A multiscale observational approach for tracking AMD from source to sink: application to Rosia Poieni copper mine, Romania

Mining operations are generally accompanied by mobilization of various contaminants that are transferred via water or air to various environmental compartments such as soil, sediments, or biota. Understanding their origins, trajectories, sinks, and spatial distribution is essential for effective risk assessment, site management, recovery planning, and transparent communication with regulators and communities. Rosia Poieni, a porphyry copper deposit that has been excavated as an open-pit mine for the past five decades, has been selected as a test site to integrate remote sensing and ground observation methods for assessing the environmental conditions in the mining and surrounding areas in the EU-funded MOSMIN project. The abundance of pyrite in the host rocks favours the generation of severe acid mine drainage, resulting in pH levels ranging from 2 to 3 and a high load of heavy metals in the streams emerging from the waste rock dumps.

A combination of satellite and UAV multispectral and hyperspectral data, and targeted ground truthing (mineralogical, geochemical, and spectroscopic analyses) has been used to acquire validated and interoperable data products that delineate contaminant sources, transport pathways, and accumulation zones across terrestrial, fluvial, and atmospheric compartments.

Hyperspectral satellite data from PRISMA and EnMAP enabled the discrimination of alteration assemblages and secondary mineral phases that control acid mine drainage (AMD) processes and associated metal release. When supported by site-specific spectral libraries derived from hyperspectral scanning of hand specimens, supervised spectral unmixing approaches produced mineralogical maps with a high degree of consistency relative to field observations and laboratory XRD results. These outputs allow the differentiation of alteration facies with contrasting environmental relevance, such as pyrite-rich phyllic and argillic materials with high acid-generating potential versus propylitic and potassic assemblages characterised by greater buffering capacity.

Multispectral satellite data, particularly from Sentinel-2, were shown to be essential for spatial and temporal scaling of hyperspectral results. Sentinel-2 band ratios and mineral proxy indices, despite their limited spectral resolution, provided robust first-order indicators of iron oxidation state, hydroxyl-bearing alteration, and AMD-impacted materials.

At the local scale, UAV-based multispectral and hyperspectral imaging proved critical for resolving AMD distributions in narrow drainage corridors and near-field impact zones that are not adequately captured by satellite sensors.

The implemented workflows show that we can advance EO-based contaminant assessment from qualitative mapping to quantitative, validated, and reproducible information products. The integration of multiple sensors ensures internal consistency across spatial scales and establishes a robust technical foundation for subsequent data fusion and modelling activities.

Acknowledgement: This work was financially supported by the European Union Agency for the Space Programme under Project 101131740—MOSMIN—HORIZON-EUSPA-2022-SPACE.