Uncertainty Analysis of Flooding Drivers in the City of Savannah by an Urban Flooding Model

This study assesses flood variability during different hurricane events along the Savannah River’s fluvial system. To achieve real-time estimations of water surface elevations (WSE) across river channels and reaches, we developed a Physics-Informed Geostatistics Approach (PIGA) that combines reduced spatial correlations derived from a fluvial hydrodynamic model with a Reduced Geostatistical Approach (RGA) based on observational data. A 1D-2D HEC-RAS model was constructed as the physical model for the Savannah River Basin, an area highly susceptible to riverine flooding during hurricane seasons. The performance of the PIGA was compared to the HEC-RAS model across multiple hurricane events, including Hurricanes Matthew, Irma, Dorian, and Idalia. Our study explores the transferability and variability of reduced spatial correlations across different hurricane events, assessing the capability of one event’s spatial data to predict WSE during other events. Results indicate that spatial correlations derived from one hurricane event can reliably predict WSE for other events, with the correlations from Hurricane Irma demonstrating the greatest versatility for our study site. However, significant variability in spatial correlations between hurricane events was also observed, underscoring the need to carefully select the optimal reference event to achieve accurate predictions.

Research sponsored by the USGS/NIWR Water Resources Act Program 104b Grant (Fiscal Year 2023-2024)

Report Title

Uncertainty Analysis of Flooding Drivers in the City of Savannah by an Urban Flooding Model

Principal Investigator(s)/Authors and Affiliation

Dr. Jian Luo
School of Civil and Environmental
Georgia Institute of Technology

Abstract

This study assesses flood variability during different hurricane events along the Savannah River’s fluvial system. To achieve real-time estimations of water surface elevations (WSE) across river channels and reaches, we developed a Physics-Informed Geostatistics Approach (PIGA) that combines reduced spatial correlations derived from a fluvial hydrodynamic model with a Reduced Geostatistical Approach (RGA) based on observational data. A 1D-2D HEC-RAS model was constructed as the physical model for the Savannah River Basin, an area highly susceptible to riverine flooding during hurricane seasons. The performance of the PIGA was compared to the HEC-RAS model across multiple hurricane events, including Hurricanes Matthew, Irma, Dorian, and Idalia. Our study explores the transferability and variability of reduced spatial correlations across different hurricane events, assessing the capability of one event’s spatial data to predict WSE during other events. Results indicate that spatial correlations derived from one hurricane event can reliably predict WSE for other events, with the correlations from Hurricane Irma demonstrating the greatest versatility for our study site. However, significant variability in spatial correlations between hurricane events was also observed, underscoring the need to carefully select the optimal reference event to achieve accurate predictions.

Report

Final-Report_FY23_LUODownload