Abstract
Imaging spectroscopy (IS) can identify target materials at both mineralogical and geochemical levels. Therefore, in environmental applications, it can be used to assess contamination derived from mining activities, moving from contamination sources along pathways to receptors as acid mine drainage (AMD). This can be based on the spectra of specific assemblages of minerals from spectral libraries, which can indicate pH values at the time of their generation and the subsequent acid-generating potential. Alternatively, field spectral measurements can be used as input data for mapping algorithms. This study presents a new methodological approach to improving the results for mapping contamination sources and pathways, by combining multisource spectra from both these approaches at different scales. In addition to the mineralogical libraries and field spectra already mentioned, additional end-member spectra that are extracted from IS data are used so as to highlight particular site phenomena otherwise undetected by the two previous approaches. The highly correlated spectra are then used as input to the Spectral Angle Mapper algorithm, to establish a map of local field spectra and also one from image end-members. The intersection of the two maps results in an improved map, assigned in terms of correlation ≥0.8 of mineralogical assemblages focused on AMD indicators. This methodology was tested in the abandoned S. Domingos Mine, in southeast Portugal's Iberian Pyrite Belt, with AMD caused by long-term exploitation of volcanogenic massive sulphide deposits. Data from the HyMap™ sensor covered the area, and field spectroradiometric measurements were undertaken and analysed for mineralogical and geochemical content. A flightline containing the open pit was processed according to the aforementioned methodology, focusing directly on the target of interest and minimizing errors. The final map displays the mineralogical assemblage correlations ≥0.8 of variable pH indicators, particularly isolating a low-pH combination of significance to the contamination in the area.
Acknowledgements
The data set was collected in the framework of the MINEO project in the fifth Framework Programme of the European Union (IST-1999-10337). We are grateful to Dr Hartmut Mollat from Bundesanstalt für Geowissenschaften und Rohstoffe, responsible for the field spectroradiometric measurements as well as for his suggestions during data capture. Two anonymous reviewers contributed to the improvement of the work. This work was partially funded through the Foundation for Science and Technology of Portugal (Grant BD/17257/2004). This article is published with the approval of the Executive Director, British Geological Survey.
Notes
1. Referred to as the cosine in ENVI (ITT Citation2010) and not as inverse, as formally defined (Chang Citation2007). Although different angles have been tested, the option was to maintain a standard value, equal for all the spectra, and rely on a correlation value indicator of the input spectra to improve the classification.
2. Coefficient of variation (CV) = σ/aver, whereas σ is the standard deviation and aver is the average of data distribution.