loading page

Unsupervised classification identifies coherent thermohaline structures in the Weddell Gyre
  • +2
  • Dan Jones,
  • Shenjie Zhou,
  • Maike Sonnewald,
  • Isabella Rosso,
  • Lars Boehme
Dan Jones
British Antarctic Survey

Corresponding Author:dannes@bas.ac.uk

Author Profile
Shenjie Zhou
British Antarctic Survey
Author Profile
Maike Sonnewald
Princeton University
Author Profile
Isabella Rosso
Scripps Institution of Oceanography
Author Profile
Lars Boehme
Scottish Oceans Institute, University of St. Andrews
Author Profile


The Weddell Gyre is a dominant feature of the Southern Ocean and an important component of the climate system; it regulates air-sea exchanges, controls the formation of deep and bottom water, and hosts upwelling of relatively warm subsurface waters. It is characterized by extremely low sea surface temperatures, active sea ice formation, and widespread salt stratification that stabilizes the water column. Studying the Weddell Gyre is difficult, as it is extremely remote and largely covered with sea ice; at present, it is one of the most poorly-sampled regions of the global ocean, highlighting the need to extract as much value as possible from existing observations. Thanks to recent efforts of the EU SO-CHIC project, much of the existing Weddell Gyre data, including ship-based CTD, seal tag, and Argo float profiles, has been assembled into a coherent framework, enabling new comprehensive studies. Here, we apply unsupervised classification techniques (e.g. Gaussian Mixture Modeling) to the new comprehensive Weddell Gyre dataset to look for coherent regimes in temperature and salinity. We find that, despite not being given any latitude or longitude information, unsupervised classification algorithms identify spatially coherent thermohaline domains. The highlighted features include the Antarctic Circumpolar Current, the central Weddell Gyre, and the Antarctic Slope current; we also find potential signatures of the inflow of Weddell Deep Water and export pathways of Antarctic Bottom Water. We show how varying the statistical, machine learning derived representations of the data can reveal different physical structures and circulation pathways that are relevant to the delivery of relatively warm waters to the higher-latitude seas and their associated ice shelves.