H.R., 1953. Oxford Expeditions to NordaustlandeL Arctic, v. 6, p. 213-222. Thorarinsson , S., 1964. Sudden advance of Vatnajpkull outlet glaciers, 1930--1964. J9kull, v. 14, p. 76---89. Thorarinsson , S., 1969. Glacier surges in Iceland, with special reference to the surges of Bruarjpkul l. Canadian Journal of Earth Sciences, v. 6, p. 875- 882. van der Veen, C.J ., 1989. A numerical scheme for calculating stresses and strain rates in glaciers. Mathematical Geology, v. 21, p. 363-377. Vaykmyae, R.A., Gordiyenko, F.G., Zagorodnov, V.S. Mikhalov, V.I., Punning, Ya.-M.K. and Rayamyae, R.A. , 1977. Iztopnyye, geokhimicheskiye i stratigrfischeski issledovaniy na ledorazdele lednikov Grenf'ord i Frit'of (o . Zapadnyy Shpitsbergen). Materialy Glyatsio logicheskikh Isseldovaniy. Khronika. Obsuzhdeniya, v. 30, p._ 77-87. Vinje, T., 1985. The physical environment, western Barents Sea: drift, composition, morphology and distribution of sea ice fields in the Barents Sea. Norsk Polarinstitutt Skrifter, Nr. 179-C. Vivian, R. and Bocquet, G., 1973. Subglacial cavitition phenomena under the Glacier d' Argentiere, Mont Blanc, France. Journal of Glaciology, v. 12, p. 439-451. Voigt, U., 1965. Die Bewegung der Gletscherkunde das Kongsvegen (Kingsbay, Westspitsbergen). Petermanns Geographische Mitteilungen, v. 129, p. 1- 8. Page 274 References Walder, J.S., 1986. Hydraulics of subglacial cavities. Journal of Glaciology, v. 32, p. 439-445. Walters, R.A. and Dunlap, W.W., 1987. Analysis of time series of glacier speed: Columbia Glacier, Alaska. Journal of Geophysical Research, v. 92, p. 8969- 75. Weertman, J., 1957. On the sliding of glaciers. Journal of Glaciology, v. 3, p. 33- 38. Weertman, J., 1964. Theory of glacier sliding. Journal of Glaciology, v. 5, p. 287-303 Weertman, J., 1969. Water lubrication mechanism of glacier surges. Canadian Journal of Earth Sciences, v. 6, p. 929-942. Weertman, J., 1973. Creep of ice. In: E.Whalley, S.J. Jones and Gold, L.W. (editors) Physics and chemistry of ice, Royal Society of Canada, Ottawa, p. 320-337. Weertman, J. and Birchfield, G.E., 1983. Basal water film, basal water pressure, and velocity of travelling waves on glaciers. Journal of Glaciology, v. 29, p. 20-27. Werner, A., 1990. Lichen growth rates for the north west coast of Spitsbergen, Svalbard. Arctic and Alpine Research, v. 22, p. 129- 140. Whillans, I.M. and Bindschadler, R., 1988. Mass balance of Ice Stream B, West Antarctica. Annals of Glaciology, v. 11, p. 187-193. Whillans, I.M., Bolzan, J. and Shabtaie, S., 1987. Velocity of Ice Streams B and C, Antarctica. Journal of Geophysical Research, v., 92, p. 8895-8902. Wilbur, S.W., 1986. Surging versus non-surging glaciers: a comparison using morphometry and balance. Chapman Conference on Fast Glacier Flow, 4-8 May 1986, Whistler, British Columbia, American Geophysical Union. p. 12-13. Wilbur, S.W., 1988. Surging versus non-surging glaciers: a comparison using morphometry and balance. Unpublished MS Thesis, University of Alaska. Williams, L.D. and Knight, P.G., 1987. A computer program for plane strain analysis. University of Aberdeen Department of Geography Discussion Paper, No. 10. Wilson, A.T., 1964. Origin of ice ages: an ice shelf theory for Pleistocene glaciation. Nature, v. 201, p. 147- 149. Winsnes, T.S., 1988. Bedrock map of Svalbard