Ocean State Estimation for Climate Research

OCEAN STATE ESTIMATION FOR CLIMATE RESEARCH

T. Lee(1), D. Stammer(2), T. Awaji(3), M. Balmaseda(4), D. Behringer(5), J. Carton(6), N. Ferry(7), A. Fischer(8), I. Fukumori(1), B. Giese(9), K. Haines(10), E. Harrison(11), P. Heimbach(12), M. Kamachi(13), C. Keppenne(14), A. Köhl(2), S. Masina(15), D. Menemenlis(1), R. Ponte(16), E. Remy(7), M. Rienecker(14), A. Rosati(17), J. Schroeter(18), D. Smith(19), A. Weaver(20), C. Wunsch(12), Y. Xue(5)

(1)NASA Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr, Pasadena, CA 91109, USA. [email protected], [email protected], [email protected] (2)Institut für Meereskunde, KlimaCampus, University of Hamburg, Bundesstr. 53, 20146 Hamburg, Germany. [email protected], [email protected] (3) Division of Earth and Planetary Sciences, Graduate School of Science, Kyoto University, Kyoto 606-8502, Japan. [email protected] (4)European Centre for Medium-Range Weather Forecast, ECMWF, Shinfield Park, Reading RG2 9AX, UK. [email protected] (5)NOAA/National Center for Environmental Prediction/NOAA, 5200 Auth Rd, Camp Springs, MD 20746-4304, USA. [email protected], [email protected] (6)Department of Atmospheric and Oceanic Science, University of Maryland, 3413 Computer & Spaces Sci. Bldg., University of Maryland, College Park, MD 20742, USA (7)Mercator-Océan, 8-10 rue Hermès, 31520 RAMONVILLE ST AGNE, France. [email protected], [email protected] (8)Inter-governmental Oceanic Commission, UNESCO, 1 rue Miollis, 75732 Paris Cedex 15, France. [email protected] (9) Department of Oceanography, Texas A&M University, College Station, TX 77843-3146, USA. [email protected] (10)Reading University, Marine Informatics and Reading e-Science Centre, ESSC, Harry Pitt Bld, 3 Earley Gate, Reading University, Reading RG6 6AL, UK. [email protected] (11)NOAA/Pacific Marine Environmental Laboratory/OCRD, 7600 Sand Point Way NE, Seattle, WA 98115, USA. [email protected] (12)Massachusetts Institute of Technology, 77 Massachusetts Avenue, Department of Earth, Atmospheric and Planetary Sciences, MIT, MA 02139 USA. [email protected], [email protected] (13)Oceanographic Research Department, Meteorological Research Institute, 1-1 Nagamine, Tsukuba 305-0052, Japan. [email protected] (14)NASA Goddard Space Flight Center, Global Modeling and Assimilation Office, Greenbelt, MD 20771, USA. [email protected], [email protected] (15)Centro Euro-Mediterraneo per i Cambiamenti Climatici and Istituto Nazionale di Geofisica e Vulcanologia, Viale A. Moro 44, 40127 Bologna, Italy [email protected] (16)Atmospheric and Environmental Research, Cambridge, Massachusetts, USA. [email protected] (17)NOAA Geophysics Fluid Dynamics Laboratory, Princeton University Forrestal Campus 201 Forrestal Road, Princeton, NJ 08540-6649, USA. [email protected] (18)Alfred-Wegener-Institute for Polar and Marine Research, Postfach 12 01 61, 27515 Bremerhaven, Germany. [email protected] (19)Met Office Hadley Centre, FitzRoy Road, Exeter, EX1 3PB, UK. [email protected] (20) Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique, 42 avenue Gaspard Coriolis, 31057 Toulouse, France, [email protected]

1 are difficult to infer from observations 1. ABSTRACT alone, such as oceanic heat transport.

