How Many N Rate Trials Do We Need?.

A large number of maize N rate response trials have been conducted in the U.S. Corn Belt in recent years, and this activity continues. The rationale for continuing to add to this substantial database recognizes the large amount of variability in N response over growing seasons and soils, and perhaps over cultivars. But incorporating this variability into economic and environmental optimization models remains challenging. We used nonlinear mixed effects models approaches to analyze 70 site-years of N fertilizer response data from each of two rotations at seven Illinois sites over ten years. Our analyses focused on two research questions: 1) how many site-years of data are needed to adequately characterize and predict maize fertilizer N response from the perspective of reducing producer risk?; and 2) how does maize N response vary, at large spatiotemporal scales, with weather and soil properties? Post-hoc power analysis indicated that, at an α level of 0.05, ten years of study data could estimate site-level mean values of MRTN (maximum return to N) within a 17% margin of error. At α = 0.20, MRTN site means were estimated within a 10% margin of error. Corn response to N fertilizer was best described for our data by an asymptotic nonlinear mixed effects model with random shifts to intercept, and a fixed effect of previous crop (corn or soybean). Random effects for this model were predicted (F_1,68 = 21.4, p < 0.0001, R² = 0.24) using site data on soil physical properties (depth to soil limiting layer) and site-year data on thermal time, namely growing degree days (base 10 C) accumulated from 30 days before planting through 30 days after planting. Current data from on-station corn N response trials appears to be sufficient to aid efficient site-year N fertilizer applications. Extending this predictive approach to other parts of the Corn Belt will require regional-scale verification efforts.