Can Machine Learning Improve Tropical Cyclone Forecasts?

  • 16 June 2021
Can Machine Learning Improve Tropical Cyclone Forecasts?

Scientists have been searching for decades for breakthroughs in tropical cyclone intensity modeling to provide more accurate and timely tropical cyclone warnings. Now recent research, funded in part by CPO’s Climate Observations and Monitoring (COM) program, showcases how machine learning methods can help reduce errors in tropical cyclone intensity forecasts to prevent damages and loss of life. The study’s machine learning-based predictive model outperformed some of the most skillful operational models.

Published in the journal Weather and Forecasting, the study, led by PNNL and CIRA, developed a deep learning-based intensity prediction model that provides 24-hour and 6-hour intensity forecasts for the North Atlantic. Specifically, it forecasts the change in maximum wind speed for tropical cyclones in the Atlantic Basin. Deep learning, or deep neural networks, is a type of machine learning suited for working with complex, nonlinear relationships involving many predictors—such as the type of relationships exemplified by the processes that constitute a tropical cyclone. 

The model was developed and tested with the predictors from SHIPS (Statistical Hurricane Intensity Prediction Scheme), a well-known model from NOAA’s National Hurricane Center. The authors found that the study’s model outperformed some of the most skillful operational models in producing 24-hour intensity forecasts and predicting rapid intensification events by 5-22%. The 6-hr “light-weight” model also showed promising results.

This study adds to a growing body of evidence that machine learning-based approaches have the potential to improve operational tropical cyclone intensity forecasts.

Read the study »




1315 East-West Highway Suite 100
Silver Spring, MD 20910


Americans’ health, security and economic wellbeing are tied to climate and weather. Every day, we see communities grappling with environmental challenges due to unusual or extreme events related to climate and weather.