A Neighboring Pixel Ensemble Technique for NWP-Forced Flash Flood Forecasts

Uncertainty in Quantitative Precipitation Forecasts (QPFs) from Numerical Weather Prediction (NWP) models manifests in errors in the amounts of rainfall, storm structure, and storm location among other precipitation characteristics. In flash flood forecasting applications, these errors in the QPFs can translate into significant uncertainty in forecasts of surface water flows. In particular, the location and structure errors of QPFs result in errors on flow paths, which can be highly detrimental for forecasting the location of flash flood impacts. To account for this type of uncertainty, a Neighboring Pixel Ensemble Technique (NPET) was designed and implemented as a post-processing algorithm using deterministic outputs from a distributed hydrologic model. NPET defines a search box centered on each forecast pixel and from which surface water flow values are extracted to build the ensemble. As a constraint, only pixels with a drainage area similar to that of the center pixel are selected. The size of the NPET box and the degree of drainage area similarity are adjustable parameters that are calibrated to account for specific QPF error characteristics. The utility of NPET is demonstrated on a probabilistic flash flood prediction application over the Continental United States with short and long-range NWP QPF products.