1 
EX-SITU CONSERVATION OF PLANT DIVERSITY IN THE WORLD’S BOTANIC GARDENS  1 
Ross Mounce1, Paul Smith2 & Samuel Brockington1 2 
 3 
1Department of Plant Sciences, Tennis Court Road, Cambridge, CB23EA, UK 4 
2Botanic Gardens Conservation International, Descanso House, 199 Kew Road, Richmond, Surrey TW9 5 
3BW, UK  6 
 7 
ABSTRACT 8 
Botanic gardens conserve plant diversity ex-situ and can prevent extinction through integrated 9 
conservation action. Here we quantify how that diversity is conserved in ex-situ collections across the 10 
world’s botanic gardens. We reveal that botanic gardens manage at least 105,634 species, equating to 30% 11 
of all plant species diversity, and conserve over 41% of known threatened species. However, we also 12 
reveal that botanic gardens are disproportionately temperate, with 93% of species in the northern 13 
hemisphere. Consequently, an estimated 76% of species absent from living collections are tropical in 14 
origin. Furthermore, phylogenetic bias ensures that over 50% of vascular genera, but barely 5% of non-15 
vascular genera, are conserved ex-situ. While botanic gardens are discernibly responding to the threat of 16 
species extinction, just 10% of network capacity is devoted to threatened species. We conclude that botanic 17 
gardens play a fundamental role in plant conservation, but identify actions to enhance future conservation 18 
of biodiversity. 19 
INTRODUCTION 20 
Plants are essential for life, capturing solar energy, and creating the biomass that underpins the biosphere. 21 
Plants underpin ecological processes such as climate regulation, carbon dioxide absorption, soil fertility 22 
and the purification of water and air 1, and provide the food, medicines, building materials and fuel that 23 
sustain human life. Yet an estimated 20% of plant diversity is threatened with extinction 2. The extinction 24 
threat is largely anthropogenic, including habitat degradation, invasive species, resource over-exploitation, 25 
and climate change 3. It is estimated that 75% of the planet’s land surface is experiencing human pressures 26 
such as expansion of built environments4, with approximately 40% given to agriculture 5. Even in 27 
wilderness areas, plant populations are vulnerable to invasive species, pests, diseases and a changing 28 
climate 6. For plants with natural distributions within transformed environments, ex-situ conservation may 29 
be the only way they can survive in the short, medium and even long-term7. Crucially, threatened plant 30 
diversity may also hold the key to solving our major challenges in areas of food security, energy 31 
availability, water scarcity, climate change, and habitat degradation8.  32 
Botanic gardens are managed for many purposes, but offer the opportunity to conserve plant diversity ex-33 
situ, and have a major role in preventing species extinctions through integrated conservation action 7. 34 
Recognising the unique position of botanic gardens for plant conservation, the first Botanic Gardens 35 
Conservation Strategy was published in 1989, developing the role of botanic gardens in conservation 36 
throughout the 1990’s 8. Then, in 1998, Botanic Gardens Conservation International (BGCI), a consortium 37 
of 800 botanic gardens in >100 countries, launched an international consultation process to update the 38 
Strategy, taking into account the Convention on Biological Diversity (CBD). The consultation culminated 39 
in the adoption of the Global Strategy for Plant Conservation (GSPC), which seeks to halt the loss of plant 40 
diversity and to secure a sustainable future where human activities support plant diversity, and where the 41 
diversity of plants support human livelihoods and well-being 9. The strategy outlines sixteen targets 42 
encompassing knowledge, conservation, sustainable use, awareness and capacity building activities. 43 
Botanic gardens contribute to meeting all targets, but as the main institutions for ex-situ plant conservation, 44 
are key to achieving GSPC Target 8, which calls for “at least 75% of threatened plant species in ex-situ 45 
collections, preferably in the country of origin, and at least 20% available for recovery and restoration 46 
programmes by 2020.”  47 
BGCI recently published its vision for a botanic garden-centered, cost-effective, rational global system for 48 
the conservation and management of all plant diversity 10. Two assertions lie at the core of the central role 49 
of botanic gardens in the conservation and management of plant diversity. First, that there is no technical 50 
reason why plant species should become extinct, given the array of ex-situ and in-situ conservation 51 
2 
techniques such as seed banking, cultivation, tissue culture, assisted migration, species recovery, and 52 
ecological restoration 11,12. And second, that as a professional community, botanic gardens possess a 53 
unique skill set that encompasses finding, identifying, collecting, conserving and growing plant diversity 54 
across the taxonomic spectrum 10. While it is difficult to prove a plant species cannot be conserved 55 
vegetatively or as seed, it is possible to evaluate the potential for ex-situ conservation by assessing the 56 
extent of the plant diversity, including threatened species, that botanic gardens are already conserving and 57 
managing ex-situ.   58 
In this paper, we explore how plant diversity is currently conserved across the world’s botanic gardens, 59 
and how well botanic gardens are performing with respect to plant conservation priorities. We define the 60 
extent of the global network, and examine biases in the distribution of botanic gardens and the availability 61 
of digitised collection data. We estimate the minimum holdings of the global network of botanic gardens 62 
with respect to plant diversity, determine the impact of the biogeographic distribution of botanic gardens 63 
for conservation goals, and identify significant biogeographic and phylogenetic gaps in ex-situ collections. 64 
Finally, we quantify the number of threatened species within ex-situ collections and assess whether the 65 
global network of botanic gardens is discernibly responding to the threat of species extinction. We 66 
conclude by discussing how to build on these findings to further engineer a botanic garden-centered global 67 
system that can prevent species extinctions in perpetuity. 68 
 69 
RESULTS AND DISCUSSION 70 
 71 
Quantifying the Extent And Content of Botanic Gardens 72 
To evaluate the geographic extent of the botanic garden network, and the degree to which digital collection 73 
