Comparing hydrological frameworks for simulating crop biomass, water and nitrogen dynamics in a tile drained soybean-corn system: Cascade vs computational approach

Citation

Smith, W., Qi, Z., Grant, B., VanderZaag, A., Desjardins, R. (2019). Comparing hydrological frameworks for simulating crop biomass, water and nitrogen dynamics in a tile drained soybean-corn system: Cascade vs computational approach. Journal of Hydrology X, [online] 2 http://dx.doi.org/10.1016/j.hydroa.2018.100015

Plain language summary

Agricultural models are needed for assessing science-based mitigation options to improve the efficiency and sustainability of agricultural cropping systems. It is crucial that they can accurately simulate soil hydrology and nutrient flows which strongly influence crop growth, biogeochemical processes and water quality. The purpose of this study was to compare the performance of the DeNitrification DeComposition model (DNDC), which utilizes simplified hydrologic processes, to a more comprehensive water flow model, the Root Zone Water Quality Model (RZWQM2), to determine which processes are sufficient for simulating water and nitrogen dynamics and recommend improvements. Both models were calibrated and validated for simulating soil hydrology, nitrogen loss to tile drains and crop biomass using detailed observations from a corn-soybean rotation in Iowa, with and without cover crops. DNDC performed adequately across a wide range of metrics in comparison to a more hydrologically complex model. Soybean and corn yield, and corn biomass over the growing season were well simulated by both models. Soybean yields were also very well simulated by both models; however, soybean biomass was over-predicted by RZWQM2 in the validation treatments. The magnitude of winter rye biomass and N uptake was well simulated but the timing of growth initiation in the spring was inaccurate at times. The annual and monthly estimation of tile flow and nitrogen loss to tiles drains were well simulated by both models; however, RZWQM2 performed better for simulating soil water content, and the dynamics of daily water flow to tile drains. DNDC overestimated soil water content near the soil surface and underestimated it in the deeper profile. We recommend that developments be carried out for DNDC to include improved root density and penetration functions, a heterogeneous and deeper soil profile, a fluctuating water table and mechanistic tile drainage. However, the inclusion of computationally intensive processes needs to be assessed in context to improved accuracy weighed against the model’s broad applicability.

Abstract

Biophysical agricultural models are needed for assessing science-based mitigation options to improve the efficiency and sustainability of agricultural cropping systems. It is crucial that they can accurately simulate soil hydrology and nutrient flows which strongly influence crop growth, biogeochemical processes and water quality. The purpose of this study was to compare the performance of the DeNitrification DeComposition model (DNDC), which utilizes simplified hydrologic processes, to a more comprehensive water flow model, the Root Zone Water Quality Model (RZWQM2), to determine which processes are sufficient for simulating water and nitrogen dynamics and recommend improvements. Both models were calibrated and validated for simulating soil hydrology, nitrogen loss to tile drains and crop biomass using detailed observations from a corn (Zea mays L.) -soybean (Glycine max (L.) Merr.) rotation in Iowa, with and without cover crops. DNDC performed adequately across a wide range of metrics in comparison to a more hydrologically complex model. Soybean and corn yield, and corn biomass over the growing season were well simulated by both models (NRMSE < 25%). Soybean yields were also very well simulated by both models (NRMSE < 20%); however, soybean biomass was over-predicted by RZWQM2 in the validation treatments. The magnitude of winter rye biomass and N uptake was well simulated but the timing of growth initiation in the spring was inaccurate at times. The annual and monthly estimation of tile flow and nitrogen loss to tiles drains were well simulated by both models; however, RZWQM2 performed better for simulating soil water content, and the dynamics of daily water flow to tile drains (DNDC: NSE −0.32 to 0.24; RZWQM2: NSE 0.35–0.69). DNDC overestimated soil water content near the soil surface and underestimated it in the deeper profile. We recommend that developments be carried out for DNDC to include improved root density and penetration functions, a heterogeneous and deeper soil profile, a fluctuating water table and mechanistic tile drainage. However, the inclusion of computationally intensive processes needs to be assessed in context to improved accuracy weighed against the model's broad applicability.