Early warning indicators (EWIs) of tipping points are typically derived from insights gained from simple dynamical systems. Whether these indicators provide robust and reliable warnings when applied to the climate system remains an open question. In this study, we use climate models with known tipping points to investigate the behavior of classical EWIs associated with increasing memory and variance. In addition, we explore an alternative EWI based on changes in spatial correlations.A key challenge in applying EWIs is determining when a change is significant enough to constitute a warning. How large must a deviation be relative to background variability? How should this background variability be defined—using an early segment of the simulation or an unforced control experiment? How much temporal smoothing should be applied to the indicators? And what is the associated risk of false positives?We address these questions for several tipping elements, including the Atlantic Meridional Overturning Circulation (AMOC), the subpolar gyre region, and sea ice. Our analysis is based on simulations from both the CMIP6 ensemble and the OptimESM/TipESM ensembles.Our results indicate that EWIs are generally sensitive to methodological choices and, in some cases, exhibit significant changes only after the tipping point has occurred.

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.

The NAO’s correlation with precipitation in Norway and the Iberian Peninsula is well established, yet its explanatory power diminishes across much of Europe. Other patterns may drive precipitation variability in these regions, but traditional methods for identifying circulation-precipitation relationships have limitations. Prescribed indices assume a causal link between specific sites and physically coherent structures, whereas EOF methods impose linearity and orthogonality constraints that atmospheric circulation does not obey. This study identifies weather regimes associated with precipitation extremes across representative European regions using a non-linear, non-orthogonal, data-driven approach.Specifically, we employ a machine learning approach that discovers weather regimes directly from mean sea level pressure fields, without prescribing their structure a priori. The method builds upon Spuler et al. 2024 & 2025, with changes that allow for larger, higher-resolution input domains. It identifies distinct atmospheric states associated with different precipitation intensities at target locations, linking discovered patterns directly to their impacts. Importantly, once regimes are identified, indices analogous to traditional teleconnection indices can be derived, enabling comparison with established frameworks while capturing dynamics they may miss.We apply this method to daily ERA5 fields, targeting precipitation in selected European regions with contrasting dynamical drivers, including Bergen, the Iberian Peninsula, and Copenhagen. This allows us to present teleconnection patterns identified through this approach over the entire Northern Hemisphere as well as relevant sub-regions, including the North Atlantic and Arctic, focusing on extreme precipitation drivers. We find multiple regimes that resemble different flavors of the well-known NAO pattern, alongside circulation states consistent with blocking-like structures. Comparisons with traditional EOF analysis highlight the effects of relaxing linearity and orthogonality constraints. Correlation maps are produced for both methods, enabling direct evaluation of how the data-driven regimes compare to established EOF-based patterns.The non-linear, data-driven framework remains physically interpretable and avoids the limitations of linear orthogonal decomposition. Though currently applied to ERA5, the approach transfers directly to CMIP6 historical and scenario runs, enabling assessment of how regime frequencies and precipitation associations may shift under climate change. Overall, this study illustrates how ML-based approaches can complement traditional synoptic climatology by allowing circulation–impact relationships to emerge directly from the data.BibliographySpuler, Fiona R. et al. (2024): Identifying probabilistic weather regimes targeted to a local-scale impact variable. Environmental Data Science, 3, e25.Spuler, Fiona R. et al. (2025): Learning predictable and informative dynamical drivers of extreme precipitation using variational autoencoders. Weather and Climate Dynamics, 6, 995-1014.

The North Atlantic Oscillation (NAO) is a dominant mode of variability in the Northern Hemisphere with strong impacts on temperature, precipitation, and storminess. The predictive skill of the NAO on annual to decadal scales is therefore an important topic, which is often studied using (initialized) climate models. The temporal structure of a time series is closely related to its predictability, and on inter-annual time scales the observed NAO is frequently described to have power at 2–10 years and sometimes with a distinct peak around 8 years. However, the observational record is brief, and such estimates have high uncertainty. Here, we present a thorough study to address the following questions: (1) Is the winter mean NAO distinguishable from white noise? (2) Does the temporal structure of the NAO differ between observations and historical experiments with contemporary climate models (CMIP6)? To this end, we use a range of statistical tools in both the temporal and spectral domains: Power-spectra, wavelet-spectra, autoregressive models, and various well-known time series test statistics. Overall, for both the observed and modelled NAO we find little evidence to reject the null-hypothesis of white noise. For observations, the peak in the power spectrum near 8 years is, taken alone, significant in the period after 1950 but not before. However, considering the complete spectrum, significant peaks will often occur at some frequencies, even for white noise. The large CMIP6 multi-model ensemble is statistically very similar to an ensemble of similar size of white noise. These results suggest limited decadal predictability of the NAO.

