• 论文 •

基于地理探测器的区域土壤耕层有机碳含量 多元复合模型构建——以珠三角核心区为例

1. （1．中山大学地理科学与规划学院//广东省城市化与地理环境空间模拟重点实验室，广州 510275； 2．华南农业大学资源环境学院，广州510642；3．中山大学 新华学院，广州 510520）
• 出版日期:2018-07-05 发布日期:2018-07-05
• 作者简介:任向宁（1978―），男，河北正定人，博士生，主要研究方向为土地资源开发利用与保护，（E-mail）xnren@scau.edu.cn。

Construction of Multivariate Composite Calculation Model of Soil Organic Carbon Content in Plough Horizon Based on Geodetector

REN Xiangning1,3，DONG Yuxiang1,2

1. （1．Guangdong Provincial Key Laboratory of Urbanization and Geo-simulation，School of Geography and Planning，Sun Yat-sen University，Guangzhou 510275，China；2．Xinhua College of Sun Yat-sen University，Guangzhou 510520，China；3．College of Natural Resources and Environment，South China Agricultural University，Guangzhou 510642，China）
• Online:2018-07-05 Published:2018-07-05

Abstract: The accurate calculation of organic carbon content in regional soil plough is very important for the study of global carbon cycle, but its influence factors are many, the spatial variability is stronger, and the accuracy of the existing interpolation calculation method is low. The Geodetector provides a new method of spatial differentiation and factor detection, which can effectively measure the contribution of the spatial differentiation of soil organic carbon content in plough layer. Combining the traditional Cooperative-Kriging interpolation method with the Geodetector, according to the detection results of the contribution of the geographical detector to the influencing factors, this paper constructs a multivariate composite model based on hierarchical Cooperative-Kriging to calculating regional soil organic carbon content in plough layer. The core area of the Pearl River Delta is used as the study area. Through the setting of soil plough organic carbon content interpolation set and verification set, Ordinary Kriging, Geographically weighted regression-Kriging, BP neural network model-Kriging and Multielement composite model are used to cross validation. The results showed that: 1) the spatial variation of soil organic carbon content in the core area of the Pearl River Delta is related to topography, hydrology, soil and farmland utilization, and the contribution of different factors is different, the contribution of various factors (q Statistics) is from 0.076 to 0.201, among which the contribution of soil physical and chemical properties and farmland utilization methods is greater than that of topography, hydrology. The objective difference of contribution of different factors has an important impact on the accurate estimation of soil organic carbon content. 2) On the basis of the detection of the geographic detector, the factor contribution sequence is introduced into the Multielement composite model, which inhibits the interpolation noise generated by the interference among the factors, and effectively avoids the low value overestimation or the high value underestimation. Through cross validation, the mean error (ME), mean absolute error (MAE) and root mean square error (RMSE) between the results of soil organic carbon content calculated by Multielement composite model and the validation set are less, the correlation (Pearson R) and consistency (Kappa) are higher, and the accuracy of Ordinary Kriging, Geographically weighted regression-Kriging, BP neural network model-Kriging is 16.62%, 23.86% and 37.33%, respectively, lower than that of Multielement composite model. 3) Multielement composite model fully considers the effects of structural and random factors such as topography, soil physical and chemical properties and farmland utilization on the spatial differentiation of soil organic carbon content in plough layer. It breaks through the limitation of the number of auxiliary factors in the existing algorithms, and can cooperate more auxiliary variables in the calculation of soil organic carbon content in plough layer. At the same time, by setting weights in the model, the contribution of different factors to spatial variation is reflected, the order and coordination in the process of regional soil organic carbon content are taken into account, the uncertainty is reduced, the spatial difference is depicted more meticulously and effectively reveals the spatial variability of soil organic carbon content in plough layer. The construction of Multielement composite model has made a positive exploration for further improving the prediction accuracy of soil organic carbon content in plough layer, which provides a new research idea for the study of soil organic carbon in plough layer with strong variation characteristics under complex geographical environment.