Tuesday, Aug 6: 3:05 PM - 3:20 PM
3206
Contributed Papers
Oregon Convention Center
Storm surge is the unusual rise in sea level caused by a storm's winds pushing water onshore. Due to high damage to roads, buildings, and lives, accurate estimation of storm surge high quantiles (e.g., r-year return level) and associated uncertainty are essential. Purely data-driven approaches to these tasks pose a challenge, as the number of hurricanes occurring near any single location is limited. A physically-driven approach, utilizing high-fidelity hydrodynamic computer simulations, provides an alternative by leveraging well-known physics to model how the surge level responds to hypothetical storms, where these simulated storms are parameterized by a few key characteristics. This approach requires the following tasks: 1) estimate the joint distribution of storm characteristics; 2) emulate the computer model input-output relationship via surrogate models; and 3) integrate out the input distribution to obtain the surge output distribution, then determine the synthetic data for high quantile (e.g. 1-in-100 year return level) estimation. A case study of this approach in South West Florida using the computer model ADCIRC will be presented to illustrate the proposed workflow.
Gaussian Processes
Computer Model
Quantile Estimation
Hurricanes and Storm Surges
Main Sponsor
Section on Statistics and the Environment