Time: 16:00-17:30pm, Thursday, April 16, 2026

Venue: E2-313, Yungu Campus

Host: Dr. XinYuan, AssociateProfessor, Westlake University

Language: Chinese

Speaker:

Zuchuan Li

Assistant Professor

Duke Kunshan University

Biography:

Zuchuan Li's research focuses on machine learning, Bayesian inference, and their applications to environmental data science. His current research interests include scientific machine learning, Bayesian inference, and their applications to constrain the mechanisms of global carbon cycle and primary productivity. His research has been published in leading academic journals, including Nature, Nature Climate Change, Nature Communications, and NeurIPS. He is a recipient of the NASA Earth and Space Science Fellowship and the WHOI Postdoctoral Scholarship. He holds a Ph.D. from Duke University.

Abstract:

In this talk, I will present our recent work on learning representations of biogeochemical dynamics in the world’s ocean. We model the spatiotemporal dynamics of the surface ocean by pretraining masked autoencoders on large-scale satellite remote sensing data, including ocean color, wind speed vectors, and sea surface temperature. We evaluate our framework against strong baselines, including tree-based models and prior-data fitted networks across seven downstream tasks in ocean biogeochemistry. Our preliminary results demonstrate that the model captures key oceanic biogeochemical dynamics by implicitly learning tracer conservation equations, as a result outperforming the strong baselines.

Contact:

Linghui Xiang

xianglinghui@westlake.edu.cn