Projected changes in atmospheric blocking and associated extreme weather events are marked by significant uncertainties due to climate model biases, the complex nature of the phenomenon, and its strong natural variability. Paleoclimate proxy records, suggesting shifts in blocking activity, extend observational records and may help reduce these uncertainties. However, extracting synoptic-scale signals from low-temporal-resolution paleoclimate records is challenging. In this talk I will present a novel approach to solve this problem, using deep learning. I will show that the deep learning models are constrained implicitly by the paleoclimate proxy records, even without directly incorporating them or having any knowledge of their specific locations. The DL blocking reconstructions reveal new insights into tropical-extratropical climate interactions and provide a new pathway for improved understanding of the response of extreme weather to past climate changes.

Christina Karamperidou is a Professor of Atmospheric Sciences at the University of Hawaii at Mānoa. She holds a PhD from Columbia University, an M.S. in Environmental Protection & Sustainable Development and a Diploma (BS+MS) in Civil & Environmental Engineering from the Aristotle University of Thessaloniki, Greece. Christina’s research focuses on El Niño-Southern Oscillation dynamics and predictability, extreme events in response to large-scale climate variability, paleoclimate modeling, and machine learning applications in environmental sciences.