Climate models are known to have difficulty in simulating the present day climate in the Arctic. Many studies, including the most recent Inter-governmental Panel on Climate Change Sixth Assessment Report (IPCC AR6), report that, comparing to reanalysis datasets such as ERA5 as reference, climate models participating the past phases of the Coupled Model Intercomparison Project (CMIP) simulate a too cold Arctic. However, recent studies reveal substantial warm biases over sea ice surface in global atmospheric reanalyses due to missing representation of physical processes such as the snow layer on top of the sea-ice. In this work we revisit the so-called long-standing climate model bias in the Arctic by using a new, satellite-derived near surface air temperature (T2m) dataset for the Arctic sea ice region as an alternative reference to the commonly used reanalysis data ERA5. This observational T2m dataset is derived from the satellite based on DMI/CMEMS daily gap-free (so called L4) sea surface temperature and sea ice surface temperature climate data record, spanning from 1st January 1982 to 31st May 2021, covering the Arctic region (> 58 ◦N). We show that, in comparison with the new observational dataset, the ERA5 reanalysis exhibits widespread warm biases exceeding 2℃ over sea ice in the central Arctic, particularly during winter when the warm bias may be as large as 6-10℃. In contrast, the CMIP6 model ensemble demonstrates reasonable performance, with an annual mean bias less than ±1℃ in the same region. We also find that the CMIP6 model mean slightly outperforms the ERA5 in capturing the observed warming trend over the central Arctic region where is fully covered by sea ice with concentration of more than 70%. Outside of this region, it is evident that ERA5 aligns well with observations, while CMIP6 models show large cold bias in the North Atlantic marginal ice zone, consistent with the well-documented results in the past. Our results challenge the current assessment of climate models in the central Arctic, suggesting that relying on existing reanalyses datasets as a reference may underestimate the climate models creditability in the region. It is thus imperative to integrate new observational data for benchmarking climate models in the Arctic region.

Despite tremendous efforts made to improve model performance over the past several decades, climate models exhibit systematic errors or biases in the simulation of many aspects and regions of the climate system. These systematic biases indicate the misrepresentations of physical processes in the models, which can be amplified by feedbacks among different processes and/or climate components. The magnitude of the model biases is often similar to the magnitude of the climate change that have been observed in the past several decades in some regions and for some parameters. This gives rise to large uncertainties in the climate predictions and projections. Furthermore, using climate models with such biases for assessing future climate change implies an assumption that these biases are stationary over time. However, the assumption may not be justified due to the internal variability and the evolving background state of the climate system.In this study, we investigate model biases of key climate variables with the aim of understanding their links with biases of the same or other variables at remote locations. We analyze the multi-model multi-member ensemble (MME) of CMIP6 historical runs. We first focus on the model bias of the Atlantic meridional overturning circulation (AMOC) and its links with remote oceanic biases (e.g., sea surface temperature, sea surface salinity), North Atlantic deep water formation, etc.) and sea ice extent, as well as the atmospheric biases. We assess to what extent these linkages may affect the AMOC change. We further explore the representations of the circulation modes (e.g., North Atlantic Oscillation) in the CMIP6 MME relative to the observations, with emphasis on understanding how the internal variability influences the representation of the circulation modes and their relationship with other climate variables, and how these relationships in turn impact the climate predictability.

