Explore the vast range of titles available.

Anthropogenic warming has led to an unprecedented year-round reduction in Arctic sea ice extent. This has far-reaching consequences for indigenous and local communities, polar ecosystems, and global climate, motivating the need for accurate seasonal sea ice forecasts. While physics-based dynamical models can successfully forecast sea ice concentration several weeks ahead, they struggle to outperform simple statistical benchmarks at longer lead times. We present a probabilistic, deep learning sea ice forecasting system, IceNet. The system has been trained on climate simulations and observational data to forecast the next 6 months of monthly-averaged sea ice concentration maps. We show that IceNet advances the range of accurate sea ice forecasts, outperforming a state-of-the-art dynamical model in seasonal forecasts of summer sea ice, particularly for extreme sea ice events. This step-change in sea ice forecasting ability brings us closer to conservation tools that mitigate risks associated with rapid sea ice loss. Accurate seasonal forecasts of sea ice are highly valuable, particularly in the context of sea ice loss due to global warming. A new machine learning tool for sea ice forecasting offers a substantial increase in accuracy over current physics-based dynamical model predictions.

The ECMWF OCEAN5 system is a global ocean and sea-ice ensemble of reanalysis and real-time analysis. This paper gives a full description of the OCEAN5 system, with the focus on upgrades of system components with respect to its predecessors, ORAS4 and ORAP5. An important novelty in OCEAN5 is the ensemble generation strategy that includes perturbation of initial conditions and a generic perturbation scheme for observations and forcing fields. Other upgrades include revisions to the a priori bias correction scheme, observation quality control and assimilation method for sea-level anomalies. The OCEAN5 historical reconstruction of the ocean and sea-ice state is the ORAS5 reanalysis, which includes five ensemble members and covers the period from 1979 onwards. Updated versions of observation data sets are used in ORAS5 production, with special attention devoted to the consistency of sea surface temperature (SST) and sea-ice observations. Assessment of ORAS5 through sensitivity experiments suggests that all system components contribute to an improved fit to observation in reanalyses, with the most prominent contribution from direct assimilation of ocean in situ observations. Results of observing system experiments further suggest that the Argo float is the most influential observation type in our data assimilation system. Assessment of ORAS5 has also been carried out for several key ocean state variables and verified against reference climate data sets from the ESA CCI (European Space Agency Climate Change Initiative) project. With respect to ORAS4, ORAS5 has improved ocean climate state and variability in terms of SST and sea level, mostly due to increased model resolution and updates in assimilated observation data sets. In spite of the improvements, ORAS5 still underestimates the temporal variance of sea level and continues exhibiting large SST biases in the Gulf Stream and its extension regions which are possibly associated with misrepresentation of front positions. Overall, the SST and sea-ice uncertainties estimated using five ORAS5 ensemble members have spatial patterns consistent with those of analysis error. The ensemble spread of sea ice is commensurable with the sea-ice analysis error. On the contrary, the ensemble spread is under-dispersive for SST.

The two main large-scale features of Arctic sea-ice drift are the Beaufort Gyre and the Transpolar Drift Stream. They exhibit strong intraseasonal and interannual variability. Winter 2016/17 showed increased cyclone activity, leading to the collapse of the Beaufort Sea high and the reversal of the Beaufort Gyre. Winter 2020/21 displayed decreased cyclone activity and intense anticyclonic ice transport in the Beaufort Gyre. Here we show that the European Centre for Medium-Range Weather Forecasts’s (ECMWF) extended-range (46 days) retrospective forecasts were able to predict the ice motion during these cases. The initial contrasts in sea level pressure, surface winds and ice drift were well captured, and their temporal evolution—including the reversal of the usual drift direction—well reproduced by the forecasts initialized about a week before the event. Sea-ice thickness in the forecast exhibited initial errors even greater than 1 m that persisted throughout the forecast and negatively affected the ice speed forecast. Despite these shortcomings, the dynamic forecast outperformed the persistence and climatology forecast and represented the observed relation between surface winds and ice drift well. The benefit of dynamic forecasts is especially clear in cases that differ from climatology, like the one we focus on.

