Retrieval Of Rocky Desertification In Shilin County Of Yunnan Province Based On Geometric-Optics Model

Karst rocky-desertification is one of the most serious problem in karst regions,southwest China.Monitoring the rocky desertification by means of remote sensing is the basis for studying the causes and development trend of rocky desertification.Using the Landsat OLI data of November 20,2015,the traditional vegetation indices(VIs)and linear spectral unmixing(LSU)are assessed to extract the key indicators of karst rocky-desertification,and endmember selections are assessed too.Then historical rocky-desertification of Shilin Country from 1996 to 2015 was retrieved The results are as follows:1.the semi-automatic PPI method and hourglass tool are batter than SMACC.2.The field spectral endmember showed a certain feasibility in the remote sensing extraction of rocky desertification information,the relative error of bare rock rate was 22.7%,the coefficient R2 was 0.382.Compared with it,the linear spectral unmixing based on image endmember is the best in all methods,the absolute error is 0.003,the relative error is 1.2%,RMSE is 0.174,and the R2is 0.467.3.The NDVI is the best vegetation index for detecting karst green vegetation,and the correlation coefficient is R2=0.54.SAVI and MSAVI show that soil has little effect on rocky desertification(R2=0.50 and R2=0.49).But NDVI has lower linear relationship with rock fraction and soil fraction,R2 of 0.166 and 0.021.4.The rocky desertification area of Shilin decreased from 813.1842 to 367.8201 km2,and the area ratio decreased from 47.10%to 21.30%.The potential rocky desertification decreased by 100.96 km2,Mild rocky desertification reduced by 151.51 km2,moderate rocky desertification decreased by 112.71 km2,and severe rocky desertification decreased by 80.18 km2.