Predicting the winter climate over Eurasia is challenging as both the initialized predictions and uninitialized climate projections show limited skill in reproducing observed variability on multi-annual to decadal timescales. This study presents a novel approach to constrain variability in projection simulations over Eurasia by exploiting the spatial patterns of teleconnection between the North Atlantic Oscillation (NAO) and the near-surface air temperature (SAT). The constraint evaluates simulated and observed NAO-temperature teleconnection patterns during 20 year windows prior to making predictions. The resulting top ranking members are used to make predictions for the next 10 years. We find here that the constrained ensemble is skillful in predicting winter SAT on multi-annual to decadal timescales over Eurasia. We also find that the constrained multi-annual predictions yield higher correlation skill compared to both the unconstrained ensemble and the initialized predictions, especially up to 5 year mean winter SAT predictions. Therefore, we argue that for the Eurasia region, the constraint could provide skillful winter SAT predictions on multi-annual timescales with minimal cost compared to the state of the art initialized decadal climate predictions for which added value from initialization wanes quickly after the first forecast year.

We present the first results of the TIPMIP ESM experiments using the post-CMIP6 model EC-Earth3-ESM. The main objective of the TIPMIP ESM is to study the risks and consequences of potential tipping events in the Earth system, as well as the potential reversibility of triggered events - as a function of e.g., magnitudes and durations of different global warming levels (GWLs) before cooling the climate back to lower GWLs and pre-industrial climate.Following the TIPMIP ESM protocol, we performed a set of idealized emission-driven simulations, including 1) ramp-up runs with a constant CO2 emission that drives a global mean surface air temperature (GMSAT) warming rate of 0.2 K/decade; 2) stabilization runs with zero CO2 emission at multiple GWLs; and 3) ramp-down runs to the pre-industrial climate with a negative CO2 emission (same magnitude with the ramp-up run) branched after 50 years of stabilization runs.We present and discuss the simulated responses in the large-scale features of the different components of the earth system in the ramp-up, stabilization, and first test runs of the ramp-down experiments, with a focus on, e.g., GMSAT, AMOC, sea ice, carbon pools/fluxes, and Greenland Ice Sheet. Our findings suggest that some fast climate system components are reversible, e.g., sea ice, but Arctic summer sea ice can show some delays in recovering at high GWLs. AMOC linearly declines during the ramp-up phase, stabilizes during the stabilization phase, and recovers during the ramp-down phase (with an overshoot). Melting in the Greenland Ice Sheet accelerates during the ramp-up phase, its mass loss continues with somewhat slower speeds during stabilization and hardly reverses during the ramp-down phase when branched at high GWLs.

We review how the international modelling community, encompassing integrated assessment models, global and regional Earth system and climate models, and impact models, has worked together over the past few decades to advance understanding of Earth system change and its impacts on society and the environment and thereby support international climate policy. We go on to recommend a number of priority research areas for the coming decade, a timescale that encompasses a number of newly starting international modelling activities, as well as the IPCC Seventh Assessment Report (AR7) and the second UNFCCC Global Stocktake. Progress in these priority areas will significantly advance our understanding of Earth system change and its impacts, increasing the quality and utility of science support to climate policy.We emphasize the need for continued improvement in our understanding of, and ability to simulate, the coupled Earth system and the impacts of Earth system change. There is an urgent need to investigate plausible pathways and emission scenarios that realize the Paris climate targets - for example, pathways that overshoot 1.5 or 2 degrees C global warming, before returning to these levels at some later date. Earth system models need to be capable of thoroughly assessing such warming overshoots - in particular, the efficacy of mitigation measures, such as negative CO2 emissions, in reducing atmospheric CO2 and driving global cooling. An improved assessment of the long-term consequences of stabilizing climate at 1.5 or 2 degrees C above pre-industrial temperatures is also required. We recommend Earth system models run overshoot scenarios in CO2-emission mode to more fully represent coupled climate-carbon-cycle feedbacks and, wherever possible, interactively simulate other key Earth system phenomena at risk of rapid change during overshoot. Regional downscaling and impact models should use forcing data from these simulations, so impact and regional climate projections cover a more complete range of potential responses to a warming overshoot. An accurate simulation of the observed, historical record remains a fundamental requirement of models, as does accurate simulation of key metrics, such as the effective climate sensitivity and the transient climate response to cumulative carbon emissions. For adaptation, a key demand is improved guidance on potential changes in climate extremes and the modes of variability these extremes develop within. Such improvements will most likely be realized through a combination of increased model resolution, improvement of key model parameterizations, and enhanced representation of important Earth system processes, combined with targeted use of new artificial intelligence (AI) and machine learning (ML) techniques. We propose a deeper collaboration across such efforts over the coming decade.With respect to sampling future uncertainty, increased collaboration between approaches that emphasize large model ensembles and those focussed on statistical emulation is required. We recommend an increased focus on high-impact-low-likelihood (HILL) outcomes - in particular, the risk and consequences of exceeding critical tipping points during a warming overshoot and the potential impacts arising from this. For a comprehensive assessment of the impacts of Earth system change, including impacts arising directly as a result of climate mitigation actions, it is important that spatially detailed, disaggregated information used to generate future scenarios in integrated assessment models be available for use in impact models. Conversely, there is a need to develop methods that enable potential societal responses to projected Earth system change to be incorporated into scenario development.The new models, simulations, data, and scientific advances proposed in this article will not be possible without long-term development and maintenance of a robust, globally connected infrastructure ecosystem. This system must be easily accessible and useable by modelling communities across the world, allowing the global research community to be fully engaged in developing and delivering new scientific knowledge to support international climate policy.