data is available, we applied the most widely accepted definition of a botanic garden, as an institution 74 
‘holding documented collections of living plants for the purposes of scientific research, conservation, 75 
display and education’ 9. BGCI have accumulated data on botanical institutions and have assembled a 76 
digital directory of the world’s botanic gardens within a database called ‘GardenSearch’ 77 
(https://www.bgci.org/garden_search.php). Applying this definition to the ‘GardenSearch’ database, we 78 
estimated that there are over 3269 botanical collections in 180 countries around the world (BGCI, 2012) 79 
(Fig. 1a). Of these 3269 institutions, BGCI has amassed collection data from 34% or 1,116 institutions, in 80 
the ‘PlantSearch’ database (https://www.bgci.org/plant_search.php), the most comprehensive list of 81 
botanic garden accession names, containing 1,330,829 records of 481,696 taxon names. We analysed the 82 
PlantSearch database set against the most comprehensive list of plant taxa, ‘The Plant List’, and applied 83 
rigorous cleaning to these 481,696 ‘PlantSearch’ taxa, removing invalid taxon names, deceased 84 
accessions, and horticultural cultivars. We can only present a minimum estimate of the diversity held in 85 
botanic gardens and associated seed banks, as our digitised data is derived from one third of documented 86 
botanic gardens within the GardenSearch database (See Fig. 1b). But we show that, of the 350,699 87 
accepted plant species (TPL 2013), 105,634 or 30% are held within the living collections of the global 88 
botanic garden network (Fig. 2a). These numbers equate to 59% of all plant genera (Fig. 2b), 75% of all 89 
embryophyte plant families (Fig. 2c) and 93% of tracheophyte plant families (Fig. 2d), indicating a 90 
remarkable degree of taxonomic coverage within ex-situ collections (Supplementary Table 1).  91 
Biogeographic Distribution of Ex-Situ Collections and Data 92 
The relative number of species records in each of the 1,116 BGCI member institutions, is depicted in Fig. 93 
1B where the diameter of each bubble is scaled to the number of species recorded at an institution. It is 94 
evident that there are biases both in the distribution of botanic gardens (Fig. 1a), and the extent to which 95 
the data that has been uploaded to the ‘PlantSearch’ database (Fig. 1b). The absence of digital data does 96 
not necessarily equate to species absence, but in evaluating global targets and defining species 97 
conservation priorities, absence of a species and absence of data can be an equivalent problem, and here 98 
they are treated in the same way. Fig. 1A and 1B show that the most dominant world-wide bias in the 99 
distribution of botanic gardens, and availability of associated digitised collection data, is a phenomenon 100 
termed positive latitudinal bias13. Several countries in the southern hemisphere, such as South Africa, 101 
Australia, and New Zealand, are major contributors of digital collection data. Still, 91% of recorded 102 
accessions, and 93% of recorded species are documented from ex-situ collections in the northern 103 
hemisphere (Fig. 3a). This bias is due to the primary determinants of the geographical distribution of 104 
botanic gardens and species richness in botanic gardens, including socioeconomic factors such as GDP and 105 
metropolitan population size14. But although explicable, it remains essential that biogeographic gaps in 106 
3 
digital collection data are filled, to provide the robust cyber-infrastructure needed for coordinated ex-situ 107 
plant conservation.  108 
A positive latitudinal gradient, where botanic garden species diversity increases in temperate latitudes, runs 109 
counter to natural latitudinal gradients, where tropical ecosystems harbour the bulk of plant species 110 
diversity15. The consequences of this skewed latitudinal distribution of botanic gardens (Fig. 3A) for plant 111 
conservation has not been quantified on a global scale. Here we made that assessment, asking how the 112 
latitudinal distribution of a species affects the likelihood of its representation within the botanic garden 113 
network. We retrieved species occurrence data for 236,904 accepted plant species, calculated the median 114 
of the latitudinal range for each species, cross-referenced these data with recorded presence or absence of 115 
within the botanic garden network, and visualized these data in Fig. 3B (Supplementary Table 2). We then 116 
refined the dataset to species with at least five geo-referenced occurrences, whose latitudinal range is either 117 
temperate or tropical. Analysis of these tropical and temperate splits, showed that a temperate species has a 118 
60% probability of ex-situ cultivation in the botanic garden network, but just 25% for a tropical species. 119 
Indeed from this dataset, 66,905 or 76% of species absent from the botanic garden network, are tropical 120 
species. On the one hand, to harbor 60% of all the temperate species in our dataset, reveals the 121 
extraordinary capacity of the world’s botanic gardens. But on the other hand, ex-situ conservation of 122 
tropical taxa in temperate climates is unfeasible on a scale that is meaningful for conservation, in part due 123 
to limited space and high energy costs of glasshouses. Given the shortage of data from tropical regions, the 124 
tropical-temperate disjunction may not be as severe as we imply here, but it is clearly vital that the 125 
temperate network, with its associated conservation skills and resources, is extended to tropical latitudes, 126 
where many of the world’s conservation priorities lie. 127 
Identifying and Targeting Under-Represented Lineages 128 
We then refined our understanding of how phylogenetic diversity is captured. We mapped all 10,133 129 
genera, known to be represented in botanic gardens by at least one species, on a genus-level phylogenetic 130 
tree comprising 14,126 genera or 83.5% of all accepted land plant genera16. These results, depicted in Fig. 131 
4, reveal striking macroscopic biases in ex-situ conservation of the land plant phylogeny. Whereas 132 
angiosperms, gymnosperms, and ferns enjoy 62.8%, 96.6% and 54.0% generic coverage respectively, the 133 
non-vascular early-diverging land plant lineages - Bryophyta, Marchantiophyta, Anthocerotophyta  - are 134 
almost completely undocumented with less than 5% generic coverage across the global botanic garden 135 
network. Our visualization of this disparity is stark, revealing a weakness in the delivery of ex-situ 136 
conservation goals for the plant kingdom as a whole. The lack of coverage for ‘Bryophyte’ taxa denies 137 
