Preprints
https://doi.org/10.5194/egusphere-2023-1385
https://doi.org/10.5194/egusphere-2023-1385
18 Jul 2023
 | 18 Jul 2023

MAESSTRO: Masked Autoencoders for Sea Surface Temperature Reconstruction under Occlusion

Edwin Goh, Alice R. Yepremyan, Jinbo Wang, and Brian Wilson

Abstract. This study investigates the use of masked autoencoders (MAE) to address the challenge of filling gaps in high-resolution (1 km) sea surface temperature (SST) fields caused by cloud cover, which often results in gaps in the SST data and/or blurry imagery in blended SST products. Our study demonstrates that MAE, a deep learning model, can efficiently learn the anisotropic nature of small-scale ocean fronts from numerical simulations and reconstruct the artificially masked SST images. The MAE model is trained and evaluated on synthetic SST fields and tested on real satellite SST data from VIIRS sensor on Suomi-NPP satellite. It is demonstrated that the MAE model trained on numerical simulations can provide a computationally-efficient alternative for filling gaps in satellite SST. MAE can reconstruct randomly occluded images with a root mean squared error (RMSE) of under 0.2 °C for masking ratios of up to 80 %. It has exceptional efficiency, requiring three orders of magnitude (a factor of 5000) less time. The ability to reconstruct high-resolution SST fields under cloud cover has important implications for understanding and predicting global and regional climates, and detecting small-scale SST fronts that play a crucial role in the exchange of heat, carbon, and nutrients between the ocean surface and deeper layers. Our findings highlight the potential of deep learning models such as MAE to improve the accuracy and resolution of SST data at kilometer scales. It presents a promising avenue for future research in the field of small-scale ocean remote sensing analyses.

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.
Edwin Goh, Alice R. Yepremyan, Jinbo Wang, and Brian Wilson

Status: closed (peer review stopped)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on egusphere-2023-1385', Anonymous Referee #1, 18 Oct 2023
    • AC1: 'Reply on RC1', Jinbo Wang, 17 Jan 2024
  • RC2: 'Comment on egusphere-2023-1385', Anonymous Referee #2, 11 Dec 2023

Status: closed (peer review stopped)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on egusphere-2023-1385', Anonymous Referee #1, 18 Oct 2023
    • AC1: 'Reply on RC1', Jinbo Wang, 17 Jan 2024
  • RC2: 'Comment on egusphere-2023-1385', Anonymous Referee #2, 11 Dec 2023
Edwin Goh, Alice R. Yepremyan, Jinbo Wang, and Brian Wilson
Edwin Goh, Alice R. Yepremyan, Jinbo Wang, and Brian Wilson

Viewed

Total article views: 814 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
490 285 39 814 29 22
  • HTML: 490
  • PDF: 285
  • XML: 39
  • Total: 814
  • BibTeX: 29
  • EndNote: 22
Views and downloads (calculated since 18 Jul 2023)
Cumulative views and downloads (calculated since 18 Jul 2023)

Viewed (geographical distribution)

Total article views: 792 (including HTML, PDF, and XML) Thereof 792 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 22 May 2024
Download
Short summary
Our research used an AI model to fill in missing parts of sea temperature maps caused by cloud cover. We found MAE can recreate missing SST with less than 0.2 °C error, even when 80 % is missing, and does this 5000 times faster than conventional methods. This can enhance our ability in monitoring global small-scale ocean fronts that affect heat, carbon, and nutrient exchange in the ocean. The method's accuracy and efficiency are promising for future ocean research.