| This study took the southern area of Beijing’s Daxing district as the study areas. After processing the data of cropland utilization, soil type, soil testing and formula fertilization data and other soil survey data, there were 3 different data forms generated, which were point data, area data and soft data of SOM content. Based on those different forms data, constructed different estimation methods to estimate the soil organic matter content of cropland and Evaluated the accuracy of different estimation methods, and chose the best one. Objective is to provide the scientific guidance and technical processes for the quantitative and accurate study on the spatial variation of regional cropland SOM content. The results were as follows:1 These estimation methods based on the sampling point data, area data and soft data predicted the spatial distribution of SOM were reasonable. And they effectively showed the spatial distribution characteristics of SOM.2 Based on the cropland sampling points data, the study estimated the SOM of cropland though the methods of Ordinary Kriging, Residual Kriging and Point-and-Area Kriging. The accuracy analysis which estimated the sampling points in the validation polygons showed:1) The Index of agreement and Pearson correlation coefficient’s merits order was Point-and-Area Kriging(PAK), Residual Kriging(RK) and Ordinary Kriging(OK). The RMSE in ascending order arranged PAK, RK and OK, which illustrated that PAK was the best methods of three methods.2) The proper supplementary of environmental variables obviously increased the estimation accuracy of SOM.3) Based on the analysis on accuracy of estimation methods of the map spot area data, evaluation indexes were not consistent. Therefore, taking area data of validation polygons to evaluate the cropland SOM content accuracy of methods which constructed by the data of sampling points was not suitable.3 Based on area data of cropland soil organic matter, using the methods of Area Kriging and the Sequential Gaussian Condition Simulation to estimate the SOM content of cropland. The result showed that the optimal method was Sequential Gaussian Condition Simulation. Based on the sampling points data and area data of validation polygons, the accuracy of two validation methods were consistent. It turned out that the Sequential Gaussian Condition Simulation was better than Area Kriging, whether the Index of agreement, Pearson correlation coefficient or RMSE. The multiple random simulation and fitting variogram function accuracy was the reason that Sequential Gaussian Condition Simulation was better than Area Kriging. Both two validation methods were feasible for evaluating the accuracy of methods built by area data.4 Based on soft data of cropland soil organic matter, using the methods of "GIS linkage based on soil type" and Nearest-Neighbor Area to estimate the SOM content of cropland. The study showed that based on the accuracy validation of the sampling point of validation polygons, the "GIS linkage based on soil type" was better than the Nearest-Neighbor Area, but the same when estimation methods based on area data of validation polygons. By comparing the soft data models with Kappa coefficient, "GIS linkage based on soil type" was slightly better than the Nearest-Neighbor Area. The result showed that the optimal estimation method was "GIS linkage based on soil type" when constructing estimation method based on soft data of cropland SOM. It is limited by using factors such as soil utilization to improve the estimation method accuracy, and still necessary to explore more assistant variables and estimation methods to improve the accuracy of soil organic matter content. |
PDF Full Text Request