Lessons learned from a detailed exploration of APEX as a tool to represent corn residue management and cover crops

CONTEXT

Agricultural management practices to improve the regulation of water, sediments, or nutrients, make farming decisions and operations more complex. This extra complexity often stems from the use of multiple species and farm heterogeneity so that species can complement each other, and different fields (and the space between fields) can provide alternative benefits, like biomass or nutrient regulation. Mechanistic crop and farm models provide tools to explore the effect of these practices.

OBJECTIVE

The study goal was to assess the capability of a mechanistic crop model (the Agricultural Policy/Environmental eXtender Model, APEX) to represent the impacts of cover crops and corn residue on plant growth, water, erosion, and nutrient flow.

METHODS

Using Southern Ontario conditions, a simplistic corn–cover crop rotation strategy was implemented using APEX and hundreds of variables dynamically updated by the model were analyzed. The model's documentation and source code were analyzed to understand the connections among the variables.

RESULTS AND CONCLUSIONS

The model reproduced corn and cover crop growth patterns observed in Southern Ontario and the positive effects of cover crops and residue on water, sediments, and nutrient control. The model suggested that these practices generate important differences in nutrient dynamics and patterns of vertical accumulation of soil nutrients. Issues with the model are reported and ways to avoid them discussed. There were inconsistencies and unrealistic responses in the outputs when simulating two crops growing together or multiple fields, including small mass balance discrepancies, which —in complex numerical models like APEX— can generate hard-to-track differences and may be amplified when multiple fields are simulated over several years. Users should be aware of these limitations when assessing the role of diversified farming practices.

SIGNIFICANCE

The results highlight the importance of carefully reviewing the internal consistency of mechanistic models beyond validating a few key outputs, especially when the intended use of a model is to extrapolate the impacts to novel conditions or to infer processes not directly validated. These findings could open the conversation for more robust modelling and validating approaches when using crop models.