WeatherIA
FRGBMA
climat

Ocean Warming: New Geodetic Method Reveals Double-Digit Discrepancies with Climate Models

Researchers from the University of Bonn have developed a geodetic technique that shows ocean temperatures sometimes deviate by more than ten percent from climate model predictions, raising significant questions about the accuracy of global simulations.

WE

Rédaction Weather IA

vendredi 2 octobre 2026 à 15:387 min
Partager :Twitter/XFacebookWhatsApp
Ocean Warming: New Geodetic Method Reveals Double-Digit Discrepancies with Climate Models
Oceans Warm Faster Than Expected: New Geodetic Method Reveals Major Discrepancies

A new geodetic method reveals that the differences between observed ocean temperatures and climate model predictions sometimes reach double-digit percentages. This finding, published by researchers from the University of Bonn, highlights a major uncertainty in our global climate predictive models.

What the study reveals: significant discrepancies between observations and models

By analyzing variations in ocean volume using satellite data from Copernicus missions, the team converted these height changes into heat storage estimates. The results show that, depending on the basin, real temperatures are sometimes higher or lower than values simulated by ECMWF models and other global models, with differences regularly exceeding double-digit percentages. These discrepancies are not uniform; some regions, like the tropical Pacific, exhibit more pronounced divergences.

The report emphasizes that these discrepancies are not merely random fluctuations; they reflect systematic biases in how models integrate convection processes, heat exchanges between atmosphere and ocean, and the representation of deep currents. In other words, predictive models underestimate or overestimate oceans' ability to absorb excess heat, which has direct implications for sea-level rise projections and extreme phenomena.

How: Geodetic method and neural networks to measure warming

The researchers combined satellite geodetic measurements – mainly radar altimeters that capture ocean surface height – with a neural network trained on archives of atmospheric and oceanographic data. This machine learning model was calibrated using the Copernicus Marine Service dataset, providing high-resolution temperature and salinity fields. By linking height variations to water density, the network could estimate heat storage with greater precision than traditional methods based solely on in-situ measurements.

The process also relies on predictive models from the ECMWF consortium, offering reference simulations for comparing observations. By overlaying both datasets – geodetic observations and simulations – the group could quantify discrepancies and identify regions where models fail most often. This hybrid approach, combining satellite data, artificial intelligence, and climate models, represents

Was this article helpful?

Commentaires

Connectez-vous pour laisser un commentaire