We investigate the role of the subpolar North Atlantic (SPNA) for downstream predictability, using two decadal climate prediction systems. We use the subpolar extreme cold and fresh anomaly event developing in winter 2013/2014 as initial conditions and evaluate ensemble predictions of the two systems in the following decade. In addition, we perform ensemble pacemaker experiments where the models are forced toward observed ocean temperature and salinity anomalies in the SPNA from November 2014 through December 2019. The pacemaker experiments show improved skill along the Atlantic Water pathway, compared with the standard decadal predictions, and we therefore conclude that the correct description of the ocean in the SPNA is the key. The enhanced skill is most prominent in subsurface salinity in the form of propagating anomalies. Observations show that ocean anomalies propagate across the North Atlantic and further north along the Norwegian coast. Anomalous ocean temperatures can modify heat exchange between ocean and atmosphere and influence the circulation of the atmosphere. Since such anomalies can persist for months, it potentially opens a door to seasonal prediction of the atmosphere. However, climate models used for seasonal and decadal prediction fail to predict these propagating anomalies. In our study we aim to find causes for this lack of predictability. We use the record low temperatures observed in the North Atlantic in 2015 as a test case. We employ two climate prediction systems and make two types of predictions: one type is a standard prediction initialized with winter 2014/2015 ocean conditions; the other type is initialized in the same way and in addition, the model is kept close to observations in the subpolar gyre during the whole prediction, in the region where the cold blob was formed. Comparing the two types of predictions, we see that keeping the model close to observations in the subpolar gyre increases the predictability along the Norwegian coast. This points to the subpolar gyre as a key area where models need to be improved. Knowing temperature and salinity in the subpolar gyre increases skill downstream along the Atlantic Water pathway Salinity anomalies propagate northward, while temperature anomalies are more difficult to trace A pacemaker experiment has been used for decadal predictions for the first time

The IPCC’s 2021 assessment suggested that substantial emissions reduction and limiting global temperature rise to well below 2.0°C could prevent the complete loss of Arctic sea ice in this century. However, these assessments come with large uncertainties. Recent research projects a summer ice-free Arctic by the 2050s even under a low emission scenario by constraining future sea ice area with satellite-derived sea ice concentration (SIC) since 1979. Notably, the climate models in these assessments commonly underestimate the accelerated Arctic warming and the pace of sea ice melting, particularly over the last two decades. Moreover, recent studies indicate that in a warming climate, the thinning of sea ice and snow over sea ice may intensify surface warming, thereby accelerating the melting. In this study, we leverage the increasing availability of observations and recent reanalysis data for Arctic-wide sea ice to investigate the link between changes in sea ice thickness (SIT), sea ice concentration (SIC), and Arctic warming. We employ these datasets to evaluate biases in historical periods and uncertainties in future scenarios within the CMIP6 multi-model ensemble for SIT and SIC. We further investigate the relationship between the thinning of sea ice and the snow layer on sea ice and surface temperature changes on a basin or regional scale. The findings are then used to constrain projected Arctic changes. Our study aims to gain some insights into the impact of model biases in the Arctic on projected climate projections, crucial for decision-making in a changing climate.

