Tackling supply chain ‘blind spots’ via end-to-end visibility, traceability, and transparency: Conceptualising the role publicly data and big data analytics Wei Nie (wn230@cam.ac.uk) University of Cambridge Mariel Alem Fonseca University of Cambridge Dr Naoum Tsolakis University of Cambridge Dr Minette Bellingan University of Cambridge Dr Mukesh Kumar University of Cambridge Abstract This study explores the nature of publicly available data, big data analytics, the knowledge derived from such analytics, and their interconnections in the context of achieving supply chain visibility, traceability and transparency to uncover the blind spots in the end-to-end supply chain. This study conducts a scoping review and evaluates the relevant academic works using publicly available data. From the reviews, we compiled a list of relevant data sources which can be categorised into 6 groups. Several data concerns like data source selection and data storage remain to be discussed. Finally, a conceptual framework is proposed to guide the implementation. Keywords: Supply chain visibility, Publicly available data, Operations management Introduction Modern slavery, human trafficking, environmental sustainability issues are becoming urgent issues in the supply chain management (SCM) field due to the lack of supply chain visibility, traceability and transparency (SCVTT). SC visibility (SCV) pertains to the degree to which entities involved in the SC possess timely and precise information regarding the operations upstream and downstream (Barratt and Oke, 2007). Transparency is concerned with information disclosure to both internal and external SC stakeholders about the compliance in its SC operations and products whereas traceability is the capability of ascertaining provenance (Sodhi and Tang, 2019). The modern SC has become increasingly complex, resembling networks more than linear chains. This opaque structure can create blind spots, leading to a lack of SCV and exposing companies, governments and other stakeholders to several challenges, particularly in the hidden deep tiers. For example, cobalt-mining is related to human rights violations and environmental negligence, especially in the upstream SC (van den Brink et al., 2020). However, there is a lack of visibility into these areas of the SC, making it challenging to identify and eliminate the issue in the deep tiers and therefore ‘blind spots’. The challenge lies in the fact that managing the blind spots that we cannot measure or visualise is difficult. Therefore, the quest for accurate, timely and complete information in a useful format has gained significant interest (Barratt and Oke, 2007). Several big IT companies like Google have offered SC digital infrastructures solutions to monitor and control companies’ activities and performances; however, they are usually tailored for companies’ operations management with limited end-to-end data capabilities and thus limited visibility (e.g., only to tier-2 suppliers). An approach to mitigating these concerns involves leveraging publicly available data, which requires only minimal agreements to be reached between companies and relevant authorities. The term "publicly available data" generally refers to data that can be easily accessed and downloaded, encompassing various categories such as open data and commercially available data (Cooper and Coetzee, 2020). This approach has the potential to alleviate the burden of limited visibility while providing insights into the SC ecosystem. Yet, the role of publicly available data in SCVTT has not been thoroughly established. Moreover, Publicly available data can be vast, fast, and diverse, indicating the potential for utilizing BDA to support the development of SCVTT. The value of BDA in SC operations and management has been largely recognised among scholarly works like Brinch (2018). This paper adopts a scoping review approach to answer the question: “How can publicly available data be used with big data analytics to develop end-to-end supply chain visibility, traceability and transparency?” The aims include: (1) identify existing publicly available data that can be linked to SCVTT; (2) understand the exploitation of publicly available data using BDA for the creation of SCVTT, and their challenges. This research contributes to the Operations Management field. We proposed BDA implementation details using publicly available data with a three-level conceptual framework. This help to create the knowledge and data capabilities for companies and governments to tackle the blind spots. The rest of the paper is organised as follows: we firstly describe the research method, then summarise the findings and discussion which address the two aims. Finally, conceptual framework is detailed, and the paper is ended with a conclusion. Method This paper applies the scoping review following the six-stage process outlined by Arksey and O’Malley (2005) (Table 1). Based on the research question, we conducted a rigorous search with a search string in Web of Science which is one of the most comprehensive databases for academic journal works. Due to the huge volume of papers in this area, we selected the results by applying several filters. A total of 2380 articles were obtained by 26th January, 2023. To further identify articles that are relevant to the research question, we performed title and abstract screening