Functionality-based evaluation of three crop models with different key process simulation approaches across maize–soybean cropping systems in Nebraska, USA

Using numerical crop models that simulate fundamental plant-related processes is the most efficient way to get insights into crop responses to future potential climate-weather-environmental conditions. This is because numerical crop models can be easily manipulated while focusing on single or multiple factors. Based on functions (empirical relationships) and equations (physical representation of the processes) derived from experimental observations, such models are our most advanced attempts to predict crop “behavior” under future conditions. The current standard practice is to run as many crop models as possible and then use an ensemble of these model outputs to predict an “averaged” change in yield production and crop quality metrics in the future. However, even though tens of different crop models are often being used in the ensemble, the differences among the models can be reduced to very few core functionality processes being simulated differently in such models. Functionality-based model evaluation involves evaluating the model's ability to simulate the underlying processes that determine crop yield rather than just comparing the model output to observed data. This approach can help identify the sources of model discrepancies and improve the accuracy of crop yield projections.

Here, we used three crop models with different functionality-based approaches (DSSAT, WOFOST, and Gcros) to assess biophysical parameters, including leaf area index, aboveground biomass, and grain yield, in a maize–soybean cropping system in Nebraska, USA. We calibrated the models using field data from the US-Ne Mead site, acquired through the AmeriFlux net, as well as soil information derived from the POLARIS soil properties dataset (30 m spatial resolution). We run the models with the 4km GRIDMET weather dataset for maize and soybean across Nebraska to examine the conditions (meteorological, climatic, and other static factors) that drive the change in the results of the different crop models. We aimed to select the most suitable model for best representing the impacts of future climate and environmental changes on these crops in the area per local conditions. We present essential discrepancies among the models and attribute such differences to the functionality-based representation of key processes in the models.