Many climate models simulate near-surface air temperatures that are too low in the Arctic compared to the observation-based ERA5 reanalysis data, a bias that was noted in the Inter-governmental Panel on Climate Change Sixth Assessment Report (IPCC AR6). Here we present a high-resolution, satellite-derived dataset of near-surface air temperatures for the Arctic sea-ice region (1982–2020). We use it as a benchmark to reevaluate climate reanalyses and model simulations in CMIP6 (Coupled Model Intercomparison Project 6). We find that the CMIP6 simulations in the central Arctic, with generally thicker ice and snow, align well with satellite observations, with an annual mean bias of less than ± 1 °C over sea ice. By contrast, climate reanalyses like ERA5 exhibit widespread warm biases exceeding 2 °C in the same region. We conclude that reliance on ERA5 reanalysis as a reference may have led to an underestimation of climate model reliability in the Arctic region.

Abstract. In this study, we address a persistent positive bias in Arctic sea ice (concentration and thickness) in the global climate model EC-Earth3 (ECE3) by including a modulating factor to the surface sensible heat flux over regions with sea ice concentrations above 70 %, so-called ECE3L. We performed two pairs of 50-year simulations with repeated seasonal cycles: one pair replicating a cold climate and the other a warmer climate, with the latter characterised by thinner ice and weaker atmospheric boundary layer stability during winter. We show that modified heat flux can significantly alter surface air temperatures in the Arctic, with no substantial impact on lower latitudes. The changes are more pronounced in the cold climate, particularly during Arctic winter. We extended our comparison to two CMIP6 historical ensembles in a transient climate (1980–2014). We found that the mean sea ice states in the changing climate for the ECE3 (ECE3L) ensemble mean closely resembled the mean states in the cold-climate experiment. However, the reduction in sea ice area and volume achieved by ECE3L was nearly four times greater in the cold climate experiment than in the transient climate, reflecting the diminishing role of sea ice leads in a changing climate with decreasing occurrences of stable stratification in winter. Finally, our comparisons with satellite observations and reanalysis datasets demonstrated that ECEL3 significantly improves the local amplification ratio in the marginal ice zone of the Arctic, underscoring the importance of atmospheric stability shaped by central Arctic pack ice and its impact on Arctic amplification.

The acceleration of Greenland ice sheet melting over the past decades is raising concern regarding the impacts on ocean circulation in the North Atlantic and Arctic regions. Global climate models struggle to assess these impacts as they do not include a realistic amount of meltwater from the Greenland ice sheet. Using an extended observation-based dataset of runoff and solid ice discharge for the recent historical period (1920–2019), we force the EC-Earth3 climate model following the ensemble approach and protocol of a previous study using a different model. We observe a slight increase of the ensemble mean AMOC with a large spread in the response: 0.20± 0.81 Sv for the maximum AMOC at 45 ^∘ N. We notice that members with a strong initial AMOC state (18 Sv) show a strengthening of the AMOC, while members with intermediate AMOC strength between 16 and 18 Sv are not affected by the freshwater forcing. Weaker initial AMOC members respond with a mean weakening of - 0.21± 0.58 Sv of the AMOC. The AMOC ensemble spread is reduced by half in the freshwater forced ensemble at the end of the experiment and the negative trend is stronger compared to historical simulations without the forcing. This suggests the possible ability of the freshwater to constrain AMOC variability on multi-decadal time-scales. Comparison of the two ten member ensembles with ORAS5 reanalysis shows a reduction of the surface temperature and salinity biases in the freshwater forced ensemble in key regions such as the North Atlantic subpolar gyre and the Beaufort Gyre relative to the historical simulations. Recent trends are also more aligned with reanalysis in those regions. Arctic Ocean Atlantic Water subsurface core temperature is also closer to reanalysis but still strongly biassed. Members with a high AMOC initial state display an unrealistic Atlantic water layer core depth.

