The Dynamics and Predictability of Tropical Cyclones

Through methodology unique for tropical cyclones in peer-reviewed literature, this study explores how the dynamics of moist convection affects the predictability of tropical cyclogenesis. Mesoscale models are used to perform short-range ensemble forecasts of a non-developing disturbance in 2004 and Hurricane Humberto in 2007; both of these cases were highly unpredictable. Taking advantage of discrepancies between ensemble members in short-range ensemble forecasts, statistical correlation is used to pinpoint sources of error in forecasts of tropical cyclone formation and intensification. Despite significant differences in methodology, storm environment and development, it is found in both situations that high convective instability (CAPE) and mid-level moisture are two of the most important factors for genesis. In the gulf low, differences in CAPE are related to variance in quasi-geostrophic lift, and in Humberto the differences are related to the degree of interaction between the cyclone and a nearby front. Regardless of the source of CAPE variance, higher CAPE and mid-level moisture combine to yield more active initial convection and more numerous and strong vortical hot towers (VHTs), which incrementally contribute to a stronger vortex. In both cases, strength differences between ensemble members are further amplified by differences in convection that are related to oceanic heat fluxes. Eventually the WISHE mechanism results in even larger ensemble spread, and in the case of Humberto, uncertainty related to the time of landfall drives spread even higher. It is also shown that initial condition differences much smaller than current analysis error can ultimately control whether or not a tropical cyclone forms. Furthermore, even smaller differences govern how the initial vortex is built. Differences in maximum winds and/or vorticity vary nonlinearly with initial condition differences and depend on the timing and intensity of small mesoscale features such as VHTs and cold pools. Finally, the strong sensitivity to initial condition differences in both cases exemplifies the inherent uncertainties in hurricane intensity prediction. This study illustrates the need for implementing advanced data analysis schemes and ensemble prediction systems to provide more accurate and event-dependent probabilistic forecasts.