Bridging AI, remote sensing, and climate modeling to understand polar climate variability and change

Abstract: Sea ice is crucial to Earth’s climate, Arctic communities, and ecosystems, yet climate change is driving significant losses, threatening polar stability. Quantifying the long-term impacts of a declining sea ice cover requires tools which improve climate-timescale prediction and bring new understanding of climate interactions. In this talk, I discuss how meeting this challenge requires a multi-disciplinary approach. Climate models, while essential, suffer from systematic biases due to missing or inaccurate physics, leading to uncertainty in future projections. I show how data assimilation (DA) offers a statistical framework for integrating satellite observations with climate models to quantify systematic sea ice model errors. Using convolutional neural networks (CNNs), we can learn these errors based on the model’s atmospheric, oceanic, and sea ice conditions—what I term a state-dependent representation of the error. This approach enables real-time corrections to subsequent model simulations, which systematically reduces global sea ice biases. I highlight key successes and challenges in developing this hybrid ML+climate modeling framework, including transfer learning to enhance online generalization of ML models, and new methods for integrating Python-based ML frameworks with Fortran climate model code. Finally, I introduce GPSat, a scalable Gaussian process-based tool for reconstructing complete sea ice fields from sparse satellite altimetry data. Together, the DA+ML framework and GPSat offer future opportunities for improving targeted model physics errors for more robust climate simulation.

Bio:  My research focuses on the development of machine learning (ML) methods to improve the way we measure, understand, and predict the polar climate system. As a postdoctoral researcher based at Princeton University and the NOAA Geophysical Fluid Dynamics Laboratory (GFDL), I have developed a sea ice data assimilation program in Python which is now being used in the GFDL coupled climate model for integrating satellite observations. I am also leading the development of a hybrid version of the coupled climate model SPEAR, which embeds neural networks into the fortran-based sea ice component to improve the fidelity of numerical sea ice forecasts. Another focus of my work leverages Gaussian process models. During my PhD at University College London, I published one of the first studies to apply ML algorithms to Arctic sea ice forecasts, using a combination of cluster analysis and Gaussian process models. I also led the first study to derive complete maps of daily sea ice thickness fields from satellite altimetry. This remote-sensing application has formed the basis of ongoing work into scalable Gaussian process models for interpolating high-resolution sea ice altimetry data. 

*This seminar will be held in hybrid format, in the IACS Seminar Room or on Zoom*

Join Zoom Meeting

https://stonybrook.zoom.us/j/95916050624?pwd=jJYAO2BZJXTAoNUlblYyM8EuXSFmvA.1

Meeting ID: 959 1605 0624

Passcode: 622950