as well as a full-text screening by using the listed set of inclusion and exclusion criteria. Eventually, a total of 56 papers were selected for full paper appraisal. Table 1 – Scoping review processes Database Web of Science Search field Topic Search string "supply chain*" AND ("simulation" OR "statistic*" OR "big data" OR "data-driven" OR "analytics" OR "machine learning" OR "data science" OR "data analy*" OR "operations research" OR "optimization" OR "optimisation" OR "open data" OR "public* data*" OR "database" OR "public* available data*" OR "data mining" OR "business intelligence") Filters Publication time: Last 5 years Document type: Articles Language: English Journal: Only ABS listed and with 3/4/4* scores Title, Abstract and Full-text screening Inclusion criteria: · Application paper rather than review or conceptual notes · Applied in SC operation context · Used publicly available data as the data source · Applied any of the mentioned techniques/models · Illustrated what and how publicly available data is used Exclusion criteria: · Studied the importance/capability/influence/culture of the mentioned techniques (especially using some qualitative research methods like survey and interview) Findings & Discussion Aim (1): Publicly available data related to SCVTT To unpack the different ways in which SCVTT is used in SC research, we categorised publicly available data that have been used in the reviewed literatures into 6 dimensions. This categorisation (Table 2) developed based on a previous framework of SC transparency information platforms (Gardner et al., 2019), describes the themes and types of the publicly available data used, lists typical sources of the information, assesses their intended uses and limitations. Although it does not provide an exhaustive list of all related platforms, it contributes to draw some useful conclusions on the state of using publicly available data for SCVTT research from this finding. It is comprised of: · Traceability information Traceability information provides transparency in connections among SC actors and locations, and life cycle inventory information of SC processes and flows. The former is usually characterised with freely accessible trade databases for the country-level import and export data. Although trade data are not originally used in the context of SC, they are commonly seen in the research to imply the country level connections. Life cycle assessment (LCA) databases are most seen in the reviewed papers, they provide energy and material flow information on the environmental impacts associated with the entire life cycle of a product or service. · Supply network structure information Supply network structure information provides visibility into downstream and upstream SC actors and their relationships through information like buyer-supplier dyadic information (Geng et al., 2022). · Supply chain activity information SC activity information creates visibility by reporting actions taken by SC actors in terms a series of activities like production, sales and purchasing. Nevertheless, existing sources of publicly available data used in the reviewed literatures are usually not in the context of SC due to the confidentiality or timeliness of the information. Research has exploited data from other disciplines and hence additional translation is required. For example, using Google trends search patterns to imply the demand pattern (Nikolopoulos et al., 2021). · Financial information Financial information is mainly from commercial databases like Bloomberg with a resolution of corporate level. Information like total assets provide visibility into SC actors’ characteristics and helps identify the actors with main powers in the SC. · External environments information Environments external to the SC can be Earth environments (land, atmosphere, and ocean) and socio-economic environments. These data are rich in contents and formats, for example, historical and projected geospatial data of various forms. Unlike information of other dimensions, external environments information is usually not directly linked to SC, and additional data translation is needed to interpret these data for use in SC research. · Environmental, Social and Governance (ESG) policy and commitment, activities and effectiveness information These data are usually processed information (e.g., indicators) published by public institutions or NGOs to evaluate the sustainability performance by aggregating information from public domain (e.g., news) or voluntarily disclosed by the entities. These data provide transparency on any differences in the levels and strengths of policies and activities adopted by different SC actors. Aim (2): Exploiting the publicly available data in existing research While the listed database platforms belong to the same dimension that is linked to SCVTT, they can be used in diverse contexts for various purposes. We summarise some of the example data exploitations from the reviewed literatures following big data analytics processes: (1) data source & collection; (2) data pre-processing; (3) data storage and management; and (4) data analysis and application. We also identify challenges associated with these steps from the review. · Data source & collection Data definition is the first step when leveraging publicly available data for SC research as these data can from various disciplines