their importance, as they represent key stages in land plant evolution, occur in endangered habitats such as 138 
peatland 17, host diverse microbiota 18, and play a central role in nutrient cycling 19. Given the vascular 139 
plant emphasis of botanic gardens, this finding is unsurprising, however the magnitude of deficit calls for 140 
action. Many living collections host incidental collections of ‘Bryophytes’, and an increase in 'Bryophyte’ 141 
representation could be achieved by documenting existing taxa, as well as through specific acquisition 142 
strategies and horticultural innovation. 143 
Of the 34 missing vascular plants families, twelve are monotypic and thirteen monogeneric, with the 144 
majority restricted endemics, tropical trees, or parasites (Supplementary Table 3), indicating how species 145 
paucity, endemism, and life history can limit ex-situ conservation. The cultivation of certain plants can 146 
pose a challenge, and this may be especially true for the estimated 4000 species of parasitic angiosperms 20 147 
However, below the rank of family, phylogenetic mapping provides a framework to target acquisitions to 148 
fill collection gaps. We exemplify this idea using two approaches. First, for all missing genera, we 149 
calculated the amount of evolutionary distinctiveness (ED; Isaac et al 2007) represented by each genus. 150 
We then ranked all genera according to the amount of ED that would be captured, if each genus was 151 
accessioned into ex-situ collections (Supplementary Table 4). Here, it is notable that many of the most 152 
important genera are also from early diverging land plant lineages, emphasizing the importance of 153 
conserving these taxa. In a second approach, we computationally searched for clusters of closely related 154 
but absent genera, below the taxonomic rank of family, to identify phylogenetic islands of evolutionary 155 
history, not captured within ex-situ collections. We list the top ten clusters in terms of numbers of absent 156 
genera e.g. the Grammitioideae, a subfamily of the fern family Polypodiaceae, of tropical distribution, with 157 
thirteen out of sixteen (81%) genera missing, and the Helieae tribe, within Gentianaceae, which occupy 158 
highly restricted ranges in the New World, with ten out of twelve (83%) of genera missing (Supplementary 159 
Table 5). Most absent clusters are tropical, emphasizing that latitudinal bias impacts on phylogenetic 160 
representation.  161 
4 
Through these gap analyses, we have generated resources that enable targeted acquisition, including a list 162 
of genera missing from gardens (Supplemental Table 6), and a list of all families ranked by their 163 
percentage of genera represented (Supplemental Table 7). Targeted acquisition strategies have potential to 164 
enhance the value of ex-situ collections, not just for conservation, but for research and education more 165 
generally. For example, comparative genomics depend on ready access to living material to sequence 166 
phylogenetically pertinent taxa, and cultivation of key phylogenetic lineages can provide essential material 167 
to teach evolutionary transitions. However, phylogenetically targeted strategies are just one approach to 168 
enhance the value of living collections, and future studies should also explore under-representation of 169 
environmental niches, life histories, and medicinal, ethnobotanical or crop plants. 170 
Evaluating Progress Towards GSPC Target 8 171 
BGCI ‘ThreatSearch’ database, is the most comprehensive list of threatened plants, incorporating global, 172 
regional and national threat assessments (https://www.bgci.org/threat_search.php). Here, ‘Threatened’ is 173 
defined as species, which fall into the categories of ‘Vulnerable’, ‘Endangered’, and ‘Critically 174 
Endangered’, as per IUCN criteria, or their equivalent designations, in the case of non-IUCN 175 
methodologies. By cross-referencing two data sources, an early release version of the ‘ThreatSearch’ 176 
database and BGCI ‘PlantSearch’, we assessed progress towards achieving GSPC Target 8, which calls for 177 
“at least 75% of threatened plant species in ex-situ collections, preferably in the country of origin”, First, 178 
we asked how many threatened species are present in the global network of botanic gardens and show that, 179 
currently, the global network is over half way towards achieving GSPC Target 8, with about 13,218 180 
threatened species held in at least one ex-situ collection, equating to 41.6% of all plant species assessed as 181 
threatened (Fig. 5A). As with the total diversity estimates, our figures are likely an underestimate of 182 
threatened plant diversity held in botanic gardens, as only a third of gardens are analysed here (Fig. 5B). 183 
Unsurprisingly, the extent to which ex-situ collections contribute to these overall numbers varies 184 
considerably, from as little as one threatened species, to over five thousand, with a median number of 185 
threatened species per garden of 38 (Fig. 5C). Nonetheless, these figures are impressive, as threatened 186 
species are often range-restricted, harder to find, and more difficult to cultivate and manage in ex-situ 187 
collections. Although over 41% of all threatened species are currently held in ex-situ collections, there is 188 
scope to improve these global efforts. Of the 1,330,829 records in ‘PlantSearch’, 134,771 or about 10% 189 
are threatened species, with 90% of ex-situ collections devoted to species not yet identified to be at risk of 190 
extinction. If the network can hold over 41% of threatened species, with just 10% of current network 191 
capacity, there is potential to hold a greater proportion of threatened species. Furthermore, if ex-situ 192 
collections of threatened species are to be of value for in-situ restoration programs, it is imperative that 193 
large populations are maintained ex-situ to provide the necessary intra-specific genetic diversity for viable 194 
populations and species recovery. Such a goal will require the network to devote more collection capacity 195 
to conservation priorities.  196 
Evaluation of GSPC Target 8 is problematic as it calls only for a percentage of threatened plants to be 197 
represented in ex-situ collections, and yet the focus of the threat assessments varies considerably across the 198 
plant phylogeny. For example, of the 89,810 assessed species in our BGCI ‘ThreatSearch’ dataset, 80,990 199 
species of angiosperms (26%) have been assessed for extinction risk, compared with 3611 pteridophyte 200 
