Flashiness Intensity-Duration-Frequency: A New Approach to Quantify Flash Floods

Flash floods are one of the deadliest weather-related natural disasters, yet our ability to observe them in fine spatiotemporal detail or to quantify them remains elusive. In basins with streamgages, a flashiness variable can be derived from the observed hydrograph to describe the slope of the rising limb of a flood wave. This concept has been used in prior studies to highlight regions of the US that are particularly susceptible to flash flooding given heavy rainfall. In this study, we expand the concept by conducting a statistical analysis of flashiness in gaged basins across the US to yield intensity-duration-frequency (IDF) curves, similarly to what is typically done with rainfall accumulations. This enables us to map flashiness annual exceedance probabilities for durations of 1-6 hr. Next, we evaluate correlations of basin attributes to flashiness, revealing characteristics of the soils, land cover, anthropogenic factors, hydroclimatological, and physiographic features that lead to flashy responses to rainfall.

The flashiness concept is then expanded to ungaged basins using a machine-learning approach and a physics-based hydrologic model that’s been forced with a decade of rainfall from the MRMS archive. The models can yield hydrographs similar to streamgages, thus enabling the computation of flashiness at all 1km-resolution grid points across the US. A statistical analysis follows to generate IDF curves of flashiness from simulated data. The simulated flashiness IDF analyses are compared to the observed maps and to observations of flash flooding provided by the National Weather Service, showing excellent agreement. Insights will be provided into how the static maps can be used for applications in water resources management and risk planning. Model-forecast flashiness can be generated in real-time and thus provide potentially powerful information for quantifying the intensity, duration, and frequency of an impending flash flood event.