A Study On Comprehensive Assessment Of Desertification Status Based On Remote Sensing And Principle Component Analysis

Land desertification, one of the most seriously ecological and environmentalproblem on the globle scale, is seriously threatening the ecological security in ourcountry, restricting social and economic sustainable development. The desertificationevaluation is an important content of land degradation research, and also a main taskof environmental monitoring in arid areas. Remote sensing technology with its largeobservation area and rich information, has a monopoly advantage in desertificationresearch. Aiming that some of the desertification evaluation indexes couldn’t bedirectly obtained by means of remote sensing, and simple visual interpretation ordirect classification evaluation methods were so strongly subjective, a comprehensiveevaluation method has been put forward in this research, which combined principalcomponent analysis with remote sensing technology.First, a typical desertification region in Huang river area was selected as a studyarea, Landsat ETM+images as data sources. On the basis of image radiationcorrection and743color composite, severe desertification area were distinguished byvisual interpretation based on the sand dunes texture and vegetation coveragesituation, and the sample area were arranged as priori data.Second, the modified soil regulating vegetation index, surface albedo, landsurface temperature and soil humidity, which were closely related to soil andvegetation features, were selected as the desertification evaluation indexes of thisresearch area, comprehensively considering the representative objectivity,comprehensiveness, diversity, and the maneuverability of the remote sensing imagesautomatically extracting of evaluation indexes; These indexes were inversed by theappropriate remote sensing methods that were selected considering features of thedesertification region.Third, the principal component analysis was carried out on the sample area indexinversion data. According to principal component contribution rate and thecumulative contribution rate, the principal components were selected as new evaluation factors, and a comprehensive evaluation model of desertification wasestablished by factor weighted summation, the characteristic values as weights.Finally, each pixel value referring to desertification level was computedaccording to the comprehensive evaluation model in the study area; desertificationlevel distribution was obtained by average distance method. Precision test and erroranalysis were carried out by the method of ground area of interest，and the overallevaluation accuracy reached88.96%, classification results were satisfactory.These desertification evaluation factors selected in this study were directlyobtained by remote sensing means, not only reflecting the soil and vegetationinformation comprehensively, but also decreasing the workload of field investigation;At the same time, the evaluation factors and weighted coefficients of thecomprehensive evaluation model were calculated on the statistics of some sample data,which was a kind of objective empowerment method, solving the problem ofstrong subjectivity of visual interpretation.