干旱区生态环境知识资源中心

Based on unmixing model and using the Thematic Mapper (TM) image as well as obtaining end-members by ground spectral measurements, quantitative retrieval of information on sparse vegetation coverage in oasis-desert transitional area in Minqin, Gansu was done. The results showed that a wide band of TM images can be used in extracting sparse vegetation coverage of arid regions. Three components were needed to build unmixing model and there includes light, vegetation - not light and vegetation - bare soil. And the unmixing model has a high correlation with measured vegetation coverage. Methods used to choose the end-members had certain influence on estimating the vegetation coverage. Monte Carlo method recorded 0.64 as coefficient of determination (R2) 3.8 as root mean square error (RMSE) and therefore was more accurate than the root mean square error minimization method.