and not originally collected for the use in SC context. For example, Nikolopoulos et al. (2021) use search trend data to estimate the demand. Defining data also ensures the validity of using this data to imply the SCVTT, helping to identify relevant data sources. Limited works have mentioned the reasons of data source selection. Despite the common reasons such as credibility of the data provider and data comprehensiveness (Ingrao et al., 2019), Baker et al. (2022) choose to use Compustat database because it provides the desired data update frequency. This is particularly important when exploiting the publicly available data because similar data can have different providers. As an illustration, Srinivasan et al. (2019) and Nagendra et al. (2022) both employ regional projected rainfall data in their respective studies, but they adopt different data sources, namely, the WorldClim database with global coverage and the Indian Meteorological Department that solely offers regional data. Data richness, consistency and original data collection methods can be very different with these two providers, and potentially affect the research result. Data source selection is also concerned with data quality and validity. Furthermore, our review identifies that the biggest challenge associated with the data source is the data heterogeneity. The data can be different in aspects like formats (csv file, image), resolutions (sub-national level, country level), units (bales, metric tons), frequency (real-time, annual), size (single data point, big data). Such heterogeneity brings challenges when multiple data sources are used in the study and there is a need to integrate all of them together for analysis and knowledge generation. Heterogeneity also brings the necessity to have data triangulation to avoid potential data collection biases introduced by the original data providers. However, only van den Brink et al. (2020) specifically 4 10 Table 2 – Publicly available data used in the existing SCVTT research Theme of publicly available data Type of publicly available data Information used Publicly available data platforms used in the reviewed literatures Who primarily produces the data? Intended use of the information in the research Limitation and unintended consequences Example Ref. Traceability information Traceability information linking supply chain places and actors Trade information and statistics (e.g., import and export information) Organisation for Economic Co-operation and Development (OECD) Statistics, The United Nations Comtrade Database Public institutions Risk management, identify relationships between supply chain actors Usually based on aggregated data and therefore lacking information on sub-national original of traded commodities as well as the trading companies; Coverage limited to certain part of a supply chain (Nikolopoulos et al., 2021) Life cycle inventory information of supply chain processes and flows Supply chain process flow information (e.g., material and energy flow), socio-economic accounts, input-output economics models EcoInvent, GaBi, Exiobase3, World Input-output Database (WIOD), NREL U.S. Life Cycle Inventory Database, Eora Database, Agri-footprint Database, U.S. Bureau of Economic Analysis (BEA) Input-Output Accounts Most from private providers for commercial use, some are from governments and public institutions but usually only contains regional data Impacts and efficiencies monitoring, policy effectiveness assessment, policy design Data coverage and granularity is often a trade-off; Based on aggregated sample data and lacking sub-national information; Coverage limited to certain industries (Ingrao et al., 2019) Supply network structure information Information on supply chain downstream and upstream actors and their relationships (i.e., buyer-supplier relationships) Information on focal firm’s rivals, tier-1/tier-2 suppliers, main customers Bloomberg's Supply Chain Relationships (SPLC) Database, FactSet Supply Chain Relationships Database, Standard and Poor's Compustat Database (Segment data) Private providers Usually used to construct the supply chain structure for traceability study Usually not freely available in the public domain; Little information is easily available; Unconnected to downstream players (Geng et al., 2022) Supply chain activity information Information on activities of supply chain actors, including supply/production, sales, purchasing, processing, and consumption activities Commodity/ product production, sales and consumption data, search trend data, deposit information Food and Agriculture Organisation (FAO) Ecocrop, Google Trends, U.S. Energy Information Administration, British Geological Survey, U.S. Geological Survey, Geological Survey of India Public institutions Sourcing decisions, supply chain design Usually not directly linked to supply chain activities and additional interpretations and translations are needed (Nagendra et al., 2022) Financial information Information on firm’s patterns of financial performance and transactions Cost of goods sold (COGS), total assets, total liabilities, sales, net income, ownership status Bloomberg, Thomson Reuters Eikon, Worldscope, Standard and Poor's Compustat Database Private providers Usually used as control variables in supply chain visibility studies Usually not freely available in the public domain; Limited to companies greater