Flooding is a major hazard in central Texas and impacts the City of Austin, Texas. The combination of intense and heavy precipitation coupled with hilly terrain in urban and peri-urban areas poses flood hazards, justifying the name, flash flood alley. Flood management planning personnel and decision-makers rely on Vflo, a physics-based distributed (PBD) hydrologic model coupled with radar and rain gauge inputs to make time-critical decisions as storms evolve over watershed drainage areas throughout the City. Floodplain mapping and LiDAR digital terrain models help to enhance the hydraulic configuration of the modeled drainage network making stage forecasts possible for predictive inundation mapping in near real-time. With hydraulic and infiltration parameters, Vflo transforms detailed rainfall information into forecast stage and discharge at locations throughout the City.
While many factors affect the accuracy of the quantitative precipitation estimates (QPE) and its usefulness for flood forecasting, the density of the rain gauge network in combination with radar makes it possible to provide actionable flood information in real-time. A recent flood-producing events including Tropical Storm Hermine, demonstrated the importance of knowing when and where flooding would occur with sufficient lead-time to take action. An added benefit of radar is that it improves accuracy of stage forecasts and lead-time, which is better than can be obtained with rain gauge-only input. QPE is produced using local bias adjustment of Level II reflectivity converted with an initial Z-R relationship that is constantly updated. The system tracks accuracy among other statistics of the QPE, which show great variability during storm events. While quantitative precipitation forecasts (QPF) from numerical weather prediction can provide more lead-time, it generally lacks specificity and accuracy, limiting its use for the site-specific flood forecasts needed by the City. Nowcasting provides short-term QPF used to provide advanced notice of approaching storms, and to make ready for emergency operations. Flood forecasting lead-time is associated with the hydraulic lag of surface runoff traveling through the drainage network to each forecast location within a basin. Evaluated forecast accuracy shows an 80% reduction in timing error using QPE derived from bias-corrected radar as opposed to gauge-only input. Accuracy of peak stage forecasts was found to be comparable to the rainfall accuracy of the radar product at 22 locations throughout thirteen watersheds. During heavy rainfall events, real-time hydrologic forecasts continue to be reliable and supportive of emergency response to rapidly evolving flood situations at locations distributed throughout the City.