In this paper we describe SEAS5, ECMWF's fifth generation seasonal forecast system, which became operational in November 2017. Compared to its predecessor, System 4, SEAS5 is a substantially changed forecast system. It includes upgraded versions of the atmosphere and ocean models at higher resolutions, and adds a prognostic sea-ice model. Here, we describe the configuration of SEAS5 and summarise the most noticeable results from a set of diagnostics including biases, variability, teleconnections and forecast skill.An important improvement in SEAS5 is the reduction of the equatorial Pacific cold tongue bias, which is accompanied by a more realistic El Niño amplitude and an improvement in El Niño prediction skill over the central-west Pacific. Improvements in 2 m temperature skill are also clear over the tropical Pacific. Sea-surface temperature (SST) biases in the northern extratropics change due to increased ocean resolution, especially in regions associated with western boundary currents. The increased ocean resolution exposes a new problem in the northwest Atlantic, where SEAS5 fails to capture decadal variability of the North Atlantic subpolar gyre, resulting in a degradation of DJF 2 m temperature prediction skill in this region. The prognostic sea-ice model improves seasonal predictions of sea-ice cover, although some regions and seasons suffer from biases introduced by employing a fully dynamical model rather than the simple, empirical scheme used in System 4. There are also improvements in 2 m temperature skill in the vicinity of the Arctic sea-ice edge. Cold temperature biases in the troposphere improve, but increase at the tropopause. Biases in the extratropical jets are larger than in System 4: extratropical jets are too strong, and displaced northwards in JJA. In summary, development and added complexity since System 4 has ensured that SEAS5 is a state-of-the-art seasonal forecast system which continues to display a particular strength in the El Niño Southern Oscillation (ENSO) prediction.

This study combines state-of-the-art reanalyses such as the fifth-generation European Re-Analysis (ERA5) and the Ocean Reanalysis System 5 (ORAS5) with novel observational products to present an updated estimate of the coupled atmosphere–ocean–sea ice Arctic energy budget, including flux and storage terms covering 2001–17. Observational products provide independent estimates of crucial budget terms, including oceanic heat transport from unique mooring-derived data, radiative fluxes from satellites, and sea ice volume from merged satellite data. Results show that the time averages of independent estimates of radiative, atmospheric, and oceanic energy fluxes into the Arctic Ocean domain are remarkably consistent in the sense that their sum closely matches the observed rate of regional long-term oceanic heat accumulation of ~1 W m−2. Atmospheric and oceanic heat transports are found to be stronger compared to earlier assessments (~100 and ~16 W m−2, respectively). Data inconsistencies are larger when considering the mean annual cycle of the coupled energy budget, with RMS values of the monthly budget residual between 7 and 15 W m−2, depending on the employed datasets. This nevertheless represents an average reduction of ~72% of the residual compared to earlier work and demonstrates the progress made in data quality and diagnostic techniques. Finally, the budget residual is eliminated using a variational approach to provide a best estimate of the mean annual cycle. The largest remaining sources of uncertainty are ocean heat content and latent heat associated with sea ice melt and freeze, which both suffer from the lack of observational constraints. More ocean in situ observations and reliable sea ice thickness observations and their routinely assimilation into reanalyses are needed to further reduce uncertainty.

The Atlantic Meridional Overturning Circulation (AMOC) is a main driver for predictability at decadal time scales, but has been largely ignored in the context of seasonal forecasts. Here, we show compelling evidence that AMOC initialization can have a direct and strong impact on seasonal forecasts. Winter reforecasts with SEAS5, the current operational seasonal forecasting system by the European Centre for Medium-Range Weather Forecasts, exhibit errors of sea-surface temperature (SST) in the western part of the North Atlantic Subpolar Gyre that are strongly correlated with decadal variations in the AMOC initial conditions. In the early reforecast period 1981–1996, too warm SST coincide with an overly strong AMOC transporting excessive heat into the region. In the ocean reanalyses providing the forecast initial conditions, excessive heat transport is balanced by additional surface cooling from relaxing towards observed SST, and therefore the fit to observations is acceptable. However, the additional surface cooling contributes to enhanced deep convection and strengthens the AMOC, thereby establishing a feedback loop. In the forecasts, where the SST relaxation is absent, the balance is disrupted, and fast growth of SST errors ensues. The warm SST bias has a strong local impact on surface air temperature, mean sea-level pressure, and precipitation patterns, but remote impact is small. In the late reforecast period 2001–2016, neither the SST in the western North Atlantic nor the AMOC show large biases. The non-stationarity of the bias prevents an effective forecast calibration and causes an apparent loss of skill in the affected region. The case presented here demonstrates the importance of correctly initializing slowly varying aspects of the Earth System such as the AMOC in order to improve forecasts on seasonal and shorter time scales.

