Geological evidence from West Antarctica demonstrates that polar regions supported temperate rainforest ecosystems during Cretaceous greenhouse conditions, implying substantially reduced meridional temperature gradients and a fundamentally different climate state than today. Plausible projections of such future transitions into very different climate states require a profound understanding of the dynamics and changes in the Earth's past, e.g., the greenhouse climates of the Late Cretaceous (∼ 94–86 million years ago (Ma)) and Eocene (∼ 56–34 Ma), as well as transitions into or out of them. So far, reliable polar data from such periods have been particularly scarce, even though these regions react most sensitively to climatic changes. Previous work (DeConto and Pollard, 2003; Zachos et al., 2008) has shown that Antarctic Ice Sheet inception and retreat are strongly linked to atmospheric CO₂ thresholds, ocean gateways, and Southern Ocean circulation, but direct polar evidence from greenhouse climates remains sparse. Against this background, two AWI-led interdisciplinary studies based on paleo-samples from RV Polarstern expedition PS104 provided new constraints on Antarctic warm climates under high-CO₂ conditions (Klages et al., 2020, 2024).

First, the unexpected discovery of original remains of a former temperate rainforest only ∼ 900 km away from the South Pole (Klages et al., 2020) illustrates the severity of global climate conditions during greenhouse states – a fully vegetated South Pole region with annual mean temperatures of ∼ 12 °C presumes a drastically different global climate under high-CO₂ atmospheric concentrations (1120–1680 ppmv, i.e., 4× to 6× higher than pre-industrial levels) and associated oceanic and atmospheric circulation patterns. Second, a unique sedimentary drill core record from West Antarctica's Pacific margin combined with novel coupled climate–ice-sheet modeling puts the timing and location of permanent Antarctic Ice Sheet inception ∼ 34 million years ago in a new light (Klages et al., 2024). The study defined the size and location of the Antarctic Ice Sheet (AIS) during the last fundamental climate shift from the Eocene greenhouse to the current icehouse climate (Fig. 1) and demonstrates how regionally different the AIS reacted to external forcing during this last drastic climatic transition (Fig. 2).

While East Antarctica was already covered by a large and thick ice sheet, West Antarctica started to glaciate about 7 million years later, when atmospheric CO₂ levels dropped to values between 2× and 1× preindustrial CO₂ (560–280 ppmv). This result adds a regional perspective to the broader view that East and West Antarctica can respond differently to CO₂ forcing because of contrasting bed topography, ocean access, and ice-sheet dynamics. This emphasizes the fragility of the West Antarctic Ice Sheet and ramifications for global sea level and is essential knowledge to better understand climate dynamics during past and potential future climate reorganizations. Both studies therefore significantly help understanding polar responses to higher-than-present CO₂. With this research we constrain Earth-system models, which aim at simulating future ice-sheet and climate characteristics under CO₂-driven warmer-than-present conditions more reliably than currently possible.

Later in Earth history (about 3 million years ago, in the Pliocene), we find that the Antarctic Circumpolar Current (ACC) accelerated as the Earth gradually cooled (Lamy et al., 2024; Wu et al., 2021). The ACC is a key regulator of Southern Ocean heat, carbon, and nutrient exchange and mediates the connection between wind forcing, ocean stratification, and global carbon storage (e.g., Rintoul et al., 2001; Marshall and Speer, 2012). Against this established background, recent proxy and modeling work has refined our understanding of how ACC strength varied across glacial–interglacial and Pliocene climate states. The acceleration is consistent with an enhanced meridional temperature gradient between low and high southern latitudes, which strengthened the Southern Hemisphere westerlies, a primary driver of the ACC. However, in the late Pliocene the ACC began to slow. The switch came at a time when the climate and the circulation in the atmosphere and the ocean experienced major changes linked to broad expanses of the Northern Hemisphere ice sheets and the expansion of the Antarctic Ice Sheet. This is attributed to changes in ocean currents, set off by tectonic processes, together with a long-term cooling of the ocean and decreasing atmospheric CO₂ levels. More ice-proximal drill records from the International Ocean Discovery Program (IODP) Expedition 379 and seismic analysis provide strong indication for an intensively dynamic West Antarctic Ice Sheet in the Amundsen Sea sector during a newly defined Pliocene Amundsen Sea Warm Period (4.2–3.2 Ma) with prolonged ice-sheet retreats (Gohl et al., 2021).

