Spatial Mapping Of Soil Organic Matter And It's Driving Factors Study

Spatial prediction is an important approach to obtain location-specific values of soil organic matter(SOM),which is an important figure of soil fertility and farmland management properly.This study comparing different digital soil spatial mapping methods of SOM to get better prediction accuracy,revealing the spatial non-stationarity characteristics of environmental covariates and the spatial scale of different environmental covariates at the same time.In this study,we use a digital soil spatial mapping method that is combination of multiscale geographically weighted regression model with simple kriging of the residuals(MGWGK)for mapping SOM in the study area where Gaoping city.Compared the performance of MGWRK with those of Ordinary Kriging(OK),Multiple Linear Regression(MLR),Regression Kriging(RK),Geographically Weighted Regression(GWR),Geographically Weighted Regression Kriging(GWRK),Multiscale Geographically Weighted Regression(MGWR),and to explore the relationship between influence factors and SOM on the influence degree and space scale.Height,slope,aspect,plan curvature,terrain position index(TPI),terrain ruggedness index(TRI),topographic wetness index(TWI),net primary productivity(NPP),evapotranspiration(ET),normalized difference vegetation index(NDVI),annual average precipitation,annual average temperature were selected as environmental covariables in the modeling,through the stepwise regression method.In the multiple linear regression(MLR),the formula had statistical significance and pass the multicollinearity test.The following results can be obtained:1.The spatial distribution of the residuals of the MGWR model is more different than that of the residuals of the other two models.MGWR model effectively reduces the residual level of MLR model and GWR model from the perspective of range.2.The overall soil organic matter in the study area showed a trend of high in the east and low in the west,which was specifically manifested as low in the western banding area,high value aggregation in most of the eastern areas,and the highest level of organic matter in the lowest elevation(central region)of the study area.3.MLR model has the worst spatial prediction effect and the greatest distortion degree among the seven models.RK,GWRK and MGWRK have the same effect on soil mapping.4.Radius indexes from large to small are: MLR model >GWR model >MGWR model >MGWRK model >RK model >OK model >GWRK model.According to Radius index,MLR model has the worst spatial prediction accuracy,while GWKR model has the best spatial prediction accuracy.The spatial prediction accuracy of MGWRK was 94.63% of RK,93.88% of OK,and 83.58% of GWRK.5.MGWR model is more practical and accurate in describing the relationship between environmental covariables and organic matter than GWR model,and MGWR is better than GWR model in describing spatial non-stationary characteristics.The MGWR model reduces the residual level of the GWR model,and the GWRK model has the highest spatial prediction accuracy among all models.The MGWR model performs better than the GWR model in terms of the spatial non-stationarity of the environmental covariables.