In general, agricultural drought always occurs under the circumstance of the comprehensive interactions among the factors of nature, economy and society. The loss due to agricultural drought in China is huge every year. Therefore the timely monitoring of agricultural drought is critical to help reduce the loss. The information of agricultural drought early warning is helpful for local governmental officials and farmers in preparation for coping with the likely happening drought. The paper presents an approach and findings of an early warning of agricultural drought which has been successfully conducted in the semiarid and rainfed farming area in Ningxia autonomous region in the northwest of China. The approach of the early warning of agricultural drought presented in this paper is based on the theory of natural disaster system that may break the disaster system down into three parts, composing of environmental factors potentially raising drought, factors of causing drought and drought affected bodies. 11 indicators have been selected as key factors for the early warning of agricultural drought, based on the previous studies and literatures and the consideration of the characters of agricultural drought occurrences in this region and the availability of the indicators. The indicators of the environmental factors potentially raising drought were chosen from the natural and social factors that affect mainly the distribution of rainfall and soil water, and finally include altitude, slope degree, slope aspect, soil type, the drought risk index and the input to farming. The indicators of factors of causing drought were chosen from the factors that affect the supply of soil water and the severity of the previous drought. The indictors of drought affected bodies were crop type, crop yield and the sensitivity of water stress in different stages of crop development. Crop type classification was done by the means of the supervised classification using 1km resolution EOS MODIS and SPOT VEGETATION data. Winter wheat, corn, potato, grass land, forest and desert were discerned. The drought risk index and the supply of crop water were calculated based on the meteorological data records, soil moisture and soil parameters from 1981 to 2008 in this area. The sensitivity of water stress was decided according to the knowledge and experience of experts. A spatial interpolation of these indicators was explored based on GIS. The contribution rates of these indicators to agricultural drought were quantified by the means of AHP. Furthermore, the early warning model was set up by multiply these indicators by their respective contribution rates with the support of GIS software. The final drought index was calculated based on this model and categorized into different classes representing the early warning of agricultural drought. The findings showed that the trend and distribution of agriculture drought in the study region was captured by the model. Compared to the current index of meteorological drought, the accuracy of this model is much higher and moreover this model has more ability to capture the processes of occurrence, development and relief and end of agricultural drought, and reflect the spatial distribution of the different categories and the extent that drought actually occurs as well as the circumstance that a severe meteorological drought may be occurring but the agricultural drought does not occur or is light. The early warning model of agricultural drought is expected to be expanded in the further study and to able to provide the technological supports for the prevention and reduction of agricultural disasters by producing the informed early warning and drought trend prediction.
