AOSC Seminar by Jennifer Sleeman, 04/16/2026
AOSC Seminar
Jennifer Sleeman
Johns Hopkins University APL
Title
Learning the State Space of Earth System Tipping Points with AI
Abstract
Predicting abrupt and potentially irreversible transitions in Earth system components remains a major challenge because the regions of state space that lead to tipping are poorly characterized. Traditional numerical models simulate forward trajectories but provide limited insight into which initial conditions lead to critical transitions or how basin boundaries are structured. As a result, identifying proximity to tipping thresholds and understanding the mechanisms that drive them remains difficult. In this study, we describe an AI-assisted methodology for Earth system tipping point discovery and early warning forecasting that reframes this problem as one of learning and exploring the structure of the system’s state space. Included in this methodology are three core AI contributions. The first is a deep conditional generative adversarial network that learns the distribution of initial conditions associated with distinct system outcomes. This enables the identification of regions in state space that result in tipping outcomes. The second is a transformer-based surrogate model that emulates system dynamics and forecasts the timing of critical transitions. The third is a multi-agent large language model system that supports structured exploration of the hypothesis space through coordinated model execution. We apply this framework to three interacting tipping elements including the loss of multi-year Arctic sea ice, the collapse of the Atlantic Meridional Overturning Circulation (AMOC), and critical die-off in coral reef systems, each represented using a reduced-order model. We present results demonstrating that our system identified previously uncharacterized transition pathways that could suggest Atlantic–Pacific interactions leading to a weakening of the Atlantic circulation. In addition, we present an approach for exploring intervention strategies using a Sample-ThenOptimize Batch Neural Thompson Sampling method to identify candidate interventions that reduce tipping likelihood. Finally, we discuss how these components can be integrated into an Earth system digital twin, enabling interactive exploration of tipping dynamics and intervention scenarios. This work demonstrates how AI can be used to probe the structure of the Earth system state space by uncovering previously uncharacterized subspaces associated with tipping behavior and identifying early warning signals, as well as exploring the intervention space for more informed mitigation strategies.
Bio
Dr. Jennifer Sleeman is a Senior AI Research Scientist at the Johns Hopkins Applied Physics Laboratory (APL) and an Adjunct Associate Professor at the University of Maryland, Baltimore County (UMBC). She has over 18 years of experience in artificial intelligence research and earned her Ph.D. in Computer Science from UMBC. Since 2018, her work has focused on applying AI to complex Earth system challenges, including atmospheric composition, air quality forecasting, wildfire emissions, and extreme events. More recently, she has investigated how deep learning can help identify potential climate tipping points and support early warning through AI-enabled Earth system digital twins. She is the principal investigator on a NASAfunded project in this area. Her research aims to bridge advanced machine learning with physically grounded Earth system science to better understand and anticipate high impact Earth system changes.
Contact
If you are not subscribed to seminar announcements and need a link to this online seminar please contact aosc-helper. You can also subscribe to weekly links and announcements below.
AOSC Seminar
Pre-seminar refreshment: N/A
Seminar: 3:30-4:30pm, Room: ATL 2400(only when in-person)
Meet-the-Speaker: 4:30-5:00pm, Room: ATL 3400(only when in-person) [For AOSC Students only]
[ Subscribe to receive seminar announcements | Email aosc-deptseminar@umd.edu to give a seminar ]