With regard to the last 800 000 years, in which the atmospheric CO₂ levels varied from 170 to 300 ppmv, we were able to identify a close connection between ACC strength and glacial cycles: during warm periods, in which the atmospheric CO₂ levels rose, the flow speed increased by up to 80 % compared to the present; during ice ages, it decreased by up to 50 %. At the same time, during transitions between interglacials and glacials, there was a shift in the ACC's position and therefore in the upwelling of nutrient-rich deep water in the Southern Ocean, as geochemical sediment analyses revealed. They show that the silicate shells of diatoms – the most important phytoplankton in the Southern Ocean – were deposited on the seafloor farther north in ice ages than in warm periods. A weaker ACC and lower atmospheric CO₂ levels during the ice ages of the Pleistocene indicate less pronounced upwelling and more stratification in the Southern Ocean, that is, more capacity for CO₂ storage (Lamy et al., 2024; Wu et al., 2021). Evaluation of these processes is essential for our understanding of climate–carbon feedback under warmer climates and highlights the necessity to equip climate models with additional components of the Earth system and to allow sufficient spatial resolution to capture processes that resolve the necessary dynamics. These findings provide important constraints on how Southern Ocean circulation, carbon storage, and Antarctic ice sheets may respond to future warming.

Together, these studies demonstrate that Antarctic Ice Sheet evolution and Southern Ocean circulation are strongly climate-state dependent. Geological evidence, proxy reconstructions, and coupled modeling indicate that the response of the Antarctic cryosphere cannot be described by a single CO₂ threshold. Instead, East and West Antarctica exhibit distinct sensitivities arising from differences in topography, ocean connectivity, and feedback processes. At the same time, changes in Southern Ocean circulation emerge as a key mechanism linking ice-sheet stability, carbon storage, and global climate evolution across both past and future warm climates.

Reconstructing the distant past is a challenging task, as the thickness of glaciers and sea levels cannot be measured directly. As a result, climate researchers must rely on various forms of evidence for a complete picture of past climate changes. The challenge is that different types of evidence can lead to different interpretations of the past. Previous reconstructions of Last Glacial Maximum sea level and continental ice volume were difficult to reconcile, resulting in a long-standing discrepancy known as the “missing ice problem” (Clark et al., 2009; Lambeck et al., 2014). It describes the incongruity of two different scientific approaches that sought to reconcile sea level height and glacier thickness at the Last Glacial Maximum (LGM, ∼ 21 thousand years before present). The “missing ice problem” is based, on the one hand, on an analysis of sediments from core samples collected from the seafloor in the tropics. These contain traces of corals. Because corals only live in well-lit waters near the ocean's surface their remnants can tell us to what extent the sea level rose or fell over the millennia. The sediment cores in the tropics imply that the glacial sea level was roughly 130 m lower than today. On the other hand, previous models have suggested that the glacial masses were not large enough to explain such a low sea level. For the sea level to be that low, on a global scale an additional volume of water with twice the mass of the Greenland Ice Sheet (corresponding to ∼ 14 m of sea level change) would have to have been deposited on land; hence the “missing ice problem” (Gowan et al., 2021). In a project together with the excellence cluster “The Ocean Floor” at MARUM, we apply data assimilation to an ice sheet–solid-Earth model to provide an independent solution for ice volume changes based on oxygen isotopes in sediments. The approach integrates different key indicators such as the steepness of the ice surface, glacier flow directions and the resistance of rocks and sediments to basal ice flow. In addition, the ice sheet presses down on the Earth's crust in the respective areas, which is determined by the underlying pattern of mantle viscosity, resulting in LGM ice masses that can be reconciled with the sea level height reconstructions (Gowan et al., 2021). The resolution of the missing ice problem highlights the growing importance of data assimilation approaches that combine geological constraints with physically consistent ice-sheet and solid-Earth models.

