A near real-time crop-specific drought monitoring system that combines crop modelling and a geographical information system (GIS) has been developed in South Africa. The system is intended to provide decision support for resource managers concerned with drought aid. Individual crop growth simulations are run for cells representing an area of 14 km(2). Values of the weather elements used to drive the models are obtained from interpolation of ground observations and processing of weather satellite imagery. Monitoring is undertaken throughout a production season, with updates provided on a fortnightly basis. Predictions of expected yield at the end of the season are made by using observed data up to the current date and completing the season with surrogate historical weather data. Appropriate surrogate scenarios are chosen, based on the synoptic situation in the current season. Maps and tabulated information are produced showing the spatial distribution of drought-stricken ar eas and the intensity of the drought in these areas. Favourable comparisons were obtained for recorded average maize yields and simulated average maize yields per magisterial district. The root mean square error of these comparisons was 221 kg ha(-1), whereas the mean absolute difference was 191 kg ha(-1). Copyright (C) 1996 Elsevier Science Ltd
