Evaluation of bogus vortex techniques with four-dimensional variational data assimilation

The effectiveness of a four-dimensional variational data assimilation (4DVA R) technique for creating "bogus" vortices in numerical simulations of hurr icanes is evaluated in this study. A series of numerical experiments is con ducted to generate initial vortices for Hurricane Georges and Bonnie (1998) in the Atlantic Ocean by assimilating bogus sea level pressure and wind in formation into a mesoscale numerical model (MM5). Several different strateg ies are tested for investigating the sensitivity of the initial vortex repr esentation to the type of bogus information. While some of the results in this study confirm conclusions made in previou s studies, some significant differences are obtained regarding the role of bogus wind data in creating a realistic bogus vortex. In contrast with prev ious studies in which the bogus wind data had only a marginal impact on cre ating a realistic hurricane, this study concludes that the wind information is very important because 1) with assimilation of only bogus sea level pre ssure information, the response in wind field is contained largely within t he divergent component, with strong low-level convergence leading to strong upward motion near the center; and 2) with assimilation of bogus wind data only, an expected dominance of the rotational component of the wind field is generated. In this latter case, the minimum pressure is also adjusted si gnificantly, although the adjusted sea level pressure does not always match the actual hurricane minimum pressure. The generated vortex offers a smoot h start to the forecast and leads to a significant improvement in the forec ast. Only when both the bogus sea level pressure and wind information are a ssimilated together does the model produce a vortex that represents the act ual intensity of the hurricane and results in significant improvements to f orecasts of both hurricane intensity and track. As the 4DVAR experiments are performed with relatively coarse horizontal gr id resolution in this study, the impact of vortex size on the structure of the initial vortex is also evaluated. The authors find that when the scale of the specified bogus vortex is smaller than that which can be resolved by the model, the assimilation method may result in structures that do not co mpletely resemble observed structures in hurricanes. In contrast, when the vortex is sufficiently large for it to be resolved on the horizontal grid, but not so large as to be unrealistic, more reasonable hurricane structures are obtained.