Prior to the LGM, reconstructions of continental ice sheets are further complicated, since the glacial geomorphological record is mostly erased by the following ice-sheet expansion. Hence, considerable uncertainties exist in the initial ice-sheet configuration for the final glaciation phase (equivalent to a sea level drop of around 50 m) starting at ∼ 38 ka (upper panel of Fig. 3). Given such fundamental ice-sheet differences at 38 ka, the question remains as to how these differences impact the North American Ice Sheet evolution toward the LGM. In the framework of the BMBF-funded PalMod Project (https://www.palmod.de, last access: 30 December 2025), we have developed and used a coupled climate–ice model (the Earth-system model AWI-ESM, Ackermann et al., 2024) to study the dynamics of ice-sheet inception and evolution. The study simulates the build-up of the North American Ice Sheet before the LGM and shows which factors influenced its growth. With our new Earth-system model with interactive ice sheets, it is possible to adequately capture ice-sheet feedback mechanisms, which are not yet included in the current scenario simulations that produce projections of future climate. We find that distinct initial North American (Laurentide) ice-sheet configurations at 38 ka converge toward a configuration consistent with the LGM (lower panel of Fig. 3). The rapid expansion of the North American Ice Sheet can be explained by summer snowfall, which was caused by the transport of moisture from warmer lower latitudes of the Atlantic into the interior of the North American continent. This mechanism not only led to more ice but also to greater snow cover over North America, which reflected sunlight more effectively and thus additionally reduced melting (Niu et al., 2024). These feedbacks provide an effective mechanism for ice growth for a range of different initial ice-sheet states that converge to an LGM ice-sheet configuration and may also explain the rapid North American ice-volume increase during the last ice age and potentially driving growth during previous glacial periods.

The transition from the cold LGM to warmer interglacial conditions is driven by gradual changes in Earth's orbit but also by abrupt shifts in global ocean circulation. These shifts amplify gradual external forcing by altering the exchange of heat and carbon between the ocean and atmosphere, facilitating deglaciation. However, it remains unclear why these interactions during other glacial periods did not lead to a successful termination and a new interglacial. To address this, we analyzed global hydrography changes using models and data (Knorr et al., 2021). Our findings show that during the LGM, ocean stratification was notably different, particularly in the deep South Atlantic, where increased salt stratification reduced the cooling effect of upwelled waters in response to an abrupt weakening of the Atlantic overturning circulation. This unique stratification led to a warmer high-latitude surface ocean, as the Southern Ocean tapped into deeper, less cold, and saltier waters than at other times in the glacial cycle. This mechanism amplified the Antarctic seesaw warming response, doubling the rate of warming across Antarctica on a multi-centennial scale compared to interglacial conditions. This helps explain the large magnitude and rapidity of glacial terminations in the late Quaternary.

A key uncertainty in understanding sea level variations and paleoclimate is the freshwater balance and interpretation of oxygen isotope records. Evidence suggests a major ice shelf existed in the Central Arctic Ocean during glacial periods, a hypothesis supported by seafloor morphology (Jakobsson et al., 2016). Our study provides sedimentary geochemical evidence for this ice shelf and identifies traces of widespread freshwater at depths up to 2500 m, detected using advanced uranium-series isotope analysis at AWI (Geibert et al., 2021). The finding has stimulated extensive discussion because it implies a substantially different Arctic Ocean state during glacial periods. A more isolated and freshened Arctic Ocean would influence water-mass exchange with the North Atlantic, ocean circulation, carbonate chemistry, and the interpretation of oxygen-isotope records. The implications of such an “Arctic Mediterranean” are now being explored in both proxy studies and climate models.

The interpretation critically depends on robust age models. Consequently, much of the subsequent work has focused on evaluating chronological uncertainties and testing independent dating approaches. Radiocarbon chronologies have been reassessed (Spielhagen et al., 2022; Geibert et al., 2022), while microscopic inspection of microfossils (Wollenburg et al., 2023), cosmogenic isotopes, and additional radiogenic tracers (Nicolas et al., 2024) provide complementary constraints on sediment age and depositional history.

One example of how natural variability can temporarily mask or amplify warming trends is the air temperature evolution on the Greenland Ice Sheet. Until now, the impact of global warming on the elevated parts of the ice sheet (up to 3000 m above sea level) has remained unclear due to the very strong natural climate variability and the lack of long-term observations. This uncertainty has fueled discussions of a “warming hiatus” – a delay or pause in the signature of global warming in Greenland, quite similar to the much discussed “pause” of global warming in the early 2000s (e.g., Box et al., 2009). By utilizing the unique infrastructure of the AWI (Polar-5 airplane) and benefiting from advancements in the ice core proxy understanding, we re-drilled and updated existing ice core records from central and northern Greenland. This allowed us to reconstruct air temperatures on the Greenland Ice Sheet over the past millennium with unprecedented quality (Hörhold et al., 2023), which complements the previously used Arctic-wide temperature reconstruction (Fig. 4). The analysis of these unique data shows that modern temperatures on Greenland are 1.5 °C warmer than those of the 20th century. It shows that anthropogenic-induced global warming has reached the high-elevation areas of central and northern Greenland. Our data also enabled us to generate an estimate of the meltwater runoff anomalies for Greenland over the past millennium to put the recent runoff anomalies into a long-term context. The findings suggest that current decadal meltwater runoff anomalies are unprecedented over the last thousand years and are likely to impact firn densification and meltwater storage capacity, with further implications for the ice sheet's mass balance.

