Increased risk of near term global warming due to a recent AMOC weakening

Projections from Earth System Models (ESMs) are crucial for climate change mitigation and adaptation. Equilibrium Climate Sensitivity (ECS) and Transient Climate Response (TCR) are key metrics for estimating ESM sensitivity to anthropogenic CO2 emissions. However, some CMIP6 models display higher ECS and TCR values than the IPCC AR5 likely ranges, projecting greater 21st-century warming. This raises questions about the realism of these models compared to historical observations and how to interpret their high sensitivity in relation to past climate change. Studies have attempted to constrain CMIP6 projections with observed warming trends, focusing on recent decades to minimize the impact of aerosol forcing. However, understanding the model spread and exploring the role of internal variability is essential for a comprehensive interpretation. Low-frequency internal variability can temporarily mask or amplify the effects of external forcing on climate change, significantly influencing model-observation comparisons. This study uses the IPSL-CM6A-LR model, characterized by high sensitivity and internal variability, along with its extended historical simulation ensemble (IPSL-EHS) to investigate the impact of multi-centennial internal climate variability on global near-surface air temperature (GSAT) warming since the mid-20th century and its implications for future warming projections.

Previous research highlighted discrepancies between some CMIP6 models' high climate sensitivity (ECS and TCR) and historical warming observations. Several studies tried to reconcile these differences by constraining CMIP6 projections with observed warming trends, primarily focusing on recent decades to minimize aerosol forcing effects. However, these approaches often neglect the significant influence of low-frequency internal climate variability, which can temporarily mask or enhance long-term trends. Existing studies on this topic mostly focus on analyzing ensemble means, thereby overlooking the potentially crucial information embedded within individual ensemble member variations and their inherent spread. This study builds upon previous work by explicitly addressing this gap, focusing on the role of internal variability of the Atlantic Meridional Overturning Circulation (AMOC) in shaping the observed warming trends.

The study uses an ensemble of 32 extended historical simulations (IPSL-EHS) produced by the IPSL-CM6A-LR model, known for its relatively high climate sensitivity (ECS of 4.5 K and TCR of 2.4 K) and substantial internal variability. The authors calculate historical climate sensitivity (S_hist) and transient climate response (TCR_hist) for each ensemble member using a method similar to previous studies, comparing surface air temperature changes between the preindustrial period (1850–1879) and the present-day (1999–2018). The spread in S_hist and TCR_hist values helps assess the influence of internal variability. The study investigates the relationship between GSAT warming trends and AMOC trends using observational data (HadCRUT4-CW and Berkeley Earth datasets) and AMOC reconstructions (Caesar index). They identify ensemble members with GSAT trends closely matching observations and examine their AMOC characteristics. The relative contributions of internal variability and external forcing to AMOC weakening are evaluated using a linear model. To further validate their findings, the authors analyze several AMOC fingerprints (Caesar index, Atlantic Multidecadal Variability (AMV) index, Ocean Heat Content (OHC) anomaly, and interhemispheric temperature difference (ΔITA)) from observations and compare them to the identified ensemble members and the ensemble mean. Finally, a multi-model analysis is conducted using other CMIP6 models with large historical ensembles (MPI-ESM1.1, CNRM-CM6A-1, and CanESM5) to assess the generality of the findings. The statistical significance of correlations and regressions is evaluated using appropriate methods, accounting for serial correlation where necessary.

The IPSL-CM6A-LR ensemble mean shows greater warming than observed historically. However, some members closely match observed GSAT trends, exhibiting a substantial internally-driven AMOC weakening over recent decades. Members with the best GSAT representation also show AMOC weakening consistent with observations. This suggests that AMOC internal variability masked a portion of anthropogenic warming during the historical period. A strong positive relationship (r²=0.82) exists between GSAT and AMOC trends over 1940-2016. Analysis of AMOC fingerprints (Caesar index, AMV index, AOHC index, and ΔITA) confirms that ensemble members with the best GSAT match also exhibit AMOC variations consistent with observational data, particularly member #14. A multi-model analysis shows a similar relationship between AMOC and GSAT trends in several other CMIP6 models. Specifically, the subset of IPSL-EHS members displaying a strong AMOC weakening during 1951-1990 also shows comparatively lower warming than the ensemble mean during the same period. In contrast, the same subset is projected to show an internally generated AMOC strengthening and consequently a greater warming rate (~0.36 K per decade) in the coming decades than the ensemble mean (~0.34 K per decade). The study's findings indicate that the AMOC weakening, primarily driven by internal variability, might have masked a portion of anthropogenic warming in recent decades, increasing the projected warming in the near future. The results suggest that emergent constraint approaches should consider individual ensemble members' internal variability rather than solely relying on ensemble means.

This study's findings highlight the crucial role of low-frequency internal variability, particularly AMOC weakening, in influencing the observed global warming trends over the past few decades. The results suggest that a portion of anthropogenic warming might have been masked by this internal variability, which could manifest in the coming decades. The significant positive relationship between AMOC and GSAT trends underscores the importance of considering internal variability in climate model projections. This study emphasizes the limitations of solely relying on ensemble means in assessing the impacts of climate change and the importance of considering individual ensemble members and their intrinsic spread. The implications are significant as ignoring the effects of internal variability in AMOC could lead to an underestimation of future warming and the associated risks.

This study demonstrates that the observed AMOC weakening since the mid-20th century may be largely attributed to internal multi-centennial variability, potentially masking a portion of anthropogenic warming. This implies that estimates of transient climate sensitivity based on recent observational records might be underestimated. The authors propose that emergent constraint methods should explicitly account for low-frequency internal variability by analyzing individual ensemble members, rather than ensemble means, to enhance the accuracy of future climate warming projections. Future research could investigate the combined impact of Atlantic and Pacific ocean variability on observational constraints of future warming levels and refining methodologies to better capture and model the interplay of internal and external forcings in shaping long-term climate trends.

The study's reliance on a single high-sensitivity model (IPSL-CM6A-LR), although supported by a multi-model analysis, might limit the generalizability of the findings. The realism of the model's multi-centennial internal variability remains uncertain, despite comparisons with observational AMOC fingerprints. While the study utilizes several datasets and methods, uncertainties associated with observational data and AMOC reconstructions remain, which could influence the interpretation of the findings. Further research is needed to explore the limitations of the employed methods for estimating S_hist and TCR_hist and to investigate the relative contribution of internal variability and external forcings to AMOC changes.