Sea ice and snow volume are essential variables for polar predictions, but operational systems still struggle to accurately capture their evolution. Satellite measurements now provide estimates of sea ice freeboard and snow depth. The combined assimilation of sea ice concentration (SIC), along-track altimetry radar freeboard data from Cryosat-2 and observations of snow depth from Cryosat-2 and SARAL is implemented in a multivariate approach in a global 1/4 degrees ocean/sea ice coupled NEMO4.2/SI3 model. A multivariate experiment, performed on two full seasonal cycles 2017-2018, is compared to a free (no assimilation) and a SIC-only assimilation simulations. The multivariate technique increases the sea ice volume, even in the absence of freeboard and snow measurements during summer, and rapidly changes the spatial patterns of ice and snow thicknesses in both hemispheres, in accordance with the assimilated observations. The sea ice volume from the multivariate approach compares better with independent (not assimilated) estimates from ICESat-2 and CS2SMOS or SMOS in both hemispheres. The multivariate system performs better in the Arctic than in Antarctica where the ice and ocean separate analyses are not designed to handle properly the strong interactions between upper oceanic layers and sea ice cover in the Southern Ocean. These results also confirm the importance of using variable snow and ice densities in a freeboard assimilation context. This study shows promising results for enhancing the capacity of assimilation systems to monitor the volume of sea ice and snow and paves the way for future satellite missions.

Since October 2016, and in the framework of Copernicus Marine Service, Mercator Ocean International delivered in real-time daily services (weekly analyses and daily 10-day forecasts) with a global 1/12° high resolution (eddy-resolving) system. In this system, oceanic observations were assimilated in the model using a reduced-order Kalman filter method. Along track altimeter Sea Level Anomaly (SLA), satellite sea surface temperature (SST) and sea ice concentration, and in situ temperature and salinity vertical profiles were jointly assimilated to estimate the initial conditions for numerical ocean forecasting. A 3D-VAR scheme was also used to better control the slowly evolving large-scale biases in temperature and salinity. A major release of this analysis and forecasting system is available since November 2022 with the following main changes and updates: * A new version of NEMO ocean and sea ice models (new numerical schemes, coherent bulk formulation with the atmospheric forcing, multi-categories sea ice model); * Higher spatial and temporal resolution (1/10° - 1 hour) atmospheric forcing from IFS ECMWF analyses and forecasts; * A new assimilated SST observation (assimilation of L3 ODYSSEA SST high resolution product instead of L4 OSTIA gridded product); * A new Mean Dynamic Topography for SLA assimilation; * A different parametrization of the model error covariance with a new anomalies base deduced from the Mercator Ocean reanalysis at 1/12°; * A 4D extension of the data assimilation scheme allowing a better spatiotemporal continuity of mesoscale structures; * The assimilation of “super-observations” to filter out noisy data and scales that the model does not resolve; * The use of satellite-based monthly estimates of the Global Mean Sea Level to better constrain the ocean mass and the steric height. This presentation shows how some identified weaknesses present in the previous system have been improved. It also highlights the new system’s performance in terms of analysis and forecast skills, of representation of mesoscale activity, of mass/steric distribution and of representation of the equatorial dynamics. The new system is very close to SLA observations with a forecast RMS difference below 5 cm (best analysis is around 4 cm). The description of the ocean water masses is also very accurate and departure from in situ temperature and salinity observations are generally below 0.3 °C and 0.05 PSU. In addition, a global comparison with independent (not assimilated) velocity measurements shows that the location of the main currents is accurately represented.