species (34.4%), 4303 bryophyte species (12.2%), and 986 gymnosperm species (89.3%). In the context of 201 
a variable number of assessments and hence threatened species across major lineages, conserving a 202 
percentage varies in its significance. But with respect to GSPC Target 8, only gymnosperms meet the 203 
target threshold, with 89% of threatened species held ex-situ (Fig. 5D). Gymnosperms are a successful ex-204 
situ conservation story as: they are the least speciose of the major plant lineages rendering the percentage 205 
based GSPC Target 8 more feasible; they have an international conifer conservation programme; like most 206 
botanic gardens are broadly temperate, and; they have horticultural value as evergreen collections. In stark 207 
contrast, the bryophytes, which have the poorest overall assessment rate of 12.2%, are similarly 208 
impoverished with respect to ex-situ conservation, such that only 2.6% of threatened bryophytes are 209 
documented in the botanic garden network. Evidently, poor performance of ex-situ collections with respect 210 
to non-vascular plants will further undermine ex-situ conservation goals for these important but under-211 
represented plant groups.  212 
We then sought to evaluate progress towards the clause in GSPC Target 8, which asks that threatened 213 
plants should be held “preferably in the country of origin”. Here, we mapped the ex-situ location of all 214 
globally and regionally threatened plants within ‘ThreatSearch’. As visualised in Fig. 5E, a relatively small 215 
number of nations are holding an exceptional number of threatened species, consistent with the skewed 216 
5 
distribution of botanic gardens. Furthermore, using a set of IUCN-assessed threatened endemic species we 217 
found that 2780 country-endemic, threatened species are present in the botanic garden network with 1231 218 
or 44% are held in ex-situ collections within their country of origin, and 56% or 1549 species are only held 219 
in ex-situ collections outside of their country of origin (Supplementary Table 8). While dispersed 220 
collections provide some security against extinction, if endemic species are held solely outside of their 221 
natural range, it seems less likely that they will be available for species recovery, and again, large ex-situ 222 
populations are needed to provide genetic diversity for viable populations. 223 
Measuring Response to Species Extinction Risk  224 
Threatened species lists are established tools that provide a scaled assessment of extinction risk, which can 225 
guide conservation actions 21.While scale of threat is not sufficient to define priorities21, if botanic gardens 226 
are actively responding to perceived extinction risk, one might find signal of this response within 227 
collections themselves. Here, we looked for evidence of that response using a dataset of IUCN-globally 228 
assessed species.  Ideally this question would be answered by a time series analysis, however the present 229 
study is the first global assessment of ex-situ conservation for threatened plant species, and as such, there is 230 
no historic data against which to compare. Consequently, to address this question here, we first asked 231 
whether threatened species at a higher risk of extinction were more likely to be found in at least one ex-situ 232 
living collection. We found that 39% of critically endangered species were held in ex-situ collections 233 
compared with 35% of endangered species, and 27% of vulnerable species, indicating that a greater 234 
proportion of higher risk species are held within the botanic garden network (Fig. 6A). Here, the relative 235 
proportion of each red list category held by botanical gardens differs significantly from the proportions 236 
held on the red list (X22 = 76.67, Nobs, = 3454, p<0.01) suggesting an active response to increasing threat 237 
status for threatened species, as a whole. We then assessed whether threatened species at a higher 238 
extinction risk were more likely to be accessioned multiple times across the botanic garden network. Here, 239 
we found that 11% of IUCN red-listed species, were documented in just one institution, with a median 240 
representation of three. But we found that there was no relationship between elevated extinction risk, and 241 
the number of institutions that hold any given threatened species (X220 = 28.63, Nobs, =3454, p>0.05) (Fig. 242 
6B), a result that suggests no coordinated shared global response to the extinction risk posed to individual 243 
species.   244 
A signal of a global response to extinction risk is confounded by the fact that only a small fraction of 245 
capacity, 10%, is currently devoted specifically to conservation. Furthermore, most IUCN globally 246 
assessed species are centred in the tropics (Fig. 6C), and as global collections are deficient in tropical 247 
species, a tropical-temperate disjunction could underestimate any response signal. We therefore explored 248 
whether threatened species were more likely to be included in the botanic garden network if they were 249 
temperate in origin, rather than tropical, see Fig. 6C. Here we used a dataset of globally assessed 250 
threatened species with at least five geo-referenced occurrences, which had a latitudinal range that is either 251 
temperate or tropical (Supplementary Table 9). We find that the probability of ex-situ conservation for a 252 
globally threatened temperate species is 77% (a 17% increase relative to temperate species as a whole), but 253 
probability of ex-situ conservation for a tropical species fell to 24% (a 1% drop relative to tropical species 254 
as a whole). These findings suggest a differential response to threatened plants in temperate versus tropical 255 
environments. We further found that the odds of conservation of temperate threatened species is 1.8 times 256 
that of a near-threatened temperate species (p<0.01), but the odds of conservation of threatened tropical 257 
species is 0.35 times that of a near-threatened tropical species (p<0.001). Together these analyses indicate 258 
that botanic gardens are discernibly responding to threatened temperate species, but less so for threatened 259 
tropical species.  260 
CONCLUSIONS 261 
The global network of botanic gardens conserves an astonishing array of plant diversity, holding 105,634 262 
species, equating to 30% of species diversity, 59% of plant genera, 75% of land plant families, and 93% of 263 
all vascular plant families. These numbers are all the more remarkable as they represent a minimum 264 
