To give you the best possible experience, this site uses cookies. Continuing to use this site means you agree
to our use of cookies. If you'd like to learn more about the cookies we use please find out more

Session 3.1 Abstracts

Symposium home | Session 2 | Session 3 | Session 4 | Abstracts by Author |


3.1 Ocean Modelling and Data Assimilation: Underpinning Ocean Forecasting and Reanalysis

Session conveners: Eric Chassignet and Masafumi Kamachi

The table below lists all abstracts for Session 3.1 by author. To read the full abstract click on the title-link.

The unique reference number (ref. no.) relates to the abstract submission process and must be used in any communications with the organisers.

All abstracts from session 3.1 are available for download - pdf.

 Ref.NoPrimary AuthorAffiliationCountryAbstract titlePosters
S3.1-01Agoshkov, ValeryINM RASRussian FederationThe Black Sea as a natural test laboratory of operational oceanography 
S3.1-02Aguiar, AlessandroFederal University of BahiaBrazilUpwelling Processes along a Western Boundary Current: the Abrolhos-Campos Region(15°S - 23°S), Brazil Poster-pdf
S3.1-03Amorim, FabiolaREMOBrazilWind and Western Boundary Currents driven the Eastern Brazilian Shelf Circulation (10S - 16S): a numerical study Poster-pdf
S3.1-04Backeberg, BjornNansen-Tutu Centre, Univ. Cape TownSouth AfricaAssimilating along-track SLA data using the EnOI in an eddy resolving model of the Agulhas systemPoster-pdf
S3.1-05Bertino, LaurentNERSCNorwayModeling and data assimilation developments of the TOPAZ system in support of operational oceanography in the Arctic 
S3.1-06Bourdalle-Badie, RomainMercator OceanFranceA 2-equation vertical mixing scheme in NEMO ocean code Poster-pdf
S3.1--7Brassington, GaryCAWCR/Bureau of MeteorologyAustraliaMulti-cycle ensemble ocean forecasting 
S3.1-08Candille, GuillemLGGE/CNRSFranceTowards an ensemble data assimilation strategy for oceanography: a stochastic approach developed for research and operational applications (MyOcean-2) Poster-pdf
S3.1-09Carton, JamesUniversity of MarylandUnited StatesDevelopments in ocean climate analysis 
S3.1-10Chevallier, MathieuMercator OceanFranceArctic Ocean circulation and water masses: Validation and sensibility to the horizontal resolution 
S3.1-11Garnier, FlorentLGGE/Equipe MEOMFranceTowards data assimilation in a state-of-the-art physical-biogeochemical model of the North Atlantic: Estimation of model uncertainties using Stochastic parametrizations 
S3.1-13 Best posterGuinehut, StephanieCLSFranceMonitoring the Ocean from observations Poster-pdf
S3.1-14Kuragano, TsuraneMRI-JMAJapanImprovement of assimilation procedure for sea surface height data by removing nonsteric seasonal height Poster-pdf
S3.1-15Lea, DanielMet OfficeUnited KingdomUse of GOCE MDT and error information in NEMOVAR, a variational data assimilation scheme for NEMO Poster-pdf
S3.1-16Li, ZhijinJet Propulsion LabUnited StatesCoastal Ocean Data Assimilation and Forecasting Using a Multi-Scale Three-Dimensional Variational Data Assimilation Scheme 
S3.1-17Lima, LeonardoOceanographic Modeling and Observation Network (REMO)BrazilAssimilation of sea surface temperature data into HYCOM in the Atlantic Ocean Poster-pdf
S3.1-18Mignac, DaviOceanographic Modeling and Observation Network (REMO)BrazilArgo data assimilation into HYCOM with an EnOI method over the Atlantic Ocean Poster-pdf
S3.1-19Oke, PeterCSIROAustraliaTowards a dynamically balanced eddy-resolving ocean reanalysis Poster-pdf
S3.1-20Paiva, AfonsoCOPPE/UFRJBrazilHigh-resolution Simulations of the South Atlantic Ocean Poster-pdf
S3.1-21Penny, SteveUniversity of MarylandUnited StatesAdvancements in NCEP’s Global Ocean Data Assimilation System 
S3.1-22Penton, JenniferUWAAustraliaThe effects of wind forcing on surface currents on the continental shelf surrounding Rottnest Island Poster-pdf
S3.1-23Rahaman, HasiburINCOISIndiaFresh water content variability over Bay of Bengal in a nested high resolution regional model 
S3.1-24Richman, JamesNRLUnited StatesThe Atlantic Meriodional Overturning Circulation and Mesoscale Eddy Statistics in an Eddy Resolving Ocean Reanalysis (1992-2010) 
S3.1-25 Best posterRykova, TatianaCSIROAustraliaA comparison of boundary current eddies in an ocean model and an ocean reanalysis Poster-pdf
S3.1-26Siva reddy, SanikommuINCOISIndiaInter-annual variability of thermal inversion in the Bay of Bengal using Global Ocean analysis 
S3.1-27Smith, ScottNRLUSAThe Impact of Velocity Data Assimilation from Drifters Relative to Altimeter Observations 
S3.1-28Srinivasan, AshwanthFlorida State University (COAPS)United StatesGulf of Mexico multi-scale data assimilation 
S3.1-29Tanajura, ClementeREMO-UFBABrazilREMO Ocean Data Assimilation System into HYCOM (RODAS.H1) Poster-pdf
S3.1-30Vernieres, GuillaumeSSAI/NASAUnited StatesImpact of assimilation of Aquarius sea surface salinity data in the GEOS system 
S3.1-31Waters, JenniferMet OfficeUnited KingdomImplementation of a variational data assimilation system in the Met Office's 1/4 degree global ocean model Poster-pdf
S3.1-32Yan, ChangxiangInstitute of Atmospheric PhysicsChinaOcean reanalysis in the Indian Ocean and Pacific Ocean 
S3.1-33Zuo, HaoECMWFUnited KingdomGlobal ocean reanalysis and data assimilation in NEMOVAR system Poster-pdf


ID 3.1-01

The Black Sea as a natural test laboratory of operational oceanography

V.I. Agoshkov1, V.A. Ivanov2, G.K.Korotaev2, V.B. Zalesny1, A.G. Zatsepin3

1 Institute of Numerical Mathematics RAS, Moscow, Russia

2 Marine Hydrophysical Institute NASU, Sevastopol, Ukraine

3 Institute of Oceanology RAS, Moscow, Russia


Modern operational oceanography technologies offer efficient means of marine environment monitoring and forecasting. Achievements of operational oceanography were stimulated by significant progress of numerical mathematics, by increasing capacity of multiprocessor computational platforms, and by development of new measuring technologies. Operational or near-operational systems have been set up worldwide permitting to provide analysis and forecast of three-dimensional marine and ocean circulation and ecosystems.

Further increase of accuracy and term of marine forecasts is an acute problem of operational oceanography. It depends on improvement of the circulation model physical content and the skill of data assimilation methods. Available global observing system is suitable for the resolving of mesoscale ocean dynamics but it is too coarse to support development of models and data assimilation methods. Moreover it is too expensive to improve observing system on global scales for pure scientific goals. Development of denser observations in some part of the ocean is also problematic as it requires control of the open boundary conditions. Therefore the selection of a suitable test area can be considered to be an alternative. The Black Sea looks a suitable object to study marine dynamics predictability. It is rather deep, practically closed ocean type basin with simple configuration and connection with the Marmara and Azov seas only by two narrow straits. The Black Sea has permanent and seasonal pycnoclines, intense jet-like Rim current. Its mesoscale variability includes stream-jet meandering, development of mesoscale eddies and Rossby waves. The basin has broad north-western shelf. Space imagery analysis shows manifestation of submesoscale structures containing rather significant energy. Thus, the Black Sea presents all major features of oceanic dynamics and can be considered as the prototype of the ocean.

