PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
June 1, 20260 citationsOpen Access

Probabilistic prediction of Atlantic Meridional Overturning Circulation collapse with Glass-Atlantic

View Full Paper
RBRobin Bisht

Key Points

  • This research aims to predict the likelihood of collapse of the Atlantic Meridional Overturning Circulation (AMOC) using a novel approach called Glass-Atlantic.
  • Introduced Glass-Atlantic, a spectral neural operator for climate modeling.
  • Trained using satellite altimetry, ECCO reanalysis, and 15 CMIP6 models.
  • Utilized advanced techniques like divergence-free spectral projection and multiscale coupling for enhanced predictions.
  • Achieved a 0.76 anomaly correlation at a ten-year lead time, outperforming the CMIP6 multi-model mean by 40%.
  • Assigned a 54±8% probability of AMOC collapse before 2050 under SSP5-8.5 scenarios.
  • Identified largest uncertainties in the predictions localized to the subpolar gyre due to divergent eddy parameterizations.

Abstract

The Atlantic Meridional Overturning Circulation (AMOC) is a potential tipping element in Earth's climate system, yet Earth System Models disagree by a factor of three on projected weakening by 2100. Here we introduce Glass-Atlantic, a spectral neural operator that learns the governing equations of rotating stratified flow while enforcing mass, momentum, and energy conservation by architectural construction. A divergence-free spectral projection eliminates numerical drift; renormalization-group multiscale coupling captures energy transfer across five decades of spatial scale; and an evidential uncertainty head decomposes predictive variance into intrinsic and structural components. Trained on satellite altimetry, the ECCO reanalysis, and 15 CMIP6 models, Glass-Atlantic achieves an anomaly correlation of 0. 76 at ten-year lead time, 40% above the CMIP6 multi-model mean, and assigns a 548\% collapse probability before 2050 under SSP5-8. 5, with the largest structural uncertainty localised to the subpolar gyre where eddy parameterisations diverge most strongly among state-of-the-art climate models.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Robin Bisht (2026) studied this question.

synapsesocial.com/papers/6a1d234302fbce9130638e73https://doi.org/10.5281/zenodo.18508751
Ask AI
Helpful
Bookmark
Share
View Full Paper