Distinct Characteristics of Hurricane Ensemble Forecasts using Physical Parameterizations vs. Stochastic Perturbations

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 4 February 2014: 4:45 PM
Room C202 (The Georgia World Congress Center )
Shuyi S. Chen, Univ. of Miami/RSMAS, Miami, FL; and F. Judt, J. Berner, C. Y. Lee, M. Curcic, C. Snyder, and R. Rotunno

Ensemble model predictions are not only valuable for probabilistic forecasting with quantitative uncertainty estimates but also can be used for understanding predictability of weather systems. However, ensembles generated with various techniques often represent rather different aspects of the physical system. This study examines characteristics of ensemble hurricane forecasts, using the weather research and forecasting (WRF) model, generated from two different ensemble techniques to better understand the predictability of hurricane intensity. One is a physical parameterization ensemble in which each member uses a different model physical parameterization scheme, which represents the uncertainty due to parameterized physical processes such as microphysics, atmospheric boundary layer, and surface momentum and enthalpy fluxes. A second ensemble uses a stochastic kinetic energy backscatter (SKEBS) perturbation method, which adds stochastic forcing to the u, v, and temperature tendency equations at each time step, representing uncertainty due to unresolved processes such as turbulence. Both ensembles are 7-day forecasts using the WRF model with triply nested vortex-following domains. The innermost nest has a 1.3 km grid spacing that is cloud resolving. The ensembles are analyzed in terms of the storm track, maximum wind speed, storm size and asymmetry, and evolution of mean error kinetic energy. The two ensembles display distinct characteristics in hurricane structure and intensity forecasts. For example, the spread in the intensity forecasts in the physics ensemble is much larger through out the entire forecast period than that of the SKEBS ensemble, suggesting that the parameterized processes have a dominant impact on the model forecasted hurricane structure evolution. On the other hand, the SKEBS ensemble shows a greater spread during the rapid intensification of the hurricane than other time periods, indicating its sensitivity to various dynamic processes associated with the hurricane, which is less evident in the physics ensemble. The spatial scale of the stochastic perturbations can be specified in the SKEBS ensemble, which can be used to understand forecast uncertainty related to error growth at different scales from convection within a hurricane, the hurricane vortex, and the storm environmental flow. It is found that, while the error growth on the small scale limits forecast of convective features within a hurricane to hours, the model predictability of the hurricane vortex properties such as the vortex mean wind speed, storm size, and asymmetry is much longer and dominated by the storm environmental flow. These properties are much less clear in the physics ensemble.