Joint project of Russia and Ukraine “The Black Sea as a prototype of the ocean” is dedicated to application of extended observing system in the basin for careful test of circulation models and data assimilation algorithms. The observing system will be based on the broad range of space observations including altimetry, visible, IR and SAR imagery supplemented to arrays of profiling floats and surface drifting buoys with thermistor chain, ships of opportunity sections and temperature, and salinity time series from a set of stable platforms. Specially designed cruises and optionally glider experiment will supplement operational observation network. A set of circulation models: MHIC (Ukraine), INMOM (Russia), NEMO (EU), POM (USA) are installed in the basin and are under improvement for the submesoscale resolution and testing of different mixing parameterizations. Among other models INMOM is based on combination of implicit splitting methods and adjoint equations. The essence of the method consists in the representation of a complicated problem as a set of simpler subproblems. The prognostic model can be simplified or physically enriched by neglecting or adding particular modules. The method of adjoint equations allows formulate and solves inverse problems including estimation of sensitivity of solutions and four-dimensional data assimilation. All available models will be applied in future for the multimodel forecast, careful analysis of error dependence on the term of forecast and conclusions concerning the increase of predictability of marine dynamics as a result of the numerical models improvement. The project is not finished yet, thus some preliminary results will be presented.

ID 3.1-02

Upwelling Processes along a Western Boundary

Current: the Abrolhos-Campos Region ( 15 ° S - 23 ° S) , Brazil

A. L. Aguiar1,2,3, M. Cirano2,3, J. Pereira2,3, M. Marta-Almeida3

1 Universidade Federal da Bahia (UFBA), Programa de Pós-Graduação em Geofísica, Salvador, Brazil

3 Grupo de Oceanografia Tropical (GOAT), Instituto de Física, Universidade Federal da

Bahia (UFBA), Salvador, Brazil

3 Rede de Modelagem e Observação Oceanográfica (REMO), Brazil


Upwelling events observed over the years of 2003 to 2011 are analyzed. Our focus is on the Abrolhos Campos Region - ACR (15°S - 23°S), located along the Brazilian Margin and influenced by a Western Boundary Current, the Brazil Current (BC). Satellite Sea Surface Temperature, NCEP and GFS wind data complemented with the results of a high-resolution regional model are used to investigate the occurrence and the responsible mechanisms for these events along the region. Upwelling events were more frequent from 20°S to 23°S. Wind-driven upwelling represented more than 90% of the upwelling events. The Ekman transport was found to be more important for the region from Prado (17°S) to Marataízes (21°S), whereas the Ekman pumping had a more important role from 22°S to 23°S region. Current-driven upwelling processes associated with the location of the BC, its velocity, transport and meso-scale activity were also analyzed. Results show that these mechanisms are highly influenced by the local topography. Topographic effects, via the BC acceleration, are more noticeable in southern ACR while, in Prado, lateral movements of the BC axis tend to be more relevant. Moreover, eddy-driven upwelling increases southward after the passage of the Vitória-Trindade Rigde (20°S), an important submarine chain, which acts to constrain and modulate the southward BC flow.

Keywords: Sea Surface Temperature, wind-driven circulation, Ekman transport, Ekman Pumping, Brazil Current, Regional Modelling


Figure 1: Cross-shore section of the temperature during a) a previous stage, b) an intermediate stage and c) a peak stage of the casestudy upwelling event along the Marataízes section (21°S). The interval between isotherms is 2 ° C. Cross-shore section of the associated alongshore velocity in d) a previous stage, e) an intermediate stage and f) a peak stage of the same event. Negative (positive) velocities are southward (northward). The interval between isotachs is 0.05 m s 1

ID 3.1-03

Wind and Western Boundary Currents driven the Eastern Brazilian Shelf Circulation (10oS - 16oS): a numerical study

F.N. Amorim1, 2, M. Cirano1, 2, 3, M. Marta-Almeida2, H.B. Oliveira2

1 Grupo de Oceanografia Tropical, Instituto de Física, Universidade Federal da Bahia, Brazil

2 Rede de Modelagem e Observação Oceanografica (REMO), Brazil

3 South Australia Research and Development Institute, Australia


The Regional Ocean Modelling System (ROMS), with embedded nesting capabilities based on AGRIF was configured with a refined grid (1/36o) and realistic forcings (6-hourly winds and surface fluxes, daily large scale oceanic forcing and tides) to describe the influence of the wind and Western Boundary Currents (WBC) on the Eastern Brazilian Shelf (EBS) seasonal circulation. The EBS is a passive tropical margin, where the dynamics is strongly influenced by the seasonal latitudinal excursion of the South Equatorial Current (SEC) bifurcation and the large-scale seasonality of the trade winds, both influenced by the north-south displacement of the marine Inter-Tropical Convergence Zone (ITCZ). The model results show that for the northern region (10oS) the northward flow is the dominant pattern throughout the year and the southward flow is confined to the top 50 m of the water column during the spring/summer seasons. The surface circulation at the inner and mid shelves is mostly driven by the wind while the currents at the shelf-break are mainly driven by the currents at the slope.

In the middle (14oS) and southern (16oS) regions, an alternate dominance of the southward/northward flows for the first 150 m is observed, despite the annual net transport is oriented southwards. In the 150-400 m layer the dominance of the northward flow is clear. At 14oS the wind forcing mainly drives the inner-shelf circulation and the mid-shelf circulation is forced by both the wind and the flow over the slope. At 16oS the inner and mid shelf currents are mainly driven by the wind forcing, while the shelf-break currents present a poor correlation with the wind and a strong influence of the WBC dynamics.

These results represent the first effort toward the implementation of an operational numerical modelling system for EBS, since the main forcing mechanisms were well understood and described.

ID 3.1-04

Assimilating along-track SLA data using the EnOI in an eddy resolving model of the Agulhas system

Bjorn Backeberg1,2,3, François Counillon3 and Johnny Johannessen3,4

1 Nansen-Tutu Centre for Marine Environmental Research, Cape Town, South Africa
2 University of Cape Town, South Africa
3 Nansen Environmental and Remote Sensing Center, Bergen, Norway
4 Geophysical Institute, University of Bergen, Norway


The greater Agulhas Current is one of the most energetic current systems in the global ocean. It plays a fundamental role in determining the mean state and variability of the regional marine environment, affecting its resources and ecosystem, the regional weather and the global climate on a broad range of temporal and spatial scales. In the absence of a coherent in-situ and satellite based observing system in the region, modeling and data assimilation techniques play a crucial role in both furthering the quantitative understanding and providing reliable forecasts of this complicated western boundary current system. In this study we use a regional implementation of the Hybrid Coordinate Ocean Model and assimilate along-track satellite sea level anomaly (SLA) data using the Ensemble Optimal Interpolation (EnOI) data assimilation scheme. Compared to a free model run over a two-year period, data assimilation reduces the error compared to SLA observations that have not yet been assimilated. Mesoscale features are placed in more consistent agreement with drifter trajectories, and the error calculated from independent drifter measurements for eddy kinetic energy and surface velocities is reduced. However, the assimilation introduces a sea surface temperature bias in the Agulhas Return Current, which is associated with the correction of a sea surface height bias in the historical ensemble of the EnOI. Overall however, the data assimilation system produces a more realistic representation of the mesoscale dynamics in the greater Agulhas region.

ID 3.1-05

Modeling and data assimilation developments of the TOPAZ system in support of operational oceanography in the Arctic

Laurent Bertino1, François Counillon1, Sylvain Bouillon1, Pierre Rampal1, Timothy Williams1, Jon Bergh1

1 Nansen Environmental and Remote Sensing Center, Bergen, Norway


We present the TOPAZ4 modeling and data assimilation system, based on the Nansen Center’s version of the HYCOM model (at horizontal resolution of about 12 km) and an Ensemble Kalman Filter (EnKF), integrating a dynamical ensemble of 100 members. The multivariate properties of the EnKF allow the TOPAZ system to assimilate several ocean and sea ice data types simultaneously, both in real-time forecasts applications (exploited at met.no) and in reanalysis mode. The TOPAZ system is the core Arctic component of the MyOcean system (http://www.myocean.eu).

We will analyze the results from a 20-years TOPAZ reanalysis, showing the good stability of the EnKF used in realistic settings and its ability to provide physically consistent error estimates for most variables assimilated. The reanalysis also pointed to limitations of the sea ice model in terms of sea ice drift and motivates the further developments of new sea ice rheology models for the Marginal Ice Zone and the ice pack.

ID 3.1-06

A 2-equation vertical mixing scheme in NEMO ocean code

R. Bourdallé-Badie1, G. Reffray1, C. Calone1,2, G. Garric1, C. Bricaud1, Y. Drillet1

1 Mercator Océan, Ramonville Saint Agne, France

2 LOCEAN, Paris, France


Mercator Ocean, the French ocean forecast service provider, operates a hierarchy of analysis and forecasting systems (regional to global scale). Recently, the R&D department, dedicated to improve and develop the future systems, have been working on the general representation of the vertical processes in the system’s modelling component. We have studied different closure schemes for the vertical mixing parameterisation to potentially improve the representation of the upper layers stratification in the global Mercator configurations. We have principally tested second order schemes against the TKE closure scheme widely used in the NEMO users community.

