Work

Surfzone Water Depth Estimation From Surface Flows Using a Data‐Driven and Physics‐Informed Deep Conditional Generative Model

Year: 2026

Type: article

Abstract: A physics‐informed deep conditional generative model driven with remotely sensed surface currents is shown to estimate surfzone water depths (bathymetry). The model encodes measured flow data as latent Gaussian parameters and decodes these distributions to estimate water depths over the domain, progressively refining its predictions via a loss‐minimization strategy. The model performance is evaluated on in‐distribution and out‐of‐distribution data sets collected in Duck, North Carolina, demonstrating promising site‐specific results, especially given the limited training data set used here (6 bathymetries with a total of 8 flow realizations). However, broader applicability requires transfer learning across a wider range of bathymetric observations.

Source: Journal of Geophysical Research Machine Learning and Computation

DOI: 10.1029/2025jh000906

Authors: Reza Salatin, Steve Elgar, Britt Raubenheimer

Institutions: Northeastern University, Woods Hole Oceanographic Institution

Language: English

Topic: Model Reduction and Neural Networks

Subfield: Statistical and Nonlinear Physics

Field: Physics and Astronomy

Domain: Physical Sciences

SDGs: Life below water, Clean water and sanitation

Open Access status: gold

Estimated APC paid: $2,500