Speaker:	Dr. Daniel B. Wright
Institution:	Civil & Environmental Engineering
	University of Wisconsin-Madison
Seminar Date:	Thursday, May 6th, 2021, 3:30-4:50 PM

Predicting the frequency and severity of floods has been a longstanding challenge in hydrologic research and practice. Despite the fact that every flood is a unique combination of multiple physical processes (rainfall, snow, groundwater, and surface runoff, to name just a few), the estimation of key metrics such as the 100-year flood has generally been treated as a statistics problem. This view limits the opportunities to use our evolving understanding of how actual floods work in order to improve such estimates. This limitation has become particularly problematic in an era of rapid climatic and hydrologic changes that can influence flood frequency and severity.

In this presentation, I argue that long-term flood prediction can benefit from deeper consideration of the physical processes that cause floods, and particularly from decades of progress in the observation and simulation of these processes. I briefly describe a process-based approach that we use to generate estimates of flood frequency and severity. This approach is flexible, making it is straightforward to integrate recent advances in rainfall measurement, flood simulation, and climate modeling. I will show results from two locations: 1.) an agricultural watershed in Iowa that is undergoing rapid hydrologic change; and 2.) a mountainous watershed in the Colorado front range that exhibits complex flood seasonality. These examples show that the approach can produce results with comparable or superior accuracy to more conventional statistical methods and can provide insights into how changes in physical processes lead to changes in flood hazard and risk.