The Atlantic Meridional Overturning Circulation (AMOC) is a key component of the climate system, yet it exhibits a large inter-model spread and its future evolution remains uncertain. Here, we investigate the sensitivity of the AMOC to air–sea turbulent flux parameterizations using the IPSL climate model with five different bulk formulations. Differences in AMOC strength across the simulations reach up to 2 Sv (≡106 m3 s−1). They are driven by the relative strength of the subtropical and subpolar gyres, which controls heat and salt transport in regions of deep water formation. We quantify the respective contributions of the parameterization and feedbacks from wind, temperature, and specific humidity to the simulated flux differences. The behaviour of the surface roughness length at high wind speeds plays a key role and triggers both local and remote feedbacks. We show that these feedbacks dominate surface fluxes and subpolar gyre differences in the North Atlantic, leading to opposite results compared to the direct effect of the wind stress parameterization. This study emphasises the role of the interplay between heat and momentum fluxes and the way they are mediated by wind speed. Several aspects emerge from this study that could guide developments of air–sea turbulent flux parameterizations and help to understand the results in a coupled model.

Abstract. Quantitative reconstruction of ocean surface density during the Last Glacial Maximum (LGM) offers valuable insights into the ability of climate models to simulate past climate conditions, when global temperatures were about 4.5 °C to 6 °C colder than today. We assess the performance of the LGM climate simulations, as part of the 3rd and 4th phase of the Paleoclimate Modeling Intercomparisons Project, using a recent ocean surface density reconstruction based on the δ¹⁸O of foraminiferal calcite (δ¹⁸Oc). We consider the differences between the LGM and the preindustrial climates and each period separately, at both global and regional scales. Because surface density reflects the combined effects of temperature and salinity, we also examined sea surface temperature (SST) to better identify the processes underlying model–data differences. Surface density reconstructions show greater variability than simulated surface density. Models therefore struggle to reproduce the spatial variability of the density difference (LGM – pre-industrial (PI)), but part of the mismatch may arise from the uneven spatial distribution of reconstructions, which are mostly located near coastal areas. Density anomaly (LGM – PI) differences between data and models are largely controlled by sea surface salinity (SSS), with SST contributing to a lesser extent. This influence of SSS is directly linked to the reduction in tropical precipitation during the LGM: models that best match the large-scale density anomalies also simulate the strongest reductions in reconstructed low-latitude precipitation during the LGM, highlighting the key role of hydrological cycle changes in shaping surface density. On a global scale, 100 % of model simulations show a statistically significant relationship with surface density reconstructions, looking at LGM and PI separately. However, on a regional scale, some features are poorly simulated, leading to weaker agreement between data and model simulations, particularly in the North Indian and Southern Oceans. Our analysis concludes with a focus in the Indo-Pacific Warm Pool. Past reconstructions indicate a LGM weakened Indian ocean west–east surface density gradient, but only 7 out of 14 models (50 %) reproduce this feature. These results highlight the need to better constrain regional hydrological cycle changes in models, as improving their representation is crucial to reduce uncertainties in both paleoclimate simulations and future climate projections.

