Neural Networks Predicting Submesoscale Tracer Dispersion

• Published in: Journal of geophysical research. Machine learning and computation Vol. 2; no. 3
• Main Authors: Bijay, Mayank Kumar, Thomas, Jim
• Format: Journal Article
• Language: English
• Published: 01.09.2025
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2025JH000655

In this paper we examine the possibility of directly predicting passive tracer dynamics from flow fields. We use a two‐vertical‐mode model for generating flows ranging from low to high Rossby numbers and advect the passive tracer with such flows. While typically tracer dynamics are obtained by time integrating a tracer advection equation, here the emphasis is on predicting the tracer dynamics from the flow field. We experiment with popular architectures such as Autoencoder, UNet, and GAN, and also develop a novel model, that we call LoConv, to make predictions. The LoConv model uses custom Local Convolution layers that allows convolution with spatially varying weights and this model outperforms usually used architectures such as Autoencoder, UNet, and GAN based on various metrics. Autoencoders with very few trainable parameters were unsuccessful in making good predictions even at large‐scales. GAN and UNet predictions were biased towards large‐scale features, unfavorable for capturing small‐scale tracer dispersion, especially at high Rossby numbers. Overall, the LoConv model with some physics‐informed training produced the best fine‐scale tracer predictions, along with tracer flux and related derived quantities. More broadly, the results of this study point towards successful direct ways of predicting tracer fields from the flow, overcoming the computational cost of long numerical integration of tracer advection equations. Oceanic tracers such as dissolved oxygen and carbon are important from a climate and ecological perspective. Typically these tracers are obtained computationally by integrating differential equations along with the flow in ocean models, which is computationally expensive and time‐consuming. In this study we explore the possibility of using deep learning tools to make predictions about tracers. We trained neural network models to predict tracers from flow generated in different regimes, similar to oceanic flows whose scales vary from hundreds of kilometers to a kilometer. We experiment with some popular architectures and also develop some new neural net algorithms to make these predictions, with the aim of accurately predicting fine‐scale tracer features. This makes analysis of tracer fields fast and highly efficient, by instantaneously obtaining tracer predictions instead of waiting for long computations. With these experiments, we come to a better understanding of the strengths and weaknesses of the various neural network architectures. Our developed model is particularly seen to make excellent fine‐scale tracer predictions, outperforming popular architectures in various metrics. We use neural network models to predict passive tracer fields directly from flow fields across a broad range of Rossby numbers Popular convolution based neural network architectures are seen to miss small‐scale features in the predictions A new neural network architecture is developed and it successfully predicts submesoscale tracer dispersion features very well