Insights gained from modeling grain yield, nitrate leaching, and soil nitrogen dynamics in a long-term field experiment with spring cereals on fertilized and unfertilized soil over 35 years

Abstract

Crop models are useful tools for predicting changes in yield and nitrogen losses in response to changes in agricultural management practices and climate. We used a soil-crop model (CoupModel) to interpret trends in yields, drainage, and nitrate leaching observed for two contrasting treatments (fertilized and unfertilized cereals) in a long-term field experiment (35 years) on a sandy loam in southern Sweden. The model was calibrated using a Monte Carlo-based method, in which the 30 best simulations of 10,000 model runs were identified based on multiple criteria. The posterior distributions differed significantly between the two treatments for 6 of the 16 parameters included. For example, the decomposition rate coefficient of the slow organic matter pool was significantly larger in the unfertilized treatment. The model simulated yearly drainage and nitrate leaching well overall, but did not fully capture between-year variations. Although the simulated mean annual nitrate leaching was 1.4 times greater in the fertilized treatment, N leached per unit of N harvest was twice as large in the unfertilized plot. The model simulated substantial decreases in yield for both treatments in 2018 in response to an extremely hot and dry summer, although not as large as that observed. The range in simulated annual N mineralization due to parameter uncertainty was wider in the fertilized treatment. We conclude that model calibration strategies require careful attention to how different management practices may influence decomposition and long-term N balance components in agroecosystems and that more data on especially belowground biomass would help in reducing uncertainties.