Atmosphere Ocean Science Colloquium

---

Exploring Predictability from Stratosphere-Troposphere Coupling Using AI Models

Speaker: Zheng Wu, Albany

Location: Warren Weaver Hall 1302

Date: Wednesday, October 8, 2025, 3:30 p.m.

Synopsis:

Stratosphere-troposphere coupling has been identified as an important source of subseasonal to seasonal (S2S) predictability for both the stratosphere and the surface. However, current forecasting systems are limited in predicting extreme stratospheric events beyond 1-2 weeks, which can hinder the predictive skill of surface weather. The rise and rapid development of Artificial Intelligence (AI) techniques offer a new opportunity to investigate the role of this coupling in forecasting extreme stratospheric and surface events. In this talk, I will present the application of several AI models, ranging from Random Forests to large-scale AI weather emulators. I will show the performance and biases in various AI models in predicting stratospheric extremes and the associated dynamical processes and surface impacts. I will also discuss AI model explanation techniques to interpret these predictions and identify forecasts of opportunity from the stratosphere-troposphere coupling. The talk will discuss the potential of using an explainable AI framework as a new tool to advance our understanding of the underlying dynamics.