| Description |
Over the last several decades, ensemble forecasts of atmospheric phenomena have become increasingly popular, not only because they provide an improved mean forecast of various events, but also because they render an estimate of the accompanying forecast uncertainty. Research into high-resolution ensembles based in the Tropics and in terms of tropical cyclone (TC) genesis mechanisms has been relatively sparse, even though such disturbances are notoriously difficult to forecast. In this study, we couple several popular ensemble perturbation methods to the mesoscale Weather Research and Forecasting (WRF) model at high resolution to examine the predictability of genesis, error growth characteristics, and underdispersion issues in forecasts of Hurricane Ernesto (2006) and Typhoon Nuri (2008). In order to examine the effects of model resolution on TC genesis forecasts, a downscaled 5-km resolution regional control ensemble, based on a downscaling of the National Centers for Environmental Prediction's Global Ensemble Forecast System (GEFS), is compared against the standard GEFS simulations. To analyze the effect of the various perturbation methods on genesis and forecast characteristics, we compare results from the regional GEFS-based simulation to several implementations of the breeding of growing modes (BGM), wherein we vary the variables perturbed, cycling period durations, and boundary conditions. While the global GEFS forecast failed to predict a well-developed Ernesto in any of its members, the high-resolution GEFS-based ensemble contained several intense TCs by actual genesis time. Based on a sample of 154 ensemble member forecasts, the impact of environmental precursors on TC genesis likelihood is investigated. Despite the large number of easterly waves that do not develop into TCs and the large amount of water vapor in the summer Tropics, we find that the strength of the preexisting wave and initial 850 hPa water vapor are significant determining factors for TC genesis. Finally, we create several ensemble forecasts of Ernesto using the stochastic kinetic-energy backscatter scheme (SKEBS) and find that the standard SKEBS ensemble has more dispersion per unit error compared with both the BGM and GEFS-based ensembles. In addition, SKEBS shows notably lower vapor bias and larger theta bias compared with the initial condition-based ensembles. |
