Multiobjective evolutionary optimization for quantifying corn yield and drainage nitrate load tradeoffs of fertilizer management decisions

Achieving 45% nutrient loss reduction goals in Illinois and the Mississippi River Basin will require farmers and land managers to adopt multiple best management practices. The current statewide estimates for nitrogen (N) load reductions of individual practices, however, cannot be added together because of nonlinear practice interactions. Our objective is to fully explore the synergies and tradeoffs of combined fertilizer management decisions by coupling the USDA's Root Zone Water Quality Model 2 (RZWQM2) with a multiobjective evolutionary algorithm. To initially develop and test the optimization framework, we use the calibrated model from Jeong and Bhattarai (2018) for two sites in east-central Illinois during the study period 1993 to 2000. The feasible ranges for decisions variables are based on historical fertilizer rates and application dates from the site management records. To calculate the profit and cost effectiveness of seasonal management decisions, we collected historical economic information for central Illinois, including market corn prices, fertilizer costs, and costs of shifting fertilizer timing from fall to spring, over the study period. With the vector-valued objective function to minimize N loads and cost effectiveness and maximize profit and corn yields, we implement the Strength Pareto Evolutionary Algorithm 2 with RZWQM2 under historical weather to generate nondominated sets of fertilizer rate, timing, and method decisions for eight growing seasons. We directly use these results to quantify the benefit of optimal management by comparing outcomes between optimized, rule-based, and historical management scenarios.