We describe a new Earth system model (ESM) experiment protocol, as part of the international Tipping Points Modelling Intercomparison Project (TIPMIP) project. We propose this as a protocol for the Coupled Model Intercomparison Project 7 (CMIP7). The protocol requires ESMs to run in CO2-emission mode, with atmospheric CO2 a predicted variable. Forcing for the protocol consists solely of a constant emission of CO2, based on each model's transient climate response to cumulative emissions of carbon dioxide (TCRE) value, to give a common global mean surface warming rate of 2 °C per century. This positive emission (ramp-up) experiment is started from the pre-industrial state of a given model. When the ramp-up run first exceeds a specified level of global warming (2 and 4 °C) relative to the model's pre-industrial global mean surface air temperature (GMSAT), CO2 emissions are set to zero and the positive emission run is branched into a zero-emission run. The zero-emission runs continue for 300 years. 50 years into each zero-emission run, CO2 emissions are set to the negative of the positive emission rate and the model run until GMSAT cools below the original pre-industrial value. Additionally, when the negative emission run started from the global warming level (GWL)=4 °C first drops below GWL=2 °C, a zero-emission run is branched off this, completing the set of experiments. Using this protocol, we are able to control the rate of global warming, and potentially also the rate of cooling, across participating models. TIPMIP experiments will support a range of analyses, including; an assessment of abrupt/rapid Earth system change under net zero CO2 emissions at a range of global warming levels, the long-term Earth system response to net zero CO2 emissions at these warming levels, the response to net negative CO2 emissions and the efficacy of negative emissions to drive cooling, and the reversibility of Earth system change under a pathway of positive (warming), zero, and negative (cooling) CO2 emissions.

Abstract Reconstructing global temperatures from sparse proxy data remains a key uncertainty in paleoclimate science. Here, we apply an emergent constraint framework that combines statistical relationships from paleoclimate model simulations with proxy‐based estimates to infer unknown climate variables. We compiled a new multiproxy data set of tropical land temperature reconstructions for the Last Glacial Maximum (LGM), which implies tropical terrestrial cooling of −4.2°C (−5.0 to −3.5°C, 95% Confidence Interval, CI). Applying our emergent constraint framework to this terrestrial temperature reconstruction, we derived updated estimates of tropical sea surface temperature (−2.6°C; −3.3 to −2.0°C, 95% CI) and global mean surface temperature (−5.4°C; −6.5 to −4.4°C, 95% CI). These observation‐based estimates support LGM cooling toward the colder end of published ranges, implying that very low equilibrium climate sensitivity (ECS < ∼3°C) is unlikely. Our method is adaptable for different time periods and climate variables, providing a flexible framework for integrating proxy and model information.

Given that the Arctic could be ice-free in summer within the next ten to 20 years, accurately predicting low-ice states is of crucial importance. Paleo-evidence shows that the strong orbitally-induced high latitude insolation anomaly at 127 000 years ago (127 ky), of around +70 Wm-2 in the Arctic during spring-summer, led to warm conditions and an Arctic that was occasionally or often ice-free during summer. Building on two Coupled Model Intercomparison Projects (CMIPs): the Sea-Ice Model Intercomparison Project and the Paleoclimate Modelling Intercomparison Project, we propose an Assessment Fast Track experiment, abrupt-127k, focusing on this seasonally ice-free, or near ice-free, Arctic at 127 ky. The abrupt-127k experiment is initialised from a piControl simulation and abruptly imposes observed values for the insolation distribution and greenhouse gas forcing at 127 ky. It provides a new opportunity to evaluate models used to compute climate projections, both against paleo-evidence and each other, during a known low Arctic sea ice state. As CMIP models are not usually tuned to paleo observations, abrupt-127k represents a true "out-of-sample" test. The abrupt-127k experiment has four key scientific objectives, to: ascertain the simulated Arctic sea ice state, including the presence and characteristics of last-ice areas; evaluate the simulated climates using Arctic paleo-evidence; characterise the central Arctic surface energy budget; and analyse the ice budget including ice melt, growth, and transport. We show that a large Arctic ice response will manifest within the first 30 years of the simulation, thus a single 100-year long run is sufficient for these objectives. Modelling groups are requested to follow standard CMIP output protocol for analysis, including the use of standard "fixed-length" output. Given abrupt-127k is similar in setup to abrupt-2xCO2 and abrupt-4xCO2 CMIP7 experiments, combined analysis of these abrupt-experiments will facilitate understanding of the impacts of instantaneous radiative forcing in the Arctic.

