Development of Real-Time Tools for Hurricane Risk Assessment

摘要

This paper examines the development of real-time hurricane/storm surge risk assessment tools utilizing existing databases of high-fidelity simulations. Foundation of the approach is the formulation of a surrogate model that can provide an approximation to storm responses utilizing the information in any such database. Kriging metamodeling is adopted here for this purpose as it facilitates a matrix-based, computationally efficient evaluation of the response over an extensive coastal region, including a large number of points (nodes) for which this response is evaluated. To alleviate problems related to the high-dimensionality of this output, principal component analysis is considered as a dimensional reduction technique, contributing significantly to computational efficiency (lower memory requirements and faster evaluations). Additionally, a correction stage is introduced to address problems for nodes that have remained dry for certain storms in the initial database. The developed surrogate model can be ultimately utilized for real-time predictions. The approach is illustrated for estimation of the surge risk in the New Orleans region.