Development of a Wheat Yield Prediction Model 1

This research was motivated by a need to predict crop yields on a large‐area basis without benefit of direct measurement of plant characteristics. Mathematical models were developed that require only a modest amount of historical data for application on a real‐time basis to geographical regions other than the one where the models were developed. The models, one for winter wheat ( Triticum aestivum L.) and one for spring wheat ( Triticum aestivum L. and Triticum durum Desf.), predict wheat grain yields from observation of meterological and agronomic variables. Use of standardized yields, simulated water budgets, and measurement of weather events over simulated phenological stages helped to create a broad data base for model development. Weather scenarios for low and high yields can be deduced from the models as can the contribution of applied N, improved cultivars, and increased irrigation to long‐term yield increases. The models include soil factors to help explain regional differences in yield. The models also allow for reduced yield estimates due to episodic events (diseases, freezes, etc.). Root‐mean‐square errors of differences between the model and USDA estimates of yield for the U.S. Great Plains for 1955 through 1976 were 1.2 and 1.3 quintals/ha for winter and spring wheat, respectively