Delta yield predicts nitrogen fertilizer requirements for corn in US production systems

Predicting crop nitrogen (N) fertilizer needs is a major challenge in contemporary agriculture. Despite the success of current N recommendation tools, environmental concerns over N pollution from agriculture, and the adoption of improved corn (Zea mays L.) technologies with enhanced N efficiencies highlight the need for more accurate N fertilizer recommendation systems. Here, we aimed to develop a methodology to predict corn N requirements based on delta yield (dY = maximum yield−unfertilized yield). To develop this delta yield-based nitrogen (dY-based N) tool, we selected 486 quadratic-plateau corn yield response to N curves (from 732 N rate trials across northern US) to calculate dY and N fertilizer required to reach the yield plateau (N_x). The economic optimum nitrogen rate (EONR) was calculated using different fertilizer:crop price ratios (PR). The response curve outputs were then partitioned into calibration and validation sets. The calibration set was used to select linear models to predict N_x based on dY, resulting in nine state, agroecosystem region, and irrigation-specific sub-models. These sub-models predicted N_x of the validation set with a mean absolute error (MAE) of 33.0 kg N ha⁻¹. Predicted values from the site-year quadratic-plateau response fits were used to improve further predictions’ outcomes. Predictions of EONR based on dY had a lower MAE than the predictions of N_x, ranging between 19.9 and 25.4 kg N ha⁻¹ depending on the PR, highlighting the system's predictive power. The exclusion of non-responsive and linear-response trials in our proposed dY-based approach enables future model refinement to improve EONR prediction accuracy across a broader range of yield responses to fertilizer-N rates. The proposed dY-based N system, which integrates both economic and agronomic inputs (including management, environmental effects on soil N supply, and maximum yields), could help to reduce N losses and provide functional benefits for N optimization.