Deep Learning for Sea Temperature Eddy signature Classification

Mesoscale eddies are oceanic vortices with radii of tens of kilometers, which live on for several months or even years. They carry large amounts of heat, salt, nutrients, and pollutants from their regions of formation to remote areas, making it important to detect and track them. Using satellite altimetric maps, mesoscale eddies have been detected via remote sensing with advancing performance over the last years [1]. However, the spatio-temporal interpolation between satellite track measurements, needed to produce these maps, induces a limit to the spatial resolution (1/12&#176; in the Med Sea) and large amounts of uncertainty in non-measured areas.Nevertheless, mesoscale oceanic eddies also have a visible signature on other satellite imagery such as Sea Surface Temperature (SST), portraying diverse patterns of coherent vortices, temperature gradients, and swirling filaments. Learning the regularities of such signatures defines a challenging pattern recognition task, due to their complex structure but also to the cloud coverage which can corrupt a large fraction of the image.We introduce a novel Deep Learning approach to classify sea temperature eddy signatures [2]. We create a large dataset of SST patches from satellite imagery in the Mediterranean Sea, containing Anticyclonic, Cyclonic, or No Eddy signatures, based on altimetric eddy detections of the DYNED-Atlas [3]. Our trained Convolutional Neural Network (CNN) can differentiate between these signatures with an accuracy of more than 90%, robust to a high level of cloud coverage.We furtherly evaluate the efficiency of our classifier on SST patches extracted from oceanographic numerical model outputs in the Mediterranean Sea. Our promising results suggest that the CNN could complement the detection, tracking, and prediction of the path of mesoscale oceanic eddies.[1] Chelton, D. B., Schlax, M. G. and Samelson, R. M. (2011). Global observations of nonlinear mesoscale eddies. Progress in oceanography, 91(2),167-216.[2] E. Moschos, A. Stegner, O. Schwander and P. Gallinari, "Classification of Eddy Sea Surface Temperature Signatures Under Cloud Coverage," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 13, pp. 3437-3447, 2020, doi: 10.1109/JSTARS.2020.3001830.[3] https://www.lmd.polytechnique.fr/dyned/

Download Full-text

Deep Learning of Sea Surface Temperature Patterns to Identify Ocean Extremes

Remote Sensing ◽

10.3390/rs13040744 ◽

2021 ◽

Vol 13 (4) ◽

pp. 744

Author(s):

J. Xavier Prochaska ◽

Peter C. Cornillon ◽

David M. Reiman

Keyword(s):

Deep Learning ◽

Surface Temperature ◽

Sea Surface Temperature ◽

Sea Surface ◽

Space Distribution ◽

Western Boundary ◽

Learning Modules ◽

Ocean Science ◽

The Many ◽

Future Work

We performed an out-of-distribution (OOD) analysis of ∼12,000,000 semi-independent 128 × 128 pixel2 sea surface temperature (SST) regions, which we define as cutouts, from all nighttime granules in the MODIS R2019 Level-2 public dataset to discover the most complex or extreme phenomena at the ocean’s surface. Our algorithm (ULMO) is a probabilistic autoencoder (PAE), which combines two deep learning modules: (1) an autoencoder, trained on ∼150,000 random cutouts from 2010, to represent any input cutout with a 512-dimensional latent vector akin to a (non-linear) Empirical Orthogonal Function (EOF) analysis; and (2) a normalizing flow, which maps the autoencoder’s latent space distribution onto an isotropic Gaussian manifold. From the latter, we calculated a log-likelihood (LL) value for each cutout and defined outlier cutouts to be those in the lowest 0.1% of the distribution. These exhibit large gradients and patterns characteristic of a highly dynamic ocean surface, and many are located within larger complexes whose unique dynamics warrant future analysis. Without guidance, ULMO consistently locates the outliers where the major western boundary currents separate from the continental margin. Prompted by these results, we began the process of exploring the fundamental patterns learned by ULMO thereby identifying several compelling examples. Future work may find that algorithms such as ULMO hold significant potential/promise to learn and derive other, not-yet-identified behaviors in the ocean from the many archives of satellite-derived SST fields. We see no impediment to applying them to other large remote-sensing datasets for ocean science (e.g., SSH and ocean color).

Download Full-text