Geospatial Data and Deep Learning Expose ESG Risks to Critical Raw Materials Supply: The Case of Lithium

• DOI: 10.3389/esss.2024.10109
• Publication Date: 2024-07-04T08:50:05Z
• Authors: Christopher J. M....; Marcus Haynes; Bijal Chudasama; Kathryn Goodenough; Toni Eerola; Artem Golev; Steven E. Zhang; Junhyeok Park; Eleonore Lèbre
• Keywords: critical mineral; Dynamic and structural geology; geoscience; machine learning; sustainable development; QE500-639.5; mineral potential; sustainability; 01 natural sciences; 0105 earth and related environmental sciences

Disruptions to the global supply chains of critical raw materials (CRM) have potential delay or increase cost renewable energy transition. However, for some CRM, primary drivers these chain disruptions are likely be issues related environmental, social, and governance (ESG) rather than geological scarcity. Herein we combine public geospatial data as mappable proxies key ESG indicators (e.g., conservation, biodiversity, freshwater, energy, waste, land use, human development, health safety, governance) a dataset news events train validate three models predicting “conflict” disputes, protests, violence) that can negatively impact CRM chains: (1) knowledge-driven fuzzy logic model yields an area under curve (AUC) receiver operating characteristics plot 0.72 entire model; (2) naïve Bayes AUC 0.81 test set; (3) deep learning comprising stacked autoencoders feed-forward artificial neural network 0.91 set. The high demonstrates accurately predict natural resources conflicts, but show machine results biased by population density underestimate conflict in remote areas. Knowledge-driven methods least impacted bias used calculate rating is then applied lithium occurrences case study. We demonstrate giant brine deposits (i.e., >10 Mt Li 2 O) restricted regions with higher spatially situated risks relative subset smaller pegmatite-hosted yield ratings lower risk). Our reveal trade-offs between sources lithium, resource size, risks. suggest this type broadly applicable other mapping prior mineral exploration has improve outcomes government policies strengthen chains.