Conventional soil maps are the major data source for information on soil spatial distribution which is essential in watershed management and eco-hydrological modeling. Due to the limitations of conventional soil mapping techniques，the level of spatial detail and attribute accuracy in conventional soil maps are not high. With the development of geographic information technology and the accumulation of high quality detailed spatial data on environment，the question about how to update the conventional soil maps using these new techniques and high quality data comes up. Based on the assumption that soil polygons in the conventional soil maps do contain basic relationships between soil types and environment，we have developed a new method to extract knowledge on soil-environment relations embedded in conventional soil maps and then update the soil maps with the knowledge using the detail environmental data. The method consists of the following four steps，1） Fuzzy c-means clustering of environmental data to obtain the combinations of environmental factors；2） Relating the combinations of environmental factors to soil types based on conventional soil maps；3） Extraction of knowledge on the soil-environment relationships；4） Prediction of soil spatial distribution using the SoLIM （Soil-Land Inference Model）. This method was applied in Wakefield，New Brunswick，Canada for updating its 1∶20 000 conventional soil map. It was showed that the updated soil map had more spatial details than the original one. And verification at independent validation points in the field indicated that the digital soil map depicting spatial distribution of soil associations with drainage classes was about 20% higher in accuracy than the conventional soil map. It is therefore concluded that the proposed method is an effective way to update conventional soil maps.