A Systems Modeling Approach to Forecast Corn Economic Optimum Nitrogen Rate.

Historically process-based cropping systems models have been used to assess N management decisions after the growing season (i.e. ex-post), when it is too late to make in-season adjustments. In-season model application may represent a significant opportunity to improve N management decisions, however, the ability of the models to switch from ex-post to an in-season N and yield forecasting needs to be investigated. We used the Agricultural Production Systems sIMulator (APSIM) model previously calibrated with 15-years of yield response to N fertilizer data for continuous corn (CC) and soybean-corn (SC) rotation systems from central Iowa. Our first objective was to evaluate the accuracy and uncertainty of corn yield and economic optimum N rate) EONR predictions at four forecast times relative to end-of-season prediction (weather known). The forecasting times were planting, corn 6^th leaf (V6), corn 12^th leaf (V12), and silking (R1). Our second objective was to determine whether the use of a selection of historical weather data as opposed of using a complete historical 35-year dataset could improve the accuracy of yield and EONR predictions at each forecast period. The scenarios included weather years with similar precipitation and temperature patterns during summer period or during the whole year, and use of the last 5, 10 and 20 years of weather data. Results indicated that across years and cropping systems, the model simulated end of season yields and EONR at planting time with a relative root mean square error (RRMSE) of 13% and 33% respectively. Predictions did not substantially improve towards the end of the growing season. Prediction of yields and EONR values at either forecast time was more accurate using a 35-year historical weather file than use of selected weather scenarios. We concluded the use of this approach could improve year-to-year predictability of corn yields and optimum N rate, therefore, assisting farmers and informing other N management tools (e.g. yield-goal approach, maximum return to N, and others) used in US Midwest. Further improvements in model predictions and set-up are discussed to better simulate EONR in extreme weather years with the most significant economic loss and environmental impact from under- or over-application of N input.