Identification Of Spatial Patterns And Influencing Factors Of Soil Organic Carbon Density In Shanxi Province

Soil is the largest source of organic carbon in terrestrial ecosystems,and the content of soil organic carbon not only affects global climate change,but also plays an important role in improving soil fertility and food production.Shanxi Province is located in the eastern part of Loess Plateau,which is an important energy and food production area in China,and a key area for soil erosion control and sand control in China.Understanding the distribution and influencing factors of soil organic carbon in Shanxi Province can not only enhance the soil carbon sequestration capacity of the region,but also alleviate soil erosion and improve ecosystem productivity.Based on the Generalized Additive Models(GAM)and Geographically Weighted Regression(GWR),this study proposes a GWR with smooth terms(SGWR)by considering both global spatial non-linearity and local spatial non-stationarity.The GWR with smooth terms(SGWR)was proposed to study Soil Organic Carbon Density(SOCD)in Shanxi Province,using topographic,climatic,vegetation and human activity factors as environmental variables,and applying multiple the effects of environmental variables on SOCD in Shanxi Province were investigated by using Multiple Linear Regression(MLR),GAM,GWR and SGWR.Then,Ordinary Kriging(OK)and Geographically Weighted Regression-Kriging(GWRK)were combined to jointly map the distribution of SOCD in Shanxi Province.Finally,the optimal model was screened using ten-fold cross-validation and SOCD prediction maps.The main results are as follows:(1)The overall SOCD in Shanxi Province is not high and unevenly distributed.The distribution of SOCD in Shanxi province can be predicted as follows: the SOCD gradually increases from west to east,which is consistent with the topographic direction;the SOCD is higher in higher terrain such as mountains,and lower in lower terrain such as basins.(2)Elevation,Slope,Normalized Difference Vegetation Index(NDVI),Gross Primary Productivity(GPP),Annual Precipitation(AP)and Soil Moisture(SM)were finally selected as modeling variables for MLR,GAM,GWR and SGWR by Pearson correlation coefficient and variance inflation factor.The analysis of SGWR variables showed that Elevation and Slope were local linear variables;NDVI,GPP,AP,and SM were global non-linear variables.In terms of the effects of individual variables,NDVI and GPP had the greatest influence on the distribution of SOCD,and Elevation and SM had the least influence on the distribution of SOCD.(3)MLR,GAM,GWR,SGWR,OK and GWRK,jointly predicted the spatial distribution of SOCD in Shanxi Province,and the validation results of the ten-fold cross-validation method showed that SGWR had the highest prediction accuracy,MLR had the lowest prediction accuracy,and GAM was similar to that of GWR and GWRK.(4)The SGWR method proposed in this study showed high advantages in the case study of SOCD in Shanxi Province,and is a more accurate method for exploring the influencing factors and spatial distribution of soil properties.