Ensemble Hydrometeorological Hindcasts of the 2013 Colorado Floods

The Colorado Front Range floods of 2013 were the product of a number of hydrometeorological and physiographic factors coming together to focus heavy rainfall and runoff in a steep mountain front region. The event lasted over 5 days, caused an estimated $2 billion dollars in damages, destroyed over 450 miles of roads and was associated with 9 fatalities. Unlike many warm-season flood events in the Colorado Front Range the 2013 floods covered a region spanning multiple river basins and 10s of thousands of square kilometers whereas most flash flood events in the region are isolated to single watersheds. Despite the extreme nature of the event and its large regional coverage there was very little in the way of meaningful hydrologic forecasts produced prior to the start of the event. In this talk we present detailed analyses of ensemble precipitation and hydrologic hindcasts generated using the Weather Research and Forecasting (WRF) model and its hydrological modeling extension package, WRF-Hydro. The ensemble skill of the WRF model precipitation hindcasts is highly linked to model resolution where fine resolution is required to properly simulate the location and intensity of sustained orographic precipitation. In addition to heavy rainfall, a diversity of physiographic features such as, steep terrain and land cover disturbances also contributed to the severity of local flooding. This process diversity necessitates a spatially-explicit, process-based approach to hydrologic modeling. Using operational quantitative Precipitation Estimates (QPEs), the WRF-Hydro system demonstrated meaningful simulation skill of the flood event in capturing these features although there were significant uncertainties in characterization of vegetation recovery from areas burned in recent wildland fires. This reasonableness is evidenced by the validation of modeled peak flows across a range of watershed sizes from small headwater catchments to the South Platte River. Ensemble WRF-Hydro hindcasts driven by the ensemble quantitative precipitation hindcasts demonstrate performance characteristics that are consistent with the underlying bias and uncertainty structures in precipitation from the WRF model. Nevertheless, the ensemble hydrometeorological hindcasts using WRF and WRF-Hydro show that it is possible to provide meaningful predictive guidance on potential flood hazards with lead times up to 24 hours in advance of severe flooding from this event.