Sea ice is a key element in our climate system, and it is very sensitive to the current observed climate change. Sea ice volume is a sensitive indicator of the health of Arctic although very challenging to estimate precisely since it is a combination of sea ice area and sea ice thickness. Arctic sea ice volume has decreased by as much as 75% at the end of the summer season if compared with the conditions 40 years ago. The ongoing decline of Arctic sea ice exposes the ocean to anomalous surface heat and freshwater fluxes that can have potential implication for the Arctic region and beyond, for the general oceanic circulation itself. For more than a decade, Mercator Ocean International develops and produces Global Ocean Reanalysis with a 1/4° resolution system. Based on the NEMO modelling platform, observations are assimilated by a reduced-order Kalman filter. In-situ CORA database, altimetric data, sea surface temperature, and sea ice concentration are jointly assimilated to constrain the ocean and sea ice model. In previous reanalysis, long-term sea ice volume drift has been observed in the Arctic. To obtain a better constraint on the sea ice thickness, Cryosat-2 radar Freeboard data are assimilated jointly with the sea ice concentration in a multidata/multivariate sea ice analysis. The coupled ocean and ice assimilation system runs on a 7-day cycle, using IAU (Incremental Analysis Update) and a 4D increment. The “white ocean” is modelled with the multi-categories LIM3.6 sea ice numerical model. The aim of this study is to initiate the development of the future operational multi-variate and multi-data sea ice analysis system with freeboard radar assimilation. After describing this global sea ice reanalysis system, we present results on the abilities of this configuration to reproduce sea ice extent and volume interannual variability in both hemispheres. Comparisons between experiments with and without assimilation show that the joint assimilation of CryoSat-2 radar freeboard and sea ice concentration reduces most of model biases of sea ice thickness, e.g., in the north of the Canadian Arctic Archipelago and in the Beaufort Sea in the Arctic. Moreover, radar freeboard assimilation does not hinder the good results in simulating sea ice extent previously obtained with the assimilation of only sea ice concentration. Validation with non-assimilated satellite data and in-situ data supports these findings. Lastly, snow depth significantly influences the Freeboard measurement: this study also reveals the importance of including snow information on freeboard retrieval and on the ice volume assimilation methodology. These experiments take place in a context of increasing interest in polar regions and prepare the launch of Copernicus Sentinel expansion satellite missions.

Canada has the longest coastline in the world and includes diverse ocean environments, from the frozen waters of the Canadian Arctic Archipelago to the confluence region of Labrador and Gulf Stream waters on the east coast. There is a strong need for a pan-Canadian operational regional ocean prediction capacity covering all Canadian coastal areas in support of marine activities including emergency response, search and rescue, and safe navigation in ice-infested waters. Here we present the first pan-Canadian operational regional ocean analysis system developed as part of the Regional Ice Ocean Prediction System version 2 (RIOPSv2) running in operations at the Canadian Centre for Meteorological and Environmental Prediction (CCMEP). The RIOPSv2 domain extends from 26∘ N in the Atlantic Ocean through the Arctic Ocean to 44∘ N in the Pacific Ocean, with a model grid resolution that varies between 3 and 8 km. RIOPSv2 includes a multivariate data assimilation system based on a reduced-order extended Kalman filter together with a 3D-Var bias correction system for water mass properties. The analysis system assimilates satellite observations of sea level anomaly and sea surface temperature, as well as in situ temperature and salinity measurements. Background model error is specified in terms of seasonally varying model anomalies from a 10-year forced model integration, allowing inhomogeneous anisotropic multivariate error covariances. A novel online tidal harmonic analysis method is introduced that uses a sliding-window approach to reduce numerical costs and allow for the time-varying harmonic constants necessary in seasonally ice-infested waters. Compared to the Global Ice Ocean Prediction System (GIOPS) running at CCMEP, RIOPSv2 also includes a spatial filtering of model fields as part of the observation operator for sea surface temperature (SST). In addition to the tidal harmonic analysis, the observation operator for sea level anomaly (SLA) is also modified to remove the inverse barometer effect due to the application of atmospheric pressure forcing fields. RIOPSv2 is compared to GIOPS and shown to provide similar innovation statistics over a 3-year evaluation period. Specific improvements are found near the Gulf Stream for all model fields due to the higher model grid resolution, with smaller root mean squared (rms) innovations for RIOPSv2 of about 5 cm for SLA and 0.5 ∘C for SST. Verification against along-track satellite observations demonstrates the improved representation of mesoscale features in RIOPSv2 compared to GIOPS, with increased correlations of SLA (0.83 compared to 0.73) and reduced rms differences (12 cm compared to 14 cm). While the RIOPSv2 grid resolution is 3 times higher than GIOPS, the power spectral density of surface kinetic energy provides an indication that the effective resolution of RIOPSv2 is roughly double that of the global system (35 km compared to 66 km). Observations made as part of the Year of Polar Prediction (2017–2019) provide a rare glimpse at errors in Arctic water mass properties and show average salinity biases over the upper 500 m of 0.3–0.4 psu in the eastern Beaufort Sea in RIOPSv2.

