Abstract
Karst rocky desertification is a special kind of land desertification developed under violent human impacts on the
vulnerable eco-geo-environment of karst ecosystem. The process of karst rocky desertification results in simultaneous
and complex variations of many interrelated soil, rock and vegetation biogeophysical parameters, rendering it difficult to
develop simple and robust remote sensing mapping and monitoring approaches. In this study, we aimed to use Earth
Observing 1 (EO-1) Hyperion hyperspectral data to extract the karst rocky desertification information. A spectral
unmixing model based on Monte Carlo approach, was employed to quantify the fractional cover of photosynthetic
vegetation (PV), non-photosynthetic vegetation (NPV) and bare substrates. The results showed that SWIR (1.9-2.35μm)
portions of the spectrum were significantly different in PV, NPV and bare rock spectral properties. It has limitations in
using full optical range or only SWIR (1.9-2.35μm) region of Hyperion to decompose image into PV, NPV and bare
substrates covers. However, when use the tied-SWIR, the sub-pixel fractional covers of PV, NPV and bare substrates
were accurately estimated. Our study indicates that the "tied-spectrum" method effectively accentuate the spectral
characteristics of materials, while the spectral unmixing model based on Monte Carlo approach is a useful tool to
automatically extract mixed ground objects in karst ecosystem. Karst rocky desertification information can be accurately
extracted with EO-1 Hyperion. Imaging spectroscopy can provide a powerful methodology toward understanding the
extent and spatial pattern of land degradation in karst ecosystem.