This present work is decomposed in three parts:

  • A 1D approach with an ideal case (the Kato-Philipps experiment) to evaluate various closure schemes. A second order closure scheme gives always better results than the TKE solution. Thanks to this experiment, we have then retained the k-ε scheme for our further investigations.
  • In a second step, we have used a more “real case” by using the PAPA (north-eastern Pacific; 45°N, 145°W) station conditions to evaluate NEMO1D configuration over an annual cycle. Far from coastlines and known to have weak horizontal currents, this location is an ideal case to study the vertical processes. Two experiments, one performed with the TKE closure scheme and the other with the k-ε scheme have been performed over 410 days with ECMWF forcing and initial state from PAPA observations. These experiments have also shown better results with k-ε scheme compared to TKE, especially a better stratification during summertime at the top layers and less and better heat penetration during the rest of the year.
  • The last step of this work consisted in performing two 6-years global ¼° experiments with the NEMO3.4 platform, one performed with the TKE closure scheme and the other with the k-ε scheme. The k-ε scheme exhibits an improvement compared to the TKE solution in the mean temperature profile. At the surface, a clear improvement is noticeable in the northern Pacific Ocean when the k-ε scheme is used instead of the TKE solution. This is mainly due to a better representation of the stratification during summer time in the northern mid-latitudes areas.


ID 3.1-07

Multi-cycle ensemble ocean forecasting

Gary B. Brassington1

1 CAWCR/Australian Bureau of Meteorology, Sydney, Australia


Modelling the ocean, a nonlinear dynamical system, leads to rapid and non-uniform error growth from errors in the initial conditions and applied atmospheric forcing. A single deterministic forecast is therefore one member of a population of forecasts. When the error growth rates increase a single deterministic forecast becomes unreliable and a statistical description of the population becomes essential. Ensemble forecasting is a popular technique for estimating the statistics of the forecast population. However, ensemble forecasting is at present computationally prohibitive for operational ocean forecasting. At a similar stage of development for numerical weather prediction, time-lagged ensembles were examined to estimate the population but it was found that the use of sequential forecasts was ineffective at the generation of independent random errors. An extension to this concept that results in an increase in independence is the use of a multi-cycle time-lagged ensemble. Multi-cycle refers to the fact that each cycle is independent from the previous forecast cycle i.e., the background field is not from the previous cycle but from an earlier forecast integration. For an M-cycle system the background field for each cycle is from a model integration M units of time earlier where each unit of time is the minimum period between two independent cycles. The errors in the background therefore have a larger period to grow compared with a conventional sequential system. However, the increased independence in the forecast errors provides ensemble averages with reduced RMSE and favourable spread-error relationships. The Bureau of Meteorology, Ocean Model, Analysis and Prediction System (OceanMAPS) has been implemented as a four-cycle system. Although the number of cycles is small it is found to provide some early insights into error growth and forecast uncertainty. For example there is an increasing probability with model integration period that the latest forecast is not the best. This technique is particularly relevant whilst global ocean forecasting remains a computationally intensive application delaying the adoption of more conventional ensemble forecasting techniques.


Figure 1: The percentage improvement of RMSE for sea surface height anomaly for a 4-cycle ensemble average over the latest forecast. The percentage improvement is shown as a distribution from a 6 month hindcast. Distributions are shown for -096, -048, 000 and 048 hrs.

ID 3.1-08

Towards an ensemble data assimilation strategy for oceanography: a stochastic approach developed for research and operational applications (MyOcean-2)

G. Candille1, J.-M. Brankart1, P. Brasseur1

1 LGGE/CNRS, Grenoble, France


SANGOMA (Stochastic Assimilation for the Next Generation Ocean Model Applications) is a project funded by EU over the 2012-2015 period to support the development of advanced methods suitable for addressing non-linear and non-Gaussian assimilation problems in oceanography.

In this framework, a realistic circulation model of the North Atlantic ocean at 1/4° resolution (NATL025 configuration) has been adapted to include effects of unresolved scales on the dynamics. This is achieved by introducing stochastic perturbations of the equation of state to represent the associated model uncertainty (Brankart 2013). Assimilation experiments are designed using altimetric data from the Jason missions (Jason 1, but also Envisat mission) and temperature/salinity profiles (WOD05, GTSPP and Argo) to better control the Gulf Stream circulation, focusing on the frontal regions which are predominantly affected by the non-resolved dynamical scales.

An ensemble based on such stochastic perturbations is then produced and evaluated against satellite observations and T/S profiles. In this presentation, we will demonstrate the statistical properties of the pdfs: dispersion analysis and reliability rank histogram and RCRV, and global probabilistic properties through CRPS (Candille and Talagrand 2005, Candille et al. 2007). Further, we will discuss the relevance of using non-Gaussian extensions of conventional assimilation methods for re-analyses and real-time operational assimilation systems based on NATL025 or higher resolution configurations operated in the MyOcean-2 framework.

Brankart J.-M., 2013: Impact of uncertainties in the horizontal density gradient upon low resolution global ocean modelling. Ocean Modelling, 66, 64-76.

Candille G., and O. Talagrand, 2005. Evaluation of probabilistic prediction systems for a scalar variable. Quart. J. Roy. Meteor. Soc. 131, 2131-2150.

Candille G., C. Côté , P. L. Houtekamer, and G. Pellerin, 2007. Verification of an ensemble prediction system against observations. Mon. Wea. Rev. 135, 2688-2699.

ID 3.1-09

Developments in ocean climate analysis

Jim Carton

University of Maryland / NOAA


The Simple Ocean Data Assimilation (SODA) effort has been a long term effort to apply data assimilation techniques to the reconstruction of ocean climate variability. SODA, along with related activities at GFDL and the NOAA National Meteorological Center, began in the 1980s as efforts to reconstruct variability of the tropical upper ocean. The NCEP atmospheric reanalysis of Kalnay et al. [1996] accelerated this process by providing the surface forcing required to expand these ocean reanalysis efforts to the global ocean. Through the succeeding decades the expansion of the ocean observing system, both in situ and satellite observations, have provided the observations needed to constrain initially upper ocean temperature, and gradually other state variables. More recently efforts to reconstruct early 20th century meteorology combined with the growth of the World Ocean Database has made it practical to attempt to reconstruct the ocean variability during the first half of the 20th century. In this paper we review the history of ocean reanalysis/synthesis, its parallels to the developments in atmospheric reanalysis, and discuss future directions. The talk will conclude outlining a program of upcoming developments including developments in estimation techniques, the transition to eddy resolution, coupled atmosphere/ocean/ice models, and the inclusion of biogeochemical processes.

ID 3.1-10

Arctic Ocean circulation and water masses:

Validation and sensibility to the horizontal resolution

M. Chevallier1,2, G. Garric1, R. Boudallé-Badie1, C. Bricaud1, C. Régnier1

1 Mercator-Océan, Ramonville saint Agne, France

2 CNRM-GAME, UMR 3589, Météo-France/CNRS, Toulouse, France.


Mercator Ocean, the French ocean forecast service provider, develops and operates ocean analysis and forecasting systems based on state-of-the-art Ocean General Circulation Models assimilating observations of the Global Ocean Observing System. The mandate of Mercator Ocean in the European/GMES (Global Monitoring for Environment and Security) context is to cover the global ocean at eddy resolving resolution.

In this context, the R&D department developed few years ago a global ocean-sea ice high-resolution model configuration at 1/12° horizontal resolution.

This presentation is focused on the ability of this configuration to represent the circulation and the water masses in the Arctic Ocean. Based on the ORCA-type grid, this configuration has two to six kilometers resolution in the Arctic basin but is really eddy resolving in the deep interior of the Canada basin.

The model is based on the NEMO3.2 code version with the LIM2 sea ice model. The 50-level vertical discretization retained for this configuration has 1m resolution at the surface, decreasing to 450m at the bottom, and 22 levels within the upper 100 m. Initial conditions for the water masses come from Levitus (2005) climatology and the model is driven at the surface by ERA-Interim reanalysis atmospheric data over the 1999-2010 hindcasts period. Skill of the Arctic Ocean 1/12° simulation is compared to a twin experiment performed at lower (1/4°) resolution. The quality of the simulated Arctic Ocean circulations is assessed using various data including CTD data collected during the International Polar Year and the DAMOCLES project. Main results obtained until now can be summarized as follows: Atlantic water enters in the Arctic with better properties and penetrates more deeply into the Arctic Ocean with the 1/12° configuration; better representation of summer Pacific waters properties is obtained with the 1/12° configuration and the eddy kinetic energy penetrates deeper in the water column.

