November 18, 2010 Meeting Summary
Hua Chen from UMD presented some of her dissertation work on "Numerical Prediction of Hurricane Wilma (2005)" Hua began with the motivation for her work and reason for selecting Hurricane Wilma for this study. Wilma (2005) broke many records including for deepening rate, central pressure for an Atlantic storm, and small eye size. Wilma also produced a large storm surge and high rainfall rates on the Yucatan peninsula, and the forecast for Wilma also showed large track errors and underprediction of intensity. Next, Hua detailed her model configuration. She used WRF version 3.1 integrated from 00 UTC on 18 October to 00 UTC on 21 October 2005. This covered the 18 hours prior to the onset of rapid intensification (RI), the 18 hours of RI, and the 36 hours of weakening associated with an eyewall replacement cycle (ERC). Initial and boundary conditions were taken from the GFDL forecast with 55 vertical levels. Four domains were used with 27km grid size for the largest domain, 9km grid size used for the second largest domain, 3km used for the second smallest domain and 1km used for the innermost/smallest domain.
Hua then described her model verification of track and intensity. The solid line represented the model track and intensity while the dashed line represented observed track and intensity. For track, the model followed observations for the most part in capturing the NW movement of Wilma, but the model exhibited a slight NW bias. Hua explained that this bias was caused by the bogus vortex in GFDL which was large enough to alter the environmental flow and impact the track. For intensity, the model captured the overall features well, but model RI occurred 3 hours earlier than what was observed and decreased sooner than what was observed likely due to a fictitious double eyewall cycle in the model. Overall, the model max intensity was weaker than observed. Next, Hua discussed environment and eye soundings from her work. For the environment soundings, two (A and B) were taken from the dry air intrusion NW of the storm and two (C and D) from Wilma's rainbands. Dry air can be seen in the upper two (A, B) soundings, while the lower two (C,D) were much more moist. For the eye soundings, the upper panels show what was observed and the lower panels show model results. Hua explained that these sounding are a good example of how the model underpredicted intensity. Early in the forecast, the model does a good job capturing the early stages of intensification, however, later in the forecast, the model produces a thicker PBL depth indicating more weakening than what was observed.
Next, Hua looked at some inner core structures comparing SSM/I satellite images on the left to model-predicted reflectivity on the right. Satellite images were used because observed radar reflectivities were not available. The top two panels show Wilma near its peak intensity while the bottom two panels show Wilma near the end of an ERC. Overall, the model simulations are able to capture the inner structure of the eye and rain bands as well as overall storm size. In her description of the radial profile of tangential wind, Hua noted some differences between the observations (on the left) and the model simulation (on the right). Most notably, the simulated inner eye is larger than what was observed and the simulated outer eyewall contracted more slowly than observations. However, the model was able to capture the ERC pretty well. Looking at the axisymmetric storm structure, Hua noted several features including a fictitious outer eyewall which she believed was due to the microphysics used. This eyewall caused less of a contraction in the inner eyewall, which was seen in the tangential wind images. Hua also showed an animation of equivalent potential temperature at 1km (animation on the left) and 4km (on the right). 5 minute time intervals were used with no convection scheme, just microphysics. In the animations, the fictitious double eyewall can be seen and boundary layer clouds are very clearly captured.
Hua then presented some images on the upper level warm core, which was a unique feature. In the time (x-axis) vs. height (y-axis) cross section, the upper level warm core is clearly visible. Hua explained that this feature relates to the intensity changes. The warm core is also clearly visible on isosurfaces. At 18h, the isosurface is mostly flat, however by 36h concavity has developed. To conclude, Hua summarized her presentation by noting that the model she used did a reasonable job with track and intensity for Wilma. The model simulations also capture the inner-core structure changes as Wilma's intensity increased and decreased. However, the model produced an eye size twice as large as that observed and was not completely able to capture the depth of the PBL during RI.