Abstract: Prediction and simulation of water and salt transfer in freezing and thawing soils have always been a considerable attention and difficult research on the physics of frozen soils and soil water science. It is considered that the movement and distribution rule of frozen soil water and salt is a significant factor on the evolution and transition of soil salinity and alkalinity. Based on field measurements of water, salt and temperature for five years(1994～1999) overwintering in Inner Mongolia Hetao Irrigation District, the model of artificial neural network on soil water and salt transfer under freezing-thawing conditions was established. The results show that BP network of ten factors i.e. groundwater table, water content, salt content, irrigation amount and average monthly soil temperature from November to April next year, can reflect the coupling relation between soil water and salt under freezing and thawing conditions. The BP network may forecast field coupling transfer of soil water and salt under freezing and thawing conditions accurately. This study offer an effective and feasible method for manages irrigation district and forecast of soil water and salt transfer under freezing-thawing conditions.