Improving the CROPGRO-Tomato Model for Predicting Growth and Yield Response to Temperature

Parameterizing crop models for more accurate response to climate factors such as temperature is important considering potential temperature increases associated with climate change, particularly for tomato (Lycopersicon esculentum Mill.), which is a heat-sensitive crop. The objective of this work was to update the cardinal temperature parameters of the CROPGRO-Tomato model affecting the simulation of crop development, daily dry matter (DM) production, fruit set, and DM partitioning of field-grown tomato from transplanting to harvest. The main adaptation relied on new literature values for cardinal temperature parameters that affect tomato crop phenology, fruit set, and fruit growth. The new cardinal temperature values are considered reliable because they come from recent published experiments conducted in controlled-temperature environments. Use of the new cardinal temperatures in the CROPGRO-Tomato model affected the rate of crop development compared with prior default parameters; thus, we found it necessary to recalibrate genetic coefficients that affect life cycle phases and growth simulated by the model. The model was recalibrated and evaluated with 10 growth analyses data sets collected in field experiments conducted at three locations in Florida (Bradenton, Quincy, and Gainesville) from 1991 to 2007. Use of modified parameters sufficiently improved model performance to provide accurate prediction of crop and fruit DM accumulation throughout the season. Overall, the average root mean square error (RMSE) over all experiments was reduced 44% for leaf area index, 71% for fruit number, and 36% for both aboveground biomass and fruit dry weight simulations with the modified parameters compared with the default. The Willmott d index was higher and was always above 0.92. The CROPGRO-Tomato model with these modified cardinal temperature parameters will predict more accurately tomato growth and yield response to temperature and thus be useful in model applications.