基于主成分分析的协同克里格插值模型对土壤铜含量的空间分布预测

发布时间：2019-02-21 15:15

【摘要】：以湖北省宜都市红花套镇的柑橘产区为例,选取土壤重金属全铜含量为研究对象,利用Pearson相关系数方法选择与土壤铜含量具有较高相关性的土壤因子(全K、全Cr、CEC、全Al、全N(P0.05))作为辅助变量,利用主成分分析(PCA)对辅助变量进行降维求总得分处理后,与协同克里格插值相结合构建土壤铜含量空间模型(COKPCA),同时构建土壤铜含量的普通克里格插值模型(OK)和以全K作为辅助变量构建协同克里格插值模型(COKK),对不同的空间模型进行模拟精度的对比和分析之后提出最优空间模型方法,进一步分析土壤铜含量在研究区域的空间分布特征。结果表明:普通克里格插值(OK)、协同克里格插值(COKK)和与主成分相结合的协同克里格插值(COKPCA)3种模型的RMSE分别为7.044、6.907和6.309,R2分别为0.716、0.743和0.852,赤池信息量准则(AIC)分别为101.591、96.908和87.203。综合比较,COKPCA具有最优的模拟插值结果,其次为COKK,而OK则相对较差。
[Abstract]:Taking the citrus growing area of Honghuaaozhen, Yidu City, Hubei Province as an example, the heavy metal total copper content in soil was selected as the research object. The Pearson correlation coefficient method was used to select the soil factors (all K, all Cr,CEC, all Al,) which had a high correlation with soil copper content. The total N (P0.05) is regarded as the auxiliary variable. The spatial model of soil copper content, (COKPCA), is constructed by using the principal component analysis (PCA) to reduce the dimension of the auxiliary variable and calculate the total score of the auxiliary variable, combined with the cooperative Kriging interpolation. At the same time, the common Kriging interpolation model (OK) and the cooperative Kriging interpolation model (COKK), with all K as the auxiliary variable were constructed at the same time. After comparing and analyzing the simulation accuracy of different spatial models, an optimal spatial model method was proposed to further analyze the spatial distribution characteristics of soil copper content in the study area. The results show that the RMSE of the general Kriging interpolation (OK), co-Kriging interpolation (COKK) and the cooperative Kriging interpolation (COKPCA) model combined with the principal components are 7.044 ~ 6.907 and 6.309, respectively. The R2 was 0.716, 0.743 and 0.852, respectively, and the (AIC) of the information content criterion of the red pool was 101.591n 96.908 and 87.203, respectively. In a comprehensive comparison, COKPCA has the best simulated interpolation results, followed by COKK, and OK is relatively poor.
【作者单位】：华中农业大学资源与环境学院;武汉大学资源与环境学院;湖北省测绘工程院;
【基金】：国家自然科学基金项目(41371227,41101192) 中央高校基本科研业务费专项资金项目(2013JC016)
【分类号】：S153.6

【参考文献】