and Jan Mayen: Geological Map, 1:1,000,000. Norsk Polarinstitutt Temakart, Nr. 3. Wolf, P.R., 1983. Elements of photogrammetry. McGraw Hill, New York. Wood, W.A., 1936. The Wood Yukon expedition of 1935; an experiment in photographic mapping. Geographical Review, v. 26, p. 228-246. Wood, W.A., 1942. The para~huting of expedition supplies. Geographical Review, v. 32, p. 36-55. Worsley, D., 1986. The geological history of Svalbard. Den norske stats oljeselskap a/s, Stavanger. Zagorodnov, V.S. and Zotikov, I.A., 1981. Kemovoya bureniye na Shpitsbergen. [Mehtods and results of deep thermal drilling of Svalbard glaciers in 1975-76.] Materidly Glyatsiologicheskikh lsseldovaniy. Khronika. Obsuzhdeniya, v. 40, p. 157-163. 0strem, G., 1975. Sediment transport in glacial meltwater streams. In: A.V. Jopling and B.C. McDonald (editors) Glaciofluvial and glaciolacustrine sedimentation, p. 101-122. Society of Economic Palaeontologists and Mineralogists, Tulsa, Oklahoma, Special Publication 23. 0strem, G. and Stanley, A., 1969. Glacier mass-balance measurements: a manual for field and office work. Department of Energy, Mines and Resources, Ottawa. Page 275 Ill I I APPENDIX A This appendix contains data concerning the short.period variations in surface velocity recorded at all the markers on Bjuvbreen during the 1989 and 1990 field seasons. The locations of the markers are illustrated in Figure 4.5. _1989 Julian day Target Horiz. velocity Deviation from Vertical velocity period m d-1 seasonal vel., % m d-1 201- 204 1 0.085 +1 -0.057 2 0.115 +56 -0.056 3 0.073 +21 -0.044 4 0.116 +65 -0.067 5 0.116 +80 -0.115 6 0.103 +61 -0.016 7 0.134 +79 0.037 8 0.118 +59 0.029 15 0.174 +87 0.026 16 0.198 +90 0.051 17 0.192 +86 0.053 18 0.232 +79 -0.058 19 0.219 +97 0.047 20 0.123 +18 -0.009 21 0.194 +88 -0.034 22 0.104 -10 -0.044 23 0.201 +91 0.009 204--211 1 0.078 -6 -0.027 2 0.085 +16 -0.021 3 0.066 +10 -0.026 4 0.073 +4 0.032 5 0.076 + 11 -0.066 6 0.070 + 11 0.007 7 0.038 -34 -0.028 8 0.069 -6 0.006 204--214 15 0.134 +44 -0.079 16 0.091 -13 -0.068 17 0.091 -12 -0.068 18 0.186 +44 - 19 0.093 -16 , -0.082 20 0.077 -23 -0.093 21 0.076 -26 -0.053 22 0.104 -11 -0.070 23 0.082 -21 . -0.061 continued ... Appendix A Julian day · Target Horiz. velocity Deviation from Vertical velocity period m d-1 seasonal vel., % m d-1 208- 213 30 - - - 31 0.122 +45 -0.008 32 0.192 +53 -0.011 33 0.118 +49 -0.008 34 0.183 +56 -0.027 35 0.233 +97 -0.047 36 0.201 +63 -0.090 37 0.231 +98 -0.055 38 0.215 +91 -0.016 39 0.201 +91 0.002 40 0.203 +91 0.006 41 -- - - 42 0.209 +57 -0.060 43 0.202 +66 -0.014 44 0.220 +80 -0.027 211- 214 1 0.099 +17 0.016 2 0.074 0 0.001 3 0.012 -81 -0.015 4 0.081 +18 -0.095 5 0.034 -50 -0.020 6 0.032 -49 -0.014 7 0.067 +15 0.016 8 0.032 -56 -0.022 213- 216 30 - - - 31 0.102 +21 0.002 32 0.125 0 0.004 33 0.083 +5 -0.050 34 0.105 -10 -0.018 35 0.1 16 -2 -0.001 36 0.183 +48 0.050 37 0.184 +82 0.055 38 0.097 -14 -0.016 39 0.091 -13 -0.048 40 0.150 +41 0.014 41 0.062 -29 0.009 42 0.099 -25 -0.038 43 0.076 -37 -0.016 44 0.075 -38 -0.028 continued . .. I A2 Appendix A Julian day Target Horiz. velocity Deviation from Vertical velocity period md-1 seasonal vel., % m d-1 214-217 1 0.096 +14 0.008 2 0.025 -65 -0.005 3 0.061 +1 0.001 ii ! 