Four active research sub-topics related to the FLASH project are described below.
Evaluation of skill relative to benchmarks
In order to demonstrate improvement, we need to know where the current tools used for flash flooding stand. We have conducted several studies to benchmark flash flood guidance and newer, gridded flash flood guidance values across the US. Analysis of these skill metrics as a function of basin scale and basin characteristics are very useful for guiding FLASH basic development.
Flash Flood Observation Database
In order to evaluate the forecasting tools, we need to use observations of flash flooding. When we first began this research, we thought we could simply download flash flood observations from a unified database. However, we soon learned that flash floods are not only difficult to predict, they are even difficult to observe! Since then, we’ve assembled flash flood observations from USGS automated discharge measurements, trained spotter reports from the NWS, and from NSSL’s Severe Hazards Analysis and Verification Experiment (SHAVE). This database is available for community research purposes. Currently we support the database in Google Earth kmz format, GIS shapefile format, and comma-delimited text files that can be easily read into Excel.
While we presently focus on rainfall forcing from NMQ/Q2 observations, lead time can be greatly increased using forecasts of rainfall from stormscale numerical weather prediction models. While these forecast products are quickly improving following enhanced grid cell resolution, explicit physical process representation, and radar data assimilation, errors with the specific locations of intense rainfall are still common. We are currently investigating the use of ensemble stormscale NWP products through involvement in the Hazardous Weather Testbed (HWT) experiment.
Ensemble Forecasting and Probabilistic Impact-Focused Outputs
Understanding and incorporating uncertainties in stormscale hydrologic forecasting is paramount. FLASH development has been conducted within an Ensemble Framework For Flash Flood Forecasting, a flexible design to accommodate multiple modeling principles, forcings, and probabilistic outputs. Details obtained from the flash flood observation database are being incorporated into forecast products to provide end-users with specific details about the anticipated location, timing, and magnitude of flash flood impacts (i.e., flooded roads, inundated crops, infrastructure).