Multiscale Processes and Predictability of Low-Probability Tropical Cyclogenesis

The understanding and accurate prediction of tropical cyclone formation (“genesis”) remains challenging. Sometimes, genesis from weak tropical waves in the western Atlantic and Caribbean occurs unexpectedly. These low-confidence forecasts give limited time for officials and the public to prepare for a potential hurricane impact.

The project is aimed at advancing knowledge of these issues using conventional and AI techniques, ensemble forecasts, and state-of-the-art modeling. The project will offer insights and suggestions to forecasters and model developers on the strengths and weaknesses of conventional and AI-based models. Another expected outcome is the identification of time windows of increased and decreased predictive confidence, and thereby a basis for enhanced, actionable forecasts.

The project will support the education of graduate and undergraduate students, and the introduction of AI and ensemble techniques in the classroom. Students will engage with professionals, advance outreach to increase public scientific literacy, and participate in mentoring programs. Results and software will be shared with the community.

The research will use reanalysis data, ensemble forecasts, and multiscale modeling to investigate the processes and predictability in the 5 days leading up to genesis and immediately after. The mechanisms of weak waves that had a low genesis probability but developed into tropical cyclones will be compared against higher-probability developers and non-developers.

While the fate of precursor disturbances is understood to depend on the environmental preconditioning and organization of convection, the specific nature of these processes remains open to question for the low-probability situations. A key hypothesis is that intense convective bursts occur on small scales and couple quickly with the wave, leading to upscale growth, rapid axisymmetrization, and genesis.

The sensitivity of genesis to the timing and location of convection is expected to be high, suggesting a short range of predictability. In higher-probability developers, this sensitivity is expected to be lower. It is expected that there will be case-by-case variability which depends on the strength and symmetry of the precursor.

The research will also include a novel AI-based wave tracker that surpasses the capability of conventional trackers in the western Atlantic and Caribbean, and investigations using the world’s first operational global ensemble that is driven by AI.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.