A sustainable collaboration with Environment Canada (CMC) and DFO (Fisheries and Ocean Canada) is on-going to develop dedicated systems and tools for Arctic Ocean and Nordic seas studies.

ID 3.1-11

Towards data assimilation in a state-of-the-art physical biogeochemical model of the North Atlantic: Estimation of model uncertainties using Stochastic parametrizations

F. Garnier1, P. Brasseur1, J.M. Brankart1, E. Cosme1, and J. Verron 1

1 CNRS/UJF/LGGE, MEOM team, Grenoble, FRANCE


In recent years, the assimilation of data into coupled physical-biogeochemical models of the ocean has been addressed primarily by investigating the potential of ocean colour observations to constrain the primary production and associated biogeochemical fluxes. In the framework of the FP7 MyOcean project, a prototype, sequential assimilation system was developed to monitor the biogeochemical state of the ocean during the SeaWiFS period, using a reduced-order state estimation scheme in a North Atlantic NEMO configuration at 1/4° (NATL025) coupled to a simple LOBSTER ecosystem (Fontana et al., 2013). In a companion study, the NATL025/LOBSTER platform was used to estimate the distribution of a number of key parameters of the biological model within biogeochemical provinces of the North Atlantic using the composite Globcolour data set (Doron et al., 2011; 2013). The implementation of those two assimilation approaches rely on the anamorphosis technique that is designed to account for the non-gaussian behaviour of errors on model variables (Brankart et al., 2012).

In this presentation, we will review the recent progress made on the NATL025 configuration.

The biological modelling platform has been complexified by substituting LOBSTER with the more elaborated PISCES formulation to ensure consistency with both operational development and research applications. We analyse here uncertainties coming from the biogeochemical model, using random processes to represent both unresolved fluctuations and missing formulations (Brankart, 2013). Further, implications of this new biogeochemical model formulation and associated stochastic parametrizations will be demonstrated in terms of prior error covariance estimations and ocean colour derived products to be considered for future data assimilation studies.


Brankart J.-M., Testut C.-E., Béal D., Doron M., Fontana C. Meinvielle M. and Brasseur P., 2012 : Towards an improved description of oceanographic uncertainties : effect of local anamorphic transformations on spatial correlations, Ocean Sci., doi:10.5194/os-8-121-2012, 8, 121–142.

Brankart J.-M., 2013: Impact of uncertainties in the horizontal density gradient upon low resolution global ocean modeling, Ocean modelling, 66 (2013) 64–76

Doron M., Brasseur P. and Brankart J.-M., 2011: Stochastic estimation of biogeochemical parameters of a 3D ocean coupled physical biogeochemical model: Twin experiments, J. Marine Syst., 87, 194-207.

Doron M., Brasseur P., Brankart J.-M. and Losa S., 2013: Stochastic estimation of biogeochemical parameters from Globcolour ocean colour satellite data in a North Atlantic 3D coupled physicalbiogeochemical model, J. Marine Syst., in revision.

Fontana C., Brasseur P., Brankart J.-M., 2013: Toward a multivariate reanalysis of the North Atlantic ocean biogeochemistry between 1998-2006 based on the assimilation of SeaWiFS data, Ocean Science, doi:10.5194/os-9-37-2013.

ID 3.1-12


L. Gaultier1,2,J. Verron1,2, P. Brasseur1,2, J.-M. Brankart1,2

1 CNRS, LGGE, F-38041 Grenoble, France

2 Univ. Grenoble Alpes, LGGE, F-38041 Grenoble, France


During the past 20 years, altimetric satellites have revealed the turbulent ocean dynamics at the mesoscale. Additionally, high resolution sensors of tracers such as the Sea Surface Temperature or the Ocean Color reveal even smaller structures at the submesoscale, which are not detected by altimetry. Therefore, the two types of complementary observations are expected to refine the estimation of the ocean circulation.

The goal of this study is to explore the feasibility of inverting tracer information to possibly control ocean dynamics emerging from altimeter data analysis. To do so, we choose an image data assimilation strategy in which a cost-function is built that aims at minimizing the misfits between some image of submesoscale flow structure and tracer images. In the present work, we have explored the extent to which the Finite-Size Lyapunov Exponents (FSLE) can be considered as a proxy characterizing the submesoscale flow structure.

A prerequisite to the study is the investigation of the sensitivity of FSLE horizontal patterns to velocity errors. Indeed, the misfit between a FSLE derived from a velocity with errors, and the FSLE derived from an error free velocity is successfully minimized. The next step is the inversion of sub-mesoscale tracer information to correct a mesoscale altimetric field using real observation (from AVISO for the velocity and from MODIS sensor for the tracer). The ocean dynamical variable to be corrected is the mesoscale altimetric velocity field using a high resolution tracer image. The strategy is similar to the one used to invert FSLE. The cost function measures the misfit between the FSLE derived from the altimetric velocity and the high resolution tracer image. Several test cases have demonstred the success of the inversion of sub-mesoscale tracer information to correct a mesoscale altimetric velocity field. A high resolution physico-biogeochemical coupled model of process and a high resolution realistic model of the Solomon sea have been used to assess the error associated with the inversion. The efficiency of the correction on the oceanic circulation has also been demonstrated using these models.

These results show the feasibility of assimilating tracer submesoscales into ocean models for the control of mesoscale dynamics and larger scales as deduced from altimetry and therefore the benefit of the joint use of tracer image and altimetric data for the control of ocean circulations.

ID 3.1-13

Monitoring the Ocean from observations

S. Guinehut, M.H. Rio, S. Mulet, and G Larnicol

CLS, Space Oceanography Division, Ramonville Saint-Agne, France


Producing comprehensive information about the ocean has become a top priority to monitor and predict the ocean and climate change. Complementary to modeling/assimilation approaches, an observation-based approach is proposed here. It relies on the combination of remote-sensing (altimetry and sea surface temperature) and in-situ (temperature and salinity profiles) observations through statistical methods.

The method uses first a multiple linear regression method to derive synthetic T/S profiles from the satellite measurements. These synthetic profiles are then combined with all available in situ T/S profiles using an optimal interpolation method. The thermal wind equation with a reference level at the surface is finally used to combine current fields from satellite altimetry with the thermohaline fields to generate the global 3D geostrophic current fields. Global temperature, salinity, absolute height and geostrophic current fields are thus available at a weekly period from the surface down to 1500-meter depth and a reanalysis is available for the 1993-2011 periods. The method has been assessed through comparison with independent in situ data sets as OVIDE sections or RAPID current meter array.

An analysis of the ocean variability using the 18-years long time series of the global 3D-fields of temperature, salinity and current has then been performed. The temperature variability of the 1993-2011 periods shows a clear warming that is visible at all depths and for all latitudes. If the variability is baroclinic with strong interannual signals in the tropics, it shows a clear long term trend at high latitude with depth consistent signals. Changes of ocean circulation are also been studied through mass transport at key regions and maximum Atlantic Meridional Overturning Circulation strength. Although high interannual variability is observed in the AMOC time series, it is not possible to extract a clear trend. Our analyses have also been compared to other observation-based approach (Roeammich and Gilson, 2009) and to outputs from numerical models (SODA, GLORYS).

Lastly, impact of the remote-sensing and in-situ observing system has also been studied using metrics derived from a Degrees of Freedom of Signal analysis applied to the optimal interpolation method.

ID 3.1-14

Improvement of assimilation procedure for sea surface height data by removing nonsteric seasonal height

T. Kuragano, Y. Fujii, T. Toyoda, N. Usui, K. Ogawa, M. Kamachi

Japan Meteorological Agency/ Meteorological Research Institute, Tsukuba, Japan


Many previous studies indicated that deviation of global mean seasonal variations between altimetric sea surface height (SSH) and thermal steric height was comparable to the global ocean mass variation, which is caused by total water flux between ocean and atmosphere and total runoff. While, in the local aspect, ocean barotropic responses, and thus redistribution of ocean mass, to wind and surface-pressure variations play an important role in SSH variation. Detecting density and mass variations in seawater column is one of the fundamental works for understanding baroclinic and barotropic ocean variation using altimeter data.

