Modification of a Soybean Model to Improve Soil Temperature and Emergence Date Prediction

Recent studies have shown that the CROPGRO-Soybean model does not predict soil temperature very well inIowa. This typically gives errors in predicted emergence date, which translates to errors in timing of development andbiomass accumulation during the remainder of the season. In order to improve the model, an energy balance-based soiltemperature model was integrated into the soybean model and compared to the original soil temperature model, whichwas driven primarily by air temperature. In the new model, temperature at the soil surface is estimated from the basicenergy balance equation at the air-soil interface and the soil temperature profile is calculated using the one-dimensionalheat flow equation. The model was calibrated using five years of bare-soil temperature data measured at an experimentalfarm in Ames, Iowa. Validation of the new model using five additional years of bare-soil temperature data from the samelocation gave slightly better predictions of soil temperature in the top 5 cm (RMSE = 3.0, R 2 = 0.86 for validation years),and responded better to surface perturbations than the original model (RMSE = 3.2, R 2 = 0.80 for validation years).Under bare soil conditions, the new model generally gave lower RMSE and higher R 2 values compared to the old modelat all soil depths. The models were also compared for accuracy in predicting emergence date. Experimental data on soiltemperature and emergence for soybeans planted on weekly intervals over an eight-week period were used to test themodels. The new model gave excellent predictions of emergence, with an average error of 0.6 day for the eight weeklyexperiments. The old model had an average error of one day. Under cool conditions, the new model gave more accuratepredictions of emergence dates. However, under warm periods, both models typically gave the same accuracy, and werewithin about one day of the measured emergence date.