Climate Change & Flooding in Snohomish County: New Dynamically-Downscaled Hydrologic Model Projections

We produced new projections of future streamflow, with a particular emphasis on flooding, for the Snohomish and Stillaguamish Rivers. Our results show that peak flows will increase in the future, with increases ranging from about 10 to 40% on average by the end of the century, depending on the river location or flood statistics considered. Funded by Snohomish County, these new results will support the County’s efforts to incorporate climate change impacts in floodplain management decisions.

Snohomish, WA. Photo credit MARVIN NAUMAN/FEMA

RESULTS   REPORT
SNOHOMISH MODEL CONFIGURATION  STILLAGUAMISH MODEL CONFIGURATION

Project Background

In the Pacific Northwest, climate change is expected to impact flooding through sea level rise, more intense heavy rains, and reduced snowpack. Snow influences flood risk by increasing the amount of precipitation falling as rain during winter storms (see, e.g.: Mauger et al., 2015). Other factors are also expected to affect flood risk, such as wildfires and increased sediment mobilization and transport in rivers. Acting together, the changes are likely to be consequential. The purpose of this project was to develop new projections of future streamflow, with a particular emphasis on flooding, for the Snohomish and Stillaguamish Rivers.

This project improved on existing models developed by the University of Washington Climate Impacts Group (UW CIG) and Western Washington University (WWU) on both the Snohomish and North Fork Stillaguamish Rivers, respectively. One set of models, funded by King County, used improved estimates of future weather conditions from new regional climate model projections and calibrated fine-scale hydrologic models of the Green and Snohomish River basins. Another set of models, funded by the Stillaguamish tribe, projected future streamflow on the North Fork of the Stillaguamish River. This project was designed to address the following gaps in previous studies on the two rivers: (1) the approaches for the two river basins were not sufficiently compatible to allow an apples-to-apples comparison of risks; (2) the meteorological inputs to the hydrologic simulations needed refinements and improvements; and (3) new regional climate model simulations were available; these can be used to improve the characterization of future conditions.

Approach

Past hydrologic studies have typically used interpolated estimates of daily weather on model grid cells (e.g., Hamlet et al. 2013). In this project we use dynamically downscaled historical and future weather data (Chen et al., 2018, Lorente-Plazas et al. 2018), as input to the hydrologic model simulations. In addition to providing better estimates of changing heavy rain events (Salathé et al. 2014), these new datasets allow us to consider hourly variations in streamflow, whereas previous approaches were limited to daily meteorology.

We used the fine-scale Distributed Hydrology Soil Vegetation Model (DHSVM, Wigmosta et al. 1994) to simulate historical and future flows in each watershed. We calibrated the model taking a step-wise approach: first testing the snow simulations, then adjusting the soil properties to improve the match with observed flows at a few key streamflow sites in each basin. We then processed the hydrologic model results to evaluate changes in peak flow statistics for four durations (one hour, one day, three day, and seven day) and return intervals (two-, five-, 10-, 20-, 25-, 50-, and 100-year events).

Results and findings

The average model projection shows an increase in flooding for all of the above durations and return intervals, without exception. Specifically: our results indicate that peak flows will increase substantially – ranging from about +10 to +40%, on average by the 2080s, depending on the river location, duration, and return interval in question. Our results show a tendency for larger changes for longer duration flood events, and the range among models tends to also be larger for longer duration events. The results do not show a clear relationship between projected changes and the return interval (e.g. two-year vs 100-year events).

It is difficult to detect trends in rare events. For this reason we expect that the projections for the two-, five-, and 10-year events will be most reliable. Similarly, our methods are not likely to provide reliable estimates of changes for the 1.01-year and 500-year events. We nonetheless provide projections for these events, for reference, but recommend against using them in planning or design decisions.

Acknowledgements

This work was supported by Snohomish County through a grant from the Washington State Floodplains by Design Program. Regional climate model simulations were provided by Ruby Leung at Pacific Northwest National Laboratory and Cliff Mass in the UW Department of Atmospheric Sciences.

Related studies and resources

Effect of Climate Change on Flooding in King County Rivers: Using New Regional Climate Model Simulations to Quantify Changes in Flood Risk models the changes in future streamflow and evaluates potential changes in peak flows on the Snoqualmie, South Fork (SF) Skykomish, and Green rivers.

Heavy Precipitation Projections for Use in Stormwater Planning includes an online tool that allows users to evaluate projections as a function of precipitation intensity, duration, and frequency.

Flood Risk PaperStormwater Planning Paper