The reconstruction also revealed strong natural decadal temperature variability (Fig. 4) in Greenland (Hörhold et al., 2023). To understand the origins of climate variations, it is essential to examine the spatial patterns of temperature variability. By leveraging a comprehensive sedimentary pollen database developed at AWI along with instrumental data, we reconstruct the spatial pattern of terrestrial temperature variability across annual to millennial timescales (Hébert et al., 2022). We show that in coastal regions, strong ocean-driven climate variability overprints the local land climate on multi-decadal and longer timescales. Due to marine influence, regions characterized by stable oceanic climate at sub-decadal timescales experience stronger long-term variability, and continental regions with higher sub-decadal variability show weaker long-term variability. This fundamental relationship between timescales offers unique insights into the emergence of a low-frequency marine-driven regime that governs terrestrial climate variability. It also highlights that the muted slow climate variability observed in standard climate model simulations likely originates from the ocean.

One possible reason for the weak regional temperature variability on multidecadal and longer timescales in models could thus be related to limited spatial resolution. This leads to inadequacies in the representation of subgrid-scale processes, e.g., ocean eddies. We find that there is a tendency for higher-dimensional models to capture variability and extremal behavior better (Contzen et al., 2022). This underlines our goal of understanding and quantifying climate variability to improve modeling by bridging observations with model performance. Marine sediment archives provide an additional perspective on large-scale climate variability and ocean circulation changes. Using planktonic microfossil assemblages, Jonkers et al. (2023) reconstructed glacial sea-surface temperature patterns and biogeographic shifts across the North Atlantic. The study helps reconcile discrepancies between paleoclimate simulations and proxy reconstructions of the Last Glacial Maximum and illustrates how biological archives provide independent constraints on past ocean circulation and climate states. Understanding how eddies respond to climate warming and how their presence influences climate change is therefore critical for many aspects of Earth-system science, which will be covered in detail below.

The Greenland Ice Sheet (GrIS) has steadily lost mass over the past decades, presently representing the second-largest single contributor to global sea level rise, next to the expansion of seawater by warming. Record melt years like 2012 and 2019, as detected by the GRACE/GRACE-FO satellites, are typically enabled by a stable high-pressure system over Greenland, causing the advection of warm air masses from lower latitudes along the western coast of the ice sheet and northern Arctic Canada (Sasgen et al., 2020). Under these conditions, glaciers in Svalbard are influenced by cold air from the north, leading to opposite temperature and melt anomalies in both regions. The GRACE/GRACE-FO data show that in 15 out of 17 years since 2003, a large ice loss in Svalbard has been accompanied by a weaker loss in northern Arctic Canada and western Greenland, and vice versa. This phenomenon has been occurring more frequently since the 2000s (Sasgen et al., 2022).

We see that the outflow of ice from the GrIS is strongly localized, with only a few ice streams draining most of the entire ice sheet. These river-like zones, where ice flow is significantly faster than in their surroundings, are assumed to be triggered by either bedrock properties, enhanced sliding, and/or by deformation mechanisms leading to shear localization. Oceanic heat transport in the fjords leading up to marine-terminating outlet glaciers can trigger glacier retreat and acceleration of ice flow, which also affects the discharge of ice transported from the inner ice sheet to the coast. Thus, understanding ice stream behavior and its temporal variability is crucially important to project future rates of sea level change. The 500 km long Northeast Greenland Ice Stream (NEGIS) is by far the largest ice stream in Greenland (Joughin et al., 2001; Fahnestock et al., 2001). It feeds its three major outlets on the coast from the region of the central ice divide, and its catchment area covers 17.2 % of the ice-sheet area. While the gates in the coastal mountain range in the northeast of Greenland determine the location of the coastal outlets, its course in the interior of the ice sheet appears not to be constrained by bed topography. We investigate temporal changes in ice stream flow from decadal to millennial timescales to decipher the processes which control or initiate changes in NEGIS. All these data are essential pieces of a larger puzzle, with which we contribute to a better understanding of how ice streams work and how fast they can respond to changing boundary conditions. This will eventually lead to improved projections of sea level rise, and especially to more accurate timing.

