Flood duration projections at King County, Washington

Regional scale data were created for the full Puget Sound region using a hydrodynamic, wave, and hydrologic model. All regional scale models were run for 3 periods: reanalysis (1940-2023), historic CMIP6 (1950–2015) and future CMIP6(2015-2050). Forcing for the reanalysis period was provided by the ERA5 dataset and by 7 CMIP6 models for the historic and future periods (HighResMIP, Haarsma and others, 2016). The Hydrodynamic Model (delft3d-FM) was used for simulating regional water levels and is detailed in Grossman and others, 2023. Waves were modelled using a SWAN lookup table with linear swell propagation as detailed in Crosby and others, 2023. In addition to winds, swell was added through the Strait of Juan de Fuca via modeled CMIP6 Waves and from data provided by ERA5. Hydrology was provided by a county level hydrological model detailed in Buitink and others, 2025. The hydrodynamic and wave models were additionally run for a variety of SLR positions (0, 25, 50, 100, 150, 200, 300 cm). Outputs were saved as station output across the full Puget Sound region, in the nearshore for wave and water level data and at a variety of elevation contours for hydrologic data.

A SFINCS model was developed following the methodology detailed in Nederhoff and others, 2025. The model was built using topographic/bathymetry data from a DEM and covered the full extent of King County. Models were forced with continuous oceanographic and hydrologic forcing (provided by the regional model) for 2 periods: a reanalysis period and future projection period. The reanalysis period was used for model validation and the projection period was used for producing CoSMoS hazard outputs. For the future projection period, the model was run continuously for 100 years at 6 SLR scenarios. Outputs of maximum water levels were saved for each simulation year.

All simulation data were downloaded from the USGS Advanced Research Computing Center’s Hovenweep Supercomputer (Falgout and others, 2024) to be post-processed. For each of the 6 SLR scenarios, the 1-, 10-, 20-, 50-, and 100-year return period hazard level was derived using empirical return intervals from the 100-year time series of annual max water levels. Output data (flood depth, water elevation, maximum depth x velocity, flood extent) were derived from the occurrence of the water level extreme. Flood depth and extent were post-processed with an attenuated bathtub to remove edge effects and therefore do not match the spatial extent of the other products. Flood duration was calculated as the maximum that occurs on average once every 'rp' years. Practically this was calculated by sampling 1000 ‘rp’ year samples, with duration being the duration wet for the sample year in hours per year. For each ‘rp’ year sample, the max duration wet was calculated. Then the average across the 1000 samples was taken. Flood durations were then smoothed using a gaussian filter with a standard deviation of 6 cells. GeoTIFFs were created and bundled by county (Flood_duration_projections-King.zip). Data are organized by storm scenario (’RP’) and SLR magnitude.