Using Crowd-Sourced Data to Improve Analyses of Flash Flood Events

Flash flooding continues to be one of the most dangerous short-fused weather events forecasted by the National Weather Service (NWS). From 1995-2012,flash flooding accounted for approximately 28% of the weather fatalities caused by short-fuse events (tornadoes, thunderstorm wind, hail, lightning, and flash flooding). Of these short-fuse weather events, only tornadoes have killed more over the same time period. Despite these statistics, many instances of flash flooding go un-documented due to low population density and lack of mandatory post-event surveys (as with tornadoes and straight-line wind damage). The lack of consistent flash flood reporting, including that of a null event, makes it difficult to properly analyze experimental flash flood tools and techniques. Experimental nowcasting techniques being examined by the NWS could help forecasters better predict the relative magnitude of events. Comparison of realtime rainfall estimates to estimated rainfall average recurrence intervals and the use of runoff routing techniques can put an ongoing event into the context of historical events for the purposes of estimating event severity. Ideally, for these techniques to be adequately examined, output should be compared to closely-spaced reports of flood severity as well as areas of no flooding. For multiple events over the last several years, anecdotal evidence of flooding was obtained from online media sources including newpaper and television websites. These reports were supplemented with anecdotes from social media, including Facebook and Twitter. The process of acquiring and organizing this information is discussed, and multiple case studies are presented.