The Research And Application Of Methods Used In Grassland Sandy Desertification Monitoring Based On TM Data

In the face of the serious desertification progress, efforts to monitor the progress and provide support for the relevant governments seem meaningful and reasonable.Although a lot of researches have been carried out, a relatively complete monitoring method system hasn't formed. In such circumstances, based on four periods of TM or ETM data, we choose the common used vegetation index and linear spectral mixture model to monitor the grassland sandy desertification in Zhenglan Banner. The main contents and results are as following:1. Comparison and selection of methods: Linear spectral mixture model retrivals the occupation of bare sand, vegetation index retrivals the vegetation coverage. The linear regression analysis results show that the correlation between vegetation component, got from LSMM, and vegetation coverage, got form field survey, is prominent, and R2 is 0.634.besides, the predicted precision of the bare sand component used for sandy desertification information extraction is more than 80%.The Log analysis model makes the vegetation index and the vegetation coverage a good correlation, and the goodness-of-fit for NDVI and MSAVI is 0.61 and 0.62 respectively. But in the independence test, we find that the desertification level in low vegetation coverage area often be overrated while be evaluated by vegetation coverage, which is got from vegetation index. Faced with this situation, we choose bare sand coverage as the main measurement, and choose the LSMM as the main method to get the bare sand coverage of each pixel.2.Apply of methods:In order to get better result, we choose the best endmember in the LSMM progress through pure pixel index (PPI) and minimum noise fraction. At last, we choose vegetation, bare sand and bare soil as the three main endmembers for LSMM. Then, according to the hierarchical system built in this paper, we can get four periods of desertification status map .Using the dynamic degree, confusion matrix and grid calculation methods, the status and spatial dynamic variation of grassland sandy desertification in Zhenglan Banner were analyzed with climate, population, and livestock data. We find that: Although the total area of desertification didn't increase before 2000,the area of the severe desertification land has increased obviously.Since then,the overall trend of grassland sandy desertification in Zhenlan Banner is reverse ,however, deteriorated area still exist in some place.The innovation of this paper:In order to reduce mistakes bring by soil coverage and vegetation environmental sensitivity, we use the bare sand coverage instead of vegetation coverage as the major indexes of the desertification evaluation.