| Abstract/Description or Keywords |
Currently, two different large-scale hydrologic modeling studies (conducted by the Climate Impacts Group (CIG) at the University of Washington and the Pacific Climate Impacts Consortium (PCIC) at the University of Victoria respectively) have projected the impacts of regional climate change for the Canadian Columbia River basin. This study evaluates key areas of convergence and divergence in the results of these studies for 18 hydroclimatic variables projected for the 2040s, including: cool-season (October – March) precipitation (P), cool-season P extremes, warm-season (April - September) P, warm-season P extremes, cool-season Temperature (T), cool-season T extremes, warm-season T, warm-season T extremes, April 1 snow water equivalent (SWE), SWE to P ratio, annual streamflow, cool-season streamflow, warm-season streamflow, July-September streamflow, August streamflow, center of timing of flow, high flow extremes, and low flow extremes. The consensus on qualitative impacts, and particularly the overall direction of change, was found to be strong for 16 of 18 variables when comparing different modeling approaches. Consensus on the percent change was strong for six variables, moderately strong for eight variables, and weak for four variables. Of the ten hydroclimatic variables evaluated spatially, six showed a strong consensus on spatial patterns (i.e. the geographic location) of impacts between different modeling approaches, one showed a moderately strong consensus, and four showed a weak consensus between different modeling approaches. Of all the metrics evaluated, hydrologic extremes, and particularly high flow extremes, showed the greatest divergence between the two modeling approaches. Significant differences were also found in the percent changes in cool and warm-season streamflow, and changes in the timing of peak flows, despite good qualitative agreement on these impacts. Substantial absolute differences in monthly flow in July, August, and September were found between the three modeling approaches due to the inclusion of glaciers in the PCIC study, however changes in August flow (as percent) were relatively consistent between the three modeling approaches, suggesting that loss of seasonal snowpack remains an important driver of summer low flow impacts with or without effects of glaciers included in the simulations. Historical baselines were often substantially different for the two studies, and different aspects of the hydrologic cycle were better simulated by different studies. Key Study Conclusions: 1. The two hydrologic modeling studies strongly agree about the general nature and overall direction of climate change impacts in the Canadian Columbia River Basin. 2. Changes in temperature (T) and precipitation (P) are broadly consistent between the two studies and downscaling methods, however the magnitude and spatial patterns of changes in warm season P extremes and cool-season T extremes are often divergent. 3. Despite the use of the same hydrologic modeling package, quantitative (numerical) differences in hydrologic variables between the two studies are frequently quite substantial, both in terms of baseline historical conditions and sensitivity to future changes. Thus, for quantitative studies using hydrologic data as inputs, differences in hydrologic modeling approach are shown to be a substantial source of uncertainty in projecting climate change impacts to the Canadian Columbia River basin. The results support the argument that multiple hydrologic modeling approaches should be used to better quantify this uncertainty in future studies. 4. Particularly strong areas of divergence include percent changes in cool season streamflow, and hydrologic extremes such as the 100-year flood and 10-year seven-day low flows (7Q10). 5. A thorough diagnosis of the divergence in quantitative results is beyond the limited scope of this study, but probable causes of divergence include: absolute differences in historical T and P data sets that serve as inputs to the model, inclusion of glacial contributions to streamflow in the PCIC simulations, and different model calibration choices which ultimately affect the sensitivity of the models to changes in climate. 6. Without bias correction, the PCIC simulations, which incorporate the effects of glaciers on late summer flow, better mimic the actual streamflow behavior in late summer than the UW simulations. However, the two models frequently have comparable percent changes in flow in August. 7. Interpretation of the differences between the studies may vary with context: a. Conclusions based on qualitative results (such as the direction of change) would likely draw very similar conclusions about the nature of impacts and identify a high level of consensus between the different modeling approaches. b. Engineering studies using these simulations as inputs (e.g. as inputs to reservoir simulation models predicting hydropower production) would likely show substantial differences in outcomes and a high level of divergence between the different modeling approaches. 8. Further study using bias-corrected streamflow simulations (not available at this time for PCIC simulations) is needed to better assess the implications for water resources modeling studies. |
