Discussion

Study Caveats

This study has several important limitations. First, only four climate scenarios were used, and only one greenhouse gas emission scenario was used. Thus, the full range of possible climate change effects on watershed hydrology is greater than that reported here. Second, constant land cover parameters were used in model calibration and in future simulations, although some changes in land cover type occurred during the base period, and more will likely occur in the future. Modeling the effects of logging, which is common in these coastal areas, on hydrology would be an excellent extension of this study. Third, a full sensitivity analysis, which could calculate the influence of the various parameters over time and space, was beyond the scope of this study. Fourth, PRMS does not consider sub-daily flow variability, which can be considerable in these steep, narrow watersheds, nor can it model geomorphology and in-stream processes. Fifth, the groundwater component of PRMS is quite simple compared to other hydrologic models such as Ground water and Surface-water FLOW (GSFLOW). It would have been difficult, though, to calibrate a more complex groundwater model in the study watersheds as groundwater head data points are extremely scarce in the coastal PNW (Marshall Gannet, U.S. Geological Survey, personal commun.). Sixth, PRMS does not consider wind as a climate input, although it can strongly affect evapotranspiration. Finally, climatic data were averaged across the watersheds, which obscures local variation due to coastal effects, orographic effects, and topographic complexity.

Key Findings

Despite the limitations of this study, the research has yielded several findings. First, the random parameter set generated in the uncertainty analysis for these watersheds systematically underestimated high flows. This finding could be attributed to limitations in the modeling of extreme flows by PRMS (fig. 31), which suggests that further research into PRMS’s ability to simulate accurately extreme flow events would be helpful. The high flow underestimation bias also may be due to underestimation of precipitation in the CIG data in the study watersheds.

Second, by comparing future outputs to reference outputs for each scenario, in each watershed, we determined that increases in autumn flow are probable in most watersheds. This change is driven directly by increases in precipitation in the same period in the NARCCAP climate data. Third, increases in the top 5 percent of flow days are likely, as are increases in late summer variability in flow and decreases in summer flow. Fourth, our uncertainty analysis demonstrated that there is more uncertainty associated with low flows. Therefore, findings regarding summer flows should be interpreted with caution. Finally, the uncertainty analysis also confirmed that the primary source of uncertainty in future flow projections is the climate models, rather than PRMS parameter values. Given that all NARCCAP climate simulations are based on a single emission scenario that does not represent the full range of possible emission levels, the uncertainty due to climate change is even greater than is indicated by these findings.

First posted February 28, 2013