SSH caused by mass variation is simulated using a barotropic global ocean model forced by seasonally varying water flux, wind stress and surface pressure. The results indicate that the model SSH well represents mass-related SSH for gyre-scale regional means, and seasonal fluctuation of the altimetric SSH corrected for the model SSH is similar to that of sea surface dynamic height whose reference level is larger than 300 dbar.

The results indicate that mass-related SSH should be removed from altimetric SSH in some ocean data assimilation procedure in which the altimetric SSH is treated as the variable reflecting subsurface temperature and salinity. In the presentation, we will also show the effect of the correction for mass-related SSH in our ocean data assimilation system MOVE/MRI.COM.

ID 3.1-15

Use of GOCE MDT and error information in NEMOVAR, a variational data assimilation scheme for NEMO

Daniel J. Lea1, Rory Bingham2, Keith Haines3, Matt J. Martin1

1 Met Office, UK

2 Newcastle University, UK

3 Reading University, UK


The GOCE mission is now producing gravity information of useful accuracy for ocean data assimilation. We test a new GOCE based mean dynamic topography (MDT) along with a new online bias correction scheme which we developed for a 3D VAR assimilation system (NEMOVAR).

NEMOVAR is a multivariate assimilation scheme which assimilates sea level anomaly (SLA), remotely sensed and in-situ sea surface temperature, profile temperature and salinity data, and sea ice concentration data. Assimilating the SLA data requires a MDT to be provided. Currently in the Met Office's operational ocean forecasting system, FOAM, we use the CNES-CLS09 MDT which is a combination of GRACE data, and a synthetic MDT based on dynamic heights and velocities from insitu observations. As GOCE data is accurate to higher resolutions than GRACE this provides the opportunity to use a purely GOCE based MDT while allowing a bias correction scheme running online in NEMOVAR to correct the smallest scales of MDT. The bias correction scheme is designed to focus the bias correction at shorter length scales less than 200km where the MDT errors are known to be larger.

Results are presented from the new Met Office NEMOVAR system running at 1/4 degree global resolution. Several experiments are performed testing different MDTs and bias correction schemes. We compare the bias corrected MDTs to alternative MDT products. The results are assessed by looking at the observation minus background errors. In addition we compare to surface drifters and by examination of transports estimated by the model along various standard sections.

ID 3.1-16

Coastal Ocean Data Assimilation and Forecasting Using a Multi-Scale Three-Dimensional Variational Data Assimilation Scheme

Zhijin Li1, James C. McWilliams2, and Kayo Ide3

1 Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA


2 Department of Atmospheric and Oceanic Sciences, University of California, Los Angeles, CA, USA

3 Department of Atmospheric and Oceanic Science, Center for Scientific Computation and Mathematical Modeling, Institute for Physical Science and Technology, and Earth System Science Interdisciplinary Center, University of Maryland, College Park, MD, USA


Coastal ocean flows are often characterized by multiple spatial and temporal scales, and mesoscale and sub-mesoscale features, such as eddies and filaments, are energetic. A multi-scale three-dimensional variational data assimilation (MS-3DVAR) scheme has been formulated for high resolution coastal ocean models that represent a wide range of spatial scales, and implemented in real-time forecasting systems in support of a few operational coastal ocean observing systems and field campaigns. This MS-3DVAR scheme uses partitioned cost functions and thus background error covariances of multidecorrelation length scales. MS-3DVAR improves the effectiveness of the assimilation of both very sparse and high resolution observations. A variety of data assimilation experiments, known as Observing System Experiments (OSEs), is performed to illustrate MS-3DVAR. These OSEs are also used to assess the relative impacts of different types of observations. The observations assimilated primarily includes satellite altimetry data and sea surface temperatures, High Frequency (HF) radar surface velocities, and vertical profiles of temperature/salinity (T/S) measured by ships, moorings, Autonomous Underwater Vehicles and gliders. The combination of high resolution HF radar surface velocities and sparse T/S profiles allows representing meso-scale systems and producing analyses and forecasts with skill. It is suggested that a potentially promising observing network may be based on satellite altimetry data and SSTs along with sparse T/S profiles, but future satellite SSHs with wide swath coverage and higher resolution may be needed.

ID 3.1-17

Assimilation of sea surface temperature data into HYCOM in the Atlantic Ocean

L.N. Lima1, C.A.S. Tanajura1,2,3, K. Belyaev1,4, A. Santana1, D. Mignac1, J. Zhu5, J. Xie 5

1 Oceanographic Modeling and Observation Network (REMO), Center for Research in Geophysics and Geology, Federal University of Bahia, Salvador, Brazil

2 Physics Institute, Federal University of Bahia, Salvador, Brazil

3 Department of Ocean Sciences, University of California, Santa Cruz, USA

4 Shirshov Institute of Oceanology, Russian Academy of Sciences, Moscow, Russia

5 Institute of Atmospheric and Physics (IAP), Chinese Academy of Sciences, Beijing, China


Sea surface temperature (SST) analyses are commonly assimilated into ocean models in operational oceanography. In the present study, SST analyses from Reynolds et al. (2007) were assimilated into the HYbrid Coordinate Ocean Model (HYCOM) with an Ensemble Optimal Interpolation (EnOI) method in the Atlantic Ocean from 78°S to 50°N. The model was configured with 1/4° of horizontal resolution and 21 vertical layers. It was forced by the NCEP Global Data Assimilation System (GDAS) analysis. In order to reduce the computational cost and prepare the system for higher resolution SST data, super-observation (superobs) was employed. To investigate the impact of SST assimilation for each 3 days, four experiments were conducted for 6 months (Jan 2011 - Jun 2011): one control run without assimilation and three assimilation runs, namely, one with a superobs for each 2 x 2 SST grid cell and homogenous observational error of representativeness (OER) (expt 22H); one with a superobs for each 4 x 4 SST grid cell and homogenous OER (expt 44H), and one similar to the latter but with heterogeneous OER (expt 44NH) (Oke and Sakov, 2008). The best results were found by the expt 22H followed by the expt 44NH and expt 44H (Fig. 1a). While the averaged root mean square error of the SST 24 h forecasts of the model free run was equal to 1.49°C, the forecast errors from the EnOI analyses were equal to 0.55°C, 0.60°C and 0.66°C, respectively. By comparing the 24 h forecasts of the assimilation runs and the model free run with Argo T-S data, maximum errors decreased from 1.27°C to 0.39°C, in the mixed layer (Fig. 1b). Investigation is being conducted to reduce the error of surface salinity (not shown), since it increased due to SST assimilation.


Figure 1: Error calculated between (a) Reynolds and SST 24 h forecasts; and (b) Argo data and temperature 24 h forecast for entire HYCOM 1/4° domain from 1 Jan to 30 Jun 2011.

ID 3.1-18

Argo data assimilation into HYCOM with an EnOI method over the Atlantic Ocean

D. Mignac1,2, C. A. S. Tanajura2,3,4, A. Santana2, L. N. Lima2, J. Zhu5, J. Xie5

1 Graduate Program in Geophysics, Federal University of Bahia (UFBA), Salvador, Brazil

2 Oceanographic Modeling and Observation Network (REMO), Center for Research in Geophysics and Geology, Federal University of Bahia, Salvador, Brazil

3 Physics Institute, Federal University of Bahia (UFBA), Salvador, Brazil

4 Department of Ocean Sciences, University of California, Santa Cruz (UCSC), USA

5 Institute of Atmospheric and Physics (IAP), Chinese Academy of Sciences, Beijing, China


The best approach to assimilate Argo data into the HYbrid Coordinate Ocean Model (HYCOM) is still an open question. The main goal of this work is to investigate the impact of Argo data assimilation into HYCOM in the Atlantic Ocean using the Ensemble Optimal Interpolation (EnOI) method. The EnOI scheme was especially made regarding the hybrid nature of the model vertical coordinate according to Xie and Zhu (2010). The Argo profiles were converted into a pseudo observed thickness (D) and observed temperature (T) and salinity (S) were projected into the model layers. These 3 variables were assimilated in different steps, so that the layer thicknesses modify the baroclinic model velocities and the temperature is diagnosed below the mixed layer through the seawater state equation. The model was set with horizontal resolution of 1/4° and 21 vertical layers, ranging from 78°S-50°N and 98°W-20°E. In order to investigate the 24-hour forecast sensitivity to the EnOI scheme, 2 experiments forced by the NCEP Global Data Assimilation System (GDAS) analysis were conducted for 3 years (Jan 2010 – Dec 2012): (i) a control run without assimilation and (ii) a run with assimilation, considering the vertical localization of D, T and S. According to Fig.1, a great reduction of the model root mean square errors (RMSE) for T and S regarding the Argo independent observations was obtained. The assimilation run was able to decrease the error from 1.7°C to 0.9°C and from 0.29 PSU to 0.13 PSU until 1500 meters, driving the model errors towards the World Ocean Atlas climatology (WOA09) errors. Considering the RMSE vertical profiles of T and S (not shown), the greatest impacts were in the thermocline region with an error reduction of 1.2°C and 0.18 PSU for the entire domain.


