Prediction of soil organic matter in peak-cluster depression region using kriging and terrain indices

• Conventional soil survey in karst areas requires a lot of time, effort and hence relatively higher budget to perform.
• Soil organic matter was predicted by using regression kriging with terrain indices and other several methods.
• The prediction precise of multiple linear stepwise regressions, ordinary kriging and regression kriging were compared.
• Regression kriging might improve soil organic matter prediction precision by up to 15%.

In an ecosystem, soil organic matter (SOM) is an important indicator of soil fertility and soil quality. Accurate information about the spatial variation of SOM is critical for sustainable soil utilization and management in karst areas. This study was conducted to evaluate and compare spatial prediction of SOM by using multiple linear stepwise regressions (MLSR), ordinary kriging (OK) and regression kriging (RK) with terrain indices. Soil organic matter was estimated by using 149 observation data for Guohua Karst Ecological Experimental Area, a 10 km2 study area in Guangxi Zhuang Autonomous Region, Southwest China. Correlation assessment between SOM and terrain indices showed that there was a significant correlation amongst 5 of the 8 pairs of indices. In the analysis of variance (ANOVA) applied in MLSR for SOM using terrain indices, two models of independant terrain indices were set to perform the models of MLSR. Relief degree of land surface (RDLS) entered into the regression equation for the first model (M1), whereas RDLS and distance to ridge of mountains (DRM) entered into the regression equation for the second model (M2). The assessment showed that the RK method combining with terrain indices obtained a lower mean predication error (ME) and root mean square prediction error (RMSE). Compared with OK, the application of RKM1 and RKM2 resulted in relative improvement (RI) of 13.87% and 15.61%, respectively. This study showed that including terrain indices in regression kriging might improve SOM prediction precision by up to 15% in the karst mountains.