Spurred by the sustained operation and The vision of global ocean state estimation new development of satellite and in-situ as a means to synthesize ocean observing systems, global ocean state observations into a dynamically consistent estimation efforts that gear towards estimate of ocean circulation was climate applications have flourished in the developed under the “World Ocean past decade. A hierarchy of estimation Circulation Experiment” (WOCE) and was methods is being used to routinely further advanced as part of the World synthesize various observations with Climate Research Program’s “Climate global ocean models. Many of the Variability and Predictability Project” estimation products are available through (CLIVAR) and “Global Ocean Data public data servers. There have been an Assimilation Experiment” (GODAE). As a increasingly large number of applications result of these programs, and with the of these products for a wide range of sustained commitment of various funding research topics in physical oceanography agencies, climate-oriented ocean state as well as other disciplines. These studies estimation efforts have flourished in the often provide important feedback for past decade. Since OceanObs’99, many observing systems design. This white ocean state estimation systems have been paper describes the approaches used by developed to routinely produce estimates these estimation systems in synthesizing of the physical state of the ocean that are observations and model dynamics, publically available through data servers. highlights the applications of their State estimation products have been used products for climate research, and to study a wide range of topics in physical addresses the challenges ahead in relation oceanography and climate science as well to the observing systems. Additional as in geodesy and biogeochemistry. applications to study climate variability using an ensemble of state estimation 3. APPROACHES products are described also by a white paper by Stammer et al. A hierarchy of estimation methods has been adopted by various groups to perform 2. INTRODUCTION ocean state estimation, ranging from various filter methods such as objective As satellite and in-situ observing systems mapping or the so-called optimal for the global ocean (e.g., altimetry and interpolation (OI), 3-dimensional Argo) progress and mature with time, there variational (3D-VAR) method, and is an ever-increasing need to synthesize the Kalman filter, to smoother methods such diverse observations into coherent as the Green’s function, Rauch-Tung- descriptions of the ocean by using them to Striebel (RTS) smoother, and the adjoint constrain state-of-the-art ocean general method (a.k.a. Lagrange multiplier, circulation models (OGCMs). The Pontryagin’s principle, 4-dimensional resulting ocean state estimation products variational or 4D-VAR). Table 1 lists the provide estimates of the time-varying, estimation methods used by various three-dimensional state of the ocean and systems, many of which have a focus on help understand the variability of ocean climate applications (for diagnostic circulation and its relation to climate. They analysis, initialization of climate also offer a tool to estimate quantities that prediction, or both).

2 In filter estimation, the estimated state at a of different surface forcing on the ocean. certain time is influenced by observations The smoother approach is adopted by the up to that time. In smoother methods, consortium for Estimating the Circulation however, the estimated state at any time is and Climate of the Ocean (ECCO) and affected by observations in the future as Japan’s K-7 project (Table 1). well as in the past and present and sources Nevertheless, fitting the model equations of model errors are often estimated as well exactly with a particular set of controls as the state itself. The filter methods as could make it more difficult for the implemented by various assimilation estimation to fit the model to certain groups are typically computationally more aspects of the observations, especially over efficient than smoother methods such as a long integration. In such case, it is the adjoint. The filter approaches allow the important to identify and implement estimated state to deviate from an exact suitable control variables that account for solution of the underlying physical model these additional errors. For instance, in by applying statistical corrections to the addition to initial conditions and surface state, which are often based on some basic forcings, mixing coefficients (e.g., physical constraints (such as preservation Stammer 2005) or an “eddy stress” can be of the water mass properties, geostrophic estimated to correct for effects of balance, etc.). These corrections are meant mesoscale eddies that are not resolved by to compensate a diverse collection of coarse resolution models. errors in the models, such as their forcings, representation of advection and mixing, Many data types are routinely synthesized lack of resolution, erroneous bathymetry, to produce ocean state estimates. The type and missing physical processes. The and volume of data used vary with estimated state is generally closer to the systems. Previous studies have shown observations than unconstrained models complementarity of different data types in are (depending on the treatment of the improving ocean state estimates. For this model and data errors) but because they do reason, all systems use data from more not explicitly correct the corresponding than one observing system. Table 1 sources of these errors, application of these summarizes the observations synthesized results for climate diagnostics can be by the various systems. The most difficult. For instance, budgets of heat, commonly used data are sea level anomaly salt, and momentum, etc, cannot be closed from altimeters (e.g., TOPEX/Poseidon without invoking some internal sources and JASON-1), in-situ temperature and sinks of these quantities. profiles (e.g., from XBT/CTD, TAO moorings, and Argo), and salinity profiles Smoother-based estimation systems often from Argo. The impact of the data on the demand the estimated state to satisfy the estimation can be seen from the reduction model equations exactly over a certain of model-data misfits for the different time interval. The optimization of the state observations as a result of constraining the within such time interval is accomplished model with the observations. Fig. 1 is an by adjusting the sources of model error or example showing the reduction of model- so-called control variables, which are data misfit as a result of the optimization typically the initial state, surface forcing, of the ECCO-GODAE system. In this and model parameters. The resulting case, altimeter data, Argo T/S profiles, consistency between the estimated ocean and SST data have relatively large impact state and its physics permits explicit on the estimation. Note that the relative closure of property budgets, which greatly impact of different data on state estimation facilitates climate analysis such as heat depends on the assumptions about data and balance and diagnosis of the relative roles model errors. The smaller impact of the

