Transforming Ocean Analysis: Learning 4D ocean field from in-situ observations via uncertainty-aware implicit representations

ICLR 2025 Conference Submission1246 Authors

Keywords: implicit neural representation, meta learning, climate, ocean gridded dataset

TL;DR: We propose a novel method to reconstruct ocean field from raw observations at a higher quality with uncertainty-aware neural network.

Abstract: A complete and accurate representation of Earth's time-evolving ocean field is crucial for understanding global warming as well as climate dynamics. However, the sparsity of current in-situ ocean measurements presents a significant challenge in estimating values in largely unobserved regions. Traditional methods, such as objective interpolation (OI), struggle with accuracy due to their reliance on discrete grids and fixed spatial correlation structures. In this paper, we propose a novel approach to reconstruct 4D ocean fields only from raw observations using implicit neural representations (INRs). Our method improves field representations by leveraging neural networks to capture continuous, complex, and nonlinear patterns inherent in ocean data. To address uncertainties in ocean measurements and the limited availability of daily observations, we incorporate uncertainty estimates and a meta-learning strategy into existing INRs. These innovations enable our approach to provide daily, resolution-free ocean temperature reconstructions, a significant improvement over monthly averaged discrete fields. Experiments demonstrate the accuracy and adaptability of our method compared with approaches, establishing our method as a transformative solution for future ocean analysis and climate monitoring.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 1246