| description abstract | This study aims to make the four-dimensional variational (4DVAR) bogus data assimilation (BDA) scheme for hurricane initialization first proposed by Zou and Xiao more objective. The BDA scheme consists of two steps: (i) specifying a bogus sea level pressure (SLP) field based on parameters observed by the Tropical Prediction Center (TPC) and (ii) assimilating the bogus SLP field under a forecast model constraint adjusting all model variables. In previous studies, specification of the bogus SLP was based on Fujita's formula, requiring the central SLP pressure (Pc), the radius of the outermost closed isobar (Rout), and the radius of the maximum SLP gradient (R0) as inputs. Although the parameters Pc and Rout are provided directly by the TPC, R0 is not. In this research, an empirical linear model designed to determine the value of R0 (the size of the bogus vortex) from the TPC observed radius of 34-kt wind (R34kt) is developed. Numerical experiments are carried out for the initialization and prediction of Hurricane Bonnie (1998) over the Atlantic Ocean. The Pennsylvania State University?NCAR nonhydrostatic mesoscale adjoint modeling system (Zou et al. 1997) is used for both the data assimilation and prediction components of the forecast. In order to study the sensitivity of hurricane initialization and prediction to the radial profile specification of the bogus vortex, the same experiment is conducted using Fujita's formula with R0 = Rmax (the radius of the TPC observed maximum wind) and another formula, Holland's formula, for the specification of the bogus SLP. The track prediction is less sensitive to the specification of the bogus SLP than the intensity prediction. The maximum track error is less than 110 km during the entire 3-day forecast for any of the three experiments using different bogus SLP specifications. However, the experiment using the linear model for the size specification required by Fujita's formula considerably outperforms the other two formulations for the intensity prediction of Hurricane Bonnie. Diagnosis of model output indicates that the 4DVAR BDA generated an initial hurricane, which allows for larger amounts of surface fluxes of heat and moisture, angular momentum, and latent heat of condensation, supporting a stronger and more realistic hurricane with more realistic intensity changes than experiments using the other two formulations. | |
