Methods Of Global Land Cover Classification Based On Climate Partition

Land cover is an important basic information of ecological system and it is very important to Research on carbon cycle, habitat and biodiversity, and public health, so as to improve hydrological and atmospheric models. Land cover plays a very important role in the exchange of material and energy in the process of the earth's ecological system. And the rapid and accurate acquisition of large scale land cover cla ssification products has important significance for quantitative remote sensing research, remote sensing application and the application of the earth science research and ecological environment.Recent years, with the rapid development of remote sensing technology, the global land cover product can be quickly acquired by remote sensing technology. And high precision land cover classification is very important for quantitative remote sensing and remote sensing applications. At present the available global land cover classification method of products mostly use supervised classification and unsupervised classification. Unsupervised classification are based on spectral statistical features, which is less human intervention, higher degree of automation, but when the spectral features of types are similar, the classification error is larger. Supervised classification has a certain human intervention, and need to carry out the selection of sample points. However, the number of sample points and the subjective factors greatly affect the accuracy of the classification. And for the world such a large scale of land coverage, with a simple supervised classification and unsupervised classification cannot meet the demand for high precision land cover products.Therefore, this paper develops a classification method of global land cover classification based on zoning, the method of which is based on the long time series, which can capture the time variation of the surface with a high time resolution, thus the classification accuracy can be improved by using the difference of the ground feature in the time dimension. Due to the land cover are closely related to precipitation, soil, temperature and so on, this paper uses a wider application global climate zoning of the Koppen-Geiger.Using this method, this paper produced the 2012 year global land cover classification, and evaluated the accuracy of the typical region. Finally get that the overall classification accuracy was 89.69%, and the kappa coefficient is 0.88. It has greatly improved and good applicability compared to the highest accuracy of the existing global products which is 75%. Because the terrain of China is more complex, the terrain is broken, the overall accuracy of global land cover products in Chinese area at present were low. Therefore, this paper develops a classification method for Chinese Vegetation based on the special terrain of China, and completed the 2012 year national land cover classification using this method. In this paper, the classification results are evaluated by the method of stratified random sampling, found that the overall accuracy and Kappa coefficient of the classified products are greatly improved, of which the overall accuracy is 90.78%, and the Kappa coefficient is 0.86. And through the comparison with MODIS land cover data products, found that the accuracy of the product in the vegetation type increased by 60.39% than the MODIS land cover data products.