The Fundamental Patterns of Sea Surface Temperature

For over 40 years, remote sensing observations of the Earth’s oceans have yielded global measurements sea surface temperature (SST).With a resolution approximately 1 km, these data trace physical processes like western boundary currents, cool upwelling at eastern and formation mesoscale sub-mesoscale eddies. To discover fundamental patterns SST on scales smaller than 10 we developed an unsupervised, deep contrastive learning model named NENYA. We trained NENYA subset 8 million cloud-free cutout images (~ 80 × 80km2) from MODerate-resolution Imaging Spectroradiometer (MODIS) sensor, with image augmentations to impose invariance rotation, reflection, translation. The 256-dimension latent space defines vocabulary describe complexity associates features. used dimensionality reduction algorithm explore cutouts interval ΔT = 0.5-1 K, identifying diverse set variance wide range scales. then demonstrated that large-scale features arise preferentially in Pacific Atlantic Equatorial Cold Tongues exhibit strong seasonal variation, while predominantly structure manifest select regions upwelling, along Antarctic Circumpolar Current. provide web-based user interface facilitate geographical temporal exploration full MODIS dataset. Future efforts will link specific dynamics (e.g., frontogenesis) examine their distribution time globe.