Minutes that Save Lives: New Warning System Boosts Flash-Flood Prediction
With a 30-minute head start, emergency managers can dramatically reduce flash-flood risk along Colorado’s Front Range. The Front Range, with its steep topography and intense summer storms, is unusually vulnerable to summertime flash floods. A new forecasting system that underwent testing and development during summer 2008 may soon provide emergency managers with information on the likelihood of a flash flood minutes or even hours before waters start rising.
Flash floods are difficult to predict because they happen suddenly, often the result of heavy cloudbursts stalling over a particular watershed. Flash floods are difficult to predict because they happen suddenly, often the result of heavy cloudbursts stalling over a particular watershed. Today’s weather forecasting systems provide relatively accurate information on 6-hour to 1-day timescales; however, flash floods typically occur in minutes to hours. Radar systems provide real-time information on storm direction, but typically don’t provide forecast information on how a storm will grow or dissipate. Merging these systems’ capabilities with watershed-specific information—soil type, topography, hydrologic conditions, etc.—could substantially advance flash flood prediction.
In an effort to give emergency managers more notice about weather conditions that could lead to flooding, David Gochis and David Yates, both scientists at the National Center for Atmospheric Research (NCAR), created and are testing a prototype system that fuses weather forecasting capabilities and radar data with hydrologic data sets.
NCAR’s flash flood system uses National Weather Service radar observations of current conditions and short-term computerized weather forecasts. The weather forecasts are generated by the NCAR-based Weather Research and Forecasting model (WRF), which produces highly detailed simulations of the local atmosphere. The system integrates weather information with data sets that describe watershed hydrology and terrain. These data incorporate information about land surface conditions, such as terrain slope, soil composition, and surface vegetation. They also include information on streamflow and channel conditions. By combining information about the land and the atmosphere, the system projects whether an intense storm is likely to stall over a specific area of the Front Range and how it could affect the flow of water on the ground. "This new system is unique in that it provides a detailed forecast of the location and duration of a severe storm, as well as the watershed's likely response to the heavy rain," says Yates. "Since flash floods are complex and fast-moving events, we need to know about both weather and ground conditions in order to predict them."
During the summer 2008 testing season, the system generated five daily weather forecasts. Each began at a different time of day and ran for 24 hours. When severe weather was imminent, countless short-term (30- to 60-minute), radar-based precipitation and streamflow forecasts were generated for specific watersheds to identify the potential for flash floods.
Currently, Gochis and Yates are evaluating the forecasts to assess model accuracy. They will keep flood and emergency managers along the Colorado Front Range—Denver, Boulder, Fort Collins—aware of the model’s progress. Once model performance is deemed adequate, the model will be shared with the flood management community in Colorado and beyond.