than certain size (Baker et al., 2022) External environments information Information on the earth environment (atmosphere, land, ocean) Historical and projected temperature, rainfall, terrain elevation, sea level elevation; ground reality WorldClim database, National Oceanic and Atmospheric Administration Climate Data Online, Europe’s Copernicus Sentinel Synthetic Aperture Radar Satellites, Indian Metrological Department, Airbus Pléiades imagery, PlanetScope Mostly from public institutions and governments, some private providers Sourcing decisions, change management Climate data (especially the projection) can be very different depending on the collection methods and models used; Not directly linked to supply chain, need additional interpretation and translation (Srinivasan et al., 2019) Information on the socioeconomic environment (e.g., demographic and economic conditions) Population intensity/concentration, commodity market price, government GDP spent on certain sectors, disaster information Government census data, Eurostat, EM-DAT: The International Disaster Database, World Health Organization's Global Health Expenditure Database, U.S. Department of Agriculture (USDA) Market News Mostly from public institutions and governments, some private providers Risk management Not directly linked to supply chain, need additional interpretation and translation (Nakano, 2021) Environmental, social and governance policy and commitment, activities and effectiveness information Supply chain actors’ policies and commitments to increase the transparency of their operations for social and governance performance Corporate social responsibility score, global slavery index, human rights index, modern slavery media coverage, corruption score, political stability index, risk score, worldwide governance indicator World Bank Worldwide Governance Indicators, Modern Slavery Registry, Social Hotspots Database (SHDB), Sustainalytics, Walk Free Foundation, Factiva, Transparency International, IHS Markit Mostly from public institutions and NGOs, some governments Risk and performance management, social responsibility and sustainability measurement Dependent on voluntary disclosure of accurate data by companies, the disclosed data can be biased (Geng et al., 2022) Environmental impacts of supply chain activities, environmental performance scores Emissions data, environmental disclosure score, water availability/resource data, water and energy consumption data Bloomberg's environmental, social, and governance (ESG) database, Emissions Database for Global Atmospheric Research (EDGAR), FAO AQUASTAT, WRI (World Resource Institute) Aqueduct database, Environmental Performance Index (EPI), EIKON database Governments, NGOs, private providers Risk and performance management, social responsibility and sustainability measurement Dependent on voluntary disclosure of accurate data by companies, the disclosed data can be biased (Doni et al., 2019) address this concern by comparing the country level mining and refining statistics provided by British Geological Survey and the United States Geological Survey, where same metric can have three times differences in values. Besides, little attention has been paid to the way data is collected. For example, whether is a manual or automatic collection, any tools used to communicate with the data source for data collection, and how often the data collection is performed if it is a repeat collection. · Data pre-processing Many authors have applied data pre-processing in their works to improve the quality of the collected data. The reviewed literature often observes the application of data pre-processing techniques to handle missing data. There are two cases of missing data: (1) missing data points (Geng et al., 2022); and (2) missing the entire data category due to availability issue of the data source (Ostroski et al., 2022). Common actions taken include removing corresponding missing entries (Geng et al., 2022) and supplement with additional sources (Ostroski et al., 2022). · Data storage and management Data storage and management involve storing and organising the collected publicly available data in a secured place like a relational database for ongoing or future operations. However, this process is hardly mentioned in the reviewed works. We observe that almost all the works use the data for one-off analysis to achieve a specific research goal. For example, proposing a new algorithm for demand forecasting (Nikolopoulos et al., 2021). Therefore, data storage and management are not as needed as other processes when performing data analytics. In the scenario where constant data update or real-time data is needed for analysis, data storage and management will be a critical step when exploiting the publicly available data. This is especially the case in the presence of data heterogeneity and large data volume. · Data analysis and application Several analysis techniques such as statistical modelling and life cycle assessment have been observed in the existing works. Even though the research contents are all relevant to SCVTT, the use of techniques is still very specific to different application purposes, ranging from environmental impact assessment (Ingrao et al., 2019) to demand forecasting (Nikolopoulos et al., 2021). There is no standard set of criteria for techniques/models selection for specific purpose. Integrating