The Impacts of Moist Convection on the Predictability of Tropical Cyclones

This study examines the mechanisms by which initial condition error can cause significant uncertainty in strength forecasts of tropical cyclones. The first case studied is a tropical disturbance that was associated with significant forecast uncertainty in the Atlantic basin during the 2004 hurricane season. The second case is Hurricane Humberto, which intensified from a tropical disturbance to a strong Category 1 hurricane in the day before it made landfall along the upper Texas coast in September 2007.

Ensemble statistics and sensitivity tests in the 2004 case reveal that the amount of convection early in the simulations, which is modulated by the initial convective instability, is instrumental in forming a deep vortex. Also, widespread cold convective downdrafts that form during the convection subsequently damp convective activity in a period that sees neither vortex growth nor decay. After the boundary layer recovers, convection reignites and stronger vortices strengthen more quickly. In the sensitivity tests, large differences in storm strength result from small initial condition differences. Finally, it is found in the sensitivity tests that the details of how the initial vortex is built also significantly depend on small initial condition differences. This is due to chaotic interactions of mesoscale features whose timing and placement significantly vary with slightly initial differences. Therefore, initial differences can more easily explain differences in subsequent area-average quantities (such as average wind speed) than absolute quantities (such as maximum wind speed).

In the case of Humberto, predictability was also very limited. The storm was marked by very large operational intensity forecast error, and no guidance predicted its intensification to a hurricane. Ensembles and deterministic simulations starting in the 12 hours before and after Humberto's genesis fail to capture the intensification of the cyclone. However, when radial velocity data from the developing storm is assimilated with an ensemble Kalman filter, subsequent forecasts capture/span the rapid intensification period. This clearly demonstrates that the inability to forecast Humberto was related more strongly to deficiencies in initial conditions than inadequacy of the forecast model. While use of data assimilation helps the ensemble to span reality, extreme ensemble spread remains, thus showing continued large uncertainty. Further work in this case will investigate the source of uncertainty and compare it to that in the 2004 case.