estimate, based on data derived from just one third of botanic gardens worldwide. Such numbers 265 
emphasize that botanic gardens possess unique skills for conserving plant diversity across the taxonomic 266 
spectrum. Furthermore, botanic gardens are discernibly responding to the threat of species extinctions, 267 
housing at least 13,218 species at risk of extinction, equating to just over 41% of the world’s known 268 
threatened flora. 269 
6 
However, our analyses reveal substantial biogeographic gaps in the representation of collections, with 93% 270 
of species occurring in the northern hemisphere. So it is essential that the network continue to incorporate 271 
institutions and collection data, particularly from tropical regions, but also from under-represented 272 
countries. The network is poorly positioned to protect tropical species, and substantial capacity building is 273 
needed here, as outlined in previous publications10-12. For example, an accessible cyber-infrastructure will 274 
be vital to collectively manage ex-situ conservation of the world’s plant diversity.  Importantly, the current 275 
global cyber-infrastructure in the form of PlantSearch is limited to taxon-level data, however effective ex-276 
situ conservation depends on high intra-specific diversity, and for this, individual accession-level data are 277 
needed. 278 
Only 10% of collections are dedicated to threatened species, and, to limit species extinction, it is essential 279 
that our full capacity is directed towards our most threatened plant species. Multiple accessions of 280 
threatened species across the network will buffer against loss of threatened species, and provide genetic 281 
diversity for ecological restoration efforts. However, 11% of globally threatened species are currently held 282 
in just one institution.  Moreover, over half of endemic threatened species are not held ex-situ within their 283 
country of origin, implying reduced availability for ecological or species restoration. Many threatened 284 
species have utility in agriculture, horticulture and forestry, with species reintroduction an important 285 
element of conservation work22-24. Botanic gardens must engage with these organizations and industries 286 
with responsibility for plant diversity in the natural landscape. Finally, it is important that coordinated 287 
international conservation of threatened species continues in the face of legislation that seeks to enforce the 288 
intellectual property rights of individual nations. 289 
Without deep sustained public support, the plant conservation movement will struggle. Fortunately, public-290 
facing botanic gardens are typically near urban areas 14, and according to data within the GardenSearch 291 
database, collectively host 500 million visitors annually. Consequently, botanic gardens can deliver the 292 
necessary education, citizen science, and information to facilitate plant conservation action across the 293 
broader society. Given the quality of the collections, and their critical importance for conservation, it is 294 
vital that we speak to the strengths of the network, and promote its unique skills and resources to policy 295 
makers and funders. Despite impressive efforts by the world’s botanic gardens, substantial investment will 296 
be required to build a fully functioning, cost-effective, rational global system for the conservation of 297 
threatened plant diversity, that can prevent species extinctions in perpetuity10. 298 
AUTHORS FOR CORRESPONDANCE 299 
Correspondence to Samuel Brockington or Paul Smith 300 
ACKNOWLEDGEMENTS 301 
We thank Nathanael Walker-Hale and Matt Castle for statistical help, Richard Smith-Unna for help with 302 
programming, Meirion Jones for help with BGCI databases, Malin Rivers for access to an early release 303 
version of the BGCI ‘ThreatSearch’ data, and Monika Bohm for compiling initial national conservation 304 
assessments that went into BGCI ‘ThreatSearch’. We thank the Brockington Lab, Nik Cunniffe, Suzanne 305 
Sharrock, and BGCI staff for useful discussion. We acknowledge the Cambridge University Botanic 306 
Garden and the National Environmental Research Council for financial support to SFB. 307 
AUTHOR CONTRIBUTIONS 308 
SFB and PS conceived the study, PS released the data, RM cleaned the data, SFB designed the analyses, 309 
RM and SFB performed the analyses, and SFB and PS wrote the manuscript. 310 
FIGURE LEGENDS 311 
Figure 1. Global distribution of ex-situ plant collections and the availability of data for the contents 312 
of these ex-situ collections. Equirectangular projection maps demonstrating (A) the location of all BGCI 313 
member institutions (B) the relative species diversity present in each of the 1,116 BGCI member 314 
institutions that share plant record data with BGCI. The diameter of each bubble is scaled to the number of 315 
species recorded at the institution (Data from BGCI ‘GardenSearch’ and BGCI ‘PlantSearch’). 316 
Figure 2. Botanic garden taxon coverage in terms of (A) all accepted land plant species names (out of 317 
350,699) (B) all land plant genera (out of 16,913) (C) all land plant families (out of 635) (D) all vascular 318 
plant families (out of 458). 319 
7 
Figure 3. Latitudinal distribution of (A) Ex-situ plant collections and the availability of data for the 320 
contents of these ex-situ collections with the number of gardens per latitudinal bin (gray, bottom y-axis) 321 
and number of digitally recorded species per latitudinal bin (red, top y-axis) (B) the latitudinal distribution 322 
of plant species (n=236,904) as recorded by the median latitude of all georeferenced GBIF records per 323 
species, with data binned per latitudinal degree (gray, top y-axis), the percentage of species found in the 324 
botanic garden network per latitudinal degree (red, bottom y-axis). 325 
Figure 4. Phylogenetic gap analysis showing land plant genus-level phylogeny16, where red edges 326 
indicate that all subtending edges and tips are present in the botanic garden network. 327 
Figure 5. Threatened land plant species in botanic collections. (A) the percentage of threatened plants 328 
held in ex-situ collections (out of 34,442) (B) the percentage of total accessions held ex-situ that are 329 
threatened species accessions (C) absolute numbers of threatened species per garden (D) the percentage of 330 
threatened species held by botanic collections by higher-level phylogenetic lineages (ANG: Angiosperms; 331 
GYM: Gymnosperms; PTE: Pteridophytes; BRY: Bryophytes). (E) Number of documented threatened 332 
species in PlantSearch, held ex-situ, per country. 333 
