For M-206, however, comparable simulation results were obtained only after calibration of COLDREP (Fig. 4). The number of observed yields greater than simulated Yp (5.4%) and percent-age of observed yields within 15% of Yp (78%) were similar to CXL745. Although the range of observed yields was relatively nar-row due to selecting the top 30% of observed yields by year, the model was able to capture differences between yields at several sites with low Yp. Adjustments to cardinal temperatures as recommended by van Oort et al. (2011) were able to reduce the correlation between temperature and prediction error, but at the cost of increasing prediction error from emergence to 50% heading (Supplemental Fig. 1). Best model performance was achievedby using default cardinal temperatures, despite some persistent error in phenology prediction (Fig. 5). Therefore, the final calibration for M-206 used the default cardinal temperatures (TBD = 8◦C,TOD = 30◦C, and TMD = 42◦C). The persistent phenology error resulted in over-prediction of days to heading in cases where heading was accelerated and under-prediction in cases where heading was delayed (Fig. 5). Average estimated Yp for CA was 12.9 t ha−1