Towards an improved methodology for modelling climate change impacts on cropping systems

Assessment of the impact of future climate change on agricultural sustainability requires a robust full system approach estimation of the interdependent soil-plant-atmospheric processes coupled with dynamic farm management. The simplification or exclusion of major feedback mechanisms in modelling approaches, such as Carbon (C) and Nitrogen (N) cycling, water and nutrient leaching, trace gas emissions and dynamic farm management, may significantly alter the final estimate. Using three case study locations in Canada and a validated full systems biogeochemical model, DNDC v.CAN, we quantified the impact of using commonly employed simplified modelling approaches to better understand their impact on the estimation of crop yields, soil organic carbon (SOC) change and N losses. These approaches included using climate with only temperature and precipitation data, annual re-initialization of soil status, fixed fertilizer application rates, and fixed planting dates. The simplified approaches were compared to a comprehensive reference approach that was employed using detailed climate drivers, dynamic planting dates, dynamic fertilizer rates, and continuous estimation of SOC, N and water budgets. We also explored alternative cultivars and rotational impacts. At the semi-arid location the fixed fertilizer, fixed planting date, and re-initialization approaches resulted in spring wheat (Triticum aestivum L.) yield estimates being reduced by 40, 25, and 29%, respectively, in the 2071-2100 period relative to the reference approach where yield estimates increased by 74% relative to the 2011-2040 period. Simulations with and without re-initialization of soil status demonstrated that long-term inter-annual feedbacks in C&N cycling sometimes significantly impacted crop yields, SOC levels and N losses. For all locations and approaches, there were significant differences in N losses relative to the reference approach. We conclude that simplified approaches often lack the necessary characterization of the feedbacks between climate, soil, crop and management that are critical for accurately assessing crop system behaviour under future climate.