| Nowadays, global climate change has become a hot issue in the world. Among lots of studies for this issue, the study of LUCC has particular significance. This is because the human activities play an important role in many factors which cause global climate change, but the effect of human activities on whole earth system is complex, and as the most obvious landscape sign of earth surface system, land cover has a far-reaching influence on biogeochemical cycle, ecological complexity, global radiation balance and water cycle, so land cover research makes revealing the rules of global ecological environment changes intuitive. To get the information of land cover conditions promptly, accurately and effectively is the basis of land cover research. With the development of remote sensing technology and the extension of remote sensing technology application, the researchers select the appropriate characteristic parameters from the mass of remote sensing data for the research on land cover classification, the application of remote sensing data in research has many advantages, such as timeliness, rapidity, high efficiency,automation, and so on. These advantages make it become hot research fields.At present, there are many systems and methods to classify land cover by using remote sensing data, most are based on their own research objectives. And there is no unified classification system and method in the world, so how to make appropriate classification systems and methods with spectrum and temporal has become a key issue.This study takes the Beijing-Tianjin-Hebei region as an example, builds a land cover classification system combined with the natural economic features of Beijing-Tianjin-Hebei by using the dichotomy principle of the United Nations Food and Agriculture Organization(FAO) and referring to the land classification systems of IGBP, UMD and Liu Yonghong. With extracting MODIS/EVI time series data, harmonic remainder and amplitude of Beijing-Tianjin-Hebei region by harmonic analysis, MODIS time series data can reflect such as farmland, woodland, shrubs, lawn, garden and other vegetation cover’s phenological characteristics. Based on extracting and calculating the first 7 bands of reflectivity data of MODIS by MOD09, this paper come up with three Characteristic bands data: NDWI(Normalized Differential Water Index), MNDWI(Modified Normalized Difference Water Index) and NDSI(Normalized Difference Snow Index).My page construct three schemes of CART decision tree to investigate the land cover classification of Beijing-Tianjin-Heb ei region.The three schemes of band combination were Scheme1-23 time phases of EVI;scheme2-Scheme1+ref1-7+MNDWI+NDSI+NDWI;and scheme3-scheme2+The harmonic parameters.Results showed that the overall classification accuracy of three schemes can reach to 85.20%, 88.78% and 92.33% respectively, their Kappa coefficient were 0.8396, 0.8932 and 0.9140 respectively.Combining with general classification situation analysis of Beijing-Tianjin-Hebei region and MODIS Land cover classification product, the results show that the classification is in line with the natural conditions of Beijing-Tianjin-Hebei region.The innovation of this paper is:(1)Set up a land cover research system for the Beijing-Tianjin-Hebei region.(2)Refactoring EVI time series data by using harmonic analysis and extracting characteristic parameters of harmonic analysis make it more favorable to analyze the phenological characteristics of vegetation cover. At the same time, it makes the classification more in line with the actual situation in the study area.(3)The use of MODIS reflectance data and calculating humidity index makes land cover classification more accurate.(4) This paper creates different CART decision trees for classifying the land cover of Beijing-Tianjin-Hebei region by using classification feature combinations, analyzes classification accuracy and selects optimal classification feature combinations, then gets the land cover classification of Beijing-Tianjin-Hebei region. |
PDF Full Text Request