Simulation of Hurricane Wind Fields for Community Resilience Applications: A Data-Driven Approach Using Integrated Asymmetric Holland Models for Inner and Outer Core Regions

Accurate modeling of the damage caused by hurricanes making landfall to physical infrastructure relies on the accurate modeling of temporally and spatially varying wind fields. However, wind field measurements from past events only include the wind field data at discrete time points separated by hours. In community-scale modeling, a hurricane wind field of a desired strength for the purpose of resilience planning (including investigation of mitigation alternatives) may be needed for a community of interest that has not experienced such a prior event, thereby necessitating simulation of a hurricane wind field of a specified strength with an arbitrary landfall location. In this context, this paper proposes a novel data-driven simulation technique to simulate temporally and spatially varying hurricane wind fields for the purposes of hindcasting and synthetic scenario analysis based on integrated asymmetric Holland models. First, the backward Holland model (i.e., Lambert $W$ function) is used to derive the time-varying model parameters from measurement data of historical events. Then the data-driven model parameters are adopted in the forward Holland model to simulate snapshots of missing times for historical events or realistic synthetic events with a synthetic track passing close to the community of interest (e.g., for the purpose of community resilience planning). To achieve high simulation accuracy, the wind fields for inner and outer core regions are modeled separately by two sets of asymmetric Holland models, whose parameters are estimated using two different branches of the Lambert $W$ function and in the end are integrated to represent the entire wind field. In addition, the sudden change of the surface wind speed due to the roughness change from water to land is explicitly modeled using a speed conversion process. In this way, the proposed technique successfully overcomes two shortcomings of the existing Holland-type models, that is, poor representation of the wind field in the inner core region and the inability to model surface wind speed change due to roughness changes. The performance of the proposed data-driven simulation technique is illustrated in examples of simulations for both historical and synthetic hurricanes.