Figure 6. Presence and absence of IUCN red-list threatened plants in ex-situ collections (A) the 334 
percentage of threatened species per threat status (B) the number of different ex-situ collections that a 335 
threatened species is held in, with Lg2 scale. Yellow for Vulnerable (VU), orange for Endangered (EN), 336 
and red for Critically endangered (CR) (C) the native distribution of just threatened plant species (as 337 
opposed to all species as shown in Fig. 3B) as recorded by the median latitude of all georeferenced GBIF 338 
records per species (n=8619), with data binned per latitudinal degree (gray, top y-axis), the percentage of 339 
threatened species found in the botanic garden network per latitudinal degree (red, bottom y-axis). 340 
METHODOLOGY 341 
Data Sources: We used BGCI ‘GardenSearch’ (www.bgci.org/garden_search.php) database (accessed 342 
2016-01-01) for the location of botanic gardens. For the presence and absence of taxa from gardens we 343 
used BGCI ‘PlantSearch’ (www.bgci.org/plant_search.php) (accessed  2016-01-01). For threatened plants 344 
we used a pre-release version of BGCI's ThreatSearch (https://www.bgci.org/threat_search.php) (accessed  345 
2016-01-01). The pre-release set of threat assessments included the official IUCN red list version 2015-4  346 
(www.iucnredlist.org) as well as the following additional regional and national lists:  Chinese Higher 347 
Plants Red List, NatureServe, Mexico Red List, Mesoamerica Red List, Brazil Tree Red List, Ecuador Red 348 
List, Threatened Plants of the Philippines, Ethiopia Eritrea RL, Andes Red List, Cuba Red List, Guatemala 349 
Red List, Caucasus Red List, Central Asia Red List, Trinidad and Tobago Red List, Vietnam Red Data 350 
Book Part II: Plants, South African Plants SANBI, South Africa Trees, Sao Tome trees list, Trees of 351 
Uganda, Red List of Korean Endemic Vascular Plants, Namibian Tree List, Malaysian Flora Database, and 352 
the Bolivian Red Book. For some analyses such as response to extinction we only used a subset of BGCI's 353 
‘ThreatSearch’, namely only the global assessments derived from the official IUCN red list version 2015-354 
4.  355 
Data Cleaning: For all datasets, records were filtered to remove assessments of taxa that were not land 356 
plants e.g fungal, algal, and animal taxa. Undescribed taxa were ignored for these analyses e.g. “Asparagus 357 
sp. nov. A”. We discarded 'orphan' BGCI plant records that were not currently associated with any gardens 358 
in the network (e.g. historical records of dead plants that are no longer held in a garden). We interpret 359 
living collections to include accessions that are maintained as part of an active cultivation cycle, and so 360 
retained seed-banked accessions held within the botanic garden network. We discarded records of 361 
horticultural taxa such as cultivars, due to the difficulties of taxonomic standardisation, and because we 362 
were interested in true biological species. We computationally-normalised the taxonomy of records using 363 
the R package Taxonstand v1.8 25 version 1.8, so that all taxa match an accepted or unresolved taxon listed 364 
by The Plant List v1.1. Raw input species names that could not be automatically matched to a species 365 
name listed at The Plant List v1.1 were manually resolved to the correct species name. By matching to 366 
TPLv1.1 in a minority of cases we were back-converting names into older ones for the sake of consistency. 367 
BGCI records were de-duplicated using the R package stringdist 0.9.4.4 using Damerau-Levenshtein 368 
distance 26,27, so that there was only one record for each unique taxon, as gardens around the world can 369 
apply different names to the same taxon. After normalisation to The Plant List (TPL) some taxa were 370 
demoted from species rank in the original assessment to subspecies rank. For consistency and 371 
comparability only species-level taxa were retained for analysis, subspecies taxa were discarded. After 372 
these data processing steps we were left with: 105,634 BGCI recorded species of TPL-normalised land 373 
8 
plants and a pre-release version of BGCI ‘ThreatSearch’ comprising 89,810 assessed species and 31,812 374 
threatened species. The subset of global threat assessments comprised 20,367 IUCN global dataset species 375 
assessments of which 11,055 species were threatened. 376 
Biogeographic Bias Analyses: Using the R package rgbif version 0.9.7 we retrieved georeferenced 377 
occurrence data for 236,904 embryophyte species with at least one geo-referenced location record. The 378 
downloaded dataset equated to 8,246,424 unique geo-located records, with a mean of 34.8 records per 379 
species. Of these 236,904 species, 89,180 species were recorded as present in gardens, and 147,724 species 380 
were recorded as absent from gardens. We applied standard cleaning techniques to filter-out corrupt data 381 
indicated by coordinates that did not match the country stated on the record, or that had coordinates in 382 
marine areas. We then took the median of the latitudes for all georeferenced occurrences for each species, 383 
to serve as a proxy for the centre of a species’ latitudinal range. The median latitude of these 236,904 plant 384 
species was then binned per latitudinal degree and plotted against the percentage of these same species, 385 
from each latitudinal bin, that are found in the botanic garden network. To mitigate against the risk of 386 
errors in single geo-located records, we then refined the dataset to 171,472 species with at least five 387 
georeferenced occurrences, and then further refined this to the 148,682 species whose latitudinal range is 388 
either temperate or tropical, and does not span both tropical and temperate latitudes. Temperate species 389 
were defined as having their latitudinal range (min, max, median) entirely between 23.440N and 66.50N 390 
and between 23.440S and 66.50S. Tropical species were defined as having their latitudinal range (min, 391 
max, median) entirely within 23.440N and 23.440S. Using this refined dataset, the percentage of species 392 
present in gardens from each latitudinal bin were averaged across all tropical latitudinal bins (between 393 
23.437040N and 23.437040S) and compared with the average percentage across all temperate latitudinal 394 
bins (between 23.440N and 66.50N and between 23.440S and 66.50S). 395 
Phylogenetic Bias Analyses. To estimate the proportion of species, genera, embryophyte families and 396 
tracheophyte families held in ex-situ collections, we used denominators from the R package Taxonstand 397 
