A Model-Data Fusion Approach for Predicting Cover Crop Nitrogen Supply to Corn

One potential benefit of cover crops (CCs) is that N mineralization from decomposing CC residues may reduce the N fertilizer requirement of a subsequent crop, but predicting this credit remains a significant challenge. This study used a model-data fusion approach to calibrate a model of CC residue N mineralization and pre-emptive competition for soil NO3- that occurs during CC growth to predict the yield response of an unfertilized corn (Zea mays L.) crop. The model was calibrated with a data set of 199 observations from four CC experiments in central Pennsylvania. The most parsimonious model explained 82% of the variation in corn yield response. Parameters representing the C humification coefficients for decomposed residues from winterkilled (epsilon(wk) = 0.00) and winter-hardy (epsilon(wh) = 0.40) CCs suggest that all winterkilled CCs resulted in net N mineralization, probably due to the longer period of time for decomposition of winterkilled residues. However, the yield response per unit of potentially mineralized N was greater for winter-hardy CCs (alpha(wh) = 0.034 with tillage, alpha(wh) = 0.020 with no-till) than for winterkilled CCs (alpha(wk) = 0.0084), probably due to the improved synchrony between corn N demand and the decomposition of winter-hardy CC residues relative to winterkilled residues. Pre-emptive competition for soil NO3- led to a reduction in the corn yield response. Because the model is based on ecological processes and can be calibrated with data sets from simple field experiments, the model-data fusion approach could be widely used to guide adaptive management of CCs and N fertilizer applications in a subsequent corn crop.