Rapid Intensification in Tropical Cyclones: Understanding Physical Processes and Forecast Uncertainly Using High-resolution Stochastic Ensembles

Forecasting rapid intensity (RI) changes in hurricanes remains a major challenge. Most major hurricanes go through RI during their lifecycle. A goal of the Hurricane Forecasting Improvement Project (HFIP) is to increase the detection rate of RI from the current ~30% to 90% for 24h forecast lead time. However, the physical processes contributing to RI are complex and not well understood, and the prediction skill of RI in both dynamical and statistical models is poor. The objective of this study is to better understand the physical processes leading to RI and systematically assess uncertainty in RI prediction. To address this complex problem, we take a new approach using three high-resolution WRF model ensembles employing the stochastic kinetic energy backscatter (SKEBS) scheme. The model is configured with triply-nested vortex-following domains with 12, 4, and 1.33 km grid spacing, respectively. The SKEBS algorithm mimics small amounts of kinetic energy that were lost to diffusion and “backscatters” them onto the resolvable scale by adding a small term to the u, v, and T tendency equations stochastically at each time step. Seven-day ensemble forecasts of Hurricane Earl (2010) are used for this study. One of the most fundamental questions is whether physical processes in the TC environment or internal to the TC vortex dominate RI. We designed the three ensembles with scale-dependent SKEBS perturbations. The ensembles have 20 members each and are constructed as follows: 1) the SKEBS-syno ensemble features synoptic-scale perturbations, (500-4200 km), which are added to all model grids, 2) SKEBS-meso ensemble perturbations are approximately on the mesoscale (24-500 km) and also added to all grids, 3) the SKEBS-conv ensemble features much smaller convective-scale perturbations (2.66-10 km), which are only added to the innermost domain while the storm environment remains unperturbed. The results show that virtually all TCs in the three ensembles have a period of RI. However, there is considerable uncertainty with respect to the timing and magnitude of RI. The timing for the start of RI varies from 3-4 days in SKEBS-syno and 2-3 days in SKEBS-conv. This result indicates that, although the environmental condition is inducive for RI in this case, a deterministic prediction of RI is still inherently difficult. Even the small-scale stochastic perturbations within a TC circulation in SKEBS-conv lead to a pronounced uncertainty with respect to RI timing of 2-3 days. Stochastic perturbations on the large-scale TC environment seem to produce the largest uncertainty in RI prediction. This study is aimed at identifying the physical processes that control the timing and magnitude of RI and how they contribute to forecast uncertainty. Composites of all TC RI events from the three ensembles will shed some lights on the conditions leading to RI. Comparisons of distinct properties in the three ensembles will help address the question of how the scale-dependent perturbations affect RI.