Maize yield prediction using seasonal weather forecasts and a crop growth simulation model.

Maize production in marginal tropical regions is at great risk from rainfall variability. Farmers would benefit from the ability to forecast production likelihood. In this study we sought to develop a simple maize production decision support tool for Masvingo by using seasonal weather forecasts and a crop production model to forecast maize yields prior to the season. Downscaled ENSO-based statistical seasonal forecasts from RAINMAN were tested against those downscaled from a Global Circulation Model (GCM) using Climate Predictability Tool (CPT). RAINMAN was found to perform better at forecasting total seasonal rainfall than CPT. RAINMAN predictions were 69 % correct in all rainfall categories for the 1991/92 - 2006/07 seasons as opposed to 44 % for CPT (p< 0.05). RAINMAN had a higher hit rate than CPT and was not biased to any rainfall category while CPT was biased towards the normal and dry/below normal rainfall categories. Monthly rainfall predictions by RAINMAN were validated. The tool explained 65 % to 81 % (p