Research

I use machine learning and explainable AI to predict and understand change in the polar oceans and cryosphere, and to help design the observing systems those predictions depend on.

AI-Based Observing System Design

Where should we put instruments in the polar oceans, and what should they measure?

Polar observations are scarce and expensive, so every instrument placement is a bet. Explainable AI can show which measurements a skillful prediction actually leans on, which means we can let the model tell us where the information is instead of guessing. Wind at the existing Arctic observing stations, it turns out, carries information about sea-ice velocity across a wide surrounding area. I am interested in investigating a range of other machine learning methods for designing observing systems.

ML and XAI for Predicting and Understanding Sea Ice

How well can we predict sea ice, and what do the models tell us about why it changes?

Sea ice is among the fastest changing parts of the climate system and is challenging to forecast. Machine learning can provide us skillful predictions, and explainable AI tells us what the model learned along the way, which can potentially teach us something about the physics. Together they turn a forecast into an explanation: why this September, why this slowdown.