AI for tropical meteorology: Challenges and opportunities

Go back to programme

AI for tropical meteorology: Challenges and opportunities

  • Watch

    * Register (or log in) to the AI4G Neural Network to add this session to your agenda or watch the replay

    • Watch

      Register

      * Register (or log in) to the AI4G Neural Network to add this session to your agenda or watch the replay

    The recent landfalls of Hurricane Ian, which led to severe destruction in coastal Southwest Florida and North Carolina; and of Tropical Storm Megi, which caused over 200 deaths due to mudslides and 90 billion USD in damages in the Philippines, emphasized the need for accurate track, intensity, and precipitation forecasts to provide enough lead time for planning evacuation and mitigation strategies. This need will only grow as more than half of Earth’s population is projected to live in the tropics by 2050 and more than a billion people worldwide could be living in low-elevation coastal zones by 2060.  

    To address this need and complement numerical weather forecasts, data-driven models are operationally used at all forecast lead times, from purely statistical forecasts to hybrid approaches that post-process numerical weather predictions. Recent progress in machine learning offers opportunities to uncover unexploited predictors, better leverage spatiotemporal patterns from data, and improve forecasts of tropical storm wind speeds and precipitation. Despite recent advances, many challenges remain for tropical cyclone forecasting, including the rarity of extreme tropical weather events; the physical complexity of the genesis and intensification of tropical storms; and the potential for changes in tropical cyclone behavior due to changes in their environment. 

    In this webinar, we will present recent research to tackle these challenges, from probabilistic predictions of cyclogenesis and tropical cyclone track and intensity, to transparent machine learning shedding new light upon the underlying physical processes. We will show how progress in atmospheric thermodynamics can help us create predictive algorithms that are robust to environmental changes, including long-term climate change, while progress in remote sensing can help us overcome data quality and quantity challenges. Finally, we will reflect on bias in AI for tropical meteorology, and how it can be alleviated via uncertainty quantification and causal inference to advance towards robust and ethical AI. 

    This live event includes a 30-minute networking event hosted on the AI for Good Neural Network. This is your opportunity to ask questions, interact with the panelists and participants and build connections with the AI for Good community.

    Share this session
    In partnership with