Why did “spin down” still occur during HWRF intensity prediction of Patricia (2015) initialized by a more realistic analysis produced by the advanced hybrid DA system?
Inner core observations collected from the recent field campaigns such as ONR TCI and NOAA IFEX have provided a unique opportunity to describe the 3-dimensional hurricane inner core structures. Our early work showed that assimilating these observations (including TCI dropsondes, IFEX SFMR, TDR, flight level data) together with the high resolution CIMSS Atmospheric Motion Vectors (AMVs) can produce realistic three-dimensional analyses using the newly developed GSI-based, continuously cycled, dual-resolution hybrid 3DEnVar DA (data assimilation) system for HWRF (Lu and Wang, 2017) for hurricane Patricia (2015).
However, the Vmax forecasts initialized from the realistic analyses produced by the hybrid DA system drops more than 10 m/s during the first 6 hours in Patricia. Meanwhile, the unrealistic initial conditions produced through VI (Vortex initialization) does not show such a significant spin-down issue. Therefore, in this study, experiments and diagnostics were conducted to understand why the significant spin-down issue occurs when initialized from the much more realistic and accurate analysis.
Diagnostics reveal that the spin-down issue is caused by the errors in the HWRF model. As a result, the HWRF model is unable to match and therefore maintain the realistic 3D structures in analysis. Further experiments with different modified model physics parameterizations revealed that errors in the HWRF model physics is one of the major contributors to such errors. In particular, a modified turbulent mixing parameterization scheme was found to significantly alleviate the spin-down issue. Diagnostics showed that the modified turbulent mixing parameterization scheme allows mixing in the cloud above the boundary layer top, which induces more stronger vertical mixing and convections in the eyewall region. The stronger and realistic secondary circulation incurred by the more accurate physics parameterization is more consistent with the initial storm analyzed by the advanced hybrid DA method and eventually leads to realistically rapid storm intensification.
Additionally, although the spin-down issue is alleviated with improved physics and therefore the peak simulated intensity is significantly improved, there is still a large gap between the simulated maximum intensity and the unprecedently observed large peak intensity for Patricia. Further experiments showed that the peak simulated intensity can be significantly improved by increasing the model resolution