Abstract: In order to establish grid scale model of agricultural drought early warning, according to the theory of natural disaster system, integrating features of dry pregnant disaster environment, hazards factors and disaster bodies, selecting the DEM, slope, aspect, soil type, history of drought risk index, agricultural inputs as influence factors of the geo-hazards stability, selecting the pre-drought, crop water satisfaction rate as hazard risk factors, selecting crop type, crop development of water sensitive coefficient, irrigation, crop yields as vulnerability factors in the drought system, the agricultural drought early warning model for Ningxia region was established by the AHP method with quantization influence coefficient based on GIS and spatial interpolation. Test results showed that compared with the measured values, more than 90% of model output levels of drought were consistent. The early warning model could accurately predict future trends and distribution of droughts. Compared with current climatic drought index prediction model, the spatial resolution of agricultural drought early warning model was significantly improved, and could better reflect the occurrence of agricultural drought and its development process. It can be used for regional drought early warning and drought trend forecast.