Comparing the performance of the DNDC, Holos, and VSMB models for predicting the water partitioning of various crops and sites across Canada

Accurately accounting for water budgets within regional agroecosystems is becoming an increasingly important practice, as both climate change and water consumption pressures have the potential for influencing agro-productivity and other water use activities. In this study, water budget measurements from 10 rainfed experimental sites across Canada were utilized to evaluate the performance of three models for their water partitioning capabilities: denitrification–decomposition (DNDC), Holos, and versatile soil moisture budget (VSMB). To assess the likely model performance at an upscaled national level, the models were applied at the site level with no water component-specific calibration. Evapotranspiration (ET) was found to be the dominate component of the water budget at the prairie sites (89%–149% of precipitation) (i.e., in comparison to runoff, tile drainage, and deep percolation), while both ET (37%–73% of precipitation) and drainage (19%–61% of precipitation) represented most of the water outflow budget at the sites in eastern and Atlantic Canada. As DNDC integrates daily crop growth dynamics with nitrogen, water, and heat stresses, in contrast to VSMB and Holos, which only utilize a water budget model, it was not surprising to find that DNDC consistently out-performed the other two models across all the statistical performance metrics considered at daily resolution.