Crop and soil management at test sites used for model calibra-tion were based on recommended “best management” practicesand, in the case of on-farm trials,on farmer experience. Such man-agement regimes seek to maximize profit by achieving highest cost-effective yield levels and are not designed to eliminate all bio-physical stresses. Therefore,it was expected that a well-calibrated model would predict Yp values greater than observed yields in most cases, consistent with work on other crops attempting to quantify Yp (Cassman, 1999; Lobell et al., 2009)On the other hand, due to sampling variability and measurement error in both yield andweather data, it can be expected that simulated Yp will be lessthan observed yields at some sites,but these cases should be a minority in a model that simulates Yp well. Unlike many calibration efforts where the goal is to minimize the RMSE or absolute error,