Geospatial monitoring and modelling of acid mine drainage and mining footprints

Abstract

Metal rich acidic waters stem from the oxidation of sulfide-bearing rock, known as acid mine drainage (AMD). Mining and excavation activities may increase sulfide exposure and exacerbate acid formation, which has little biological use. Monitoring AMD in the environment is hampered by the spatiotemporal limitations of traditional methods. The aim of this Thesis was to evaluate the scalability and limitations of using passive sensor-derived data to quantify and delineate superficial AMD deposits at a global-to-local scale, evaluate the effectiveness and constraints of unsupervised classification methods in water-inundated environments and assess AMD footprint and overlap globally. The first research chapter determines the overlap between mining activities and Key Biodiversity Areas (KBAs) by synthesizing datasets such as land cover type, USGS, and global mine polygons, while analysing associated socioeconomic factors and potential environmental impacts. The second research chapter examines regional AMD dynamics by analysing multispectral satellite imagery to identify region-specific Fe(II)/Fe(III) precipitates using band ratio proxies in Huelva mining district, SW Spain. The concluding research chapter examines the effects of using a multi-sensor, multi-altitude approach to the monitoring of AMD in a water-inundated environment through a novel same-day cross-sensor correlation study of a mine-waste repository in Cornwall, UK. Major findings include the confirmation of a 6.7% mining-KBA overlap globally, with Au as the most sought-after deposit. Iron speciation mapping identified changes in the ferrous/ferric footprint ratio between the Huelva Estuary (5:3 km2, Fe(II):Fe(III)) and its mining district (1:12 km2), highlighting region-specific precipitation and areas undergoing Fe-cycling. Cross-sensor analysis identified diminishing rates of predictability affecting AMD delineation accuracy as sensor resolution decreased (R2 = 0.23, multispectral; R2 = 0.78, hyperspectral). Increases in total mixed pixels caused by AMD-impacted waterbodies were recorded as sensor resolution decreased (UAV: 2.4%, PlanetScope PS2.SD: 3.7%, Sentinel-2: 8.5%). Sensors with finer spectral and spatial resolutions exhibited higher resilience after intense rainfall events, where surface moisture inhibited reflectance. The interpretation of remote sensing data successfully delineated superficial precipitates associated with AMD across varying spatiotemporal scales using band ratio as proxies to mineralogy, however the constraints and limitations encountered were intrinsically linked to the simplicity of the approach taken and the coarseness of multispectral data. Future research should focus on the global AMD footprint and concentration estimates, mineral composition mapping using spectral libraries and integrated supervised machine learning used in conjunction with more widely accessible hyperspectral data for enhanced spectral resolution and diagnostic accuracy.