The slowdown of the Atlantic Meridional Overturning Circulation (AMOC) due to ongoing climate change raises concerns about its impact on the global carbon cycle. Understanding how global biosphere productivity may respond to such changes is essential for predicting the future carbon cycle, yet quantifying past global biosphere productivity remains challenging. We use triple isotope composition of air oxygen ( 17 Δ) trapped in polar ice cores to reconstruct the global biosphere productivity over the period 40.5 to 38.6 thousand years before the present (ka B.P.), covering Heinrich Stadial 4, a period of weakened AMOC. Our data reveal a slight increase in productivity following a carbon dioxide (CO 2 ) rise around 39.5 ka B.P. Freshwater hosing experiments using a global climate model indicate that the CO 2 fertilization effect on the terrestrial biosphere offsets productivity decline caused by consequences of AMOC slowdown. Collectively, our results suggest that global biosphere productivity slightly increases in response to AMOC slowdown and accompanying changes.

As anthropogenic forcing increases, there is a rising concern about crossing tipping points. A key information is the critical thresholds at which the so-called tipping elements could undergo abrupt changes, and the associated early-warning signals. This is true, in particular, for boreal forests and the possible greening of the Sahara. In that perspective, testing climate models against paleoclimate allows for exploring the climate response over multi-millennial time series, while offering the possibility to compare it with past vegetation reconstructions.We consider a transient simulation from the mid-Holocene (6,000 years before present) to 2100 obtained using the IPSL general circulation model including dynamical natural-only vegetation [1]. The simulation is forced with changes in orbital parameters and trace gases, transitioning from the paleo- to the historical period, and then to the scenario SSP4.5. We focus on two terrestrial ecosystems, both identified as tipping elements: the Sahara/Sahel and boreal forests in northern Europe.We analyze the evolution of vegetation patterns and extents in the two regions along the Holocene with the objective to determine if their response to forcing conditions can be assimilated to the crossing of a tipping point. Thus, we isolate rapid shifts and investigate whether they correspond to tipping points or to centennial variability. We link them to changes in regional vegetation drivers, including vegetation feedbacks, or to AMOC variability. Then, we explore possible analogies between changes during the Holocene and in projections, allowing for the identification of sensitive regions, which may help to detect regional thresholds for tipping points.References:[1] Braconnot, P., Viovy, N., and Marti, O. (2025). Dynamic vegetation highlights first-order climate feedbacks and their dependence on climate mean state. Earth System Dynamics, 16, 2113–2136.

Monitored seabird populations have declined by up to 70% worldwide since the 1950s. Yet, data on long-term seabird population dynamics prior to the anthropogenic era are largely unknown. This limits our ability to understand future population trajectories, particularly in the Southern Ocean, where seabirds are facing multiple environmental threats. Here, we use mercury (Hg) derived from seabird guano in peatland catchments as a tracer of colony population sizes on sub-Antarctic Bird Island (South Georgia). Peat Hg flux and isotope signature results show that the first sustained seabird colonies after deglaciation were established on the island between 6800 and 6100 years ago, predating evidence for colonization on other sub-Antarctic islands by more than 1,000 y. The four subsequent periods with large local seabird populations occurred during phases of less intense Southern Hemisphere westerly winds. Our study unveils significant and repeated millennial-scale shifts in seabird abundance in response to natural climate changes, implying that the present-day increase in westerly wind intensity may lead to further declines in seabird populations in the Southern Ocean.