This paper celebrates Professor Yongqi GAO’s significant achievement in the field of interdisciplinary studies within the context of his final research project Arctic Climate Predictions: Pathways to Resilient Sustainable Societies - ARCPATH ( https://www.svs.is/en/projects/finished-projects/arcpath ). The disciplines represented in the project are related to climatology, anthropology, marine biology, economics, and the broad spectrum of social-ecological studies. Team members were drawn from the Nordic countries, Russia, China, the United States, and Canada. The project was transdisciplinary as well as interdisciplinary as it included collaboration with local knowledge holders. ARCPATH made significant contributions to Arctic research through an improved understanding of the mechanisms that drive climate variability in the Arctic. In tandem with this research, a combination of historical investigations and social, economic, and marine biological fieldwork was carried out for the project study areas of Iceland, Greenland, Norway, and the surrounding seas, with a focus on the joint use of ocean and sea-ice data as well as social-ecological drivers. ARCPATH was able to provide an improved framework for predicting the near-term variation of Arctic climate on spatial scales relevant to society, as well as evaluating possible related changes in socioeconomic realms. In summary, through the integration of information from several different disciplines and research approaches, ARCPATH served to create new and valuable knowledge on crucial issues, thus providing new pathways to action for Arctic communities.

A considerable part of the skill in decadal forecasts often come from the forcings which are present in both initialized and un-initialized model experiments. This makes the added value from initialization difficult to assess. We investigate statistical tests to quantify if initialized forecasts provide skill over the un-initialized experiments. We consider three correlation based statistics previous used in the literature. The distributions of these statistics under the null-hypothesis that initialization has no added values are calculated by a surrogate data method. We present some simple examples and study the statistical power of the tests. We find that there can be large differences in both the values and the power for the different statistics. In general the simple statistic defined as the difference between the skill of the initialized and uninitialized experiments behaves best. However, for all statistics the risk of rejecting the true null-hypothesis is too high compared to the nominal value.We compare the three tests on initialized decadal predictions (hindcasts) of near-surface temperature performed with a climate model and find evidence for a significant effect of initializations for small lead-times. In contrast, we find only little evidence for a significant effect of initializations for lead-times larger than 3 years when the experience from the simple experiments is included in the estimation.

Global climate models (CMIP6 models) are the basis for future predictions and projections, but these models typically have large biases in their mean state of the Arctic Ocean. Considering a transect across the Arctic Ocean, with a focus on the depths between 100-700m, we show that the model spread for temperature and salinity anomalies increases significantly during the period 2025-2045. The maximum model spread is reached in the period 2045-2055 with a standard deviation 10 times higher than in 1993-2010. The CMIP6 models agree that there will be warming, but do not agree on the degree of warming. This aspect is important for long-term management of societal and ecological perspectives in the Arctic region. We therefore test a new approach to find models with good performance. We assess how CMIP6 models represent the horizontal patterns of temperature and salinity in the period 1993-2010. Based on this, we find four models with relatively good performance (MPI-ESM1-2-HR, IPSL-CM6A-LR, CESM2-WACCM, MRI-ESM2-0). For a more robust model evaluation, we consider additional metrics (e.g., climate sensitivity, ocean heat transport) and also compare our results with other recent CMIP6 studies in the Arctic Ocean. Based on this, we find that two of the models have an overall better performance (MPI-ESM1-2-HR, IPSL-CM6A-LR). Considering projected changes for temperature for the period 2045-2055 in the high end ssp585 scenario, these two models show a similar warming in the Mid Layer (300-700m; 1.1-1.5°C). However, in the low end ssp126 scenario, IPSL-CM6A-LR shows a considerably higher warming than MPI-ESM1-2-HR. In contrast to the projected warming by both models, the projected salinity changes for the period 2045-2055 are very different; MPI-ESM1-2-HR shows a freshening in the Upper Layer (100-300m), whereas IPSL-CM6A-LR shows a salinification in this layer. This is the case for both scenarios. The source of the model spread appears to be in the Eurasian Basin, where warm waters enter the Arctic. Finally, we recommend being cautious when using the CMIP6 ensemble to assess the future Arctic Ocean, because of the large spread both in performance and the extent of future changes.

