In this study, the relationships between environmental variables and SOC were analysed using spatial autoregression (SAR) and geographically weighted regression (GWR) to quantitatively model the spatial variability of SOC and its environmental impact, comparing with ordinary least square (OLS) regression. The results demonstrated strong correlations between SOC and auxiliary variables. As to soil properties, bulk density and available iron played significant roles. As to topography factors and location factors, latitude, elevation, slope and roughness were the most important affecting factors. Local clustering of soil organic carbon occurred mostly on core transition zone. Plus, SAR model has a better goodness of fit than OLS regression and its estimated value showed a similar trend with the observated values of SOC. Additionally, weak spatial patterns were detected after modeling. Thanks to the flexibility to adjust the weighting function and the bandwidth, GWR model has a better detection of spatial variability of SOC than the others. On model assessment, the residual sum of squares of GWR-1 and GWR-2 were reduced by 20.717% and 8.799%, comparing with OLS's, respectively; the AIC values of SLM, SEM, GWR-1, GWR-2 were reduced by 5.108, 5.391, 19.88 and 11.751, respectively. In addition to the spatial autocorrelation, soil properties and environmental factors can significantly explain the heterogeneity of SOC. The auxiliary variables and spatial regression model used here indicated the variability of SOC may propose a certain basis for further exploring the synergies and quantitative analysis of SOC. This study may pave a way for ecological restoration, indicating changes in the environment and the planning of the typical citrus area.
|Journal||Zhongguo Huanjing Kexue/China Environmental Science|
|Publication status||Published - 1 Dec 2015|