Assessing Flood Risk from Hurricane-induced Precipitation and Storm Surge: A Bayesian Network Approach

Abstract:

Hurricanes pose a major flood hazard to communities on the U.S. Atlantic and Gulf Coasts. Over the past decade, the economic costs associated with hurricane flood damages have escalated and recent studies indicate that a large percentage of flood damages are occurring outside of FEMA-designated flood hazard areas. While FEMA recently upgraded coastal flood hazard maps using the Advanced CIRCulation (ADCIRC) Model, these maps do not consider the flood hazard resulting from the joint occurrence of precipitation over the watershed and storm surge at the coast. Instead, the two individual hazards are mapped separately, ignoring the floodplain resulting from their interaction.

In this study, a risk assessment methodology was developed to predict the damages associated with hurricane-induced flooding in the Houston Galveston Bay Area. Historical hurricanes were analyzed to derive probability distributions for storm surge height, cumulative precipitation, hurricane landfall, wind speed, angle of approach, radius to maximum winds, and forward speed. A Bayesian Network was built and used to simulate a large number of synthetic storms. The resulting 1% combinations of storm surge and precipitation were applied as boundary conditions to a hydraulic modeled and the maximum extent of flooding was compared to the FEMA-designated flood hazard areas. A high resolution GIS-based model was used to predict damages.