Analysis of temporal changes of a mining site affected by AMD. UAV based hyperspectral monitoring of the Litov tailing (Sokolov, CZ)

In this study we test and validate hyperspectral mineral mapping for acid mine drainage (AMD) detection using Unmanned Aerial Systems (UAS). The investigated area is a re-cultivated tailing and part of the Sokolov coal mine district in the Czech Republic. The dumped mining waste material bears pyrite and the consecutive weathering products. Mainly iron hydroxides and oxides affect the natural pH values of the Earth’s surface. While previous research done in this area relies on satellite and air-borne data, our approach focuses on lightweight drone systems providing ground readiness within hours. During April to September 2016, several field and flight campaigns were conducted. For validation, the waste heap was probed in-situ for pH, X-ray florescence (XRF), reflectance spectrometry and surveyed by GPS. Collected samples were measured for pH, X-ray diffraction (XRD) and XRF in laboratory conditions. A new in-house developed toolbox allowing the processing as well as the correction of topography and illumination effects, is applied on the UAS data. High-resolution point clouds and digital elevation models are build from UAS-borne data with Structure-from-Motion photogrammetric techniques. Changes of the canopy, topographic profiles and erosion values are derived from the 3D models. The classification of hyperspectral (HS) data detected the proposed minerals jarosite and goethite, which are associated with the acidic environmental conditions (mean pH = 2.9). The specific iron absorption bands are visible in the UAS-HS data, and can be confirmed with ground-truth spectroscopy. Evaluation of geochemical data shows a negative correlation for pH with sulfur content, which is in accordance with previous studies on AMD conditions. Interpolations of in-situ pH data support the UAS-based results. The evaluation of the applied methods highlights the UAS approach as a fast, non-invasive, inexpensive technique for multi-temporal monitoring.