MTMF Method for Hydromagnesite Determination Based on Landsat8 and ZY1-02D Data: A Case Study of the Jiezechaka Salt Lake in Tibet

Hydromagnesite (HM for short) is a natural carbonate mineral that is widely distributed. It is a high-quality mineral raw material for preparing flame retardants, magnesium oxides, heavy/light basic magnesium carbonates, magnesium hydroxides, and other Mg products. The evaluation of HM resources is of great significance to the development and utilization of salt lake resources. Using remote sensing technology to observe HM resources in salt lake can overcome the shortcomings of traditional prospecting methods such as discontinuous spatial data, time and effort. In addition, spectral analysis is the basis of hyperspectral remote sensing, and more detailed analysis of the spectral characteristics of HM is still lacking; therefore, we measured the reflection spectral curve of HM samples in the area of Jiezechaka by ASD FieldSpec4 short-wave infrared spectrometer and determined the mineral composition and content of HM samples by X-ray diffraction. The analysis indicated three and seven absorption valleys with high and low absorption intensities, respectively, in the reflectance spectral curves of the HM samples in the Jiezechaka area. Then, on this basis, the Landsat8 OLI multispectral data and ZY1-02D AHSI hyperspectral data were used as the basic data of remote sensing inversion. As the ZY1-02D AHSI data have 166 bands, which is much more than Landsat8 OLI data, it has a stronger ability to characterize the spectral characteristics of HM and can better meet the requirements of remote sensing inversion. The end member spectra were selected based on PPI and SMACC methods, respectively. The HM information around Jiezechaka Salt Lake in Tibet was extracted by the mixture tuned matched filtering method, and the regional distribution map of HM was made. A confusion matrix operation was used to compare the determination results of the two types of data. Among them, based on Landsat8 data, PPI method was used to obtain end members, and the overall accuracy of HM extraction results was > 69%, and the kappa coefficient was 0.688. Based on Landsat8 data, SMACC method was used to obtain end members, and the overall accuracy of HM extraction results was > 67%, and the kappa coefficient was 0.667. Based on ZY1-02D AHSI data, PPI method was used to obtain end members, and the overall accuracy of HM extraction results was > 76%, and the kappa coefficient was 0.743. Based on ZY1-02D AHSI data, SMACC method was used to obtain end members, and the overall accuracy of HM extraction results was > 73%, and the kappa coefficient was 0.728. It shows that the end members selected by PPI method can better express HM information in the image. Finally, through the overlay analysis of the four results, we concluded that HM outcrops in the Jiezechaka area are mainly distributed in the northwestern and southeastern regions of the lake. This study provides a rapid assessment technique for measuring HM information from salt lakes.