v1.8 i.e all species = 350,699; all genera = 16,913; all embryophyte families = 635; all vascular plant 398 
families =458. For phylogenetic mapping of presence and absence of genera, we used a genera-level 399 
phylogenetic tree comprising 14,126 genera or 83.5% of all accepted land plant genera 16, which provided 400 
maximal phylogenetic coverage at the generic level. We then plotted the 10133 genera known to be 401 
represented in botanic gardens, which were present in the tree by at least one species. We scored each 402 
genus tip on this tree as a binary trait according to whether the genus is documented as absent (0) or 403 
present (1) in a garden with the global network. To determine the significance of absence of genera in 404 
terms of evolutionary history, we utilized the branch length information from the tree 16 to report the 405 
Evolutionary Distinctiveness (ED)28 of each taxon in the tree, and ranked all missing genera according to 406 
ED. To detect notable clusters of absence within the large genus tree we employed an R script (available 407 
on request) to find the most absent clades in the tree with a cut off at 5 consecutive absent tips or more. 408 
Due to the wholesale absence of genera from early diverging lineages (Bryophyta, Marchantiophyta, 409 
Anthocerotophyta) the search for absent genera-level clusters was focussed solely on Tracheophyte 410 
lineages (Pteridophytes, Gymnosperms, Angiosperms). 411 
Threatened Species Representation: To estimate the total number of threatened species held in ex-situ 412 
collections, we used a pre-release version of BGCI ‘ThreatSearch’ (accessed 01/01/2016) cleaned to 413 
comprise 89,810 assessed species and 31,812 threatened species. To estimate the extent of the network 414 
capacity devoted to cultivating threatened species, we calculated the number of individual accessions of 415 
the 13,218 threatened species held in botanic gardens and expressed this as a fraction of the 1,330,829 416 
accession records held in BGCI ‘PlantSearch’. Total accession records were used as the denominator 417 
because including all taxa such as horticultural cultivars better represents the total capacity of the network, 418 
which could potentially be devoted to threatened species. We mapped the ex-situ location of all globally 419 
and regionally threatened plants within ‘ThreatSearch’ using R package ‘chloroplethr’ v3.6.1. The extent 420 
to which threatened plants are held in their country of origin was assessed using as set of 2780 IUCN 421 
globally threatened endemic species. Country-level endemicity was determined based on the IUCN data 422 
associated with each IUCN-RL assessment record. Endemics in this sense were coded as plants that are 423 
only documented to occur in one nation state according to the IUCN assessment. Presence or absence of 424 
these endemic species in ex-situ collections within their country of origin was then recorded and summed.  425 
Overall Response to Extinction Risk: For all assessments of response to extinction we used the official 426 
IUCN red list version 2015-4 (www.iucnredlist.org). We tested whether the relative abundances of 427 
critically endangered (CN), endangered (EN) and vulnerable (VU) species held by botanical gardens 428 
9 
differs significantly from the relative abundances in the IUCN red list. Here we employed an extrinsic chi-429 
squared test on the raw counts of observed number of species for each threat category held in botanic 430 
gardens versus expected number estimated from the IUCN red list. We use the term redundancy to 431 
describe when a species is held in more than one garden, such that a species that is held in more gardens 432 
exhibits greater redundancy. To determine whether there was a significant difference between the three 433 
levels of threat status (VU, EN, CR), with respect to redundancy, we represented redundancy as categorical 434 
binning from 0 to 10 gardens, and then aggregated all species redundancies in 11 to 100 gardens into a 435 
single category (>10). An intrinsic chi-squared test was then employed to assess whether there was 436 
significant independence between the three categories.  437 
Differential Response to Tropical versus Temperate Threatened Species: To test the response of ex-438 
situ conservation efforts to extinction risk in temperate versus tropical taxa, we used R package rgbif 439 
version 0.9.7 to retrieve georeferenced occurrence data IUCN threatened taxa, with at least one geo-440 
referenced location record. Geolocation data was retrieved for 8619 out of the 11,055 IUCN threatened 441 
species. We then took the median of the latitudes for all geo-referenced occurrences for each species, to 442 
serve as a proxy for the centre of a species’ latitudinal range. The median latitude of these 8619 species 443 
was then binned per latitudinal degree and plotted against the percentage of these same species, from each 444 
latitudinal bin, that are found in the botanic garden network.  To mitigate against the risk of errors in single 445 
geo-located records, we then refined the dataset to 5436 species with at least five geo-referenced 446 
occurrences, and then refined this to 4613 species whose latitudinal range is either temperate or tropical, 447 
and does not span both tropical and temperate latitudes, following the methodology outlined in the 448 
‘biogeographic bias analyses’ methodology section. Using this refined dataset, the percentage of 449 
threatened species present in gardens from each latitudinal bin were averaged across all tropical latitudinal 450 
bins (between 23.437040N and 23.437040S) and compared with the average percentage across all 451 
temperate latitudinal bins (between 23.440N and 66.50N and between 23.440S and 66.50S). To test the 452 
differential response of ex-situ conservation efforts to temperate versus tropical taxa, we implemented tests 453 
of odds ratios using the R packages ‘fmsb’ v0.6.1. We formed 2x2 contingency tables with conservation 454 
status (threatened or near-threatened) on rows and ex-situ conservation (present or absent) in columns, and 455 
calculated odds ratios, log odds ratios and associated Wald confidence intervals and p-values in R, using 456 
the ‘fmsb’ function oddsratio with p.calc.by.independence = F. 457 
Data availability: The core data sources that support the findings of this study, namely ‘ThreatSearch, 458 
PlantSearch, and GardenSearch’ were obtained from Botanic Garden Conservation International (BGCI) 459 
under a material transfer agreement. They are available from BGCI but restrictions apply to the availability 460 
