Articles | Volume 20, issue 5
https://doi.org/10.5194/os-20-1309-2024
https://doi.org/10.5194/os-20-1309-2024
Research article
 | 
24 Oct 2024
Research article |  | 24 Oct 2024

MAESSTRO: Masked Autoencoders for Sea Surface Temperature Reconstruction under Occlusion

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

Related authors

Multi-satellite U-Net for high-resolution sea surface temperature reconstruction
Ellin Zhao, Edwin Goh, Alice Yepremyan, Jinbo Wang, and Brian Wilson
EGUsphere, https://doi.org/10.5194/egusphere-2025-4847,https://doi.org/10.5194/egusphere-2025-4847, 2025
This preprint is open for discussion and under review for Ocean Science (OS).
Short summary

Cited articles

Agabin, A., Prochaska, J. X., Cornillon, P. C., and Buckingham, C. E.: Mitigating masked pixels in a climate-critical ocean dataset, Remote Sens., 16, 2439, https://doi.org/10.3390/rs16132439, 2024. a
Alvera-Azcárate, A., Barth, A., Sirjacobs, D., Lenartz, F., and Beckers, J.-M.: Data Interpolating Empirical Orthogonal Functions (DINEOF): a tool for geophysical data analyses, Mediterr. Mar. Sci., 12, 5–11, https://doi.org/10.12681/mms.64, 2011. a
Barth, A., Alvera-Azcárate, A., Licer, M., and Beckers, J.-M.: DINCAE 1.0: a convolutional neural network with error estimates to reconstruct sea surface temperature satellite observations, Geosci. Model Dev., 13, 1609–1622, https://doi.org/10.5194/gmd-13-1609-2020, 2020. a, b
Barth, A., Alvera-Azcárate, A., Troupin, C., and Beckers, J.-M.: DINCAE 2.0: multivariate convolutional neural network with error estimates to reconstruct sea surface temperature satellite and altimetry observations, Geosci. Model Dev., 15, 2183–2196, https://doi.org/10.5194/gmd-15-2183-2022, 2022. a
Belkin, I. M.: Remote Sensing of Ocean Fronts in Marine Ecology and Fisheries, Remote Sens., 13, 883, https://doi.org/10.3390/rs13050883, 2021. a
Download
Short summary
An AI model was used to fill in missing parts of sea temperature (SST) maps caused by cloud cover. We found masked autoencoders can recreate missing SSTs with less than 0.2 °C error, even when 80 % are missing. This is 5000 times faster than conventional methods tested on a single central processing unit. This can enhance our ability in monitoring global small-scale ocean fronts that affect heat, carbon, and nutrient exchange in the ocean. The method is promising for future research.
Share