Preprints
https://doi.org/10.5194/egusphere-2024-1293
https://doi.org/10.5194/egusphere-2024-1293
08 May 2024
 | 08 May 2024

Convolutional Neural Networks for Sea Surface Data Assimilation in Operational Ocean Models: Test Case in the Gulf of Mexico

Olmo Zavala-Romero, Alexandra Bozec, Eric P. Chassignet, and Jose R. Miranda

Abstract. Deep learning models have demonstrated remarkable success in fields such as language processing and computer vision, routinely employed for tasks like language translation, image classification, and anomaly detection. Recent advancements in ocean sciences, particularly in data assimilation (DA), suggest that machine learning can emulate dynamical models, replace traditional DA steps to expedite processes, or serve as hybrid surrogate models to enhance forecasts. However, these studies often rely on ocean models of intermediate complexity, which involve significant simplifications that present challenges when transitioning to full-scale operational ocean models. This work explores the application of Convolutional Neural Networks (CNNs) in assimilating sea surface height and sea surface temperature data using the Hybrid Coordinate Ocean Model (HYCOM) in the Gulf of Mexico. The CNNs are trained to correct model errors from a two-year, high-resolution (1/25°) HYCOM dataset, assimilated using the Tendral Statistical Interpolation System (TSIS). We assess the performance of the CNNs across five controlled experiments, designed to provide insights into their application in environments governed by full primitive equations, real observations, and complex topographies. The experiments focus on evaluating: 1) the architecture and complexity of the CNNs, 2) the type and quantity of observations, 3) the type and number of assimilated fields, 4) the impact of training window size, and 5) the influence of coastal boundaries. Our findings reveal significant correlations between the chosen training window size—a factor not commonly examined—and the CNNs' ability to assimilate observations effectively. We also establish a clear link between the CNNs' architecture and complexity and their overall performance.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.
Olmo Zavala-Romero, Alexandra Bozec, Eric P. Chassignet, and Jose R. Miranda

Status: final response (author comments only)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on egusphere-2024-1293', Anonymous Referee #1, 24 Jul 2024
    • AC1: 'Reply on RC1', Olmo Zavala Romero, 08 Oct 2024
  • RC2: 'Comment on egusphere-2024-1293', Michael Gray, 27 Jul 2024
    • AC2: 'Reply on RC2', Olmo Zavala Romero, 08 Oct 2024
Olmo Zavala-Romero, Alexandra Bozec, Eric P. Chassignet, and Jose R. Miranda
Olmo Zavala-Romero, Alexandra Bozec, Eric P. Chassignet, and Jose R. Miranda

Viewed

Total article views: 452 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
261 145 46 452 30 23
  • HTML: 261
  • PDF: 145
  • XML: 46
  • Total: 452
  • BibTeX: 30
  • EndNote: 23
Views and downloads (calculated since 08 May 2024)
Cumulative views and downloads (calculated since 08 May 2024)

Viewed (geographical distribution)

Total article views: 459 (including HTML, PDF, and XML) Thereof 459 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 20 Nov 2024
Download
Short summary
Deep learning is enhancing ocean science by improving data processing and forecasts. This study uses Convolutional Neural Networks (CNNs) to assimilate sea surface data in the Gulf of Mexico. Researchers conducted five experiments to evaluate the CNNs' performance across different designs and data types, revealing how training data volume and CNN design affect their effectiveness in operational ocean modeling.