of these data, and the relational use of these databases, which were used under license for the current study. 461 
Data are however available from BGCI upon reasonable request and with permission of Dr Paul Smith, 462 
Director-General of BGCI.  463 
REFERENCES 464 
 465 
1. Sukhdev, P., Wittmer, H. & Schröter-Schlaack, C. The economics of ecosystems and biodiversity: mainstreaming 466 
the economics of nature: a synthesis of the approach, conclusions and recommendations of TEEB. (2010). 467 
2. Brummitt, N. A. et al. Green Plants in the Red: A Baseline Global Assessment for the IUCN Sampled Red List 468 
Index for Plants. PLoS ONE 10, e0135152 (2015). 469 
3. Murphy, G. E. P. & Romanuk, T. N. A meta-analysis of declines in local species richness from human disturbances. 470 
Ecology and Evolution 4, 91–103 (2014). 471 
4. Venter, O. et al. Sixteen years of change in the global terrestrial human footprint and implications for biodiversity 472 
conservation. Nat Commun 7, 12558 (2016). 473 
5. Hooke, R. L. B., Martín-Duque, J. F. & Pedraza, J. Land transformation by humans: A review. GSA Today 22, 4–10 474 
(2012). 475 
6. Allen, J. A., Brown, C. S. & Stohlgren, T. J. Non-native plant invasions of United States National Parks. Biol 476 
Invasions 11, 2195–2207 (2009). 477 
7. Oldfield, S. F. Botanic gardens and the conservation of tree species. Trends in Plant Science 14, 581–583 (2009). 478 
8. Heywood, V. H. The role of botanic gardens as resource and introduction centres in the face of global change. 479 
Biodivers Conserv 20, 221–239 (2010). 480 
9. Wyse Jackson, P. & Kennedy, K. The Global Strategy for Plant Conservation: a challenge and opportunity for the 481 
international community. Trends in Plant Science 14, 578–580 (2009). 482 
10. Smith, P. Guest Essay: Building a global system for the conservation of all plant diversity: A vision for botanic 483 
gardens and Botanic Gardens Conservation International. Sibbaldia: the Journal of Botanic Garden Horticulture 0, 484 
(2016). 485 
11. Smith, P., Dickie, J., Linington, S., Probert, R. & Way, M. Making the case for plant diversity. Seed Science 486 
Research 21, 1–4 (2011). 487 
10 
12. Raven, P. & Havens, K. Ex situ plant conservation and cryopreservation: Breakthroughs in tropical plant 488 
conservation. Int J Plant Sci 175, 1–2 (2014). 489 
13. Pautasso, M. & Parmentier, I. Are the living collections of the world's botanical gardens following species-richness 490 
patterns observed in natural ecosystems? Botanica Helvetica (2007). doi:10.1007/s00035-007-0786-y 491 
14. Golding, J. et al. Species-richness patterns of the living collections of the world's botanic gardens: a matter of socio-492 
economics? Annals of Botany 105, 689–696 (2010). 493 
15. Hillebrand, H. On the generality of the latitudinal diversity gradient. The American Naturalist 163, 192–211 (2004). 494 
16. Hinchliff, C. E. et al. Synthesis of phylogeny and taxonomy into a comprehensive tree of life. P Natl Acad Sci Usa 495 
112, 12764–12769 (2015). 496 
17. Lindo, Z. & Gonzalez, A. The bryosphere: an integral and influential component of the Earth's biosphere. 497 
Ecosystems (2010). 498 
18. Kauserud, H., Mathiesen, C. & Ohlson, M. High diversity of fungi associated with living parts of boreal forest 499 
bryophytes. Bot. 86, 1326–1333 (2008). 500 
19. Turetsky, M. R. The Role of Bryophytes in Carbon and Nitrogen Cycling. The Bryologist 106, 395–409 (2003). 501 
20. Westwood, J. H., Yoder, J. I., Timko, M. P. & dePamphilis, C. W. The evolution of parasitism in plants. Trends in 502 
Plant Science 15, 227–235 (2010). 503 
21. Possingham, H. P. et al. Limits to the use of threatened species lists. Trends in Ecology & Evolution 17, 503–507 504 
(2002). 505 
22. Godefroid, S. et al. How successful are plant species reintroductions? Biological Conservation 144, 672–682 506 
(2011). 507 
23. Guerrant, E. O. in Plant Reintroduction in a Changing Climate: Promises and Perils (eds. Maschinski, J., Haskins, 508 
K. E. & Raven, P. H.) 9–29 (Island Press/Center for Resource Economics, 2012). doi:10.5822/978-1-61091-183-509 
2_2 510 
24. Dalrymple, S. E., Stewart, G. B. & Pullin, A. S. Are re-introductions an effective way of mitigating against plant 511 
extinctions? CEE review 07-008 (SR32). (Collaboration for Environmental Evidence, 2011). 512 
25. Cayuela, L., Granzow-de la Cerda, Í., Albuquerque, F. S. & Golicher, D. J. taxonstand: An r package for species 513 
names standardisation in vegetation databases. Methods Ecol Evol 3, 1078–1083 (2012). 514 
26. Damerau, F. J. A technique for computer detection and correction of spelling errors. Commun. ACM 7, 171–176 515 
(1964). 516 
27. Levenshtein, V. I. Binary codes capable of correcting deletions, insertions, and reversals. Soviet physics doklady 517 
(1966). 518 
28. Isaac, N. J. B., Turvey, S. T., Collen, B., Waterman, C. & Baillie, J. E. M. Mammals on the EDGE: Conservation 519 
Priorities Based on Threat and Phylogeny. PLoS ONE 2, (2007). 520 
  521 
 522 
ab
a b c d
Species Genera Tracheophyte Families Embryophyte Families 
30%
59% 75% 93%
a
90
80
70
60
50
40
30
20
10
0
-10
-20
-30
-40
-50
-60
-70
-80
-90
0k
0 25 50 75 100 125 150 175
25k 50k 75k 100k 125k 150k 175k
Equator
No. of digitally recorded species (in red)
N
S
No. of gardens (in gray)
La
ttit
ud
e 
(D
eg
re
es
)
N
S
La
ttit
ud
e 
(D
eg
re
es
)
Capricorn
Cancer
0k 1k 2k 3k 4k 5k 6k 7k
0 25 50 75 100
b No. of species with median range at each lattitudinal bin
% of species represented ex-situ from each lattitudinal bin
90
80
70
60
50
40
30
20
10
0
-10
-20
-30
-40
-50
-60
-70
-80
-90
Bryophyta 
Marchantiophyta 
Pteridophytes 
Gymnosperms 
Angiosperms 
Anthocerotophyta
Lycophytes
a b
1
5
10
50
100
500
1000
5000
Th
re
at
en
ed
 sp
ec
ies
 co
ns
er
ve
d 
ex
-s
itu
 (L
og
10
)
ANG GYM PTE BRY
0
25
50
75
100
 %
 th
re
at
en
ed
 sp
ec
ies
 co
ns
er
ve
d 
ex
-s
itu
c d
8000
0
e
% of Threatened Plants
held in ex-situ collections 
% of total network capacity
devoted to threatened species 
No. of Threatened Species
41%
10%
a b
VU EN CR
20
40
60
100
80
 %
 th
re
at
en
ed
 sp
ec
ies
 co
ns
er
ve
d 
ex
-s
itu
No. of ex-situ collections that a threatened species is held in (Log2)
00
90
80
70
60
50
40
30
20
10
0
-10
-20
-30
-40
-50
-60
-70
-80
-90
0 100 200 300 400 500
0 25 50 75 100
c
No. of species with median range 
at each lattitude bin
La
ttit
ud
e 
(D
eg
re
es
)
% of species represented ex-situ 
from each lattitude bin
Median=3
Capricorn
Cancer
N
S
Threat status
* *
*
1 2 4 8 16 32 64 128
100
200
300
400
500
 N
um
be
r o
f t
hr
ea
te
ne
d 
sp
ec
ies