Search

Different methods have been used to predict soil properties and obtain continuous spatial distributions. However, studies that compare the accuracy of these methods are scarce. In addition, few studies considered human activity variables as well as landscape ecology variables to predict soil properties. The objective of this research was to determine the optimal method for predicting soil properties using human activity and landscape ecology variables in regions with complex topographies. The study results are as follows. (1) Introducing land-use type as human activity variable to the soillandscape model can improve the accuracy. (2) In the study region, the mean absolute error (MAE) and mean relative error (MRE) of the soil organic matter, pH, and soil particle size predicted by the soillandscape model were lower than those predicted by ordinary Kriging method. (3) Specifically, the regions in which the soil-landscape prediction model was more accurate were all concentrated in the central and west hilly regions in the study region, whereas the regions in which the geostatistical interpolation was more accurate were mainly in the east plain region. The differences in the results are based on the topography of the study area. [ABSTRACT FROM AUTHOR]