The cascade of uncertainties in agro-hydrological model projections: implications for future water-food nexus in snow-dominated regions

Process-based agro-hydrological models have been widely used to quantify the impacts of climate change on water and food resources. Given the non-stationary nature of the agro-hydro-climate system, future projections are prone to uncertainties resulting from a variety of sources such as those related to impact models (i.e. model structure, input parameters, and data) as well as climate projections (i.e. greenhouse gas emission scenarios, global climate models (GCM), and downscaling methods). An understanding of the cascade of uncertainties depicting their sources in regional hydrological and crop yield projections is useful in non-stationary impact assessments. Here, we used the Soil and Water Assessment Tool and the ANOVA decomposition approach to analyze uncertainties arising from hydrological processes related to simulation of snowmelt, glacier melt and runoff, blue and green water resources, and crop yield and virtual water content in snow-dominated regions in western Canada. We quantified the share of uncertainty arising from temperature index (TI) and energy balance (EB) modules in the future projection of snow and glacier melt and runoff. While EB modules provide a more accurate representation of site-specific snow and ice melt processes than TI, they may increase the corresponding input data and parameter uncertainty in regional scale projections. Projected blue and green water resources showed that hydrological model contributes by up to 50% to the overall uncertainty when aggregated at large provincial scale in Alberta (~700,000 km²) . This uncertainty dominated other sources of uncertainty, i.e. GCMs, only during warm seasons in future (2040-2064) period. However, the uncertainty contribution from hydrological model significantly varied over time and space when disaggregated to sub-basin scale. Moreover, contribution from crop model to the overall cascade of uncertainty varied based on hydrological (e.g. soil and water parameters), phonological (e.g. leaf area index, harvest index, heat units, etc.), and management (e.g. fertilizer and irrigation application, planting and harvesting date) factors, as well as climate variables. Our results help to understand uncertainties associated with climate impact assessment in studying regional water-food nexus in snow-dominated areas.