data from multiple sources for analysis is a significant challenge due to the heterogeneity of publicly available data in terms of formats and specifications across various disciplines (e.g., combining satellite images with numerical and textual data). Nagendra et al. (2022) also raise the concern of the analysis result scalability, where the author applied satellite big data analytics on Indian’s humanitarian SC. This is particularly important to generate the end-to-end SCVTT where a globally dispersed SC can be expected and the analysis might need to be adapted to different areas accordingly. Conceptual Framework The review provides an overview of the use of publicly available data in recent SCVTT related research. However, there is still little guidance on what processes are needed to leverage the publicly available data for SCVTT, and how the SC companies, governments and other SC stakeholders can proceed with the processes. Therefore, we derive a conceptual framework which is comprised of 3 levels with increasing details across the levels: (1) theoretical basis which identifies the processes; (2) architecture that describes the processes; and (3) guidelines which details the processes. We firstly draw upon the information processing theory model developed by Atkinson and Shiffrin (1968) to provide the theoretical basis of using publicly available data for creating the SCVTT (Figure 1). As aforementioned, existing applications of publicly available data are mostly for one-off analysis. Without further storage and maintenance, the knowledge generated is kept as short-term memory. SCVTT is a prolonged and continuous requirement to tackle the blind spots, therefore a “long-term memory” is needed to achieve this, where constant information retrieval and analysis is required. Figure 1 – Theoretical basis (Level 1) In level 2, we further detail the process flow and their connections of implementing and exploiting the publicly available data with the reference to BDA processes. The development of this process flow is inspired by the Earth observation process which is defined as “the collection, analysis and presentation of information about planet Earth’s physical, chemical and biological systems via remote sensing technologies” (UK Space Agency, 2023) as both the end-to-end SC and Earth are exhibited as complex systems. The architecture presented in Figure 2 specifies the flow of publicly available data from its source to be knowledge generation to support the decision-makings for operations management with the presence of SCVTT. Figure 2 – The architecture of exploiting the publicly available data (Level 2) Based on the paper reviewed and a framework developed by Brinch (2018) which identifies the constructs for value creation in SC business processes using BDA, Level 3 (Table 3) is developed to further specify the implementation details and guidelines for companies, governments and SC stakeholders. This guidelines include the constructs that are needed to consider and evaluate in each of the process, and the technology/platform/technique available to support the implementation of BDA with publicly available data. Table 3 – Implementation details (Level 3) Process Construct Technology/Platform/Technique Data source Data verification Data triangulation Data translation Data characteristics (volume, variety, velocity, veracity, value) Data update Traceability information Supply network structure information Supply chain activity information Financial information External environment information ESG policy and commitment, activities and effectiveness information Data collection Data source monitoring Data collection frequency Communication with data sources Automatic vs Manual collection Internet of Things ecosystem Cloud Platforms (e.g., Google Cloud, Microsoft Azure) Data pipelines for streaming large amount of data Data Pre-processing Data integration Data interoperability Data fusion Data standardisation Data aggregation Cleaning, transformation reduction Big data tools (e.g., Apache Spark) Missing data identification and imputation Outliers removal Data scaling Pre-processing to fit the desired storing structure Data Storage & Management Data model Database design Data ingestion Data maintenance Data storage limit Data update Database communication On-premises vs Cloud database Relational vs NoSQL vs NewSQL data model (e.g., Microsoft SQL server, Neo4j, MongoDB) Data Analysis & Application Model selection. Knowledge creation Data scalability Data analytic maturity Data analysis frequency Data presentation/visualisation Automated vs Human Decision-making Strategic vs Operational Decision-making Statistical models Input-output modelling Optimisation models Network analysis Descriptive analytics Life cycle assessment Predictive analytics Business intelligence Predictive analytics Simulation models Algorithms Multi-criteria decision analysis Conclusion Companies and governments today are facing pressing issues like modern slavery brought by the SC blind spots due to the lack of end-to-end SCVTT. However, existing digital infrastructures face the challenge of having limited data visibility into the deep-tiers of the SC. Publicly available data can provide additional data relevant to end-to-end SCs, which opens unprecedented opportunities. This paper conducts a scoping review on relevant works using publicly available data to answer the research question: “How can publicly available data be used with big data analytics to develop end-to-end supply chain visibility, traceability and transparency?”