Predicting Risk from Reducing Nitrogen Fertilization Using Hierarchical Models and On-Farm Data

Current systems for developing N recommendations for corn (Zea mays L.) lack methods to quantify the effects of factors influencing yield responses to N and quantify the uncertainty in N recommendations. We utilized hierarchical modeling and Bayesian analysis to quantify the risk from reducing N to corn using on-farm observations. Across Iowa, farmers conducted 34 trials in 2006 and 22 trials in 2007. Each trial had a farmer's normal N rate alternating with a reduced rate (by about 30% less) in three or more replications. Yield losses (YLs) from reduced N were calculated at 35-m intervals. Posterior distributions were used to identify across-field and within-field factors affecting YL and to quantify the risk of economic YL (>0.31 Mg ha(-1)) from reducing N in unobserved fields. In 2006 (dry May and June), the economic YL for corn after soybean (C-S) was predicted to be 20% larger than that for corn aft er corn. Also in 2006, C-S fields with above-normal June rainfall had economic YLs 35% larger than those with below-normal June rainfall, and sidedress applications were about 20% riskier than spring applications. In 2007 for C-S, N reductions with above-normal spring rainfall were riskier than with below-normal spring rainfall. Areas with higher soil organic matter (SOM) had economic YLs about 20% smaller than those with lower SOM. Many on-farm trials can be conducted across the state and the use of the proposed statistical methodology can improve decisions on where to reduce N applications across and within fields.