Speaker: Jennifer Sleeman (Johns Hopkins Univ.)
Topic: AI Climate Tipping Point Discovery

A number of elements in our climate system are quickly approaching tipping points, described in the Intergovernmental Panel for Climate Change AR6 report in 2021. Current high performance computing environments are not sufficient to comprehensively map out the high-dimensional spaces inherent in global climate models. AI offers the promise of reducing these high dimensional spaces to areas which are likely to exhibit tipping point behavior, enabling focused model studies in these reduced-dimensional spaces. In this work we describe a novel approach for AI climate tipping point discovery using three core AI innovations: a generative adversarial network for tipping point discovery (TIP-GAN), an attention-based causal model to understand the drivers of tipping points (TIP-CAUSE) and an LLM-based climate assistant for scientific question answering (TIP-ASSIST). We share results applying this AI machinery to two tipping points: the Atlantic Meridional Overturning Circulation (AMOC) collapse and coral reef die-off, and describe how this methodology could be extended to other climate tipping points.