Water Resources Research Act Program

Details for Project ID 2020WI020G

Process-Based Prediction of Present and Future Flood Conditions and Uncertainties across the United States

Institute: Wisconsin
USGS Grant Number: G21AP10182
Year Established: 2020 Start Date: 2020-09-01 End Date: 2023-08-31
Total Federal Funds: $239,885 Total Non-Federal Funds: $239,889

Principal Investigators: Daniel Wright

Abstract: Floods result from combinations of “driving†processes, including spatially and temporallyvariable rainfall, soil moisture, evapotranspiration, snowpack, and runoff. These combinations areand will continue to change due to variations in rainfall and other processes. USGS streamflowobservations, though useful for practical estimation of flood frequency and severity, areinsufficient for understanding such changes. These observations and associated analysistechniques are thus ill-suited for understanding current and future flood conditions anduncertainties. New methods are needed to predict the frequency and severity of floods in order tosecure the long-term viability of our nation’s infrastructure and water management strategies.One promising direction is prediction via process-based modeling. A recent example is the WRFHydromodel, which underpins NOAA’s National Water Model effort to forecast floodsnationwide. It is very challenging, however, to ensure that process-based models are properly“calibrated,†i.e., that they are able to produce accurate predictions for the right reasons. Thesemodels must also be paired with the best available observations and projections of extreme rainfalland other driving factors if they are to be used to understand current and future conditions.This study will use a suite of novel tools and methods to better understand and predict floodfrequency and severity. In doing so, it would break new ground in the joint application of processbasedhydrologic models and high-resolution regional atmospheric model outputs. Informationtransfer to water experts, university students, and the public will ensure that scientific findingsreach diverse audiences. Four USGS co-PIs will take part, while the University of Wisconsin-Madison team will include a PhD student and several undergraduate students.The project has two overarching scientific research goals. Both will leverage scalable grid-basedhigh performance computing resources, software tools, and expertise at UW-Madison and theUSGS. The first research goal is to develop a unique calibration framework to support large-scalehydrologic prediction systems. The framework will combine new efficient, scalable calibrationtools from the USGS with a “sequential process-oriented†approach to calibration created at UWMadison.It will be evaluated in five distinct regions of the U.S., and will leverage new USGS dataresources and expertise in ways that have yet to be exploited in flood prediction. The calibrationframework will also provide explicit estimates of uncertainty, which are often overlooked and arecritical for making reliable hydrologic forecasts. The second research goal is to develop and applya process-based approach to predict current and future flood frequency and severity in five distinctregions of the U.S. It would do so using cutting-edge regional atmospheric model outputs, and,unlike “conventional†methods, would explicitly model the combinations and joint interactions ofrainfall, soil moisture, evapotranspiration, snowpack/melt, and runoff using modern stochasticmethods and the WRF-Hydro model. It will be used to examine changing flood conditions andrelated uncertainties across a range of hydroclimatic regions and watershed scales.The information transfer plan has two activities. First, we will create an “e-pamphlet†targetingthe general public, which clearly summarizes how floods are caused and how they are changingthroughout the nation, with an emphasis on the upper Midwest. Second, we will organize a “stateof-the-science†webinar for federal employees that will summarize the current knowledge ontrends in floods and flood-producing processes. The webinar will share results and discusspotential uses for cutting-edge models and computational tools for long-term flood prediction.