论文部分内容阅读
为快速准确获取省域尺度下土壤有机质的空间分布状况。以江西省2012年测土配方施肥项目采集的16 582个耕地表层(0~20 cm)土壤样点数据,借助四方位搜索法、地统计学和遥感影像分析技术提取环境因子和邻近信息作为辅助变量,构建基于地理坐标与辅助变量的BP神经网络模型和普通克里金法结合的方法(BPNN_OK)、基于地理坐标与辅助变量的RBF神经网络模型和普通克里金法结合的方法(RBFNN_OK)和普通克里金法(OK法)3种方法,模拟省域尺度下耕地表层(0~20 cm)土壤有机质的空间分布。对2 416个验证样点进行独立验证的研究结果显示:基于辅助变量的神经网络模型较普通克里金法有较大提升。BPNN_OK法对土壤有机质预测结果的均方根误差、平均绝对误差、平均相对误差较OK法分别降低了2.76 g/kg、2.34 g/kg、9.83%,RBFNN_OK法较OK法分别降低了2.70 g/kg、2.29 g/kg、9.61%。研究显示,基于辅助变量的神经网络模型与OK法结合的方法明显地提高了土壤有机质空间分布模拟精度,并且存在改进和提高的空间。
In order to quickly and accurately obtain the spatial distribution of soil organic matter under the provincial scale. Taking 16 582 soil sample data of cultivated land surface (0 ~ 20 cm) collected by Jiangxi Province Soil Testing and Fertilization Project in 2012, using four directions search method, geostatistics and remote sensing image analysis technology to extract environmental factors and neighboring information as auxiliary (BPNN_OK), a RBF neural network model based on geographic coordinates and auxiliary variables, and a combination of ordinary Kriging method (RBFNN_OK), a BP neural network model based on geographic coordinates and auxiliary variables, And the common kriging method (OK method) to simulate the spatial distribution of soil organic matter in the soil surface (0 ~ 20 cm) under the provincial scale. The independent verification of 2 416 verification samples shows that the neural network model based on the auxiliary variables has a higher promotion than the ordinary Kriging method. Compared with the OK method, the root mean square error, average absolute error and average relative error of BPNN_OK method decreased by 2.76 g / kg, 2.34 g / kg and 9.83%, respectively. The RBFNN_OK method decreased by 2.70 g / kg, 2.29 g / kg, 9.61%. The research shows that the combination of ANN model and OK method can significantly improve the simulation accuracy of spatial distribution of soil organic matter, and there is room for improvement and improvement.