Flood-frequency analysis is essential to water-resources management applications, including critical structure design such as bridges, culverts and floodplain mapping. A multi-phase Transportation Pooled Fund study led by the U.S. Geological Survey (USGS) is currently underway to evaluate potential nonstationarity in annual peak streamflow records and timing (seasonality) due to hydroclimatic variability in the Midwest United States which includes three discrete trend periods through water year 2020. Regional patterns for both gradual (monotonic) and abrupt changes in annual peak streamflow are found among long-term USGS streamgages corresponding to temporal transitions between flood rich and flood poor periods. Further, regional patterns are found in changes in seasonal characteristics among a subset of USGS streamgages with abrupt changes before and after the detected change points, indicating potential changes in flood generating mechanisms. A Generalized Additive Model in Location, Scale and Shape (GAMLSS) is used to model more complex patterns in annual peak streamflow records among candidate two-parameter distributions. This modeling framework is used to characterize potential nonstationarities in both the interannual variability in annual peak streamflow as well as changes in seasonality related to changes in hydroclimatic variables. This presentation will provide preliminary results and describe the benefits of using a GAMLSS model framework to model both stationary and nonstationary models when they exist.
