Soil distribution modeling using inductive learning in the eastern part of permafrost regions in Qinghai–Xizang (Tibetan) Plateau

Soil–landscape models serve as a basis for understanding the relationships between soils and landscape, and such models also allow us to study soil distribution, and classification. To understand the distribution of soils and landform relationships in permafrost regions, we developed a soil–landscape model by using See5.0 decision tree software in the Wenquan region of Xinghai County, in the eastern part of permafrost regions of Qinghai–Xizang (Tibetan) Plateau (QTP). The model was based on soil pedon data from 2009. Nine environmental factors closely related to permafrost-affected-soil formation were selected as variables for the model: land surface temperature for warm season and cool season, elevation, slope gradient, slope aspect, planform and profile curvatures, wetness index and NDVI. A 5-fold cross-validation method was applied to verify the effectiveness of the soil–landscape model. The model results were consistent with field observations, and the slope was the most strongly correlated factor with soil type of the nine environmental variables. The soils in the modeled area are mainly Ustic Cambosols and Ustic Isohumisols, which cover about 60.0% and 25.6% of the total area, respectively. Gelic cambosols occur mainly in permafrost region, while Ustic isohumosols occur in the transition region between permafrost and seasonally frozen ground. Further studies are required to utilize the soil–landscape model to predict the spatial distribution of soil types over the QTP.