GLORYS12 is a global eddy-resolving physical ocean and sea ice reanalysis at 1/12° horizontal resolution covering the 1993-present altimetry period, designed and implemented in the framework of the Copernicus Marine Environment Monitoring Service (CMEMS). The model component is the NEMO platform driven at the surface by atmospheric conditions from the ECMWF ERA-Interim reanalysis. Ocean observations are assimilated by means of a reduced-order Kalman filter. Along track altimeter sea level anomaly, satellite sea surface temperature and sea ice concentration, as well as in situ temperature and salinity vertical profiles are jointly assimilated. A 3D-VAR scheme provides an additional correction for the slowly-evolving large-scale biases in temperature and salinity. The performance of the reanalysis shows a clear dependency on the time-dependent in situ observation system. The general assessment of GLORYS12 highlights a level of performance at the state-of-the-art and the capacity of the system to capture the main expected climatic interannual variability signals for ocean and sea ice, the general circulation and the inter-basins exchanges. In terms of trends, GLORYS12 shows a higher than observed warming trend together with a slightly lower than observed global mean sea level rise. Comparisons made with an experiment carried out on the same platform without assimilation show the benefit of data assimilation in controlling water mass properties and sea ice cover and their low frequency variability. Moreover, GLORYS12 represents particularly well the small-scale variability of surface dynamics and compares well with independent (non-assimilated) data. Comparisons made with a twin experiment carried out at 1/4° resolution allows characterizing and quantifying the strengthened contribution of the 1/12° resolution onto the downscaled dynamics. GLORYS12 provides a reliable physical ocean state for climate variability and supports applications such as seasonal forecasts. In addition, this reanalysis has strong assets to serve regional applications and provide relevant physical conditions for applications such as marine biogeochemistry. In the near future, GLORYS12 will be maintained to be as close as possible to real time and could therefore provide relevant and continuous reference past ocean states for many operational applications.

In this article, Part 1 of a two-part series, we run and evaluate the skill of a regional physical–biogeochemical stochastic ocean model based on NEMO. The domain covers the Bay of Biscay at 1/36° resolution, as a case study for open-ocean and coastal shelf dynamics. We generate model ensembles based on assumptions about errors in the atmospheric forcing, the ocean model parameterizations and in the sources and sinks of the biogeochemical variables. The resulting errors are found to be mainly driven by the wind forcing uncertainties, with the rest of the perturbed forcing and parameters locally influencing the ensemble spread. Biogeochemical uncertainties arise from intrinsic ecosystem model errors and from errors in the physical state. Uncertainties in physical forcing and parameterization are found to have a larger impact on chlorophyll spread than uncertainties in ecosystem sources and sinks. The ensembles undergo quantitative verification with respect to observations, focusing on upper-ocean properties. Despite a tendency for ensembles to be generally under-dispersive, they appear to be reasonably consistent with respect to sea surface temperature data. The largest statistical sea-level biases are observed in coastal regions. These biases hint at the presence of high-frequency error sources currently unaccounted for, and suggest that the ensemble-based uncertainties are unfit to model error covariances for assimilation. Model ensembles for chlorophyll appear to be consistent with ocean colour data only at times. The stochastic model is qualitatively evaluated by analysing its ability at generating consistent multivariate incremental model corrections. Corrections to physical properties are associated with large-scale biases between model and data, with diverse characteristics in the open-ocean and the shelves. Mesoscale features imprint their signature on temperature and sea-level corrections, as well as on chlorophyll corrections due to the vertical velocities associated with vortices. Small scale local corrections are visible over the shelves. Chlorophyll information has measurable impact on physical variables.

