Seasonal Predictability of Four Major Crop Yields Worldwide by a Hybrid System of Dynamical Climate Prediction and Eco-Physiological Crop-Growth Simulation

The purpose of this study was to evaluate the prediction accuracy of a newly developed crop yield prediction system, composed of a dynamical seasonal climate prediction model (SINTEX-F2) and an eco-physiological process-based crop growth model (PRYSBI2). We explored the 3-months lead prediction accuracy of year-to-year variations in yield of four major crops (maize, rice, wheat, and soybean) in global regions and evaluated for which crops and in which areas the system performs well. The results indicated the system is more accurate for wheat relative to the other crops. Also, we found that different strategies would be useful in improving the system, depending on the crop. For winter wheat and rice, we need to improve the temperature predictions, particularly over the mid-latitudes, whereas improving rainfall predictions was more important for maize. For spring wheat and soybeans, the crop growth simulation itself should be improved. Although this study is only a first step, we believe that additional efforts to improve the system by understanding and incorporating processes of climate and crop growth will potentially provide useful prediction information to big stakeholders like global agribusiness companies and countries for improving food security in regions where crop yield is vulnerable to extreme climate shocks and where food markets are isolated from international trade.