Paper on using machine learning to improve rapid intensification forecasts released online in Weather and Forecasting – Hurricane Research Division

A new machine learning (ML) model has the potential to provide rapid intensification (RI) forecasts that are better than those currently available. The paper highlights the importance of proper predictor selection and ML development.  The approach taken in this project also sets guidelines for future relevant ML research.

Tropical cyclones (TCs) can greatly intensify within a short period of time, a process known as rapid intensification (RI), and forecasts of when and where RI will occur remains challenging.  When TCs undergo RI near the coastline, like Hurricanes Ian (2022) and Ida (2021), the time for emergency preparation and response is drastically shortened. In order to further understand RI and enhance our ability to forecast it, we developed a ML model using predictors calculated from the output of the high-resolution Hurricane Weather Research and Forecast (HWRF) model from TCs in the Atlantic and East Pacific Oceans. We analyzed the importance of the predictors to understand RI, highlighting new ones such as the relative humidity near the TC center. The results of the new model were compared with those from operational models used by the National Hurricane Center.

Major Conclusions:

• The new ML model, built upon three years of the HWRF model output, has comparable performance to SHIPS, and has better forecast skill than other models used by specialists at the National Hurricane Center (Fig. 1).
• Using information related to TC structure, such as humidity near the TC center, can improve RI forecasts.
• Since RI is a rare event, using ML methods to create additional RI data can further enhance the predictions of RI. 
• ML models are promising to provide reliable RI forecasts, and our method can be expanded to other topics.

For more information, contact aoml.communications@noaa.gov. The full text can be found at https://doi.org/10.1175/WAF-D-22-0217.1. This project is funded by National Oceanic and Atmospheric Administration.