Monday, 16 April 2018: 5:15 PM
Heritage Ballroom (Sawgrass Marriott)
An improved parametric tropical cyclone (TC) surface wind field model has been developed at AIR for hazard risk assessment. Direct, high-spatial resolution surface wind data acquired in TCs over the past twenty years have revealed details about asymmetric wind field structure in response to a vertically-sheared, environmental atmospheric flow, and these data are now routinely available from Atlantic Basin TC airborne reconnaissance missions. The primary improvements are two-fold: 1) utilization of these surface wind data; and 2) better representation of observed wind asymmetries. While numerical models of TCs are now run at spatial resolutions capable of representing fine-scale details of the surface wind field, computational resources remain prohibitive for estimating highly-localized damage risk, which requires hundreds of thousands of simulation-years for convergence. Hence, parametric wind field models still have a key role for identifying risk.
A data assimilation procedure is implemented to compliment the new parametric model. The wind field parameter error covariance is modeled in addition to mean parameter quantities, and an Ensemble Kalman Filter (EnKF) approach is used to combine the non-linearly related model parameters and surface wind observations, resulting in an optimized parametric representation of wind for an actual event. With the parametric approach, the state-space vector describing the wavenumber 0+1 wind field is quite small (11 modeled parameters total) , and so assimilating many thousands of observations over a full storm life-cycle is highly efficient, and yields wind speed error estimates consistent with known measurement uncertainty. In this talk, we will present results for major hurricanes Harvey, Irma, and Maria which impacted the United States and its territories in 2017. These storms serve as a proxy for applying the method to a stochastic model that can be used to evaluate TC risk. Companion talks will discuss the parametric model development and spatio-temporal modeling of parameter error covariance.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner