ABSTRACT:

The increase in the societal and economic impacts of flooding across the eastern United States has brought attention to the potential link between long-term increases in urban areas and changes in the watersheds’ flood response. One outstanding challenge is to isolate the effects of land cover changes from other flood-related factors. To advance our understanding of these processes and their nexus, we utilize a statistical framework in which we use different parameterizations of the Generalized Pareto distribution (GPD) to model sub-daily peak-over-thresholds (POT) events at 102 stream gauges in the following metropolitan areas across the eastern United States: Boston, New York, Philadelphia, Baltimore, Charlotte, Atlanta, Houston, and Tampa. While we keep the shape parameter constant, we allow the scale parameter to: 1) be constant; 2) depend on hourly accumulated rainfall; or 3) be dependent on a combination of hourly accumulated rainfall and the temporal changes in the percentage of the watershed’s developed land. Based on our modeling results, we select the model with the land change as predictor in only 3% of the watersheds. Moreover, the model configuration in which rainfall is the only predictor is selected the most frequently (~80% of the sites) across the eight metropolitan regions. Therefore, our findings indicate that the key flood driver in urban basins across the eastern United States is rainfall, without clear evidence linking long-term changes of impervious area (i.e., urbanization) and the watersheds’ flood response.