Abstract. The Horn of Africa, and specifically Ethiopia's Awash Basin, is experiencing accelerating hydroclimatic change, yet the mechanisms driving multi-decadal shifts in its regional water balance remain poorly understood. The terminal Afambo–Gemeri–Abhe lake system, an endorheic water body at the outlet of the Awash Basin, acts as a natural integrator of its catchment's net water balance, providing a sensitive record of regional atmospheric dynamics. This study reconstructs its hydrological history over 1985–2024 by coupling the GR2M hydrological model with a lake water-balance model, constrained by satellite observations and in situ data. The model reproduces observed lake surface area dynamics and reliably extends the record into the data-scarce pre-1998 period allowing us to relate recent lake surface area changes to large scale circulation variability and trends. The reconstructed history reveals two superimposed signals. The first is pronounced variability across interannual to interdecadal timescales, driven by the El Niño–Southern Oscillation (ENSO), the Pacific Decadal Oscillation (PDO), and the Subtropical Indian Ocean Dipole (SIOD). The second is a non-oscillatory, multi-decadal expansion: the lake tripled in surface area from ∼150 km2 in the 1980s to ∼450 km2 by 2024, with growth accelerating markedly after 2020. This systematic expansion reflects a reorganization of the regional water balance driven by increased July–September rainfall and changes in potential evapotranspiration. We attribute these basin-scale changes to large-scale atmospheric circulation shifts, including a strengthening of the Somali Jet, the Tropical Easterly Jet, and the Indian monsoon circulation, and a northward shift of the Intertropical Convergence Zone and of the African Easterly Jet, likely linked to amplified regional warming over the Arabian Peninsula and the Tibetan Plateau. This modelling framework provides a transferable tool for evaluating past and future hydroclimatic variability, supporting more robust water resource assessments in data-scarce regions.

Transient Holocene climate simulations with state-of-the-art Earth system models offers new opportunities to investigate the relationship between multidecadal to multicentennial variability, the long-term climate trends, and interannual to decadal variability (Braconnot et al. GRL, 2019). However, multidecadal to multicentennial variability is still poorly known, both because it is difficult to properly extract from proxy records and because it is at the limit of what can actually be done with Earth System models. In addition, the different feedbacks from ocean or land-surface properties that can shape its characteristics are still poorly understood. Different climate models seem to provide different centennial variability patterns between the different ocean basins that reflect either the chaotic nature of the climate system or a poor representation of these variability scales. In this presentation, we will consider new mid-to-late Holocene simulations with the IPSL Earth System model, one of which includes interactive dynamical vegetation. We will first investigate the characteristics of multidecadal to multicentenial variability in these new simulations, with reference to recent publications comparing different transient Holocene simulations and addressing temperature variability scales, the thermohaline circulation, or Atlantic Ocean patterns leading to a reduction of the African monsoon interannual variability. We will also focus on key variability events that appear in the simulations and have a substantial impact on rapid changes in the African monsoon or on Northern Europe climate and land surface conditions (snow, soil moisture or vegetation). This opens the way to new research directions as part of the Paleoclimate Modeling Intercomparison Project.

Internal variations of climate can significantly influence global warming trends, especially at the continental scale, and could contribute to the recent abnormal observed warming over Europe. Model-based studies highlight that centennial variability of the North Atlantic can strongly affect this sector. However, a lack of high-resolution paleoclimate data does not allow a proper evaluation of the real existence of such a variability mode nor its amplitude. Here, we compile a series of annual proxy-based reconstructions over Europe from diverse sources to demonstrate and confirm the presence of such multi-centennial climate variability mode and quantify its amplitude. We show that this mode is closely tied to the internal variability of the Atlantic overturning circulation (AMOC) both in proxy-based reconstructions and climate models. When combined with instrumental observations, we show that the phase of this mode is crucial to be known. Indeed, results indicate that an internally-generated strengthening of the AMOC can explain a large part of the warming in the early 20th century and the relative cooling in the second half of this last century. A change in phase of this mode since the early 2000s is able to explain the observed amplified warming over Europe, which is projected to persist until the 2050s. According to an observational-constraint approach, this mode of variability could amplify the forced projected warming in Northern Europe by more than 58% in the next three decades. These results underscore the importance of considering internal climate variability when assessing regional warming trends, in order to develop consistent adaptation strategies.

