The Study Of Methods For Mapping Hydrothermally Altered Minerals With Multi-source Remote Sensing Data In The Duolong Area

The Duolong area is in a strong potential copper mineralization zone and several porphyry copper deposits(PCDs)associated with intermediate-acid intrusions have been identified.Limited by the severe environment and poor transportation condition,it is difficult to implement regular geological survey.The PCDs are usually characterized by hydrothermally altered zones.Hydrothermally altered minerals have diagnostic spectral absorption properties in the visible and near-infrared(VNIR)through shortwave infrared(SWIR)regions.Therefore,mineral explorations using remote sensing data have irreplaceable advantage.In this study,Sentinel-2A Multi Spectral Instrument(MSI),Advanced Spaceborne Thermal Emission and Reflection Radiometer(ASTER),Hyperion and Advanced Hyperspectral Imager(AHSI)data are combined to map hydrothermally altered minerals.Three temporal Sentinel-2A MSI images are processed by integrating the principal component analysis(PCA)and the anomaly-overlaying selection method to map hydroxyl-bearing minerals and iron oxides.The results are used for evaluating the potential of Sentinel-2A MSI data in geological filed.The random forests(RF),voting-based extreme learning machine(V-ELM)and kernel-based extreme learning machine(KELM)classifiers are applied to the ASTER SWIR data,ASTER SWIR-thermal infrared(TIR)data,Hyperion SWIR data and AHSI SWIR data.Kaolinite,which is confused with muscovite in the ASTER image map,is distinguished from muscovite in the Hyperion and AHSI image maps.The main results and conclusions can be drawn as follows:(1)Sentinel-2A MSI and ASTER images are used to map the spatial distribution of hydrothermally altered minerals.The results acquired using the PCA and the anomaly-overlaying selection method can omit interference-induced false anomalies.The hydroxyl-bearing minerals in the Sentinel-2A MSI image map correspond to the Al-OH group minerals extracted from the ASTER data.Sentinel-2A MSI data can be applied for mapping minerals.When three machine learning algorithms are applied to analyze the ASTER SWIR data,the classification using the V-ELM algorithm has the highest overall accuracy(OA =98.33%),followed by the KELM(98.22%)and RF(95.18%)classifications.Compared with the classification applied to the ASTER SWIR data,the classification applied to the ASTER SWIR-TIR data can achieve a higher OA(99.01%).The ASTER TIR bands can show good identification of Mg-OH group minerals,but cannot discriminate Al-OH group minerals.(2)Hyperion and AHSI data can be used to map muscovite,kaolinite and epidote.Three machine learning algorithms are applied to analyze the Hyperion SWIR data.The classification using the V-ELM algorithm has the highest OA(OA=97.31%),followed by the KELM(96.48%)and RF(95.52%)classification.The muscovite and kaolinite in the Hyperion image map correspond to the Al-OH group minerals extracted from the ASTER data.Three machine learning algorithms are also applied to analyze the AHSI SWIR data.The classification using the KELM algorithm has the highest OA(90.60%),followed by the V-ELM(89.55%)and RF(82.26%)classifications.The AHSI data results are spatially consistent with those of the Hyperion data.(3)The distribution patterns of the hydrothermal alteration zones show a good agreement with the alteration zones identified by field investigations.The image processing results were validated by field investigations and spectral reflectance measurements.The results show that the known copper deposits are recognizable by the methods.Three mineralization prospects associated with intermediate-acid intrusions are discovered in accordance with the results.