📄 earth_science

Western North Atlantic Sea level reveals enhanced future AMOC Weakening

This study introduces a coastal sea-level gradient index as a robust proxy for Atlantic Meridional Overturning Circulation (AMOC) variability, revealing that observationally constrained climate models project a significantly stronger weakening of the AMOC by 2100 under both high and low emissions scenarios compared to previous IPCC assessments.

Original authors: Emmanuel Eresanya, Jennifer Mecking, Guillaume Hug, Patrick Mcloughlin, Yinghui He, Gerard McCarthy

Published 2026-07-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Emmanuel Eresanya, Jennifer Mecking, Guillaume Hug, Patrick Mcloughlin, Yinghui He, Gerard McCarthy

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The Ocean's Conveyor Belt

Imagine the Earth's climate system has a giant, invisible conveyor belt running through the Atlantic Ocean. This is called the AMOC (Atlantic Meridional Overturning Circulation). It moves warm water from the south up toward the north and cold water back down. This belt is crucial because it acts like a giant heater for Europe and helps regulate weather patterns all over the globe.

The problem is that scientists are worried this conveyor belt is slowing down. If it slows too much, it could cause big changes in our weather and sea levels. But, just like trying to guess how fast a car is going by looking at the smoke from its exhaust, it's hard to measure the AMOC directly. We only have direct measurements for the last 20 years (since 2004), which isn't long enough to see the full picture.

The New Tool: Using the "Shoreline" as a Gauge

This paper introduces a clever new way to measure the speed of this ocean conveyor belt by looking at sea levels along the coast.

Think of the ocean like a giant bathtub. When the water is swirling fast (a strong conveyor belt), the water piles up differently against the sides of the tub.

The Old Way: Scientists used to look at the temperature of the water surface to guess how fast the belt was moving. But temperature can be tricky because it's influenced by many things, like the sun or wind, not just the current.
The New Way: The authors looked at the height of the water along the coast. They found that when the conveyor belt slows down, the water level rises in the north and drops in the south (or vice versa, depending on the physics). This creates a "tilt" in the ocean surface.

They created a "Sea Level Gradient Index." Imagine placing a ruler between two points on the coast: one in the south (Florida) and one in the north (New Jersey). By measuring the difference in water height between these two spots, they can tell how strong the ocean current is.

The Detective Work: Checking the Models

Scientists use computer models to predict the future. However, these models are like different weather forecasters; some are very accurate, and others are a bit off.

The Test: The authors took 20 different computer models (from the CMIP6 project) and asked: "Which of these models correctly mimics the relationship between the ocean current and the sea level tilt that we actually see in real life?"
The Result: They found that some models got the relationship right, while others got it wrong. They created a "constrained" group (the good models) and an "unconstrained" group (the less accurate ones).

The Findings: The Belt is Slowing Faster Than We Thought

When they looked at the "good" models (the ones that matched real-world sea level data), the news was more serious than the general average suggested:

Historical Check: The "good" models matched the tide gauge records (water level data from the last 100+ years) much better than the average of all models. This gives us more confidence that the models are telling the truth about the past.
Future Warning: When they projected into the future (up to the year 2100), the "good" models predicted a much bigger slowdown in the conveyor belt than the standard average of all models.
- If we keep emitting high levels of greenhouse gases: The "good" models predict the belt will weaken by 53.4%. The standard average predicted only 45.3%.
- If we cut emissions significantly: The "good" models still predict a 36.0% weakening, compared to the standard 29.0%.

The Takeaway

The paper argues that by using the "tilt" of the ocean water along the coast as a ruler, we can filter out the less accurate computer models. When we do this, we see that the Atlantic Ocean's conveyor belt is likely to slow down more severely than we previously thought, especially if we don't reduce our carbon emissions.

In short: The coast is telling us the ocean current is in trouble, and the best computer models agree that the situation is worse than the general average suggests. The only way to avoid the worst-case scenario is to lower emissions.

Technical Summary: Western North Atlantic Sea Level Reveals Enhanced Future AMOC Weakening

Problem Statement
The Atlantic Meridional Overturning Circulation (AMOC) is a critical component of the global climate system, regulating heat and freshwater redistribution. However, its long-term evolution remains uncertain. While Coupled Model Intercomparison Project Phase 6 (CMIP6) models consistently project an AMOC weakening under greenhouse gas forcing, there is substantial inter-model spread regarding the magnitude, timing, and regional impacts of this decline. Previous attempts to constrain this variability using emergent constraints have relied on indirect indicators, such as surface temperature patterns, which may conflate multiple physical processes. Furthermore, direct observations of the AMOC are limited to the satellite era (since 2004), leaving a gap in understanding historical evolution and validating long-term model projections.

