Using a flow domain that we constructed from 30 m digital-elevation model data of western United States and Canada and a two-dimensional numerical model for shallow-water flow over rugged terrain, we simulated outburst floods from Pleistocene Glacial Lake Missoula. We modeled a large, but not the largest, flood, using initial lake elevation at 1250 m instead of 1285 m. Rupture of the ice dam, centered on modern Lake Pend Oreille, catastrophically floods eastern Washington and rapidly fills the broad Pasco, Yakima, and Umatilla Basins. Maximum flood stage is reached in Pasco and Yakima Basins 38 h after the dam break, whereas maximum flood stage in Umatilla Basin occurs 17 h later. Drainage of these basins through narrow Columbia gorge takes an additional 445 h. For this modeled flood, peak discharges in eastern Washington range from 10 to 20 × 106 m3/s. However, constrictions in Columbia gorge limit peak discharges to <6 × 106 m3/s and greatly extend the duration of flooding.
We compare these model results with field observations of scabland distribution and high-water indicators. Our model predictions of the locations of maximum scour (product of bed shear stress and average flow velocity) match the distribution of existing scablands. We compare model peak stages to high-water indicators from the Rathdrum-Spokane valley, Walulla Gap, and along Columbia gorge. Though peak stages from this less-than-maximal flood model attain or exceed peak-stage indicators along Rathdrum-Spokane valley and along Columbia gorge, simulated peak stages near Walulla Gap are 10–40 m below observed peak-stage indicators. Despite this discrepancy, our match to field observations in most of the region indicates that additional sources of water other than Glacial Lake Missoula are not required to explain the Missoula floods.