Abstract: A novel model integrating variable precision rough set and fuzzy neural network model for degraded land evaluation based on the knowledge reasoning technique(KR_LDAM) was built,which can diminish the impacts of subjective judgments and inconsistency of experts knowledge and uncertainty of land use data in the process of selecting indicators and performing evaluation.And then we took classic wind erosion region located in Loess Plateau as the study area to conduct a case study and validate the model.