Satellite-Based “Before-After Control-Intervention” Framework for Detecting and Attributing Coastal Mining Impacts on Water System Disruption

Our production and consumption systems fundamentally depend on, impact, and shape ecosystem state and function. Yet, the largest economic beneficiaries tend to underreport and sparsely disclose operations' impacts on biodiversity and water. Existing practices often rely on self-reporting or lack depth, lack the integration of available spatiotemporally granular datasets, and tend to avoid the attribution of changes. If we want to be ambitious about corporate impact quantification and disclosure, we need operational protocols that leverage available high-granularity and allow for comparability across assets and portfolios.

This research seeks to advance toward developing a scalable, transparent, and operational change detection and attribution workflow using satellite Earth Observation (EO). We demonstrate our workflow for assessing coastal mining impacts on surrounding water systems, a complex but important task since mining expansion driven by energy transition has spurred operations in ecologically sensitive watersheds and biodiversity hotspots with far-reaching downstream effects.

Our difference-in-differences framework compares mining site trajectories with reference areas having similar climatic, biophysical, and topographic characteristics. Counterfactuals are selected through spatial cluster analysis and satellite embedding similarity search. Using multi-sensor satellite data, we retrieve indicators covering water cycle domains, including evapotranspiration, open water surface, vegetation water content, and coastal water constituents. We quantify ecosystem displacement toward extreme multivariate values that may link to ecosystem integrity versus disruption extent and magnitude.

Our analysis reveals significant deviations in the trajectory of EO-based water indicators between impacted sites and counterfactuals. The world's largest nickel mine in Weda Bay shows a 24% shift toward more extreme values of evapotranspiration, vegetation moisture, water availability, turbidity, and total suspended matter by 2024 compared to 2018 baselines, while reference points show minimal change (4%). We discuss the framework's scalability, dissecting facility-specific impacts from regional environmental variability, and potential for benchmarking EO-derived Essential Variables. Yet, challenges remain, including (dis)agreement between BACI variants challenged by (lacking) asset data or historic imagery, spatial footprint decisions, the interpretation and comparability of relative indicators, and counterfactual search criteria. Despite these challenges, this work demonstrates advances in satellite EO-based counterfactual analyses as a scalable approach for multi-dimensional ecosystem change detection and facility-specific impact attribution.