. The review firstly summarises a list of relevant data sources and categorises them into 6 dimensions: traceability, supply network structure, supply chain activity, finance, external environments and ESG policy, activity and performance. We also map the use of publicly available data to the process of data source and collection, data pre-processing, data storage and management, data analysis and application from the review. It is observed that major challenges are associated with data source and collection, such as data definition and translation, data heterogeneity. Besides, existing works mostly use publicly available data for one-off analysis and thus data storage is rarely explored. By drawing upon information processing theory as theoretical lens, this study offers a framework which conceptualises the role of BDA and publicly available data in developing the end-to-end SCVTT, providing guidance for SC companies, governments and other stakeholders. The identified data sources also create data generation capabilities for these stakeholders to explore SC analytics. This work also paves the foundation for the development of the digital infrastructure needed to develop the SCVTT in the future. References Arksey, H. and O’Malley, L. (2005), “Scoping studies: towards a methodological framework”, International Journal of Social Research Methodology, Routledge, Vol. 8 No. 1, pp. 19–32. Atkinson, R.C., Shiffrin, R.M. (1968). "Chapter: Human memory: A proposed system and its control processes". In Spence, K.W.; Spence, J.T. (eds.). The psychology of learning and motivation. Vol. 2. New York: Academic Press. pp. 89–195. Barker, J.M., Hofer, C., Hoberg, K. and Eroglu, C. (2022), “Supplier inventory leanness and financial performance”, Journal of Operations Management, Vol. 68 No. 4, pp. 385–407. Barratt, M. and Oke, A. (2007), “Antecedents of supply chain visibility in retail supply chains: A resource-based theory perspective”, Journal of Operations Management, Vol. 25 No. 6, pp. 1217–1233. Brinch, M. (2018), "Understanding the value of big data in supply chain management and its business processes: Towards a conceptual framework", International Journal of Operations & Production Management, Vol. 38 No. 7, pp. 1589-1614. Cooper, A.K. and Coetzee, S. (2020), “On the Ethics of Using Publicly-Available Data”, in Hattingh, M., Matthee, M., Smuts, H., Pappas, I., Dwivedi, Y.K. and Mäntymäki, M. (Eds.), Responsible Design, Implementation and Use of Information and Communication Technology, Springer International Publishing, Cham, pp. 159–171. Doni, F., Corvino, A. and Martini, S.B. (2019), “Servitization and sustainability actions. Evidence from European manufacturing companies”, Journal of Environmental Management, Vol. 234, pp. 367–378. Gardner, T.A., Benzie, M., Börner, J., Dawkins, E., Fick, S., Garrett, R., Godar, J., et al. (2019), “Transparency and sustainability in global commodity supply chains”, World Development, Vol. 121, pp. 163–177. Geng, R., Lam, H.K.S. and Stevenson, M. (2022), "Addressing modern slavery in supply chains: an awareness-motivation-capability perspective", International Journal of Operations & Production Management, Vol. 42 No. 3, pp. 331-356. Ingrao, C., Faccilongo, N., Valenti, F., De Pascale, G., Di Gioia, L., Messineo, A. and Arcidiacono, C. (2019), “Tomato puree in the Mediterranean region: An environmental Life Cycle Assessment, based upon data surveyed at the supply chain level”, Journal of Cleaner Production, Vol. 233, pp. 292–313. Nagendra, N.P., Narayanamurthy, G. and Moser, R. (2022), “Management of humanitarian relief operations using satellite big data analytics: the case of Kerala floods”, Annals of Operations Research, Vol. 319 No. 1, pp. 885–910. Nakano, K. (2021), “Risk assessment for adaptation to climate change in the international supply chain”, Journal of Cleaner Production, Vol. 319, p. 128785. Nikolopoulos, K., Punia, S., Schäfers, A., Tsinopoulos, C. and Vasilakis, C. (2021), “Forecasting and planning during a pandemic: COVID-19 growth rates, supply chain disruptions, and governmental decisions”, European Journal of Operational Research, Vol. 290 No. 1, pp. 99–115. Ostroski, A., Lagos, T., Prokopyev, O.A. and Khanna, V. (2022), “Consumption-Based Accounting for Tracing Virtual Water Flows Associated with Beef Supply Chains in the United States”, Environmental Science & Technology, American Chemical Society, Vol. 56 No. 22, pp. 16347–16356. Sodhi, M.S. and Tang, C.S. (2019), “Research Opportunities in Supply Chain Transparency”, Production and Operations Management, Vol. 28 No. 12, pp. 2946–2959. Srinivasan, R., Giannikas, V., Kumar, M., Guyot, R. and McFarlane, D. (2019), “Modelling food sourcing decisions under climate change: A data-driven approach”, Computers & Industrial Engineering, Vol. 128, pp. 911–919. UK Space Agency. (2023), “Earth Observation (EO)”, GOV.UK, 18 April, available at: https://www.gov.uk/government/collections/earth-observation-eo (accessed 19 April 2023). van den Brink, S., Kleijn, R., Sprecher, B. and Tukker, A. (2020), “Identifying supply risks by mapping the cobalt supply chain”, Resources, Conservation and Recycling, Vol. 156, p. 104743. image1.png image2.png