Study On Remote Sensing Information Model Of The Vegetation In KERQIN Desertification Areas

The study is the part of the National Natural Fund project: "Research on quantitative retrieval of the main factors of desertification and assessment of remote sensing information"(No.30371192).Desertification is the globality problem, it has impacted the development of economy and the entironment, in order to father the desertification ,it is necessary to know what is the desertification and how to evaluate the degree of desertification. At present, the mostly factors for the evaluate the degree of desertification are the rate of soil moisture, vegetation factor. In order to offering the base for evaluated the desertification , the quantitative retrieval and the Remote sensing information model of vegetation factor was studied in this paper.Lots of data was obtained in this survey and measuring of the field, in Naiman country, Inner Mongolia, by applying the 3S and TM data. Deal with the RS map, the model of DEM was established. At the same time, the remote sensing image was extract pretreatment in geometry emendation and atmosphere, All of these were the stability base of the quantitative retrieval of the vegetation factor.With the quantitative retrieval of vegetation cover, the experiential model is built by NDVI and practical data. According to the theory of the mixed pixel decomposing, NDVIsoyi and NDVIveg are confirmed and the dimidiate pixel model is built. With the retrieval of biomass, according to the relative analysis of the several vegetation index and practical data , NDVI and practical data are selected to build the experiential model ,at the same time, the influence of the terrain factor ,,biology factor .meteorological factor and biomass in considered. The factors of aspect and slope are got from the DEM model ,the factors of the soil type and vegetation type are attained from TM data through visibly interpreting. The meteorological factor of local material can form the synthetical factor and built Geo-RS model of biomass. The accuracy is compared by several models. The research results showed that: The accuracy of dimidiate pixel model is better than that of the experiential model during the vegetation retrieval.The curve degree of the cubic curve fitting is the best during the three built experiential models, but the experiential model is very simple and used widely. During the retrieval of the biomass , the experiential model of NDVI and practical is simply and used more wiedly. The cubic curve fitting is better than the quadratic fitting model, the quadratic fitting model is better than unary linear regression. The Geo-RS model of biomass is not only considered the vegetation spectral ,but also the other relative information, the accuracy of this model is better than the cubic curve fitting, but workload is large and the model is complex .