Regional surge hazard map development using Gaussian Process metamodeling

Abstract

The recent, very active hurricane seasons, as well as emerging concerns related to the future effects of sea level rise, storm intensification, and increased hurricane occurrence rate projections on coastal areas, make the prediction of storm-flood hazard a key priority when discussing coastal community resilience within planning (pre-disaster), emergency management, and post-disaster settings. To address this priority, researchers have placed substantial efforts in developing improved high-fidelity, numerical models to predict surges for a given storm. For promoting computational efficiency when utilizing these models within hazard estimation applications, surrogate modeling techniques have emerged as a popular strategy. The accuracy of surrogate modeling techniques in this context has been examined so far using cross-validation (CV) or test-sample validation techniques or by testing their performance for a (very) small number of historic storms. This paper investigates this topic within a different setting, examining the resultant regional storm surge hazard maps, specifically using Gaussian Process (GP) as metamodeling technique. This is accomplished by examining the hazard products obtained by GP implementations, as well as hazard products established through alternative, simplified approaches. Examining this accuracy fills in an important knowledge gap and provides an answer to the question ﾓwhat are the real benefits in hazard estimation by using surrogate models?ﾔ, improving at the same time trustworthiness of the associated results within the context of the coastal infrastructure risk quantification. A computationally efficient framework is also presented to explicitly consider the uncertainty associated with the GP predictions to provide confidence bounds for the hazard products.