Airborne radar surveys of ice stratigraphy in the NEGIS upstream area reveal how ice stream flow has shaped isochrone layering (Jansen et al., 2024). Data from the AWI Polar-6 aircraft and ultra-wideband radar system near the EastGRIP (https://eastgrip.org, last access: 30 December 2025) drilling camp show distinct shear margin signatures in the stratigraphy. A 3-D horizon, tracing an isochrone in the upper ice column, reveals how pre-existing folds are deformed within the shear margin, forming a “kink” that indicates a two-stage deformation process. First, large cylindrical folds formed under convergent flow; they then narrowed as sharp shear margins developed. Fold amplitude changes over time suggest the shear margins were fully established around 2000 years ago, challenging the idea that streaming is triggered and sustained by long-term heat flow anomalies, instead pointing to more abrupt ice-flow reconfigurations (Fig. 5).

The signature in the radar stratigraphy also remains as a record of deformation should the velocities change. It thus provides information on past ice flow regimes. Franke et al. (2022) find such evidence further north of the current NEGIS. The ancient ice-flow regime must have had NEGIS-type shear margins reaching far inland toward the ice divide and probably became deactivated when the main drainage system was reconfigured and relocated southwards. Thus, major ice streams in Greenland might start, shift, or abruptly disappear, and these processes are very likely linked to ice-sheet geometry and volume and to processes at the outlets which influence the mass discharge. Besides the shear margins, a high strain localization zone, the third interface enveloping and enabling the streaming of inland ice, is the subsurface of this “ice highway”.

The project successfully drilled a 2668 m ice core from the Northeast Greenland Ice Stream (NEGIS), marking the first complete drill-through of an ice stream. We thus obtained unique access not only to the streaming material but to the bed properties. We find that the sediment under the ice is saturated with water, which is probably caused by basal melting of the ice sheet.

Changes in ice-stream dynamics ultimately influence the marine outlets through which ice is discharged into the ocean. Understanding the coupling between NEGIS and its outlet glaciers therefore requires both glaciological and oceanographic perspectives. More recently, the marine outlet of the NEGIS shows strong evidence of retreat. The most prominent example of this is the collapse of the floating ice tongue of Zachariæ Isstrøm (ZI). The loss of the ice tongue led to a speed-up of ice discharge into the ocean. The largest remaining floating ice tongue of Greenland is the neighboring Nioghalvfjerdsfjorden Glacier – also known as the 79° North Glacier (79NG) – which represents the second major outlet of NEGIS with considerable basal melting. Continuous thinning of the 79NG tongue by up to 30 % has occurred over the past 20 years. This process is caused by increased melting along the ice tongue base (Humbert et al., 2023; Zeising et al., 2024).

Western boundary currents (WBCs) are fast-moving ocean flows that are warming at rates 2–3 times faster than the rest of the ocean due to climate change. These currents are vital for regulating regional climates and ecosystems, with warming hotspots affecting sea level rise, marine heatwaves, and fisheries. Ocean eddies are emerging as key drivers of rapid warming in WBCs, such as the Gulf Stream and the Kuroshio Current, with recent advances highlighting their critical role in shaping ocean dynamics under climate change. Our studies using satellite data (Yang et al., 2020, 2023) and high-resolution models (Beech et al., 2022) reveal that intensifying poleward-moving eddies are the primary contributors to this rapid warming.

In our recent research using the AWI climate model (AWI-CM), part of AWI's contributions to the Coupled Model Intercomparison Project Phase 6 (CMIP6) (Semmler et al., 2020), discussed in the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (AR6 IPCC), we uncovered a significant role of mesoscale eddies in shaping ocean circulation in a warming climate (Beech et al., 2022; Yang, 2022). Leveraging the model's unstructured mesh capability to resolve eddy-rich regions, the study provides unprecedented insights into how eddy activity evolves under anthropogenic climate change. Mesoscale eddies influence western boundary current extensions such as the Gulf Stream and Kuroshio by redistributing heat, salt, and momentum. Changes in eddy kinetic energy (EKE) are therefore regionally important, but they do not necessarily imply a uniform response of all western boundary currents. In the North Atlantic, changes in Gulf Stream variability must be interpreted together with basin-scale density gradients and AMOC changes; in the North Pacific, Kuroshio variability reflects a different dynamical setting.