4 0.068 -3 -0.047 5 0.029 -57 0.058 6 0.055 -12 0.012 7 0.067 +17 0.024 8 0.003 -65 -0.022 15 0.165 +77 -0.027 16 0.048 -53 -0.066 17 0~046 -55 -0.02 18 0.221 - +75 - 19 0.035 -68 -0.023 20 0.080 -22 -0.046 21 0.058 -43 -0.022 22 0.083 -28 -0.031 23 0.073 -31 0.036 216---222 30 0.062 -30 -0.027 31 0.037 -60 -0.068 32 0.077 -38 -0.078 33 0.05 1 -35 -0.063 34 0.070 -41 -0.078 35 0.074 -37 -0.060 36 0.077 -38 -0.100 37 0.116 +14 0.003 38 0.116 +2 0.003 39 0.074 -29 -0.053 40 0.073 -31 -0.106 41 0.079 -10 -0 .088 42 0.080 -39 -0.068 43 0.140 -52 -0.075 44 0.092 -25 -0.111 217~223 1 0.062 -26 -0.008 2 0.087 +19 0.008 3 0.023 -61 -0.051 4 0.034 -51 0.039 5 0.115 +69 0.033 6 0.037 -41 -0.012 7 0.044 -24 -0.071 8 0.054 -28 -0.022 15 0.045 -52 -0.051 16 0.102 -2 -0.007 17 0.109 +5 -0.056 18 0.128 -1 -0.001 19 0.128 +15 -0.021 20 0.104 +1 0.001 21 0.135 +31 0.015 22 0.147 +26 0.016 23 0.110 +3 0.012 continued . . . Ill[ Appendix A Julian day Target Horiz. velocity Deviation from Vertical velocity period md- 1 seasonal vel., % m a-1 222-224 30 0.090 +8 0.016 31 0.105 +26 0.012 32 0.148 +19 0.030 33 0.081 +3 0.032 34 0.118 +1 - 35 0.130 +9 0.015 36 0.119 -4 0.035 37 0.086 -15 0.028 38 0.089 -22 0.060 39 0.104 -1 -0.017 40 0.125 +18 0.016 41 0~087 - -1 -0.015 42 0.148 +11 0.013 43 0.129 +7 0.009 44 0.115 -6 0.009 223--225 1 0.101 +21 -0.009 2 0.077 +5 -0.048 3 0.068 +13 -0.059 4 0.077 +11 -0.059 5 0.067 -2 - 6 0.047 -26 0.008 7 0.062 +8 -0.011 8 0.088 +19 -0.015 15 0.099 +6 0.016 16 0.101 -3 0.047 17 0.107 +4 0.079 18 0.150 +17 -0.016 19 0.097 -13 -0.039 20 0.095 -8 -0.055 21 0.129 +25 -0.015 22 0.138 +19 0.033 23 0.105 -1 0.036 224-230 30 0.102 +14 0.006 31 0.117 +39 -0.039 32 0.121 -3 -0.059 33 0.086 +8 -0.060 34 0.102 -12 -0.086 35 0.094 -21 - 36 0.104 -15 -0.093 37 0.091 -10 -0.049 38 0.121 +7 0.016 39 0.094 -10 -0.071 40 0.115 +8 -0.035 41 0.089 0 -0.056 42 0.092 -30 -0.072 43 0.089 -26 -0.076 44 0.085 -30 -0.061 continued .. . · A4 Appendix A Julian day Target Horiz. velocity Deviation from Vertical velocity period m d- 1 seasonal vel., % md-1 225-231 1 0.023 -72 -0.029 2 0.024 -66 -0.014 3 0.015 -75 -0.019 4 0.004 -80 -0.007 5 - - - 6 0.024 -61 -0.014 7 0.033 -41 -0.007 8 0.046 -37 -0.032 15 0.068 -26 -0.031 16 0.069 -33 -0.020 17 0;069 -33 -0.038 18 0.079 -38 -0.046 19 0.070 -37 -0.031 20 0.068 -34 -0.024 21 0.060 -41 -0.025 22 0.079 -31 -0.033 23 0.064 -39 -0.003 230-233 30 0.061 -32 0.015 31 0.019 -77 -0.046 32 0.064 -48 -0.009 33 0.076 -4 0.054 34 0.105 -10 -0.069 35 0.071 -40 -0.039 36 0.047 -61 -0.019 37 0.083 -17 -0.100 38 0.127 +12 0.033 39 0.109 +4 0.070 40 0.053 -50 -0.020 41 0.080 -9 0.018 42 0.101 -24 -0.007 43 0.082 -26 -0.034 44 0.077 -36 -0.009 231-234 1 0.113 +34 0.054 2 0.134 +83 0.058 3 0.134 +90 0.007 4 0.125 +78 0.008 5 0.114 +67 -0.037 6 0.154 +81 -0.039 7 0.114 +96 0.035 8 0.117 +58 0.005 15 0.137 +44 0.018 16 0.197 +89 0.023 17 0.199 +93 0.021 18 0.232 +79 0.009 19 0.183 +64 0.014 20 0.197 +91 -0.075 21 0.164 +59 0.004 22 0.125 +7 -0.004 23 0.186 +77 -0.040 Appendix A 1990 Julian day Target Horiz. velocity Deviation from Vertical velocity period m d- 1 seasonal vel., % md-1 165-172 1 0.021 -16 0.001 2 0.063 +33 -0.005 3 0.067 +17 -0.039 4 0.028 -31 -0.060 5 0.061 -47 -0.037 6 - - - 7 0.092 +2 0.005 8 0.046 -57 -0.035 9 0.082 -1 -0.041 10 0.05} -44 -0.063 11 0.069 -46 -0.007 12 0.063 -55 -0.023 13 0.077 -8 -0.035 172- 174 