Figure 1: Depth averaged RMSE of temperature (upper painel) and salinity (lower painel) until 1500 m for the control, assimilation run and WOA09 climatology regarding the Argo independent observations.

ID 3.1-19

Towards a dynamically balanced eddy-resolving ocean reanalysis

Peter. R. Oke1, Pavel Sakov2, Andreas Schiller3

1 CSIRO Marine and Atmospheric Research, Hobart, Australia

2 Bureau of Meteorology, Melbourne, Australia

3 CSIRO Marine and Atmospheric Research, Hobart, Australia


The generation and evolution of eddies in the ocean are largely due to instabilities that are unpredictable, even on short time-scales. As a result, eddy-resolving ocean reanalyses typically use data assimilation to regularly adjust the model state. In this study, we present results from a second-generation eddy-resolving ocean reanalysis that is shown to match both assimilated and with-held observations more closely than its predecessor; but involves much smaller adjustments to the model state at each assimilation. We compare version 2 and 3 of the Bluelink ReANalysis (BRAN) in the Australian region. Overall, the misfits between the model fields in BRAN3 and observations are 5-28% smaller than the misfits for BRAN2. Specifically, we show that for BRAN3 (BRAN2) the sea-level, upper ocean temperature, upper-ocean salinity, and near-surface velocity match observations to within 7.7 cm (9.7 cm), 0.68 C (0.95 C), 0.16 psu (0.18 psu), and 20.2 cm/s (21.3 cm/s) respectively. We also show that the increments applied to BRAN3 - the artificial adjustments applied at each assimilation step - are typically 20-50% smaller than the equivalent adjustments in BRAN2 (Figure 1). This leads us to conclude that the performance of BRAN3 is more dynamically consistent than BRAN2, rendering it more suitable for a range of applications, including analysis of ocean variability, extreme events, and process studies.


Figure 1: Root-mean-squared increments for temperature at 100 m depth for the period 2004-2006 from BRAN3 (left) and BRAN2 (right).


ID 3.1-20

High-resolution Simulations of the South Atlantic Ocean

Afonso M. Paiva, Mariela Gabioux, Vladimir S. Costa, Bruna F. Oliveira

Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia – COPPE

Federal University of Rio de Janeiro – UFRJ, Rio de Janeiro, Brazil


The HYCOM model has been configured in a set of nested domains with increased resolution, in order to simulated the oceanographic conditions along the shelf and slope waters off southeast Brazil. In this region (from approximately 15 to 30 degrees south and 35 to 50 degrees west) the Brazil current meanders intensely, generating eddies that are either pinched off or reabsorbed by the current, and interacts with a vigorous shelf circulation which is influenced by equatorward moving continental shelf waves, and moderate tides. Highresolution configurations (both with 1/12 and 1/24 degree horizontal resolution) were integrated for at least 10 years under synoptic atmospheric forcing, including tides. Altimeter data was incorporated in the model using the Coopper and Haines methodology, and the results were compared to free-runs not constrained by observations.

Long time series of velocity data in seven mooring sites, located over the shelf and the slope waters, and of tides at coastal stations, were made available for the model validation. Overall the model is very realistic in capturing the main oceanographic features of the simulated regions (Fig. 1), and generates energy in both high and low frequency bands that are statistically compatible with those derived from observations. The Cooper and Haines method was partially successful in driving the model results towards observations.

These model simulations were performed within the scope of the Oceanographic Modeling and Observation Network (REMO), and should form the basis for the generation of both short-term forecasts and long-term integrations, and for the development of an operational system which will give aid to the sustainable use of natural resources in Brazilian waters.

Further REMO developments should include more sophisticated assimilation techniques, and the model configuration for different regions along the Brazilian coast.


Figure 1: Observed (left) and simulated (right) sea surface temperature off southeast Brazil, showing the development of Brazil current eddies.

ID 3.1-21

Advancements in NCEP’s Global Ocean Data Assimilation System

S. G. Penny1,2, J. A. Carton1, D. Behringer2, E. Kalnay1

1 Universiy of Maryland, College Park, Maryland

2 NOAA Center for Weather and Climate Forecasting, College Park, Maryland


Data assimilation of the global ocean typically suffers from multiple sources of error, including: sparse observations, representativeness errors in observations, poor estimation of background errors, and model errors. Further, ensemble methods applied to the ocean typically suffer from under-representative ensemble sizes and collapsing ensemble spread.

A number of advancements have recently been made to NCEP’s developmental Global Ocean Data Assimilation System (GODAS) to address these errors. A new hybrid data assimilation algorithm, the Hybrid/Mean-LETKF, has been designed and applied to GODAS to improve the estimation of background errors. This hybrid algorithm combines the 3D-Var approach currently used in the operational GODAS with the Local Ensemble Transform Kalman Filter (LETKF). Experiments with simple models have shown such a hybrid approach reduces the required ensemble size to maintain stability with ensemble Kalman filter methods (EnKF). In addition, this hybrid approach was found to be more robust when using limited observation coverage compared to either 3D-Var or LETKF alone.

Additional techniques have also been applied to improve the ocean analysis. For example, as one of the primary sources of error in ocean modeling is the specification of surface fluxes, the surface flux fields are estimated along with the ocean state. The balances of the surfaces fluxes are maintained while applying a bias correction and spread correction to the surface-forcing ensemble.

ID 3.1-22

Not available for pubilcation

ID 3.1-23

Fresh water content variability over Bay of Bengal in a nested high resolution regional model

Hasibur Rahaman1, Matthew Harrison2, Stephen Griffies2, Debasis Sengupta3 and M Ravichandran1

1 Indian National Centre for Ocean Information Services (INCOIS), Ministry of Earth Sciences,Hyderabad,India

2. Geophysical Fluid Dynamics Laboratory, National e and Atmospheric Administration,Princeton,USA

3 Centre for Atmospheric and Oceanic Sciences, Indian Institute of Science,Bangalore,India


Seasonal and inter-annual variability of freshwater content (FWC) in the upper ocean over Bay of Bengal (BoB) have been studied from observation and model simulations. In this study we have used a nested multi-model approach using Modular Ocean Model (MOM4p1) at Geophysical Fluid Dynamics Laboratory's (GFDL's) global Climate Model (CM2.1) ocean resolution and a regional model with a uniform 25 km horizontal resolution. The first of its kind the regional model vertical resolution is about 1 meter near to the ocean surface and it gradually increases with depth. Model could capture the strong haline stratification over north Bay of Bengal (BoB). The observation show large fresh water content (FWC) ~ 3 m during fall (Sep-Nov) and winter (Dec-Feb) over north BoB (see figure). Model could capture these seasonal variations of upper ocean (0-30m) FWC with excellent agreement. Seasonal variation is dominant over north BoB, over southern BoB the robust seasonality is absent, however it shows signature of intraseasonal variability during fall and winter season.


Figure: Upper Ocean (0 - 30 m) Fresh Water Content (m) from model simulations and observations (IFREMER,ISHII) over Bay of Bnegal.

ID 3.1-24

The Atlantic Meriodional Overturning Circulation and Mesoscale Eddy Statistics in an Eddy Resolving Ocean Reanalysis (1992-2010)

J. G. Richman1, E. J. Metzger1, E. D. Douglass2, and O.-M. Smedstad3

1 Naval Research Laboratory, Stennis Space Center, MS, USA

2 ASEE Postdoctoral Fellow, Stennis Space Center, MS, USA

3 QuinetiQ North America, Stennis Space Center, MS, USA


The Naval Research Laboratory is performing an eddy-resolving 1993-2010 ocean reanalysis using the 1/12.5° global HYbrid Coordinate Ocean Model (HYCOM) that employs the Navy Coupled Ocean Data Assimilation (NCODA) scheme. The reanalysis and a companion simulation without data assimilation are forced by fluxes from the NCEP Climate Forecast System Reanalysis. The model has a mid-latitude resolution of ~7 km and employs 32 hybrid vertical coordinate surfaces. HYCOM contains a thermodynamic ice model, where ice grows and melts due to heat flux and sea surface temperature (SST) changes, but it does not contain advanced rheological physics. The ice edge is constrained by satellite ice concentration. Once per day, NCODA performs a 3D ocean analysis using all available observational data and the 1-day HYCOM forecast as the first guess in a sequential incremental update cycle. Sea surface height (SSH) anomalies are projected downward using synthetic profiles from the Modular Ocean Data Assimilation System.

