Simulating nitrogen management impacts on maize production in the U.S. Midwest

Abstract

Nutrient loss reduction strategies have recently been developed in the U.S. Midwest to decrease the environmental footprint associated with nitrogen (N) fertilizer use. Although these strategies generally suggest decreasing N rates and shifting the timing of N application from fall to spring, the spatiotemporal impacts of these practices on maize yield and fertilizer N use efficiency (NUE, kg grain yield increase per kg N applied) have not been assessed at the watershed scale using crop simulation models. We simulated the effects of N fertilizer rate (0, 168, 190, 224 kg N ha-1) and application timing [fall-applied N (FN): 100% N applied on 1 December; spring-applied N (SN): 100% N applied 10 days before planting; split N: 66% N applied on 1 December + 34% N applied 10 days before planting] on maize grain yield (GY) across 3042 points in Illinois during 2011-2015 using the DSSAT-CERES-Maize model. When simulations were scaled up to the watershed level, results suggest that increases in average maize GY for SN compared to FN occurred in years with higher than average winter rainfall (2011, 2013), whereas yields were similar (+/- 4%) in 2012, 2014, and 2015. Accordingly, differences in NUE for SN compared to FN were small (0.0-1.4 kg GY/kg N) when cumulative winter rainfall was < 300 mm, but increased to 0.1-9.2 kg GY/kg N when winter rainfall was > 500 mm at both 168 kg N ha-1 and 224 kg N ha-1. The combined practice of reducing N fertilizer amounts from 224 kg N ha-1 to 190 kg N ha-1 and shifting from FN to SN resulted in a wide range of yield responses during 2011-2015, with the probability of increasing yields varying from <10% to >70% of simulation points within a watershed. Positive impacts on both GY and NUE occurred in only 60% of simulations for this scenario, highlighting the challenge of simultaneously improving yield and NUE with a 15% N rate reduction in this region.

Fig 1. Comparison of observed and DSSAT simulated maize grain yield according to N fertilizer amount.

Simulations were performed for different N fertilizer amounts at six sites across two years in Illinois.

Fig 2. Comparison of observed and DSSAT simulated maize grain yield according to N fertilizer application timing.

A) fall applied N with equal amounts of N applied in two splits, B) spring applied N with equal amounts of N applied in two splits, and C) fall with spring applied N in different field experiments in 2015.

Fig 3. Model performance criteria for assessing the ability of DSSAT to predict maize grain yield.

Simulations include experiments for N fertilizer amount and timing treatments.

Fig 4. Rainfall in winter (December-April) and maize growing season (May-September) rainfall across 18 Illinois weather stations.

Fig 5. Change in average DSSAT simulated maize grain yield at the watershed scale.

Points in each watershed were averaged for spring (10 days before planting) compared with fall fertilizer N application (1 December) at two N rates (168 and 224 kg N ha^-1) during 2011–2015. Southern Illinois watersheds have been excluded as FN is not applied in southern Illinois.

Fig 6. Cumulative probability distributions for change in maize grain yield.

Individual fertilizer treatments were compared to 224 FN (224 kg N ha^-1 applied in fall) within each site during 2011–2015. Control (0 kg N ha^-1); 224 FN (224 kg N ha^-1 applied on fall); 224 SN (224 kg N ha^-1 applied spring); 168 FN (168 kg N ha^-1 applied in fall); 168 SN (168 kg N ha^-1 applied in spring); 168 Split (112 kg N ha^-1 applied in fall + 56 kg N ha^-1 applied in spring). Southern Illinois watersheds have been excluded as FN is not applied in southern Illinois.

Fig 7. Cumulative probability distributions for change in N use efficiency.

Individual fertilizer treatments were compared to 224 FN (224 kg N ha^-1 applied in fall) within each site during 2011–2015. Control (0 kg N ha^-1); 224 FN (224 kg N ha^-1 applied on fall); 224 SN (224 kg N ha^-1 applied spring); 168 FN (168 kg N ha^-1 applied in fall); 168 SN (168 kg N ha^-1 applied in spring); 168 Split (112 kg N ha^-1 applied in fall + 56 kg N ha^-1 applied in spring). Southern Illinois watersheds have been excluded as FN is not applied in southern Illinois.

Fig 8. Impacts of cumulative rainfall on N use efficiency.

Differences in N use efficiency for spring compared with fall fertilizer N application within each site at two N rates (168 and 224 kg N ha^-1) as influenced by cumulative early spring rainfall (December through April) and soil drainage class during 2011–2015 in Illinois.

Fig 9. Assessment of simultaneous changes in N use efficiency and grain yield.

Changes in N use efficiency (kg GY/kg N) and grain yield (%) for each fertilizer treatment relative to 224 FN (224 kg N ha^-1 applied in fall) within each site in Illinois during 2011–2015. 224 FN (224 kg N ha^-1 applied on 1 December); 224 SN (224 kg N ha^-1 applied on 1 May); 168 FN (168 kg N ha^-1 applied on 1 December); 168 SN (168 kg N ha^-1 applied on 1 May); 168 Split (112 kg N ha^-1 applied on 1 December + 56 kg N ha^-1 applied on 1 May). The data from southern Illinois watersheds have been excluded as FN is not applied in southern Illinois.

Fig 10. Probability of maize grain yield increase for 224 FN compared to 190 SN.

The suitability of the combined practice of reducing N fertilizer amount by 15% (224 to 190 kg N ha^-1) and delaying application time from fall to spring was based on the percentage of simulations with yield increases greater than 100 kg ha^-1 within each watershed. The data from southern Illinois watersheds have been excluded as FN is not applied in southern Illinois.