Estimating hydrocarbon content using hyperspectral remote sensing at Qaidam Basin, China

Corresponding to the Hyperion hyperspectral remote sensing image obtained in the Three Lakes region in the eastern part of Qaidam basin in which gas reservoirs located, 25 samples of soil were collected throughout the area covered by the image and the spectra of all samples were measured. A geochemical analysis was conducted in the lab for the content of acidolysis hydrocarbon in soil samples. Univariate correlative analysis was carried through between spectral variables in two types and total acidolysis hydrocarbon (TAH) content, and the linear and non-linear correlations between 7 characteristic parameters with higher correlation coefficient and TAH content were investigated using 6 univariate regressive models. Further, stepwise regressive analysis techniques were used to study the relationship between original and first-order derivative reflectance data and TAH content, the results show that estimation accuracy was significantly improved with first-order derivative spectra but with larger relative error, the regressive equation of reflectance spectra is the best estimating model for TAH content. Finally, the derived optimal estimation equation was applied to the Hyperion hyperspectral image for a distribution map of surface TAH content which was tested using measuring values.