The operational production of data-assimilated biogeochemical state of the ocean is one of the challenging core projects of the Copernicus Marine Environment Monitoring Service (hereafter CMEMS). In that framework, Mercator Ocean International is in charge of improving the realism of its global 1⁄4° coupled physical-biogeochemical simulations, analyses and re‐analyses, and to develop an effective capacity to routinely estimate the biogeochemical state of the ocean, including, amongst others, the implementation of biogeochemical data assimilation. Primary objectives are to enhance the time representation of the seasonal cycle in the real time and reanalysis systems, and to provide a better control of the production in the equatorial regions. In that framework, Mercator Ocean International has successfully updated its global biogeochemical analysis and forecasting system with an Ocean Color data assimilation capability. In this system, successfully commissioned in September 2019, the biogeochemical model (NEMO/PISCES) is offline coupled with the dynamical ocean (1/12° coarsened to 1/4° resolution) from the CMEMS global physical analysis and forecasting operational system (PSY4), at a daily frequency, and benefits from the assimilation of satellite (SSH-SST-SIC) and in situ physical data. Nevertheless, a biogeochemical climatological damping is activated in the biogeochemical model in order to mitigate the impact of some misconstrained processes (vertical velocities) from this physical data-assimilated forcing (under investigation). The dedicated assimilation of biogeochemical data relies on a simplified version of the SEEK filter, where the forecast error covariances are built from a fixed-basis - but seasonally variable - ensemble of anomalies computed from a multi-year numerical experiment (without biogeochemical data assimilation) with respect to a running mean. Regarding Ocean Colour observations, the system relies, as a first step, on the CMEMS Global Ocean surface chlorophyll concentration products, delivered in NRT. The objective of this presentation is thus to provide (1) a short description of the implementation of the aforementioned data assimilation methodology in the forecasting system; (2) a synthesis of the assessment of this global biogeochemical forecasting system, by cross-comparing the assimilated solution with various datasets, both spatial (Ocean Colour) and in situ (BGC-Argo, GLODAP), and (3) a synthetic overview of the impact/benefit of the assimilation of the Ocean Colour data.

Abstract. Ocean data assimilation systems encompass a wide range of scalesthat are difficult to control simultaneously using partial observationnetworks. All scales are not observable by all observation systems, which isnot easily taken into account in current ocean operational systems. The mainreason for this difficulty is that the error covariance matrices are usuallyassumed to be local (e.g. using a localisation algorithm in ensemble dataassimilation systems), so that the large-scale patterns are removed from theerror statistics. To better exploit the observational information available for all scales inthe assimilation systems of the Copernicus Marine Environment MonitoringService, we investigate a new method to introduce scale separation in theassimilation scheme. The method is based on a spectral transformation of the assimilation problemand consists in carrying out the analysis with spectral localisation for thelarge scales and spatial localisation for the residual scales. The target isto improve the observational update of the large-scale components of thesignal by an explicit observational constraint applied directly on the largescales and to restrict the use of spatial localisation to the small-scalecomponents of the signal. To evaluate our method, twin experiments are carried out with syntheticaltimetry observations (simulating the Jason tracks), assimilated in a1/4∘ model configuration of the North Atlantic and the Nordic Seas. Results show that the transformation to the spectral domain and the spectrallocalisation provides consistent ensemble estimates of the state of thesystem (in the spectral domain or after backward transformation to thespatial domain). Combined with spatial localisation for the residual scales,the new scheme is able to provide a reliable ensemble update for all scales,with improved accuracy for the large scale; and the performance of the systemcan be checked explicitly and separately for all scales in the assimilationsystem.

Abstract. We generate ocean biogeochemical model ensembles including several kinds of stochastic parameterizations. The NEMO stochastic modules are complemented by integrating a subroutine to calculate variable anisotropic spatial scales, which are of particular importance in high-resolution coastal configurations. The domain covers the Bay of Biscay at 1/36° resolution, as a case study for open-ocean and coastal shelf dynamics. At first, we identify uncertainties from assumptions subject to erroneous atmospheric forcing, ocean model improper parameterizations and ecosystem state uncertainties. The error regimes are found to be mainly driven by the wind forcing, with the rest of the perturbed tendencies locally augmenting the ensemble spread. Biogeochemical uncertainties arise from inborn ecosystem model errors and from errors in the physical state. Model errors in physics are found to have larger impact on chlorophyll spread than those of the ecosystem. In a second step, the ensembles undergo verification with respect to observations, focusing on upper-ocean properties. We investigate the statistical consistency of prior model errors and observation estimates, in view of joint uncertainty vicinities, associated with both sources of information. OSTIA-SST L4 distribution appears to be compatible with ensembles perturbing physics, since vicinities overlap, enabling data assimilation. The most consistent configuration for SLA along-track L3 data is in the Abyssal plain, where the spread is increased due to mesoscale eddy decorrelation. The largest statistical SLA biases are observed in coastal regions, sometimes to the point that vicinities become disjoint. Missing error processes in relation to SLA hint at the presence of high-frequency error sources currently unaccounted for, potentially leading to ill-posed assimilation problems. Ecosystem model-data samples with respect to Ocean Colour L4 appear to be compatible with each other only at times, with data assimilation being marginally well-posed. In a third step, we illustrate the potential influence of those uncertainties on data assimilation impact exercise, by means of multivariate representers and EnKF-type incremental analysis for a few members. Corrections on physical properties are associated with large-scale biases between model and data, with diverse characteristics in the open-ocean and the shelves. The increments are often characteristic of the underlying mesoscale features, chlorophyll included due to the vertical velocity field. Small scale local corrections are visible over the shelves. Chlorophyll information seems to have a very measurable potential impact on physical variables.

