Using Remote Sensing to Understand the Joint Probability of Extreme Rainfall and Flood Response

Abstract:

Floods are the products of complex interactions between the highly variable spacetime structure of extreme rainfall with land surface and drainage network features at various scales. Precise description of these interactions has proven elusive, mainly due to the lack of sufficient spacetime rainfall information and relatively short and sparse observational records. Rainfall remote-sensing data archives are now reaching sufficient length to examine these interactions in greater detail. Long-standing precipitation-based flood hazard estimation practices such as design storms and Probable Maximum Precipitation rely on simplified assumptions to describe the interactions between extreme rainfall and flood response. In this study, the validity of these assumptions are explored using RainyDay, a probabilistic stochastic storm transposition framework developed at NASA’s Goddard Space Flight Center for generating large numbers of rainfall “scenarios” using rainfall remote sensing data, each with realistic probability, intensity, and spacetime structure. RainyDay is coupled with NCEP Stage IV multisensor precipitation data and the Iowa Flood Center Model, an uncalibrated multiscale distributed hydrologic modeling platform. We study the relationship between simulated rainfall and peak discharge probability and intensity for a wide range of exceedance probabilities and for a number of nested subwatersheds of Turkey River in Northeastern Iowa, ranging in drainage area from 10 km² to 4300 km². The results demonstrate some interesting implications for the relationship between Probable Maximum Precipitation and the Probable Maximum Flood at a range of basin scales and highlight possible deficiencies in the standard approaches to compute these quantities. Satellite-based precipitation estimates with global coverage allow the extension of such understanding to data-poor regions.