Long-term simulation of soil–crop interactions in semiarid southwestern Saskatchewan, Canada

Water and nutrient dynamics of soils are strongly related to land use and crop biomass production. Selected treatments of a long-term field experiment, established in 1967 at Swift Current in southwestern Saskatchewan, Canada with different crop rotations were simulated over a period of 25 years using the integrated soil–crop model HERMES. Results of two rotations (continuous wheat and fallow–wheat) were compared in terms of crop biomass and grain yield production, water and nitrogen dynamics and nitrogen losses by leaching and denitrification. Model performance was analysed by comparing the simulated state variables in soil and crops to measured data. The model simulated crop biomass, grain yields and nitrogen uptake well, with coefficients of determination (r2) of 0.88***, 0.70*** and 0.71***, respectively. Soil water content during the growing season was simulated well with r2 of 0.80***; however, simulating soil water dynamics during the winter period was problematic. This was due, at least in part, to the model's inability to estimate surface runoff, which occurred during periods of snowmelt when the soil was still frozen. Especially the performance of the crop yield estimation dropped down significantly without re-initialisation of water contents in spring. Although simulated soil mineral nitrogen content was mostly within the standard deviation of the replicate measurements, the very high variability of the measurements resulted in small correlations between simulation and observations (r2 = 0.22***). To reflect the trends of soil organic nitrogen, it was necessary to consider the historical management starting the model in 1910.