Improving upon flash flooding forecasts for two major Great Lakes cities
Major Goals and Objectives
Accomplishments / Benefits
Flash flood prediction is often a function of how much precipitation is anticipated to fall over the next 24-48 hours in a given location. However, many factors contribute to whether or not flash flooding will actually occur. Examples of other factors to consider include: whether the pervious ground is already saturated from recent rainfall events, if the ground is impervious, or if the location is historically known to experience flash flooding given particular environmental conditions.
This project had two objectives: (1) identify areas that are most at risk to flash flooding given their static environment; (2) operationally and statistically assess the risk of flash flooding given recent rainfall compared to historical conditions when flooding occurred. The first objective is achieved with a static map that considered various environmental parameters that rarely change (e.g., land cover; slope; population; type of vegetation) and highlights areas that historically produced flash flood reports versus those areas that rarely (if ever) produced a flash flood report. Commonalities were determined under those two options in order to develop a general guidance map of areas that are more likely to flood than others. The second objective is achieved by running artificial intelligence algorithms on historical precipitation data and the timing and location of when flash flooding was reported. These routines examine how much rain fell within the 6 days prior to when a flash flood report occurred compared to 6 days of precipitation that did not precede a flash flood. These algorithms were developed at individual locations to assess the risk of a flash flood being reported given the most recent precipitation and a selected list of potential forecast rainfall amounts. The end product is a flash flood risk map that forecasters could consider in conjunction with other forecast models and data when deciding whether to issue a
flash flood alert.
Paul Roebber, University of Wisconsin-Milwaukee
National Weather Service