Methodology
This study introduces a physically grounded, observationally constrained metric based on coastal dynamic sea level (DSL) as a direct proxy for large-scale overturning variability. The methodology proceeds through several key steps:

Definition of a Sea-Level Gradient Index: The authors define a DSL gradient index based on the difference in sea level anomalies between a southern region (29–39°N, 75–55°W) and a northern region (43–53°N, 58–38°W) along the western North Atlantic boundary. This spatial pattern is derived from observed correlations between AMOC strength and DSL, characterized by positive correlations in the subtropical gyre and negative correlations in the subpolar North Atlantic.
Data Integration and Correction:
- Satellite & Reanalysis: Satellite altimetry (1993–2019) and ocean reanalysis products are used to establish the relationship between offshore sea level and AMOC.
- Tide Gauge (TG) Reconstruction: To extend the record back to the late 19th century, the authors utilize tide-gauge records from Fernandina Beach, Florida, and Sandy Hook, New Jersey. These records are adjusted to remove biases caused by vertical land motion and local coastal processes by aligning their linear trends with satellite altimetry over the common period (1993–2019).
- AMOC Reconstruction: A linear regression model, calibrated using the overlap period between satellite-derived sea surface height (SSH) gradients and reanalysis-based AMOC estimates at 26.5°N, is applied to the corrected TG gradient index to reconstruct historical AMOC variability.
Emergent Constraint Framework: The study evaluates 20 CMIP6 models based on their ability to reproduce the observed spatial relationship between AMOC and DSL. Models are weighted using a Gaussian likelihood function; those exceeding a predefined observational-consistency threshold form a "constrained ensemble," while others form an "unconstrained ensemble."
Projection Analysis: The constrained and unconstrained ensembles are compared against historical TG-derived estimates and projected forward to 2100 under two Shared Socioeconomic Pathways (SSP1-2.6 and SSP5-8.5).

Key Contributions

Direct Proxy Identification: The paper establishes coastal dynamic sea level as a more direct and robust proxy for AMOC variability than previous indirect indicators like surface temperature, as it reflects the integrated response of ocean density and mass distribution.
Historical Reconstruction: By correcting tide-gauge records for local biases, the study successfully reconstructs a longer-term AMOC record, revealing a pronounced weakening phase concentrated after 1990.
Model Selection and Weighting: The study demonstrates that models capable of reproducing the observed AMOC–DSL relationship exhibit significantly improved agreement with historical TG-derived AMOC evolution, particularly during the recent weakening period.
Revised Future Projections: The application of this constraint reveals that models with higher historical fidelity project more severe future AMOC weakening than the full CMIP6 multi-model mean.

Results

Historical Validation: The constrained ensemble reproduces the observed temporal evolution of the TG-based AMOC index with a statistically significant correlation ( $r=0.56$ for 1950–1990; $r=0.67$ for 1990–2014). In contrast, the unconstrained ensemble shows weak, non-significant correlations and underestimates the magnitude of the post-1990 decline.
Lag Analysis: Lag-correlation analysis indicates that AMOC variability leads the tide-gauge reconstruction by approximately two years, consistent with geostrophic adjustment processes.
Future Projections (2091–2100 relative to 1950–2000):
- High Emissions (SSP5-8.5): The constrained ensemble projects a mean AMOC weakening of 53.4% ± 16.0%, compared to 45.3% ± 20.3% for the full CMIP6 ensemble. The weakening trend is steeper ($-1.02$ Sv decade $^{-1}$ vs. $-0.85$ Sv decade $^{-1}$ ).
- Low Emissions (SSP1-2.6): The constrained ensemble projects a weakening of 36.0% ± 20.9%, compared to 29.0% ± 22.0% for the full ensemble.
Inter-Model Spread: The constrained ensemble consistently exhibits greater inter-model variability and steeper weakening trends, suggesting that observational constraints preferentially select models with higher sensitivity to anthropogenic forcing.

Significance and Claims
The authors claim that this study provides a physically interpretable and observationally grounded metric for evaluating AMOC behavior in climate models. By linking western North Atlantic sea-level gradients to overturning circulation variability, the proposed framework improves the representation of historical AMOC evolution and reduces uncertainty in future projections.

The paper emphasizes that the full CMIP6 ensemble may underestimate the magnitude of future AMOC weakening, particularly under high-emission scenarios. The findings reemphasize the importance of low-emission futures to avoid more extreme AMOC decline. Furthermore, the study supports the argument that historical performance—specifically the ability to reproduce observed circulation fingerprints like the AMOC–DSL relationship—should be a critical factor in assessing confidence in future climate projections. The authors acknowledge limitations, including the relatively short observational record and residual local influences in tide-gauge data, but conclude that long-term coastal sea-level observations offer a valuable tool for constraining large-scale ocean circulation change.