A key novelty of this approach lies in its ability to simulate long-term changes in eddy activity using high-resolution modeling. Unlike previous studies limited by satellite altimetry records or coarse-resolution models, AWI-CM captures the fine-scale processes driving shifts in eddy dynamics. For example, poleward shifts in EKE align with the expected movement of major ocean currents under warming, while the decline in Gulf Stream eddy activity reflects the cascading impacts of AMOC weakening. The model also shows that intensifying westerly winds over the Southern Ocean amplify ACC eddy activity without significantly altering its overall transport capacity, reinforcing the concept of eddy saturation.

We clarify that the relevant mechanism is not a simple increase in deep-water formation causing AMOC decline. Rather, regional changes in density transformation, freshwater forcing, and pathway shifts can modify where and how deep waters are formed, while the basin-scale overturning weakens. These results show that resolving mesoscale processes is not merely a technical refinement. It can change the simulated response of large-scale circulation and regional climate extremes.

Especially in the Arctic Ocean, eddy activity is set to intensify significantly in a warming world, with far-reaching consequences for the region's climate and ecosystems. Using the high-resolution global sea ice-ocean model FESOM, our study provides a novel projection of how EKE in the upper Arctic Ocean will evolve under a +4 °C warming scenario (Fig. 7).

The findings reveal a 3-fold increase in EKE, driven primarily by enhanced baroclinic instability, which fuels eddy generation. This unprecedented resolution allows us to capture mesoscale eddies, overcoming the limitations of most existing climate models, which cannot resolve these critical small-scale processes. This work represents the first quantitative assessments of Arctic eddy activity under extreme warming, offering new insights into the dynamics of the rapidly changing Arctic Ocean.

A key novelty lies in the interplay between sea ice decline and eddy damping. Despite the significant loss of sea ice, the damping of eddies through interactions with sea ice and winds – commonly referred to as “eddy killing” – is projected to maintain its annual mean effect. This challenges conventional assumptions that reduced sea ice would weaken eddy damping and further accelerate eddy activity. Instead, our findings reveal a complex balance where the processes of eddy generation and dissipation adjust dynamically, with baroclinic instability emerging as the dominant driver of increased EKE.

The implications of this study extend beyond the Arctic. The projected surge in eddy activity will enhance the transport of heat, salt, carbon, and nutrients, fundamentally altering air–sea interactions, ocean circulation, and regional ecosystems. These changes highlight the critical importance of incorporating high-resolution eddy dynamics into climate models to improve predictions of Arctic and global climate responses. By resolving these mesoscale processes, we can better understand and anticipate the cascading impacts of Arctic warming on the Earth system. One of these potential cascading effects is linked to the Atlantic Meridional Overturning Circulation.

Marine heatwaves (MHWs) have intensified in recent decades, and high-resolution climate models show that Arctic MHW variability will significantly strengthen in the 21st century, driven by sea ice retreat. With little ice cover, strong interannual variability could emerge, surpassing previous intensification levels. This increase in MHWs correlates with rapid temperature changes and enhanced stratification, which may disrupt Arctic ecosystems. Close collaboration between climate scientists and biologists (Gou et al., 2025) highlights the potential for MHWs to impact Arctic food webs, both directly through temperature effects and indirectly through nutrient shifts and species changes.

The Atlantic Meridional Overturning Circulation (AMOC) plays a central role in the global climate system in that it transports enormous quantities of heat across latitudes. Global warming is expected to weaken the AMOC, which could lead to serious climatic changes in Europe. Our current program work focuses on key factors driving the AMOC behavior. Analyses combining observations and model simulations suggest that the AMOC has undergone changes in its structure and stability (Dima et al., 2025). Climate models predict a significant slowdown of the circulation system in the future as carbon dioxide concentrations continue to increase and as melting of the Greenland Ice Sheet accelerates.

