Based on the case study of Longyou County, Zhejiang Province, an approach was introduced to deducing and expressing spatial distribution of soil organic matter.This is a kind of data mining method or classification tree modeling method, which associates soil OM content with some extensive easily observable landscape attributes, such as landform, geology, landuse and remote sensing images, thus transferring the soil OM related information into a clear, quantitative, landscape factor-associated regular system.This system can be used to predict continuous soil OM spatial distribution.By analyzing the factors such as elevation, type of the rock, type of the soil, PC₄, PC₂, land uses, PC₃, PC₁, upslope contribut ing area, slope, exposure, plane curvature and profile curvature, the classification tree can predict distribution of soil organic matter levels.Among the factors, elevation, type of rock, type of soil, landuse, PC₄ and PC₂ (two indexes of vegetation coverage) are considered as the most important variables for predicting soil OM.Results of the prediction show a quite close relationship between soil OM contents and types of the landscape sorted by the classification tree with an accuracy of 81.1%.