Two diagnostics have been performed on the reanalysis and its non-assimilative twin, a comparison of the Atlantic Meridional Overturning Circulation (AMOC) for comparison with the RAPID/MOCHA estimates and a census of mesoscale eddies to compare with the altimeter eddy census of Chelton, et al. (Prog. Oceanogr., 2011). The AMOC between April 2004 and Dec. 2009 is slightly, but not significantly, larger in both the reanalysis and the simulation (20.0Sv and 19.3Sv, respectively) compared to the RAPID/MOCHA estimate (19.0Sv) for the same time period. The variability of the monthly averaged RAPID/MOCHA timeseries is much larger than the reanalysis with standard deviation in the AMOC of 1.8 Sv for RAPID/MOCHA compared to 0.98 Sv for the reanalysis. The reanalysis has a weak decrease of 1 Sv over the 17 years, but the trend is not significant. The reanalysis has a shallow DWBC with too much Labrador Sea Water and not enough Denmark Strait Overflow Water.

A census of mesoscale eddies and their properties has been made for the reanalysis and its non-assimilative twin using two techniques from SSH assuming geostrophy and from the vorticity and strain of the surface velocity. The two techniques are broadly consistent with the SSH census of Chelton et al (2011), although the eddies in the models have a smaller mean radius compared to the altimeter maps, which may be an artifact of the different resolution of the models compared to the altimeter maps. The eddies estimated from the model velocity are more nonlinear and more compact than the eddies estimated from the geostrophic velocity. In the reanalysis and simulation, the eddies with the longest lifetimes are both more nonlinear and more compact.

ID 3.1-25

A comparison of boundary current eddies in an ocean model and an ocean reanalysis

Tatiana Rykova and Peter Oke

CSIRO Marine and Atmospheric Research, Hobart, Australia


Using output from an 18-year near-global eddy-resolving ocean model, we analyse the modeled variability in five Western Boundary Current (WBC) regions, including the extensions of the Agulhas, East Australian Current (EAC), Brazil-Malvinas Confluence (BMC), Kuroshio and Gulf Stream regions. Analysis of a canonical cyclonic and anti-cyclonic eddy for each WBC region demonstrates that eddies in the BMC region have a significant barotropic component with strong bottom velocities, owing to weak stratification. We find that the vertical displacement of isopycnals in Kuroshio eddies is small, due to strong stratification; and that EAC eddies are weaker than the other regions, with relatively weak velocities, shallow penetration, and only moderate displacement of isopycnals. The canonical Agulhas eddies are the most striking of all the WBC regions, with large displacements of isopycnals (averaging over 450 m for anti-cyclonic eddies) that occur far deeper than eddies in other WBC regions. We contrast these results to an equivalent picture of eddies in the EAC from an 18-year ocean reanalysis to investigate the importance of dynamical consistency in reconstructing eddies and their characteristics.


Figure 1 : SLA (row 1), SSTA (row 2), and meridional velocity anomaly (colour) and potential density (contours; row 3 and 4) for a canonical cyclonic eddy for each WBC region. The grey lines in rows 1 and 2 are the anomalies for all eddies, and the bold-black line is the canonical mean. Velocities and SST anomalies are computed by removing the seasonal climatology for each eddy. The near-vertical dashed lines correspond to the contour for zero meridional velocity. The longitudes shown are the average longitudes of all eddies included in the analysis.

ID 3.1-26

Inter-annual variability of thermal inversion in the Bay of Bengal using Global Ocean analysis

S Sivareddy1, M Ravichandrn1, D Behringer2, P. S. Swathi3 and KVSR Prasad4

1 Indian National Centre for Ocean Information Services, Hyderabad 500 090, India

2 Environmental Monitoring Center, National Centre for Environmental Prediction (NCEP), National Oceanic and Atmospheric Administration (NOAA), Camp Springs, USA

3 CSIR-Centre for Mathematical Modelling and computer Simulation

4 Department of Meteorology and Oceanography, Andhra University, Visakhapatnam


Global Ocean analysis of INCOIS-GODAS has been improved by (i) assimilating observed in-situ salinity, (ii) incorporation realistic river run-off, (iii) implementing strong SST relaxation and (iv) Use of satellite based gridded winds. The improvements/degradations in the quality of ocean analysis associated with these choices are quantified in the Indian Ocean.

Ocean analysis obtained from the best configuration/experiment from the above is used for understanding inter-annual variability of thermal inversions in the Bay of Bengal. It is observed that the frequency of temperature inversions is more (sparse) in the northern (southern) parts of BoB during winter. The mean, estimated based on the analysis period, winter-time Inversion Starting Depth (Ds) and Inversion Layer Thickness (ΔD) in the north BoB are between 30-40 m and mean Maximum change of temperature in the inversion layer (ΔT) is between 0.5-1.5 °C. It is found that thermal inversions show inter-annual variability in the BoB. The analysis indicates that, formation of thermal inversions, in the south BoB, is controlled by planetary wave activity, whereas in the north BoB, all the factors such as fresh water flux, surface heat flux and horizontal advection play a significant role in modulating the strength of thermal inversions at an inter-annual time scales.

ID 3.1-27

The Impact of Velocity Data Assimilation from Drifters Using the Navy Coupled Ocean 3D Variational Data Assimilation System (NCODA-VAR)

Scott Smith 1 , Gregg Jacobs1, Robert Helber1, Matt Carrier1, and Pete Spence2

1 Ocean Dynamics and Prediction, Naval Research Laboratory, USA, Scott.Smith@nrlssc.navy.mil

2 QinetiQ North America, USA.


The Navy Coupled Ocean 3D Variational Data Assimilation (NCODA-VAR) system is one of the primary tools that the Navy uses operationally to ingest, process, quality and control, and assimilate ocean observations in near-real time in order to regularly update and improve the forecast skill of several different operational ocean prediction systems. One of the new additions to NCODA is the ability to relate velocity observations to subsurface temperature and salinity structures. The primary two predominant sources of data for NCODA are SST and SSH. The spatial resolution of SSH data is typically coarse to resolve smaller eddies, and SST data lacks the vertical correlation with the subsurface to steer the analysis towards these types of features.

The capability to assimilate velocity observations has been added to NCODA, which included constructing and implementing additional error covariances to cross-correlate velocity observation with temperature and salinity throughout the water column. The construction of the new error covariances stems from a historical database of temperature and salinity; and their relation to velocity is achieved using the equation of state to calculate the covariance of geopotential and the geostrophic balance to correlate geopotential to velocity.

Results of an experiment will be presented that demonstrate that the inclusion of velocity data assimilation improves the ability to resolve eddies. This experiment uses velocity data that were inferred and assimilated from about 300 surface drifters released during the summer of 2012 in the Northeastern Gulf of Mexico as part of the GoMRI CARTHE drifter experiment. An assimilative model experiment uses all available data up to the time of the drifter launch. At this time, a second experiment is initiated from the conditions of the first. The first experiment continues on assimilating only satellite data. The second assimilates the drifter velocity data and no satellite data. The resulting analyses are compared with observations from altimetry, and the experiment assimilating only drifter data has lower errors compared to the altimeter data than does the experiment assimilating the altimeter data. This is most probably because the altimeter data is relatively sparse, and cannot constrain the mesoscale evolution as well as the high density drifter data. The results show that the surface velocity observations are properly related to the subsurface temperature and salinity structure.

