Blending Parametric and Gridded Models of Tropical Cyclone Wind and Pressure Fields

Considerable advances have been realized over the past decade or more in the representation of tropical cyclones (TCs) in numerical weather prediction models. Improvements are most evident in reductions in forecast storm track errors, although even when the track is well represented, many of these models still lack the resolution and surface boundary condition information to correctly represent storm size and intensity. This is true for forecast models and for non-forecast models used for reanalysis and climate modeling, whose horizontal grid resolutions are typically 0.25° x 0.25° or larger. A second source of data describing TC wind and pressure fields comes from meteorologists at the National Hurricane Center or the Joint Typhoon Warning Center based on their synthesis of observational data and model predictions. This data is comprised of bulk storm parameters and can be found in storm forecast advisories during events and in sources such as the International Best Track Archive (IBTrACS) for past storms. These parameters are often considered to be more accurate than the representation of TCs in individual numerical weather prediction models and are often used to evaluate the performance of the numerical models. However, the bulk storm parameters do not provide a full spatial representation of the TC wind and pressure fields and they are usually specified at 6 hour or larger time intervals. Parametric TC models provide a means for converting the bulk parameters into full spatial fields at a finer time interval, although they do not account for meteorology other than the TC.

Several parametric TC models, based on the gradient wind balance presented in Holland (1980), are integrated into the ADCIRC coastal circulation and storm surge model. The most advanced is the Generalized Asymmetric Holland Model (GAHM, Gao 2018) which removes the cyclostrophic assumption in Holland (1980) and utilizes the radial distances to the 34, 50 and 64 knot isotachs in the four storm quadrants in addition to the eye location, central pressure, and maximum wind speed, to more completely represent the TC. ADCIRC also contains a crude algorithm to blend GAM with gridded meteorological fields from the North Atlantic Mesoscale Model.

Recently, we have rewritten an improved version of GAHM as a standalone code in both Matlab and C++ and have been working to better integrate this with gridded large-scale meteorology. This involves determining a non-TC background field that is used with the TC bulk parameters and GAHM to create a spatially explicit representation of the TC wind and pressure fields. These are then blended into the large-scale meteorology. A companion talk by Shankar et al presents surge and wave results using this approach in ADCIRC+SWAN simulations of hurricanes Florence and Ida.

We expect this capability to be available as a community code and via the MetGet interface for both retrospective and forecast applications.