Hyperspectral remote sensing for mine waste characterization and critical mineral exploration

The heavy reliance on few suppliers for the provision of rare earth elements (REEs), considered as critical raw materials (CRM), is a cause of vulnerability for Europe’s economy, due to their importance in industrial applications. To counteract this issue, the European Union is resorting to the exploitation of historical mines and untapped reserves within its territory, by adopting sustainable mining practices. Therefore, legacy mining lands, which so far had been perceived as problems for their impacts on the environment and human health, now gain a pivotal role as potential secondary deposits of REEs. In this context, remote sensing emerges as a novel, non-invasive, time and cost-effective solution for geologically characterizing these deposits. This Ph.D. project is an effort to show how imaging spectroscopy can aid in the characterization of mines and their residues as both sources of REEs and as potential causes of pollution. Structured on a case study basis, the thesis leverages multispectral and hyperspectral data in the visible-near infrared (VNIR) and shortwave infrared (SWIR) spectral interval to model REE concentrations at different scales, on the one hand, and to monitor sources of acidity deriving from mine residues, on the other. In the first case study, a flexible, explainable and scalable Random Forest Regression approach was developed for quantifying individually Nd, Dy, Er, Sm and Pr, using 130 spectra of REE-bearing minerals and rocks. The Nd model’s performance was tested on hyperspectral datasets acquired in the lab and on a satellite image, confirming its scalability. The results prove to be physically meaningful as the model uses REE-related spectral absorptions during the predictive process. Moreover, the approach is independent of mineralogy and robust against unharmonized datasets. In the second case study, secondary iron-bearing minerals are mapped using a polynomial fitting technique over the historical mining district of Montevecchio-Ingurtosu, in southwestern Sardinia, using different scales and sensors. Laboratory analyses of samples collected in the field are combined with Sentinel-2 and EnMAP images to derive spatial zonations of Fe(III) hydroxides and sulfates highlighting sources of acid drainage in mine residues and compositional variations in waterways affected by neutral drainage. The continuous acquisition of Sentinel-2 data in time allows to monitor the mineral paragenesis in different seasons with changing environmental conditions. Throughout the thesis, two challenges are highlighted, whose mitigation could be the subject of potential future research: the effects of mineral masking and the effects of instrumental and atmospheric noise. Overall, this research project sets the baseline for ensuring remote sensing’s role as an essential component in efficient and streamlined processes for sustainable, safe and cost-effective mining and post-mining operations.