We show that under the influence of global warming the AMOC shows regional and disproportionately abrupt shifts, contrary to previous results that suggest a long-term and smooth decline (Gou et al., 2024). The high-resolution model study uncovers a startling twist: the AMOC may strengthen in the subarctic Atlantic due to warming, defying the widespread belief that this vital current system is uniformly weakening. Advanced climate models now reveal that, under extreme greenhouse gas emissions (RCP8.5), the AMOC could experience sharp declines in some areas, while paradoxically increasing in the Arctic. This unexpected regional strengthening occurs despite an overall weakening trend in AMOC activity (Gou et al., 2024). This leap in modeling precision exposes critical regional variations, ocean eddies, and tipping points that were previously hidden due to limitations to model dynamics of coarser models. The findings highlight the urgent need to incorporate regional dynamics, as these localized shifts could have profound impacts on climate and on marine ecosystems.

Warming of the North Atlantic region in climate history was often associated with massive melting of the Greenland Ice Sheet. To identify the meltwater's impacts and isolate these from internal variability and other global warming factors, we run single-forcing simulations including small ensembles using three complex climate models differing only in their ocean components (Martin et al., 2022). The ocean translates the step-change perturbation into a multidecadal-to-centennial signature in the deep overturning circulation. In all simulations, internal variability creates notable uncertainty in estimating trends, time of emergence, and duration of the response.

New high-resolution modeling reveals that the AMOC is particularly vulnerable to shifts in salinity in the Irminger Sea off Greenland from increasing Arctic meltwater (Ma et al., 2024). As such, this region plays a critical role in powering AMOC, a conclusion supported by recent observations. Due to a more northward part of the AMOC branch, boundary currents carry warm Atlantic Water along the Norwegian coast toward the Fram Strait and Barents Sea Opening. The heat transport keeps large parts of Barents Sea ice-free in wintertime and represents the most important heat source of the Arctic Ocean. It has long been shown that the strength of Atlantic water inflow into the Barents Sea is enhanced during positive states of the North Atlantic Oscillation (Heukamp et al., 2023). The inflow was largely absent during the period from 1995 to 2005 due to pronounced atmospheric blocking in the North Atlantic sector. This implies that an adequate representation of blocking is necessary in climate models to reliably simulate changes in poleward ocean heat transport and thus Arctic climate.

As for the role of climate change interacting with natural variability in governing the Arctic climate, we have also made considerable progress in deciphering recent drastic changes in Antarctica.

Despite global warming, the extent of Antarctic sea ice has been enormously dynamic but on average has remained surprisingly stable since satellite observations began in 1979. This stood in sharp contrast to existing climate model simulations, which typically predict a significant decline in Antarctic sea ice over the same period – a discrepancy known as the “Antarctic sea-ice paradox”. This paradox has drawn attention to the quality of 21st-century sea ice projections, a concern also emphasized in previous IPCC reports. However, since 2017 Antarctic sea ice extent has dropped sharply and the region may have entered a new stage due to upwelling of warmer waters (e.g., Purich and Doddridge, 2023; Silvano et al., 2025) challenging both previous and present understanding and future predictions of the controls on Antarctic sea ice coverage.

We demonstrated that AWI-CM with the FESOM in a high-resolution, i.e., locally eddy-permitting, ocean configuration simulates a much slower decline in Antarctic sea ice compared to its standard-resolution counterpart (Rackow et al., 2022). This suggests that coarser-resolution CMIP-type models overestimate Southern Ocean sea ice loss. Mesoscale ocean eddies, a critical feature often underrepresented in previous models, play a key role in these differences. Our high-resolution simulations reveal that Antarctic sea ice, particularly in September, could remain stable until the mid-21st century. The more accurate representation of mesoscale processes in high-resolution models enhances heat transport toward the Equator in response to global warming. This mechanism moderates the impacts of anthropogenic warming near Antarctica, effectively delaying the decline in sea ice.

Despite these results, there remains uncertainty about the competing roles of the ocean and atmosphere for the recent Antarctic sea ice changes that may have ended the Antarctic sea ice paradox. We have carried out several Polarstern cruises and remote sensing studies to verify what actual changes the ice has experienced. Results show that snow properties and the formation of superimposed ice, which provide evidence of changes in the surface energy balance in the absence of strong surface melt, have changed little (Arndt et al., 2021; Xu et al., 2024). This provides further evidence for the strong role of oceanic processes in governing Antarctic sea ice variability and change.

The Antarctic sea ice paradox underscores the importance of mesoscale processes in regulating polar climate systems. Eddies could influence the stability of the Antarctic system by mediating ocean–atmosphere exchanges and delaying feedback that could destabilize sea ice and ice sheets. These findings open new pathways for integrating mesoscale dynamics into Earth-system models and for refining predictions of polar climate impacts. Validation of the newly available models needs high-quality data and suitable analytical tools.

