Let There Be Lightning

Lightning kills roughly 2,500 people in India every year, often more than floods, heat and cold combined. An actionable alert system that saves lives needs excellent modelling. A modern weather model represents the atmosphere with about a billion numbers, estimated from a few tens of millions of scattered, indirect observations arguably the most successful inverse problem in science. We trace how this intellectual and social infrastructure was built: from Richardson's 1922 hand calculation, to neural networks trained on the reconstructed past that now match the physical models. Lightning is among the hardest problems in the field: we will go through the physics, the baselines, and the open topics. About the speaker: Dr Prasanna Bhogle is a lapsed theoretical physicist with ten years of AI/ML in German industry, and the author of a recent book on causal inference (An Unfriendly Introduction to Causal Inference). He is now building Romulan AI, which adds to India's weather-forecast infrastructure with hyperlocal, decision-grade systems. To attend online: Gmeet link: meet.google.com/cwe-jvox-dys Looking forward to seeing you!