1 - - - 2 0.015 -8 -0.054 3 0.061 +5 0.015 4 0.087 +78 0.078 5 0.182 +56 0.041 6 - - - 7 0.062 -33 -0.051 8 0.189 +76 0.048 9 0.080 -3 0.024 10 0.140 +49 0.082 11 0.110 -14 0.029 12 0.240 +70 0.083 13 0.081 -2 0.089 174-176 1 0.023 +8 -0.001 2 0.056 +16 -0.05 3 0.098 +77 0.001 4 0.071 -14 -0.050 5 0.216 +86 -0.043 6 - - - 7 0.083 -11 -0.021 8 0.175 +64 -0.062 9 0.105 +27 -0.015 10 0.103 +12 -0.037 11 0.195 +52 -0.045 12 0.207 +47 -0.023 13 0.750 -10 -0.018 continued . .. A6 Appendix A Julian day Target Horiz. velocity Deviation from Vertical velocity period m d-1 seasonal vel., % m d-1 176-181 1 0.023 -8 -0.005 2 0.014 -10 -0.013 3 0.038 -36 0.011 4 0.041 0 0.006 5 0.057 -50 0.030 6 0.079 +1 -0.013 7 0.103 +10 0.021 8 0.055 -48 -0.045 9 0.046 -45 0.000 10 0.107 +15 0.021 11 0.061 -48 0.042 12 0.067 -52 0.032 13 0.084 +1 0.002 181-184 1 0.027 +8 -0.010 2 0.080 +66 -0.014 3 0.039 -32 0.011 4 0.049 +19 -0.049 5 0.083 -28 -0.031 6 0.070 -11 0.083 7 0.052 -45 -0.039 8 0.059 -44 0.001 9 0.064 -23 0.012 10 0.055 -39 -0.016 11 0.056 -54 -0.01 2 12 0.072 -49 -0.043 13 0.052 -37 -0.079 A7 APPENDIX B Section 6.5 examined the amount of calculated basal motion occurring beneath Bjuvbreen. Basal velocities were computed using equation 6.2 assuming that the flow law parameter A was equal to 5·3 x lQ-15 s-1 kPa-3. This represents ice at O °C. Because of uncertainties in the actual temperature of ice in Bjuvbreen (section 5.5) basal velocities were re-calculated using values of A appropriate for ice temperatures of -5 °C (A = 1·7 x lQ-15 s-1 kPa-3) and -10 °C (A = 52 x 10-16 s-1 kPa-3) . The results are given in the tables below. The data used in the calculations were observed seasonal surface velocities (u), basal shear stresses ( TB) computed using the Kamb-Echelmeyer method (section 5.3.4) and ice thicknesses (h) obtained from radio echo sounding and hot water drilling. The flow law constant, n, was taken to be 3. Basal velocities were computed for targets located on the glacier centreline during the 1989 and 1990 field seasons. The location of these targets is illustrated in Figure 4.5. 1989 Target u (m d-1) TB (kPa) h (m) UB (m d-1) UB (m d-1) UB (m d-1) 5·3 X lQ-lS 1 ·7 X lQ-15 5·2 X lQ-lG 4 0-070 90-6 18 0-067 0-069 0-069 19 0-111 98-8 76·5 0-094 0-106 0-109 30 0-089 84-7 79.3 0-078 0-085 0 -089 34 0-117 83-5 92· l 0-105 0-113 0-116 38 0-092 117 · l 128-4 0-045 0-077 0-087 42 0-133 129-7 132-9 0-067 0-112 0-126 1990 Target u (m d-1) TB (kPa) h (m) UB (m d-1 ) UB (m d-1) UB (m d-1) 5·3 X lQ-lS 1 ·7 , X 1Q-l5 5·2 X lQ-lG 1 0-025 60·7 12 0-024 0-025 0 ·025 3 0-064 91 · 1 18 0-061 0·063 0·064 6 · 0-076 94·6 75·6 0-061 0-071 0-074 9 0-083 74.4 84 0-075 0-080 0-082 11 0-128 117•4 126 0-081 0-113 0-123 13 0-083 131-6 128 0-016 0-062 0-076 A8 CAMBRIDGE UNIVERSITY LIBRARY Attention is drawn to the fact that the copyright of this dissertation rests with its author. This copy of the dissertation has been supplied on condition that anyone who consults it is understood to recognise that its copyright rests with its author. In accordance with the Law of Copyright no information derived from the dissertation or quotation from it may be published without full acknowledgement of the source being made nor any substantial extract from the dissertation published without the author's written consent.