Spatio-temporal variability of snowpack and runoff response over north-western North America Under global warming

Abstract:

We analyze snowpack and runoff response over North-Western North America – a large region that contains major transboundary river basins (Yukon, Mackenzie, Fraser and Columbia) heavily dependent on the snowpack freshwater storage. We use a large ensemble (50 realizations) Canadian Regional Climate Model (CanRCM4-LE), and Variable Infiltration Capacity (VIC) hydrologic model driven with statistically downscaled GCM ensemble, and analyze the spatial and temporal variability of snow water equivalent (SWE) and runoff changes under 1.0°C to 4.0°C warming above the preindustrial global mean temperatures (GMT). Further, we employ a machine learning model to relate snowpack changes to runoff changes. The results indicate highly heterogeneous snowpack changes across the region, with steep declines in maximum SWE in the southern basins - that are closer to freeze/melt threshold - than the colder northern basins. Furthermore, under a categorical framework of below-normal maximum SWE defined as snow drought (SD) and under GMT increases, SD predominantly occur under above-normal temperature and precipitation. The implications of the snowpack changes could be seen in the runoff changes, especially in terms of higher winter flows and earlier snowmelt driven peak flows. The annual runoff responses also vary across the region, with smaller increases or no change in the southern basins, and higher increases in the northern basins. Comparing CanRCM4-LE and VIC based projections, while the directions of changes generally agree, the magnitudes of both SWE and runoff changes diverge considerably. Nevertheless, the implications of snowpack loss (e.g. especially extreme snow loss in the southern basins) could be considerable in the basins where current water demands are the highest.