Accounting for Uncertainties in Compound Flood Hazard Assessment with Data Assimilation
Abstract
Compound flood hazard assessment (CFHA) and modeling are subject to various sources of uncertainty including forcing data, model structure, model parameters, and those associated with nonlinear interactions among flood drivers. Data assimilation (DA) is an efficient method that helps account for uncertainties in many hydrological applications and has proven to be effective in water level (WL) predictions and flood forecasts. However, research to date has not yet explored the benefits of DA in coastal to inland transition zones where pluvial, fluvial, and coastal flood drivers interact. In this study, we present a DA scheme consisting of the Ensemble Kalman Filter (EnKF) technique and hydrodynamic modeling (Delft3D-FM) that provides: (i) reliable WL predictions, and (ii) accurate (near real-time) flood hazard maps (6-h update) in accordance with hurricane advisories of the National Hurricane Center. The DA scheme is tested on two well-known compound flood events and study sites in the United States, namely Hurricane Harvey for Galveston Bay, TX and Hurricane Sandy for Delaware Bay, DE. WL predictions and compound flood hazard maps are validated against observational data collected from coastal and inland gauge stations and high-water marks obtained from the U.S. Geological Survey, respectively. We show that the DA scheme can effectively account for uncertainties and reduce errors in peak WL estimates (up to 0.55 m) as well as reduce mean absolute bias in CFHA (up to 40%). We conclude that, regardless of the dominant fluvial/pluvial or coastal driver, DA can improve CFHAs in low-lying areas including coastal to inland transition zones.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFMNH21A..04M