A Regional Probabilistic Approach for Estimating Flood Probabilities

In this study, we develop a statistical modeling framework to estimate the probability of flood threshold exceedances at arbitrary river locations within an area experiencing widespread extreme rainfall. Specifically, we derive a catalog of extreme rain events — defined as the occurrence of a 500-year return period rainfall depth or greater at one or more precipitation stations — for the northeast region of the United States. The catalog is then combined with flood peak records from U.S. Geological Survey (USGS) stream-gages to develop a regional logistic regression model for estimating flood probabilities. Various flood predictors (e.g., rainfall and site altitude) are explored to assess the effects. Results show potential use to evaluate probabilities of widespread flood events.