DayCent model calibration to assess the long-term impact of the animal slurry application on grassland in Ireland: Performance, sensitivities and scope for improvement

Measurement of changes in soil organic carbon (SOC) under various management practices at the field scale poses significant challenges due to inherent spatial and temporal variability.  In comparison ecosystem biogeochemical models offer a robust framework for simulating nutrient cycling, SOC, and greenhouse gas emissions that can be used to identify and evaluate long-term effects and strengths of climate change mitigation strategies. DayCent is a coupled soil-plant dynamic model that has been widely used to simulate long-term ecosystem responses to changes in soil management and climate in the US. Its application to agricultural systems in Ireland requires a calibration and evaluation for common management practices across a range of pedo-climatic conditions. The objective of this study was therefore a) to calibrate the DayCent model with several types of field data and to evaluate its performance in simulating SOC and soil N₂O emissions and b) to explore the sensitivity of model parameters to different types of field data.  Our aim was to simulate the effects of a long-term application of dairy, pig, and mineral fertilizers on grass yields, SOC and soil organic nitrogen (N) stocks, and soil N₂O fluxes in a long-term permanent grassland experiment. To calibrate the model, the data from control and high pig slurry application treatments from 1970 to 2022 were used. The calibration was separated into two steps: a) the first step was a manual calibration for SOC and soil organic N, volumetric soil water content, and grass yield; b) the second step was an automatic calibration for soil temperature, daily N₂O emission, soil NO₃⁻ and NH₄⁺ concentrations with the PEST parameter estimation software. All remaining treatments, that varied in the rate and type of animal slurry application, were used in the independent model evaluation. Using this information the performance of the calibrated model was substantially improved for SOC stock (rRMSE=0.17, r²=0.54, d=0.78, n=102) compared to the default model (rRMSE=0.25, r²=0.29, d=0.45, n=102) across all validation treatments. Similarly, an improvement was found for soil organic N stock in the validation treatments (rRMSE=0.19, r²=0.70, d=0.78, n=102) compared to the default model (rRMSE=0.30, r²=0.64, d=0.53, n=102). Improvements in simulating daily N₂O emissions (calibrated model: rRMSE=5.30, r²=0.08, d=0.44, n=186; default model: rRMSE=2.97, r²=0.02, d=0.22, n=186), and soil NO₃⁻ and NH₄⁺ concentrations were still quite uncertain across validation treatments. In conclusion, the calibrated DayCent successfully simulated the long-term dynamics of SOC and soil organic N stocks, grass yields, soil water content, and soil temperature across varying nutrient application rates, although there were some limitations in simulating daily and annual N₂O emissions, and mineral N concentrations. While further testing under various pedo-climatic conditions is necessary, DayCent has the potential to be used as a tool for optimizing nutrient management strategies under Irish conditions.