As told to Beth Jensen
For a long time, scientists thought ice sheets like those found in Greenland and Antarctica were just white mountains that reflected sunlight and didn’t move much. We now know that this ice melts, loses mass, and becomes fresh water, which has a big impact on ocean circulation and contributes to rising sea levels.
But many of these processes remain poorly understood. I integrate physics with machine learning to help figure out how these processes affect climate change, the polar regions or cryosphere, and sea-level rise.
My interest in physics began early. As a child, I remember asking my mother why the sky was blue and feeling amazed that one could understand the workings of the natural environment. In middle school, I discovered a love for physics and its capacity to explain nature precisely and quantitatively.
As time went on and my interest in physics grew, many people, including some teachers, told me I wouldn’t be able to handle this field – that physics was too difficult for women. Because I lived in Taiwan, where good students are often urged to study medicine, I was encouraged to take that path. Hearing these comments felt discouraging at times, but the next day, I’d always wake up still wanting to focus on physics.
I became interested in combining this passion with climate science during the final year of my PhD, while I was looking to broaden the scope of my work and make an impact. I loved that I could use mathematical equations to understand how our climate system works, which can ultimately help us better predict impacts from climate change. Our ability to observe the cryosphere has only grown, which has resulted in a lot of new data that is still waiting to be understood.
One of my recent projects involved the creation of a physics-based learning model – trained on both observational data and fundamental physical laws – to describe the regions of Antarctica that will become unstable as the climate warms. To do that, I needed to know where fractures exist on the ice shelves, which are attached to land but extend into the sea. By training a machine learning algorithm, I created the first map of those fractures across the entire Antarctic ice shelves, which was exciting and wouldn’t have been possible without the use of artificial intelligence. Now that I’m at Stanford, my students here are working to do the same for the Greenland ice sheet.
We’re also using physics-informed models to determine the viscosity of ice sheets, which is impossible to measure at the ice-sheet scale, but critical to understanding how they flow. We’re now able to predict that viscosity, and our results appear more interpretable than traditional models. If that’s the case, this new understanding could easily be incorporated into existing models to gain a more accurate understanding of how climate change will affect the movement and fracture of these sheets in the future. This information is important to help us understand the potential for sea-level rise and the effect it will have on coastal cities.
I’ve only been at the Stanford Doerr School of Sustainability for six months, but I’m already talking to colleagues in many different departments whose work relates to my research interests. I’m very excited about the intersection of data-driven and physics-based models, and the potential it has to make impacts in many fields. There’s so much more to be discovered. What we’re doing now is only the tip of the iceberg.
Explore More
-
Conventional sunscreen ingredients can damage coral reefs and human health. An immunologist and a marine ecologist teamed up to develop a better approach.
-
With support from a Stanford Doerr School of Sustainability Accelerator seed grant, an interdisciplinary team has developed a groundbreaking optical sensor that measures DNA and other key molecules in seawater using light, potentially revolutionizing the study of biodiversity in the enigmatic depths below the ocean’s surface.
