Dynamics and Predictability of Secondary Eyewall Formation of Tropical Cyclones Under Vertical Wind Shear

A series of ensemble experiments with different environmental vertical wind shear conditions are conducted using the convection-permitting WRF model to study the dynamics and predictability of secondary eyewall formation (SEF) and eyewall replacement cycles (ERC) of tropical cyclones. It is found that the SEF in each of our simulations initiates from the outer rainbands that are primarily driven by the imposed environmental vertical wind shear. The probability of SEF increase with increased shear magnitude until the shear is strong enough to kill the tropical cyclone. Among the 40 simulations with a moderate 6-m/s wind shear, 5 experiments go through a clear eyewall replacement cycle (some of them have more than one ERCs), while the others undergo what may be called “partial eyewall replacement” as the inward contracting outer rainbands merge with the original eyewall. The SEF in these simulations is found to be very sensitive to the microphysics parameterizations used by the model. By removing ice processes at different time after the tropical cyclones have developed, we can see an obvious transition from the “partial replacement” or “merger” case to the more typical SEF. This result implies the importance of the ice microphysics and the stratiform dynamics in the organization of the outer-rainbands as suggested in some past studies.

A composite study of the members with clear-cut SEF/ERC shows that the super-gradient force in the boundary layer (whose importance has been suggested by some recent studies) is not the primary factor in initiating the SEF, at least for the group of the SEF simulations that we obtained. However, the agradient force may have contributed to the SEF process after the SEF has been initiated (through a positive feedback mechanism). More specifically, the super-gradient force does not lead the convective cells in our simulations. Second, strong convective cells can be found at sub-gradient wind region. Different with prior studies, the secondary wind maximum shows up at the altitude around 3km, which evidences that the lower-mid level air may play a crucial role in the SEF. Diagnosis also finds that the eddy-mean flow interaction has little contribution on the SEF.

Moreover, we find, even under the same environmental conditions, the tropical cyclone genesis and subsequent SEF/other intensity changes can be extremely sensitive to small, unobservable, random initial condition errors. The upscale error growth from small initial perturbations through random and chaotic moist convection may lead to considerable difference in the intensity variations, which is more so for environmental conditions that are only marginally favorable. Furthermore, the same level of intensity divergence can also be found between experiments with exactly the same initial conditions under the same environmental shear but are performed on different computer clusters, which is reminiscent of the “butterfly effect” that led to the development of the chaos theory half a century ago. The predictability for tropical cyclone formation and intensity change are flow dependent, the least predictable under marginally favorable environmental conditions; this lack of predictability could potentially be further exacerbated by inclusion of more spatial and temporal variations in environmental conditions.