El Niño-Southern Oscillation (ENSO) events are hard to put in one category because they differ in intensity, spatial pattern, and temporal evolution. Studies have characterized events into two main categories: central Pacific (CP) and eastern Pacific (EP) events. The indicators used to compute EP and CP events are varied, from area-averaged regions to Empirical Orthogonal Function (EOF) analysis. In the recent climatic period, they all show similar results. However, future projections show differing results when using two different methods of computing EP and CP events. Since the observational period is too short, we use paleoclimate reconstructions, which provide unique and quantitative measures of past climate changes over long time scales. We will first synthesize previous studies and discuss how they have used paleoclimate modeling and/or data to provide clues into how ENSO diversity may have been shaped in past climates. Our results indicate that many apparent inconsistencies in future projection studies are due to misleading use of ENSO diversity indicators and that investigating ENSO diversity with a climate change perspective requires assessing both changes in the climate mean state (annual mean and seasonality) and changes in variability.

We investigate how first-order albedo and water vapor radiative feedbacks are triggered by climate-vegetation interactions using mid-Holocene and pre-industrial climate simulations. The mid Holocene greening of the Sahara and northward shift of the northern tree line in the Northern Hemisphere illustrate these climate-vegetation interactions and challenge the development of Earth System models. We consider four different configurations for the IPSL Earth System model with dynamical vegetation to quantify vegetation and radiative feedbacks. They combine different parameterizations of key factors controlling vegetation functioning: bare soil evaporation, photosynthesis and associated parameters, and tree mortality. Whatever the model setup, the major differences between the mid-Holocene and pre-industrial climates are consistent with climate and vegetation reconstructions from pollen records. However, model setup differences modulate the way in which vegetation-climate interactions trigger first-order radiative surface albedo and water vapour feedbacks. Cascading effects involve both local snow-vegetation interactions and remote water vapour and long-wave radiative feedbacks. We show that the parameterization of bare soil evaporation is a key factor that controls tree growth in mid and high latitudes. Photosynthesis parameterization appears to be critical in controlling the seasonal evolution of the vegetation and leaf area index, as well as their effect on radiative feedbacks and the sensitivity of the vegetation feedback to the climate mean state. It even affects the sign of the global annual mean changes in temperature and precipitation between the mid-Holocene and pre-industrial periods. Dynamical vegetation highlights behaviours that can only be fully studied in a fully coupled Earth system model. The sensitivity of these vegetation-induced feedbacks to the mean climate state needs to be better considered when developing and tuning climate models.

The hydrological cycle plays a crucial role in the Earth’s climate and has a direct impact on human populations. Despite advances, the IPCC AR6 report highlights persistent uncertainties concerning future projections of potential changes in the hydrological cycle, in particular for low latitudes monsoonal systems. This is because numerical climate models exhibit significant spread in their projections.Traditionally, to estimate the future value of a climate variable, the distribution of projections from an ensemble of models is examined. However, this uncertainty is very high for water cycle, and the best estimates may be biased. To improve these projections, observational constraint, or emergent constraint methods, have been developed. These approaches adjust the distribution of projected variables based on observations, helping to reduce uncertainty. Furthermore, some studies show that the spatial pattern of sea surface salinity (SSS) is strongly correlated with the mean spatial pattern of the evaporation-precipitation (E-P) balance. Given that, water surface density is mainly influenced by salinity changes in region with strong precipitation and coastal runoff, both salinity and density could provide a useful tracer of the hydrological cycle.In this study, we reconstruct past sea surface density based on geochemical analyses (ẟ18Oc) on foraminifera extracted from marine sediment cores in the Bay of Bengal. Density changes in this dilution basin are mainly related to south Asian monsoon precipitation changes. We used our density reconstructions for the last glacial maximum (LGM) and Mid-Holocene (MH) as a predictor for the observational constraint method. Our goal is to reduce uncertainties in future South Asian monsoon precipitation projections in climate models by linking paleoclimatic information with future climate projections. To do so, we used PMIP and CMIP numerical climate modelling experiments.Our preliminary results show an underestimation of South Asian monsoon precipitation in the future (2000-2100) in most models, when using historical surface density and salinity (1900-2000) as a predictor. We are currently finalizing the use of LGM and MH surface density as predictor, in order to compare results when past predictors (LGM and MH) are used rather than an historical predictor.