ID 3.1-28

Gulf of Mexico multi-scale data assimilation

Ashwanth Srinivasan1 and Eric P. Chassignet2

1 Tendral LLC, Key Biscayne, Florida, USA

2 COAPS, Florida State University Tallahassee, USA


A multi-scale data assimilation technique is implemented into a 1/25 degree resolution Hybrid Coordinate Ocean Model (HYCOM) configured for the Gulf of Mexico in order to assimilate different types of observations. The model includes tides, is forced at the surface by winds, heat fluxes and precipitation derived from operational weather models, and is nested within the global HYCOM Ocean Prediction System. At analysis times, an efficient online recursive filter is used to detide the forecast state. After detiding, different types of observations are assimilated sequentially to correct the model state from coarse to fine scales (spatial and temporal). First, monthly climatological T/S profiles are used to correct the model on long time scales to reduce bias and drift. Second, all available in-situ profiles are assimilated to provide interannual and seasonal corrections to the water column density structure. The mesoscale field is then constrained using altimeter data. We compare results from a 2009-2012 hindcast using standard assimilation (altimeter and in-situ assimilation without separation of scales) and this new multi-scale ap proach. Sensitivity experiments and hindcasts results demonstrate that the multi-scale strategy, is effective in correcting model bias, improves prediction skill and produces a dynamically more consistent analysis.

ID 3.1-29

REMO Ocean Data Assimilation System into HYCOM (RODAS.H1)

C. Tanajura1,2,3, A. Santana3, D. Mignac,3,4, L. Lima3, K. Belyaev3, J. Zhu5, J. Xie 5

1 Physics Institute, Federal University of Bahia (UFBA), Salvador, Brazil

2 Department of Ocean Sciences, University of California, Santa Cruz (UCSC), USA

3 Oceanographic Modeling and Observation Network (REMO), Center for Research in Geophysics and Geology, Federal University of Bahia, Salvador, Brazil

4 Graduate Program in Geophysics, Federal University of Bahia (UFBA), Salvador, Brazil

5 Institute of Atmospheric Physics, Chinese Academy of Sciences (IAP/CAS), Beijing, China


The first version of the Brazilian Oceanographic Modeling and Observation Network (REMO) ocean data assimilation system into the Hybrid Coordinate Ocean Model (HYCOM) (RODAS.H1) is under construction and should be finalized until the end of 2013. It is based on an Ensemble Optimal Interpolation (EnOI) scheme capable of assimilating sea surface temperature (SST), temperature (T) and salinity (S) profiles from Argo and along-track sea level anomalies (SLA) data. The system is mainly based on the work by Evensen (2003), Oke and Sakov (2008), Xie and Zhu (2010) and Xie et al. (2011). RODAS.H1 assimilates first SST, 6 hours later T-S Argo profiles, and 6 hours later along-track SLA. When SST and SLA are assimilated, the model state is altered for T and/or S, zonal velocity (u), meridional velocity (v) and layer thickness (D) in all model layers. When T-S profiles are assimilated, a pseudo "observed" D is first constructed and assimilated to correct D, u and v. Then, T and S are assimilated separately. For development, HYCOM was set up for the Atlantic from 78°S to 50°N with horizontal resolution of 1/4o and 21 vertical layers. HYCOM was forced with atmospheric analyses from the National Centers for Environmental Prediction (NCEP) Global Data Assimilation Analysis System (GDAS). The first test was performed from 1 January to 31 March 2011. By comparing the 24 h forecasts of the assimilation run and the model free run with Argo data, maximum T and S root mean square errors (RMSE) decreased from 2.2 oC and 0.34 psu to 1.5oC and 0.23 psu, respectively, in the thermocline region (Fig. 1). The RMSE of the SST 24 hour forecasts showed a reduction from 1.7oC to 1.2oC over the whole domain, and from 0.9oC to 0.5oC in the tropics. SLA correlation with respect to AVISO gridded data was improved from 0.27 to 0.55. Larger SST and SLA errors were attained in the Gulf Stream and Brazil-Malvinas Confluence regions. Work is being conducted with different radius of influence and error of representativeness to decrease these errors.


Fig.1. Vertical profiles of T (oC) (left) and S (psu) (right) 24 h forecast errors from surface to 2000 m for the model free run (control) (red line) and the assimilation run (black) with respect to 3441 Argo independent profiles for 3 months.

ID 3.1-30

Impact of assimilation of Aquarius sea surface salinity data in the GEOS system

Guillaume Vernieres1,2, M. M. Rienecker2, R. Kovach1,2, S. Akella1,2, C. Keppenne1,2, A. Borovikov 1,2

1 SSAI, Lanham, USA

2 GMAO, NASA GSFC, Greenbelt, USA


We present a methodology to correct the biases and errors of along track Aquarius level 2 sea surface salinity (SSS) data (version 2.0). Observed SSS retrievals are mapped into bulk salinity and the bulk salinity data are assimilated into the GEOS iODAS system. The assimilation significantly reduces the bias and RMS observation minus forecast differences at Argo in-situ locations, especially in the tropical and Southern oceans.

The results demonstrate the complementarity of in-situ (Argo) and Aquarius SSS observations and highlight problems that arise during the assimilation of the Aquarius data.

ID 3.1-31

Implementation of a variational data assimilation system in the Met Office's 1/4 degree global ocean model

Jennifer Waters1, Daniel J. Lea1, Matthew J. Martin1, Isabelle Mirouze1, Anthony Weaver2 and James Whille1

1 Met Office,Exeter, UK

2 CERFACS, Toulouse, France


An incremental 3DVAR data assimilation scheme, NEMOVAR, has recently been implemented in the Met Office's operational 1/4 degree global ocean model, FOAM. We describe the developments and tuning of NEMOVAR for the 1/4 degree global configuration. These include implementation of sea ice concentration assimilation, inclusion of altimeter and SST bias correction schemes and development of the error covariance specification. Key developments are the inclusion of flow dependent vertical lengthscales and a new efficient look-up table method for calculating the normalisation factors at each analysis time step.

We will also present a validation of the NEMOVAR system. The performance of the new NEMOVAR system is assessed against observations and the preceding Analysis Correction (AC) data assimilation system. The main findings are that the NEMOVAR system provides significant improvements to surface fields with the insitu-SST innovation standard deviation reduced by 32% compared to the AC system and an improvement of 24% in sea ice concentration innovation standard deviation. The largest improvements are shown to occur in the regions of high variability such as the frontal and eddy shedding regions and the marginal sea ice zone. These improvements are associated with changes in background error correlation lengthscales and improved fit to observations in NEMOVAR. However, some degradation to sub-surface fields in particular regions are identified in NEMOVAR and these are discussed along with potential developments and improvements to the system.

ID 3.1-32

Ocean reanalysis in the Indian Ocean and Pacific Ocean

C.Yan1, J.Zhu1, J. Xie1, G. Lv2, L. Cheng1

1 Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

2 National Marine Environmental Forecasting Center, Beijing, China


An ocean reanalysis system is developed in the Indian Ocean and Pacific Ocean. The specific assimilation schemes depend on different types of observations. For in-situ observations, instead of direct temperature or salinity profiles. The Layer thickness is assimilated to adjust the model layer thickness, barotropic and baroclinic fields. They are based on an ensemble-based method which estimates the background error covariance matrix by the ensembles from the output of model. The model used is NERSC MPI-Parallel version of the hybrid coordinate ocean model (HYCOM) with the resolution of 0.33ox0.33 ox28 levels. The various types of observations including profiles from XBT, CTD,TAO,ARGO etc, altimetry data and remotely-sensed sea surface temperature are assimilated into the HYCOM to generate an ocean reanalysis in the Indian ocean and Pacific ocean. Some evaluations on the reanalysis are made by the comparison to independent observations including temperature, salinity, current from drifters, and other reanalysis products.

ID 3.1-33

Global ocean reanalysis and data assimilation in NEMOVAR system

H. Zuo1, M. A. Balmaseda1, K. Mogensen1, J. Waters2, P. Janssen1,

1 ECMWF, Reading, U.K.

2 UKMO, Exeter, U.K.


A new global ocean reanalysis product is under development in ECMWF for the MyOcean2 project, which is based on ¼ degree NEMO ocean model and NEMOVAR data assimilation system (3D-VAR). An ice model (LIM2) is also coupled in the system. High resolution ocean model clearly shows some improvements, i.e. in resolving Western Boundary Currents, over the low resolution operational system that was used before. A series of parameters tuning are needed with the upgrading to the high resolution model, among which the background error correlation length-scales appear to be particularly important. A new formula for the background error correlation length-scales was developed, and the sensitivity experiments have been carried out to show the importance of correlation length-scales in model assimilation (NEMOVAR - 3DVAR) results. A simplified scheme for Rossby radius of deformation dependent horizontal correlation length-scale was developed and tested against some other schemes. Model performance is improved with the new scheme particularly in the tropic regions. A mixed-layer dependent vertical correlation length-scale scheme is also tested with some sensitivity experiments, results of profile data assimilation will be discussed here.