Numerical classification of forested soils in the high-mountain region of Southwestern China

Several numerical classification methods can be used to distinguish classes of soil. The authors used principal component analysis, cluster analysis, discriminant analysis, and a graph-theoretical method (minimum spanning tree). These methods divided 30 forest soil profiles from southwestern China into six groups based on their chemical and physical properties. The numerical classifications corresponded to a traditional soil classification. The groups formed by the numerical methods also corresponded to different forest vegetation types. In this analysis by numerical classification, not only can the soil profiles be divided into different groups, but the delimitation indexes of various soil properties can be established by discriminant functions.