AI x Climate Risk: From Forecast to Decision

Wildfires arrive outside their season. Storms hit places that never planned for them. As climate risks change, better forecasts, satellite monitoring, and risk models are increasingly shaping what gets insured, what gets built, where capital goes, and when people are warned. But making a better forecast is only part of the story. What happens after the forecast? Getting from prediction to action takes a long chain of work, and each link sits with a different group. A researcher building a model, a regulator deciding what counts as evidence, an insurer pricing exposure, and an engineer supplying the compute are all working on the same problem from different ends, and rarely in the same room. This session brings that chain together: how climate data gets collected, how physical knowledge can improve models, how their outputs can be interpreted and explained, how they translate into policy and insurance, and what it takes to support these systems at scale. Through short research presentations, a moderated panel, and audience discussion, we'll unpack these questions together. This is the second session in a series by Lorong AI and Climate Change AI (CCAI) exploring how AI is shaping sustainability across different domains. More About the Sharings Dr Gianmarco Mengaldo will share on “From Physics to AI and Back.” How can existing scientific knowledge improve AI systems, and how can AI in turn help us extract new knowledge? He'll share CondensNet, a physics-enhanced deep learning approach for representing cloud physics in climate models, alongside work on explainable AI for climate applications. Together, the two examples show how AI and physical knowledge can inform each other, with implications for how we build, interpret, and improve climate models. (Technical Level: 100) Sang-Ho Yun will share on "Leveraging Radar Observations Before Reaching for AI in Climate Risk Monitoring". Robust physics and direct observations should be leveraged whenever they're available. Int