Global and regional ocean and sea ice reanalysis products (ORAs) are increasingly used in polar research, but their quality remains to be systematically assessed. To address this, the Polar ORA Intercomparison Project (Polar ORA-IP) has been established following on from the ORA-IP project. Several aspects of ten selected ORAs in the Arctic and Antarctic were addressed by concentrating on comparing their mean states in terms of snow, sea ice, ocean transports and hydrography. Most polar diagnostics were carried out for the first time in such an extensive set of ORAs. For the multi-ORA mean state, we found that deviations from observations were typically smaller than individual ORA anomalies, often attributed to offsetting biases of individual ORAs. The ORA ensemble mean therefore appears to be a useful product and while knowing its main deficiencies and recognising its restrictions, it can be used to gain useful information on the physical state of the polar marine environment.

Satellite observations show a small overall increase in Antarctic sea ice extent (SIE) over the period 1979–2015. However, this upward trend needs to be balanced against recent pronounced SIE fluctuations occurring there. In the space of 3 years, the SIE sank from its highest value ever reached in September 2014 to record low in February 2017. In this work, a set of six state-of-the-art global climate models is used to evaluate the potential predictability of the Antarctic sea ice at such timescales. This first multi-model study of Antarctic sea ice predictability reveals that the ice edge location can potentially be predicted up to 3 years in advance. However, the ice edge location predictability shows contrasted seasonal performances, with high predictability in winter and no predictability in summer. The reemergence of the predictability from one winter to next is provided by the ocean through its large thermal inertia. Sea surface heat anomalies are stored at depth at the end of the winter and influences the sea ice advance the following year as they resurface. The effectiveness of this mechanism across models is found to depend upon the depth of the mixed layer. One should be very cautious about these potential predictability estimates as there is evidence that the Antarctic sea ice predictability is promoted by deep Southern Ocean convection. We therefore suspect models with excessive convection to show higher sea ice potential predictability results due to an incorrect representation of the Southern Ocean.

We discuss the state of Arctic sea ice in the global eddy-permitting ocean reanalysis Ocean ReAnalysis Pilot 5 (ORAP5). Among other innovations, ORAP5 now assimilates observations of sea ice concentration using a univariate 3DVar-FGAT scheme. We focus on the period 1993–2012 and emphasize the evaluation of model performance with respect to recent observations of sea ice thickness. We find that sea ice concentration in ORAP5 is close to assimilated observations, with root mean square analysis residuals of less than 5 % in most regions. However, larger discrepancies exist for the Labrador Sea and east of Greenland during winter owing to biases in the free-running model. Sea ice thickness is evaluated against three different observational data sets that have sufficient spatial and temporal coverage: ICESat, IceBridge and SMOSIce. Large-scale features like the gradient between the thickest ice in the Canadian Arctic and thinner ice in the Siberian Arctic are simulated well by ORAP5. However, some biases remain. Of special note is the model’s tendency to accumulate too thick ice in the Beaufort Gyre. The root mean square error of ORAP5 sea ice thickness with respect to ICESat observations is 1.0 m, which is on par with the well-established PIOMAS model sea ice reconstruction. Interannual variability and trend of sea ice volume in ORAP5 also compare well with PIOMAS and ICESat estimates. We conclude that, notwithstanding a relatively simple sea ice data assimilation scheme, the overall state of Arctic sea ice in ORAP5 is in good agreement with observations and will provide useful initial conditions for predictions.

We assess for the first time the impact of in-situ ocean observations on European Centre for Medium-Range Weather Forecasts (ECMWF) sub-seasonal forecasts of both ocean and atmospheric conditions. A series of coupled reforecasts have been conducted for the period 1993-2015, in which different sets of ocean observations were withdrawn in the production of the ocean initial conditions. Removal of all ocean in-situ observations in the initial conditions leads to significant degradation in the forecasts of ocean surface and subsurface mean state at lead times from week 1 to week 4. The negative impact is predominantly caused by the removal of the Argo observing system in recent decades. Changes in the mean state of atmospheric variables are comparatively small but significant in the forecasts of lower and upper atmospheric circulation over large regions. Our results highlight the value of continuous, real-time in-situ observations of the surface and subsurface ocean for coupled forecasts in the sub-seasonal range.