3 QuikSCAT data may be because the Ocean state estimation products and tools assumed errors of the QuikSCAT data are have been applied to studies over a wide too large so that the difference between the range of topics in physical oceanography, estimated wind and QuikSCAT wind are for instance, the nature of sea level not weighted heavily enough. This brings variability (e.g., Carton et al. 2005, up an important point about the need to Wunsch et al. 2007, Fukumori et al. 2007, better understand the a priori errors, a Köhl and Stammer 2008a), water-mass subject that is discussed further in section pathways (e.g., Fukumori et al. 2004, 4. Many of the ocean state estimation Wang et al. 2004, Masuda et al. 2006, products are publically available through Toyoda et al. 2009), estimating surface data servers (Table 1). A few recent fluxes and river runoff (e.g., Stammer et studies have attempted to compare these al. 2004, Romanova et al. 2009), and products with a uniform set of interannual and decadal variability of the observational data (e.g., Gemmell et al. upper-ocean and heat content (e.g., Masina 2009). In the future, it would be valuable et al. 2004, Capotondi et al. 2006, Köhl et to provide more misfit diagnostics (e.g., al. 2007, Carton and Santorelli 2009). Fig. 1) for different synthesis products They have also been applied to research in calculated in a uniform way against the other disciplines such as biogeochemistry same data. Additional details of the state (e.g., McKinley et al. 2000 and 2004, estimation efforts can be found in Dutkiewicz et al. 2001 and 2006) and CLIVAR GSOP web page geodesy (e.g., Ponte et al. 2001, Dickey et http://www.clivar.org/organization/gsop/gs al. 2003, Chao et al. 2003, Gross et al. op.php. 2005). Due to limited space, here we only highlight a very limited number of examples for ocean circulation studies and discuss the implications for observing systems. More efforts should be made to provide feedback to the observing systems in a broader framework of climate variability.

Ocean state estimation products have been widely used to study the meridional overturning circulations (MOCs) as well as heat and freshwater transports, which are Figure 1. Non-dimensional model-data quantities that are difficult to measure misfits (normalized by data error) in the directly (e.g., Stammer et al. 2003, Lee and ECCO-GODAE system after the Fukumori 2003, Shoenefeldt and Schott optimization (left), and the reduction of 2006, Wunsch and Heimbach 2006, the model-data misfits as a result of the Balmaseda et al. 2007, Köhl and Stammer optimization (right). The former is the 2008b, Schott et al. 2007 and 2008, components of the so-called cost function Cabanes et al. 2008, Rabe et al. 2008, at the end of the optimization. The latter, Volkov et al. 2008). These analyses often describing the reduction of the cost have direct implications for the observing function, reflects the impact of various systems. For instance, Lee and Fukumori data on obtaining the estimate. Courtesy (2003) and Schott et al. (2007) identified of Patrick Heimbach of MIT and Ichiro the anti-correlated variability of meridional Fukumori of JPL. pycnocline transports in the western boundaries and the interior associated with 4. APPLICATIONS interannual-to-decadal variation of the

4 Pacific subtropical cells (STC). Therefore, variations of pycnocline transports in the the low-latitude western boundary currents interior and near the western boundaries (LLWBCs) and interior flow play opposite (with the interior being more dominant). roles in regulating upper-ocean heat These data, presented by Lee and content in the Pacific (with the interior Fukumori (2003) and Lee and McPhaden flow being more dominant) and an (2008), provide an effective constraint on observing system measuring one but not the estimates of pycnocline flow variability the other will not resolve the net transport. in ocean state estimation, as discussed by Such an anti-correlated variability is Lee and Fukumori (2003) and Schott et al. associated with the oscillations of the (2007). tropical horizontal gyres in the western- central Pacific Ocean in response to near- Another example of the feedback between local Ekman pumping. The oscillations of state estimation and observing system is the tropical gyres and their forcing have the study of decadal variability of the signatures in sea level anomaly as North Atlantic MOC by Wunsch and observed by altimeters and wind stress curl Heimbach (2006), Balmaseda et al. (2007), captured by scatterometers. (see Fig. 2 for Köhl et al., (2007), and Köhl and Stammer altimeter data examples). These signatures (2008). These studies discuss the complex provide some constraint on the estimated structure of the estimated decadal partial compensation of the western- variability in the MOC in depth and boundary and interior flows and thus latitude. In the ECCO-GODAE estimate tropical heat content. Nevertheless, the (Wunsch and Heimbach 2006), for satellite data have footprints that are too instance, the decadal weakening of the coarse to resolve the sharp changes near upper meridional circulation (associated the LLWBCs. Therefore, systematic with reduced northward transport above measurements of the LLWBCs, which are 1200 m) at 26°N is accompanied by a not well resolved by existing in-situ strengthening deep meridional circulation observing systems, would enhance the (i.e., the southward outflow of North Atlantic observational constraint on the state Deep Water and northward inflow of abyssal estimates. water) in the ECCO-GODAE estimate (Fig. 3). In the ECMWF operational analysis product (ORA-S3), there is a weakening northward flow (Balmaseda et al. 2007) at most latitudes of the North Atlantic basin. At 50°N the northward transport is well correlated with the intensity of the subpolar gyre (which can be derived from altimeter data) at internnual-to-decadal time scales, but the trends are opposite. Data assimilation substantially improves the estimated time-mean strength of the MOC (Fig. 4). There is apparent agreement between the ECMWF analysis with the estimates by Bryden et al. (2005) Figure 2. Interannual-to-decadal based on synoptic hydrographic sections in variability of SSH captured by the 1980s and 1990s. However, both TOPEX/Poseidon and JASON-1 altimeters Wunsch and Heimbach (2006) and imply oscillations of tropical gyres in the Balmaseda et al. (2007) discussed the large western tropical Pacific near 10°N and month-to-month fluctuations in the MOC 10°S, which result in counteracting estimate, which could cause aliasing if

