Large-scale simulation of wheat yields in a semi-arid environment using a crop-growth model A.C. Chipanshi a , E.A. Ripley b, *, R.G. Lawford c a Environmental Science Department, University of Botswana, Private Bag 0022, Gaborone, Botswana b Department of Plant Sciences, University of Saskatchewan, 51 Campus Drive, Saskatoon, SK, S7N 5A8, Canada c GCIP Project Manager, NOAA/OGP, 1100 Wayne Avenue, Silver Springs, MD 20910, USA Received 23 October 1997; accepted 2 October 1998 Abstract The ability to predict wheat yields from large-scale weather variables has bene®ts throughout the semi-arid regions of the world. In spite of the availability of numerous crop-growth models, there has been little concerted eort to analyse yields regularly at spatial scales that are relevant to agronomic decision makers. As a result many current crop- growth models are research tools only. A large-scale wheat yield assessment procedure, based on the CERES Wheat model, has been developed for the semi-arid climate of Saskatchewan. It is suitable for simulating yields at the crop- district level, an area of about 2 million hectares containing several hundred farms having dierent soils, climates and management practices. Simulations of spring wheat growth, using this procedure, have revealed two critical periods (vegetative and ear growth) when lack of moisture has the greatest impact on grain yields. Knowledge of these times could be useful in devising early warning programmes for drought amelioration, combined with reliable long-term climate forecasts. Decisions made during these critical periods would aect farm management, marketing strategy and planning for the next growing season. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: Ceres Wheat model; Crop modelling; Drought; Semi-arid climate; Yield prediction; Spring wheat 1. Introduction Precipitation variability is a dominant feature of the climates of semi-arid regions in both the trop- ics (Glantz and Katz, 1985) and mid-latitudes (Maybank et al., 1995). Where grain is grown without irrigation, such as the prairie region of Saskatchewan, drought is the main cause of year- to-year variability in yield (Phillips, 1991). This variability leads to uncertainty in agricultural planning, and lower yields lead to lost income, debt, and a depressed economy (Lawford, 1992). It is likely that farmers could reduce production risks and crop losses by basing management deci- sions on an eective interpretation of current and expected weather (Phillips, 1991). In Canada's mid-west there is recent evidence that Paci®c sea-surface temperature anomalies may in¯u- ence prairie summer climate and provide predic- tion potential at time scales of a month or more (Bonsal et al., 1993). This suggests that it would be highly desirable to develop models capable of interpreting these forecasts on a regional basis in Agricultural Systems 59 (1999) 57±66 0308-521X/99/$Ðsee front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S0308-521X(98)00082-1 * Corresponding author. Fax: +1-306-966-5015; e-mail: eripley@marathon.usask.ca