The Study Of Grassland Desertification Information Extraction From TM Data

Because of the environment degradation and the human disturbance, the Hunshandake sand has become one of main sources of sandstorm which brings about sandstorm weather conditions in the area of Beijing and Tianjin. For understanding the desertification trends quickly in Hunshandake sand, based on TM data, the data of ground investigation and the ground spectrum data, the grassland desertification evaluation system was established by the desertification spectrum characteristic analysis in this article. At first, for the purpose of getting details of the grassland desertification information, the paper studied the method of grassland desertification information extraction by way of remote sensing image enhancement technologies, such as band choice, the principal component analysis, band ratio enhancement, the tasseled cap transformation, vegetation index. The vegetation coverage analysis models were built as well for the same reason. Finally, the paper analyzed the desertification dynamics from 2000 to 2004 year.The data of vegetation index, the second components of PCA, TM3/TM4 ratio image, and siol image were combined with TM5, TM4 and TM3 band in order to carry out the unsupervised and supervised classification. Accuracy assessment showed that the total accuracy was 88.46%, which meant that the extraction of grassland desertification information was acceptable.Since the vegetation coverage is the key factor for grassland desertification evaluation, this article modeled the vegetation coverage in terms of filed observation data set: C =115.482×NDVI+20.046,The correlation coefficient was 0.803, and F test showed a more significant result. According to the model, the vegetation coverage map was generated and then used for grassland desertification evaluation. The accuracy was 85.42%. Comparing with directly calculating vegetation coverage from vegetation index, this model showed good agreement with actual condition.On the bases of specific band selection, combing with enhanced images, the grassland desertification classification may get reasonable results. Moreover, coupled with vegetation coverage and soil water condition, the extraction of grassland desertification information was accomplished simply, easily and quickly.The study on Hunshandake sand desertification dynamics from 2000 to 2004 illustrated that the heavy desertification area expanded greatly in west Hunshandake sand. In east Hunshandake sand, the heavy desertification area was shrunk. Comparing with 2000, the heavy and moderate desertification area expanded in 2004. In general, the grassland desertification trend of Hunshandake sand showed worse.