基于数据挖掘方法的陆面过程模型参数本地化分析-localization analysis of land surface process model parameters based on data mining method.docx

AbstractThe accuracy of the land surface data sets have a more significant impact on the simulation results accuracy of the Community Land Model 3.5(CLM3.5), especially on a small area in the simulation experiment, the effects using low-resolution surface data are more pronounced; this thesis goal is to build a set of high precision surface data CLM3.5 requirements, based on the Guangxi high-resolution land surface data combined with high-resolution MODIS land cover data, by applying decision tree and grid data statistics technology, the resolution of this data sets is 0.1° × 0.1°. The CLM3.5 was run to get high precision soil moisture simulated results using above surface data sets in Guangxi, through qualitative and quantitative analysis to prove the superiority and reliability of the simulation results. In this paper, the main contents are as follows:MODIS land surface data to build CLM3.5 surface data sets. This thesis first used the Guangxi high-resolution MODIS land cover data to build CLM3.5 surface. Used the decision tree classification algorithm to solve the problem the MODIS data distinguish land cover type and CLM3.5 surface cover types do not match, CLM3.5 surface data allows the coexistence of a variety of cover types on a single grid, for the mixed cover types using simple weighted processing when classification results were integrated into the grid scale process, so as to construct a 0.1° × 0.1° resolutionsurface data sets SDM; and compared with the CLM3.5 initial surface data SDC，itwas found that the SDM than the SDC more accurate.Guangxi high-resolution land cover data to build CLM3.5 surface data sets. The process of building surface data sets in Guangxi high-resolution coverage data, decision tree algorithm to classify the type of coverage too, the difference is the resolution of Guangxi high-resolution coverage data this article introduced is 30m, the classification results were integrated into the grid scale process considering a variety of land