Mercator Ocean monitoring and forecasting systems are routinely operated in real time since early 2001. They have been regularly upgraded through several systems of increasing complexity, expanding the geographical coverage from regional to global, improving models and assimilation schemes. In this chapter we give a description of the current Mercator Ocean real-time, global high-resolution system. The ocean model, the observations, and the data assimilation scheme are detailed with a particular focus to the specifics of the Mercator Ocean system. Technical details about the real-time operation of the system are given. The system is then examined through a scientific evaluation, highlighting the level of performance and the reliability of the system. User needs and evolutions of the system are finally drawn.

Since 19 October 2016, and in the framework of Copernicus Marine Environment Monitoring Service (CMEMS), Mercator Ocean has delivered real-time daily services (weekly analyses and daily 10-day forecasts) with a new global 1/12 degrees high-resolution (eddy-resolving) monitoring and forecasting system. The model component is the NEMO platform driven at the surface by the IFS ECMWF atmospheric analyses and forecasts. Observations are assimilated by means of a reduced-order Kalman filter with a three-dimensional multivariate modal decomposition of the background error. Along-track altimeter data, satellite sea surface temperature, sea ice concentration, and in situ temperature and salinity vertical profiles are jointly assimilated to estimate the initial conditions for numerical ocean forecasting. A 3D-VAR scheme provides a correction for the slowly evolving large-scale biases in temperature and salinity. This paper describes the recent updates applied to the system and discusses the importance of fine tuning an ocean monitoring and forecasting system. It details more particularly the impact of the initialization, the correction of precipitation, the assimilation of climatological temperature and salinity in the deep ocean, the construction of the background error covariance and the adaptive tuning of observation error on increasing the realism of the analysis and forecasts. The scientific assessment of the ocean estimations are illustrated with diagnostics over some particular years, assorted with time series over the time period 2007-2016. The overall impact of the integration of all updates on the product quality is also discussed, highlighting a gain in performance and reliability of the current global monitoring and forecasting system compared to its previous version.

A twin-experiment is carried out introducing elements of an Ensemble Kalman Filter (EnKF), to assess and correct ocean uncertainties in a high-resolution Bay of Biscay configuration. Initially, an ensemble of 102 members is performed by applying stochastic modeling of the wind forcing. The target of this step is to simulate the envelope of possible realizations and to explore the robustness of the method at building ensemble covariances. Our second step includes the integration of the ensemble-based error estimates into a data assimilative system adopting a 4D Ensemble Optimal Interpolation (4DEnOI) approach. In the twin-experiment context, synthetic observations are simulated from a perturbed member not used in the subsequent analyses, satisfying the condition of an unbiased probability distribution function against the ensemble by performing a rank histogram. We evaluate the assimilation performance on short-term predictability focusing on the ensemble size, the observational network, and the enrichment of the ensemble by inexpensive time-lagged techniques. The results show that variations in performance are linked to intrinsic oceanic processes, such as the spring shoaling of the thermocline, in combination with external forcing modulated by river runoffs and time-variable wind patterns, constantly reshaping the error regimes. Ensemble covariances are able to capture high-frequency processes associated with coastal density fronts, slope currents and upwelling events near the Armorican and Galician shelf break. Further improvement is gained when enriching model covariances by including pattern phase errors, with the help of time-neighbor states augmenting the ensemble spread. (C) 2016 Elsevier Ltd. All rights reserved.

