省域尺度土壤有机质空间分布的神经网络法预测

发布时间：2018-10-04 22:23

【摘要】：土壤有机质空间分布预测方法研究对指导省域尺度下土壤有机质空间插值模型选取和精度优化具有重要意义。以江西省为例,利用BP神经网络模型与普通克里金结合的方法(BPNN-OK)、RBF神经网络模型与普通克里金结合的方法(RBFNN-OK)以及普通克里金法(OK)3种方法,预测省域尺度下耕地表层(0~20 cm)土壤有机质的空间分布。16 109个土壤样点随机分成12 887个建模样点,3 222个测试样点。结果表明:在省域尺度下,BPNN-OK法、RBFNN-OK法较OK法在土壤有机质空间预测精度上有较大提升,三者的预测精度为BPNN-OKRBFNN-OKOK。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28.66%、30.71%、34.76%,RBFNN-OK法较OK法分别降低27.76%、29.74%、33.71%。在省域尺度下,神经网络模型与普通克里金结合的方法能很好地捕捉土壤有机质的复杂空间变异关系。研究结果可指导江西省土壤有机质空间插值模型选取。
[Abstract]:The prediction method of soil organic matter spatial distribution is of great significance to guide the spatial interpolation model selection and precision optimization of soil organic matter in provincial scale. Taking Jiangxi Province as an example, using the BP neural network model and the ordinary Kriging method (BPNN-OK), there are three methods, the RBFNN-OK method and the (OK) method, which are combined with the common Kriging neural network model and the common Kriging neural network model, respectively. The spatial distribution of soil organic matter on the surface of cultivated land (0 ~ 20 cm) was predicted. 16 109 soil samples were randomly divided into 12 887 pattern sites and 3 222 test sites. The results showed that the precision of spatial prediction of soil organic matter by BPNN-OK method was much higher than that by OK method at the provincial scale. The accuracy of the three methods was the root mean square error and the average absolute error of BPNN-OKRBFNN-OKOK.BPNN-OK method for soil organic matter prediction. The average relative error was decreased by 28.660.71% and 34.76% respectively compared with the OK method. The RBFNN-OK method was 27.76% lower than the OK method (29.74%) and 33.71% lower than that of the OK method. On the provincial scale, the neural network model combined with the ordinary Kriging method can capture the complex spatial variability of soil organic matter. The results can guide the selection of spatial interpolation model of soil organic matter in Jiangxi Province.
【作者单位】：江西农业大学国土资源与环境学院/江西省鄱阳湖流域农业资源与生态重点实验室;南方粮油作物协同创新中心;
【基金】：国家自然科学基金项目(41361049) 江西省自然科学基金项目(20122BAB204012) 江西省赣鄱英才“555”领军人才项目(201295)
【分类号】：S153.621

【相似文献】