Spatial Prediction of Soil Organic Matter Using Terrain Attributes in a Hilly Area

Abstract

Topography is one of major factors influencing soil properties at the landscape scale, especially in hilly areas. Derived terrain attributes based on digital elevation models (DEMs) may be used for soil spatial distribution prediction. Accurate estimate of spatial variability of soil organic matter (SOM) is critical to evaluate soil quality as well as assess the C sequestration potential. However, little is known about spatial variability of SOM in the hilly areas of the sub tropical zone of southwestern China. The current study addresses spatial distribution of SOM and its characteristics on landscape scale. SOM was significantly correlated with the terrain attributes slope (r = -0.57), elevation (r = -0.46) and topographic wetness index (r = 0.30). Geostatistical analyses indicate a moderately structured spatial dependence of SOM. The use of terrain attributes (slope and elevation) in a multiple linear regression accounts for 29.6% of the variance of SOM. Multiple linear regression (MLR), ordinary kriging (OK), and regression kriging (RK) were compared to select the best prediction method. Root mean square errors (RMSEs) show that RK outperforms MLR and OK. Compared to MLR and OK, the spatial prediction of SOM using RK is improved by up to 72.10% and 15.69%, respectively.