Recent increases in marine traffic in the Arctic have amplified the demand for reliable ice and marine environmental predictions. This article presents the verification of ice forecast skill from a new system implemented recently at the Canadian Meteorological Centre called the Global Ice Ocean Prediction System (GIOPS). GIOPS provides daily global ice and ocean analyses and 10‐day forecasts on a 1/4°‐resolution grid. GIOPS includes a multivariate ocean data assimilation system that combines satellite observations of sea‐level anomaly and sea‐surface temperature (SST) together with in situ observations of temperature and salinity. Ice analyses are produced using a 3D‐Var method that assimilates satellite observations from SSM/I and SSMIS together with manual analyses from the Canadian Ice Service. Analyses of total ice concentration are projected onto the thickness categories used in the ice model using spatially and temporally varying weighting functions derived from ice model tendencies. This method may reduce deleterious impacts on the ice thickness distribution when assimilating ice concentration, as it can directly modulate (and reverse) nonlinear processes such as ice deformation. An objective verification of sea ice forecasts is made using two methods: analysis‐based error assessment focusing on the marginal ice zone, and a contingency table approach to evaluate ice extent as compared to an independent analysis. Together the methods demonstrate a consistent picture of skilful medium‐range forecasts in both the Northern and Southern Hemispheres as compared to persistence. Using the contingency table approach, GIOPS forecasts show a significant open‐water bias during spring and summer. However, this bias depends on the choice of threshold used. Ice forecast skill is found to be highly sensitive to the assimilation of SST near the ice edge. Improved observational coverage in these areas (including salinity) would be extremely valuable for further improvement in ice forecast skill.

The GODAE OceanView systems use various data assimilation algorithms, including 3DVar, EnOI, EnKF and the SEEK filter with a fixed basis, using different time windows. The main outputs of the operational data assimilation systems, the increments, have been compared for February 2014 in various regions. The eddy-permitting systems' increments are similar in a number of the regions, indicating similar forecast errors are being corrected, while the eddy-resolving systems represent smaller-scale structures in the mid-latitude regions investigated and appear to have smaller biases. Monthly average temperature increments show significant SST biases, particularly in the systems which assimilate swath satellite SST data, indicating systematic errors in the surface heat fluxes and the way in which they are propagated vertically by the ocean models. On-going developments to the data assimilation systems include improvements to the specification of error covariances, improving assimilation of data near the equator, and understanding the effect of assimilation on the Atlantic Meridional Overturning Circulation. Longer term developments are expected to include the implementation of more advanced algorithms to make use of flow-dependent error covariance information. Assimilation of new data sources over the coming years, such as wide-swath altimetry, is also expected to improve the accuracy of ocean state estimates and forecasts provided by the GODAE OceanView systems.

In the framework of the Myocean EU (FP7 and Horizon 2020) funded projects, Mercator Ocean, the French operational oceanography center, is in charge of the development and of the production of real time analysis and forecasts and reanalysis for the global ocean at the resolution of 1/12°. The operational systems are all based on the ocean and sea ice model NEMO and the multivariate data assimilation system SAM2 (Système d’Assimilation Mercator V2). The assimilation method is a reduced order Kalman filter based on SEEK/LETKF formulation with an Incremental Analysis Update. The background error covariance is represented by an ensemble of multivariate state vectors defining a subspace of the control space. However, the SAM2 background error statistics are built using a static approach based on a prior long multi-year free simulation. In order to better represent the errors of the day, a more realistic description of the uncertainties is necessary and could be reached from ensemble simulations using stochastic modeling. In this work, a twin-experiment is carried out introducing elements of an Ensemble Kalman Filter, to assess ocean uncertainties in a high-resolution Bay of Biscay configuration. The experimental protocol presented here serves as an intermediate step in the evolution of Mercator Océan data assimilation systems, towards a Local Ensemble Transform Kalman Filter (LETKF). Initially, a long-range ensemble is performed, by applying stochastic modelling of the wind forcing. The target of this step is to simulate the envelope of possible realizations in a regional configuration and to evaluate the robustness of the method on building ensemble background error covariances. Sensitivity assimilation experiments explore the relative performance of the ensemble approach, in order to identify somes parameters of the method in future operational systems. First results indicate that the ensemble method outperforms in many cases operational methods using static covariances in the chosen experimental protocol. Thus, finding efficient methods to capture the most important aspects of ocean uncertainty is of high priority for the near future.