El Nino-Southern Oscillation (ENSO) events, whether in warm or cold phases, that persist for two or more consecutive years (multi-year), are relatively rare. Compared with single-year events, they create cumulative impacts and are linked to extended periods of extreme weather worldwide. Here we combine central Pacific fossil coral oxygen isotope reconstructions with a multimodel ensemble of transient Holocene global climate simulations to investigate the multi-year ENSO evolution during the Holocene (beginning similar to 11,700 years ago), when the global climate was relatively stable and driven mainly by seasonal insolation. We find that, over the past similar to 7,000 years, in proxies the ratio of multi-year to single-year ENSO events increased by a factor of 5, associated with a longer ENSO period (from 3.5 to 4.1 years). This change is verified qualitatively by a subset of model simulations with a more realistic representation of ENSO periodicity. More frequent multi-year ENSO events and prolonged ENSO periods are being caused by a shallower thermocline and stronger upper-ocean stratification in the Tropical Eastern Pacific in the present day. The sensitivity of the ENSO duration to orbital forcing signals the urgency of minimizing other anthropogenic influence that may accelerate this long-term trend towards more persistent ENSO damages.

Abstract. Given that the Arctic could be ice-free in summer within the next ten to twenty years, accurately predicting low-ice states is of crucial importance. Paleo-evidence shows that the strong orbitally-induced high latitude insolation anomaly at 127,000 years ago (127 ky), of around +70 W m−2 in the Arctic during spring-summer, led to warm conditions and an Arctic that was occasionally or often ice-free during summer. Building on two Coupled Model Intercomparison Projects (CMIPs): the Sea-Ice Model Intercomparison Project and the Paleoclimate Modelling Intercomparison Project, we propose an Assessment Fast Track experiment, abrupt-127k, focusing on this seasonally ice-free, or near ice-free, Arctic at 127 ky. The abrupt-127k experiment is initialised from a piControl simulation and abruptly imposes observed values for the insolation distribution and greenhouse gas forcing at 127 ky. It provides a new opportunity to evaluate models used to compute climate projections, both against paleo-evidence and each other, during a known low Arctic sea ice state. As CMIP models are not usually tuned to paleo observations, abrupt-127k represents a true ‘out-of-sample’ test. The abrupt-127k experiment has four key scientific objectives, to: ascertain the simulated Arctic sea ice state, including the presence and characteristics of last-ice areas; evaluate the simulated climates using Arctic paleo-evidence; characterise the central Arctic surface energy budget; and analyse the ice budget including ice melt, growth, and transport. We show that a large Arctic ice response will manifest within the first 30 years of the simulation, thus a single 100-year long run is sufficient for these objectives. Modelling groups are requested to follow standard CMIP output protocol for analysis, including the use of standard ‘fixed-length’ output. Given Fast Track abrupt- 127k is similar in setup to abrupt-2xCO2 and abrupt-4xCO2 experiments, combined analysis of these abrupt-experiments will facilitate understanding of the impacts of instantaneous radiative forcing in the Arctic.