Recent studies have suggested that the Atlantic water pathway connecting the subpolar North Atlantic (SPNA) with the Nordic Seas and Arctic Ocean may lead to skillful predictions of sea surface temperature and salinity anomalies in the eastern Nordic Seas. To investigate the role of the SPNA for such anomalies downstream, we designed a pacemaker experiment, using two decadal climate prediction systems based on EC-Earth3 and NorCPM. We focus on the subpolar extreme cold anomaly in 2015 and its subsequent development, a feature not well captured and predicted. The pacemaker experiment follows the protocol of the CMIP6 DCPP-A retrospective forecasts or hindcasts initialized November 1, 2014, but the models are forced to follow the observed ocean temperature and salinity anomalies in the SPNA from ocean reanalysis from November 2014 through to December 2019. Two sets of 10-year hindcasts are performed with 10 members for EC-Earth3 and 30 members for NorCPM. We here detail and discuss the design of this pacemaker experiment and present results, comparing with the initialized CMIP6 DCPP-A experiment assessing differences in decadal prediction skill outside the SPNA. We conclude that the pacemaker experiments show improved skill compared to the standard decadal predictions for the eastern Norwegian Sea, and therefore the SPNA is key for successful decadal predictions in the region.

Due to large northward heat transport, the Atlantic meridional overturning circulation (AMOC) strongly affects the climate of various regions. Its internal variability has been shown to be predictable decades ahead within climate models, providing the hope that synchronizing ocean circulation with observations can improve decadal predictions, notably of the North Atlantic subpolar gyre (SPG). Climate predictions require a starting point which is a reconstruction of the past climate. This is usually performed with data assimilation methods that blend available observations and climate model states together. There is no unique method to derive the initial conditions. Moreover, this can be performed using full-field observations or their anomalies superimposed on the model's climatology to avoid strong drifts in predictions. How critical ocean circulation drifts are for prediction skill has not been assessed yet. We analyze this possible connection using the dataset of 12 decadal prediction systems from the World Meteorological Organization Lead Centre for Annual-to-Decadal Climate Prediction. We find a variety of initial AMOC errors within the predictions related to a dynamically imbalanced ocean states leading to strongly displaced or multiple maxima in the overturning structures. This likely results in a blend of what is known as model drift and initial shock. We identify that the AMOC initialization influences the quality of the SPG predictions. When predictions show a large initial error in their AMOC, they usually have low skill for predicting internal variability of the SPG for a time horizon of 6-10 years. Full-field initialized predictions with low AMOC drift show better SPG skill than those with a large AMOC drift. Nevertheless, while the anomaly-initialized predictions do not experience large drifts, they show low SPG skill when skill also present in historical runs is removed using a residual correlation metric. Thus, reducing initial shock and model biases for the ocean circulation in prediction systems might help to improve their prediction for the SPG beyond 5 years. Climate predictions could also benefit from quality-check procedure for assimilation/initialization because currently the research groups only reveal the problems in initialization once the set of predictions has been completed, which is an expensive effort.

Abstract. The main drivers of the continental Northern Hemisphere snow cover are investigated in the 1979–2014 period. Four observational datasets are used as are two large multi-model ensembles of atmosphere-only simulations with prescribed sea surface temperature (SST) and sea ice concentration (SIC). A first ensemble uses observed interannually varying SST and SIC conditions for 1979–2014, while a second ensemble is identical except for SIC with a repeated climatological cycle used. SST and external forcing typically explain 10 % to 25 % of the snow cover variance in model simulations, with a dominant forcing from the tropical and North Pacific SST during this period. In terms of the climate influence of the snow cover anomalies, both observations and models show no robust links between the November and April snow cover variability and the atmospheric circulation 1 month later. On the other hand, the first mode of Eurasian snow cover variability in January, with more extended snow over western Eurasia, is found to precede an atmospheric circulation pattern by 1 month, similar to a negative Arctic oscillation (AO). A decomposition of the variability in the model simulations shows that this relationship is mainly due to internal climate variability. Detailed outputs from one of the models indicate that the western Eurasia snow cover anomalies are preceded by a negative AO phase accompanied by a Ural blocking pattern and a stratospheric polar vortex weakening. The link between the AO and the snow cover variability is strongly related to the concomitant role of the stratospheric polar vortex, with the Eurasian snow cover acting as a positive feedback for the AO variability in winter. No robust influence of the SIC variability is found, as the sea ice loss in these simulations only drives an insignificant fraction of the snow cover anomalies, with few agreements among models.