Mesoscale predictability of tropical storm Allison (2001) during its initiation and landfall

The focus of this study is to use short-term mesoscale ensemble forecasts to examine the predictability and dynamics of tropical storm Allison (2001) from its initiation to its landfall. Allison was a weak tropical storm that formed off the Gulf Coast of Texas on 5 June 2001 and brought record rainfall and catastrophic flooding to large areas of southeast Texas and Louisiana.

Allison's deluge brought much of the area, including the Houston metropolitan district, to a standstill. Her widespread rains dumped cumulative rainfall totals of nearly 90 cm in some isolated areas over a period of five days, resulting in nearly $5 billion in damage and claiming the lives of 22 Houston residents. In an observational study, Sippel et al. (2005) document the genesis and subsequent evolution of Allison on 5 and 6 June. That article studied the extent to which PV generation and superposition was relevant during Allison's genesis. The study found that the formation and subsequent superposition of multiple vortices was a key component through multiple scales of the cyclogenesis event.

Several sets of short-term mesoscale ensemble forecasts total of more than 100 integrations using different resolutions, different physical parameterizations and/or initialized with different analyses have been performed to examine the sensitivity of the storm to initial condition and model uncertainties. There are large ensemble spreads (and thus forecast uncertainties) within and between different ensembles in terms of both the intensity and location of the simulated storm and its associated heavy precipitation.

For example, in a 20-member ensemble that is initialized by adding small (e.g., 1m/s in u,v; 0.5K in T), balanced large-scale initial perturbations to the NCEP FNL analyses at 0000 UTC 5 June 2001, some of the ensemble members reasonably simulated the positioning of surface low at the 24-h forecast time (which corresponds to the observed heaviest precipitation in Houston). However, these same members varied significantly in both the storm intensity and precipitation characteristics. Some members do not simulate a closed low circulation near the southeast Texas coast while some members do not have any significant precipitation in the area. Our preliminary analysis shows that the extreme sensitivity in these ensemble forecasts originates from the triggering and development of the incipient convection. This is consistent with previous predictability studies of different weather systems (e.g., Zhang et al. 2002, 2005).

These ensemble forecasts are further used to derive probabilistic evaluations of the dynamics of the storm formation and landfall by examining the error covariances between different forecast variables at different times. We also plan to use an ensemble-based mesoscale data assimilation system to assimilate radar and other remote sensing observations to further examine the predictability and dynamics of this storm during its formation and landfall stages.