A Comparison of Bankfull Flow Estimation Methods at Gauged Locations Across the Contiguous United States

Floods are one of the leading causes of death from natural hazards in the United States (US). Better prediction of bankfull flow is needed in order to communicate potential flood risks and initiate response within forecasting frameworks. Traditional bankfull flow estimations are based on empirical equations whose accuracy is shown to be insufficient across a large range of streams and large geographic domains. We investigate the variability in bankfull flow estimates using well-established indices across the contiguous United States, including the National Weather Service (NWS) Action Flow. Its association with bankfull flow and utility as a bankfull flow proxy is explored. At 812 US Geological Survey (USGS) gauges where an NWS Action Flow is defined, we computed bankfull flow estimates using the minimum width-to-depth ratio, the Bench Index, and the USGS 17C flood frequency analysis. We provide a geospatial dataset of these bankfull flow estimates, the NWS Action flows, and observational bankfull flow when available (~10% of the locations). At 87% of the locations, we find a high degree of variability between the estimates produced by these methods, indicating that the method selected has a significant impact. We find the NWS Action Flow a reasonable bankfull flow proxy, performing as well as or better than the established methods considered.