Distributed Hydrologic Modeling of Coastal Flood Inundation and Damage: Nonstationary Approach

Abstract

Superstorm Sandy and Hurricane Irene on the East Coast of the United States were wake-up calls that the floodplain delineation and flood damage estimation models need major overhaul. The first step in flood-related studies is frequency analysis. A great challenge has emerged on the validity of the data stationarity assumption. In hydrologic studies, data stationarity has been doubted due to climate change and variability, and sea-level rise. This paper presents an integrated scheme for floodplain delineation and flood damage estimation in coastal areas, considering the combined effect of inland and coastal flooding. For this purpose, time series of rainfall and water level and surge are tested for stationarity. Then, nonstationary frequency analysis is performed to determine the extreme flood events. Considering the coincidence of heavy rainfalls and storm surges as a serious concern in coastal regions, the joint probability distribution of rainfall and storm surge data is also investigated. Consequently, three flood scenarios are defined. A distributed hydrologic model is then utilized for floodplain delineation. A geographic information system (GIS)–based model, using depth-damage functions, land use data, digital elevation model (DEM), and raster maps, is used to estimate flood damage for Manhattan in New York City. Results showed that by incorporating nonstationarity in frequency analysis, design values of rainfall and extreme water levels could be significantly different than those obtained under the assumption of data stationarity. The results also indicated that floodplain extent and estimation of flood damage are increased when data nonstationarity is considered. However, these values are decreased when the joint probability of rainfall and water level is incorporated into defining flood scenarios.