In brief, we compared VIIRS remote sensing data for SST on scales of ~100 km x 100 km against model outputs from the state-of-the-art ECCO LLC4320 ocean general circulation model.  Using a machine learning metric named Ulmo, we demonstrate the LLC4320 model performs well across most of the global ocean.  We highlight notable departures in the gulf stream, on the Equatorial Pacific, and in the Antarctic Cicumpolar Current.

All of the data provided here were sourced from public archives:

1. The JPL Physical Oceanography Distributed Active Archive Center (PO.DAAC, https://podaac.jpl.nasa.gov
2. MITgcm.org (using xmitgcm)

or are products of our own analysis.

This archive contains images and tables for the analyses undertaken for the purpose of assesing the LLC 4320 ocean model output of sea surface temperature (SST). Results of our investigations are reported in full in Gallmeier, Prochaska, Cornillon, Menemenlis, and Kelm 2023 https://gmd.copernicus.org/preprints/gmd-2023-39.

All code related to this project may be found on GitHub -- https://github.com/AI-for-Ocean-Science/ulmo and one may cite this DOI: https://doi.org/10.5281/zenodo.7685510