Quantitative description spatial distribution of soil texture in three dimensional is very important for the studies of land use and water and fertilizer management. In this study, we constructed 3-D Markov chain model based on observation of 109 soil profiles in piedmont alluvial plain on the North China, and then optimal prediction map and occurrence probability map of soil textural classes in 3-D were obtained based on the 100 stochastic simulations. The results indicated that one-dimensional embedded Markov chain model could well describe the spatial continuity and juxtapositional tendencies of different textural classes both in horizontal?and vertical directions. 3-D optimal distribution of soil texture could directly reflect the spatial distribution of soil textural classes. However, it existed obvious smoothing effect and could not describe the uncertainty of spatial distribution of soil textures. The probability distribution map was applied to analyze the uncertainty of spatial distribution of textural classes and it could effectively overcome this shortcoming.