For areas with no soil survey experts or soil maps available,knowledge about soil-environment relationships is a key to predictive soil mapping using knowledge-based approaches.An approach based on an unsupervised fuzzy clustering method (FCM,fuzzyc-means clustering) is recommended to acquire the knowledge.The method consists of four steps:1) define those environmental factors which play decisive roles in formation and development of soil,then build up environmental database;2) identify environmental niches (combination of environment conditions) by running FCM on the environmental database;3) allocate field efforts to relate unique environmental factor combinations to soil types;4) establish a soil-landscape model by analyzing the relationship established from the field samples.The model is then used to predict spatial distribution of soils.The method was applied in a study area in Northeast China and a soil map at the subgroup level was created for the area.The derived soil map was accurate up to 72% when verified using an independently collected field sample set.Comparing this method with the conventional soil mapping method (50%～60%) in USA in accuracy,it could be concluded that the approach is effective for prognostic soil mapping over areas with no local soil experts available.At the same time,the clustering method can improve efficiency of the field sampling.