In state of the art Earth System Models (ESM), the variables at the ocean-atmosphere interface (wind, air temperature, humidity, surface currents and SST) are linked to turbulent surface fluxes (momentum, sensible and latent heat) in a complex manner via bulk closures.Understanding how turbulent fluxes interact between them and with the ocean-atmosphere interface variables is a major scientific challenge because it connects local interactions with large scale energy and water cycles.These interactions between the different air-sea turbulent fluxes are difficult to diagnose from fully coupled ocean-atmosphere simulations due to the fact that in most modelling groups coupled and stand alone components do not necessarily use consistent forcing or representation of the air-sea fluxes. Also rigorous protocols between coupled and stand alone atmosphere and ocean simulations need to be implemented to be able to properly disentangle the role of different physical representation at the air-sea interface from global ocean-atmosphere-land adjustment feedbacks that may counteract the direct effects of air-sea fluxes modeling.Here we use an ensemble of fully coupled and stand alone simulations using a version of the IPSL ESM [1] based on the new DYNAMICO atmospheric dynamical core [2] and the ocean engine NEMO [3]. We analyse an ensemble of experiments differing by the the bulk formulation of the air-sea turbulent fluxes (NCAR, COARE3.6, ECMWF and LMDZng). The analyses will focus on the adjustment of the system in the different cases, especially on the differences in the transport of heat and water, mixed layer depth adjustement, feedback on ocean surface properties, intertropical convergence zone (ITCZ) and mid-latitude storm tracks.[1] Boucher O., Servonnat, J., Albright, A. L., Aumont, O., Balkanski, Y., Bastrikov, V., et al. (2020). Presentation and evaluation of the IPSL‐CM6A‐LR climate model. Journal of Advances in Modeling Earth Systems, 12, e2019MS002010. https://doi.org/10.1029/2019MS002010[2] Dubos, T., Dubey, S., Tort, M., Mittal, R., Meurdesoif, Y., and Hourdin, F.: DYNAMICO-1.0, an icosahedral hydrostatic dynamical core designed for consistency and versatility, Geosci. Model Dev., 8, 3131–3150, https://doi.org/10.5194/gmd-8-3131-2015, 2015. [3] “NEMO ocean engine”, Scientific Notes of Climate Modelling Center, 27 — ISSN 1288-1619, Institut PierreSimon Laplace (IPSL), doi:10.5281/zenodo.14648

The Atlantic Meridional Overturning Circulation (AMOC) is a large scale circulation of about 18 Sv and 1.2 PW at 26°N characterized by upper waters flowing northward, losing heat and becoming cold deep waters before flowing back southward. The Deep Water Formation (DWF) and the Subpolar gyre circulation are key aspects of AMOC intensity [1] but strongly depend on air-sea fluxes, thus the need to quantify their influence.To do so, we compare 5 air-sea fluxes parametrizations within the IPSL General Circulation Model (GCM) [2] based on the new DYNAMICO atmospheric dynamical core [3] and the ocean engine NEMO [4]. We show that the spread in AMOC is more than 2 Sv, confirming the high sensitivity to air-sea fluxes. Furthermore, we manage to explain these discrepancies by assessing (i) winter time buoyancy fluxes in DWF area and (ii) subtropical and subpolar gyres intensity which drives the circulation. We also analyse the ocean-atmosphere feedbacks (mainly wind and sea surface temperature) that may be responsible for changes in AMOC, hence paving the way to a better representation in GCMs. [1] Buckley, M. W. and J. Marshall (2016), Observations, inferences, and mechanisms of Atlantic Meridional Overturning Circulation variability: A review, Rev. Geophys., 54, 5–63, doi:10.1002/2015RG000493. [2] Boucher O., Servonnat, J., Albright, A. L., Aumont, O., Balkanski, Y., Bastrikov, V., et al. (2020). Presentation and evaluation of the IPSL‐CM6A‐LR climate model. Journal of Advances in Modeling Earth Systems, 12, e2019MS002010. https://doi.org/10.1029/2019MS002010 [3] Dubos, T., Dubey, S., Tort, M., Mittal, R., Meurdesoif, Y., and Hourdin, F.: DYNAMICO-1.0, an icosahedral hydrostatic dynamical core designed for consistency and versatility, Geosci. Model Dev., 8, 3131–3150, https://doi.org/10.5194/gmd-8-3131-2015, 2015. [4] “NEMO ocean engine”, Scientific Notes of Climate Modelling Center, 27 — ISSN 1288-1619, Institut PierreSimon Laplace (IPSL), doi:10.5281/zenodo.1464816