In 2023, global temperatures reached nearly 1.5 °C above pre-industrial levels, setting an alarming record (Bevacqua et al., 2025). While best-guess effects from greenhouse gas emissions, aerosol reduction, El Niño, and natural events explain much of the warming, a 0.2 °C gap remains. In our recent study, we propose a significant contributing factor: a sharp decline in planetary albedo, primarily caused by a reduction in low-altitude cloud cover (Fig. 8).

Analyzing satellite and reanalysis data, we found that planetary albedo – the fraction of solar radiation reflected back into space – reached its lowest level in 2023 since at least 1940. While the loss of Arctic and Antarctic ice contributed, it accounts for only 15 %. The key driver appears to be a substantial decrease in low-altitude clouds, including over the North Atlantic which experienced record-high sea surface temperatures. Low clouds play a vital role in cooling the planet by reflecting sunlight, and their decline attenuates this important cooling effect. Without the albedo reduction starting in 2020, global mean temperatures in 2023 would have been approximately 0.23 °C cooler.

The reasons for this cloud decline are multifaceted. Stricter regulations on marine fuel might have played a role in reducing aerosol concentrations, which serve as condensation nuclei for cloud formation. Natural variability may have also contributed. However, global warming itself may also be suppressing low-cloud formation, creating a feedback loop that amplifies warming, which is an important theme in our future studies.

If this feedback continues, future warming could exceed current average projections, pushing global temperatures past the 1.5 °C threshold sooner than anticipated. These findings suggest a relatively high climate sensitivity, highlighting an urgent need to improve the representation of low-cloud feedback mechanisms in climate models. These findings highlight the importance of improving the representation of cloud feedbacks and ocean–cloud–aerosol interactions in climate models, both for constraining climate sensitivity and for assessing future climate risks.

AWI has been at the forefront of developing and applying storylines of extreme events with global climate models, transforming the way we understand and communicate the impacts of climate change. By illustrating how recent extreme events – such as European heatwaves, marine heatwaves (see above), and extreme rainfall – would have unfolded in pre-industrial climates (attribution) and how they might evolve in future warming scenarios (+2, +3, or +4 °C projections), this methodology bridges cutting-edge climate science and actionable insights for society. Storyline analyses have quantified how anthropogenic warming has amplified the severity and frequency of heatwaves, informing risk assessments and heat adaptation measures (Fig. 9). Furthermore, investigations into marine extremes reveal their potential future impacts on ecosystems and economies, highlighting the critical role of ocean variability under a changing climate (Athanase et al., 2024a). Moreover, projections of unprecedented rainfall intensities for hydrological extremes in autumn 2024 under warming scenarios emphasize the urgent need for robust flood mitigation and planning (Athanase et al., 2024b). To this end, this approach can look into the fingerprints of extreme solar and volcanic activities in the climate system and its possible impact on population (Rimbu et al., 2024; Wirtz et al., 2024) which is a subject for future activities in the newly funded InnoPool project SOLVE.

AWI has expanded the reach of these storylines through innovative communication tools. A recently launched online platform (https://climate-storylines.awi.de/, last access: 30 December 2025) enables users worldwide to explore storylines of weather and climate events, fostering greater public understanding of complex climate risks. This prototype was successfully showcased in July 2023 in collaboration with Zeit Online, Germany's leading weekly online newspaper, demonstrating its potential to connect cutting-edge science with diverse audiences.

The storyline methodology is now a key component of the Climate Digital Twin (Bauer et al., 2021) within the EU's Destination Earth initiative, offering global and regional climate services with enhanced granularity and utility. Its inclusion in the new Innovation Pool project ACTUATE (2025–2027) underscores the importance of this approach in the program and its transformative potential for supporting adaptation strategies and increasing resilience to future climate extremes. By making climate impacts tangible, AWI's storylines empower decision-makers and stakeholders with actionable insights, bridging the gap between scientific understanding and societal resilience. This approach not only enhances climate literacy including the concept of climate change attribution but also provides a foundation for more effective and proactive climate adaptation and mitigation strategies globally.

These projections and their implications for extreme events are enabled by advanced methodologies that underpin the research, showcasing innovation and scalability. We highlight here the interdisciplinary and international year-round expedition in the Arctic. Integrating our advanced methodologies with climate insights across timescales underscores the transformative impact of our research.