5 sampled infrequently. Both studies showed that the trend in the meridional heat With their near continuous measurements transport was smaller than that of the at fixed locations, mooring observations MOC strength because surface warming have provided a valuable source of data to partially counteracted the weakening constrain and evaluate state estimation (upper) MOC. Therefore, an observing products (see the white paper by system that is capable of inferring changes McPhaden et al. for the global tropical in the volume transport alone may not be buoy array). These data also allow local adequate to monitor the heat transport. heat budget analyses near mooring sites These findings suggest that a systematic (e.g., Wang and McPhaden 2000, measurement network for the Atlantic McPhaden 2002). Although not all the MOC and heat transport at different budget terms can be measured directly, the latitudes (and different depths) beyond the analyses are helpful for evaluating the traditional synoptic hydrographic survey budget of state estimation products, and are needed. The extension of such a system they give better confidence for using these as the RAPID array is a step towards that products to study the budget on larger direction (please refer to the white paper scales, which are difficult to capture by Cunningham et al. on Atlantic MOC completely with mooring systems. The monitoring system). However, much of the studies of mixed-layer temperature balance ocean is still vastly under-sampled. The by Kim et al. (2004, 2007), Du et al. studies on decadal variation of the MOC (2008), and Halkides and Lee (2009) are re-emphasize the importance of having examples of the application of state systematic, sustained, and consistent estimation products for heat budget measurements of the global ocean analysis. In particular, the dynamical circulation in general. consistency of the smoothed estimates allows the heat budgets to be closed, permitting causal mechanisms to be analyzed.

Figure 3. Seasonal averages (3 months) of volume transport contours (m3 s−1) through time as a function of depth Figure 4. Meridional overturning estimated by the ECCO-GODAE system. circulation (MOC) variability at 26°N (in The weakening of the upper part of the Sv). The time evolution of the MOC for meridional circulation (associated with ECMWF’s ocean reanalysis (black) and the reduced Northward transport) is for the no-assimilation run (blue) is shown accompanied by a strengthening of the deeper using monthly values (thin lines) and meridional circulations (i.e., the southward annual means (thick lines). Over-plotted outflow of North Atlantic Deep Water and are the annual-mean MOC values from northward infow of abyssal water). After Bryden et al. (2005) based on synoptic Wunsch and Heimbach (2006). hydrographic sections and Cunningham et

6 al. (2007) based on RAPID mooring data quantification of model errors is only one (green circle). After Balmaseda et al. aspect. The identification of model error (2007). sources is critical for the smoothers. Some Apart from the studies of ocean model errors are attributable to multiple circulation, state estimation products and sources. For example, a biased SST tools have also many other applications. estimate in the equatorial Pacific cold For example, the estimation systems can tongue could be related to errors in wind, be used to evaluate the impact of existing surface heat flux, or mixing observations or the design of future parameterizations and advection (also observational systems (e.g., Oke and related to resolution). Determination of the Schiller 2007). The use of ocean state appropriate “controls” and accurate estimation products to initialize seasonal attribution of error sources are important to climate forecasts has become an important the fidelity of the estimation products. routine practice in operational and Moreover, assimilation groups need to experimental prediction centers. This work closely with the modeling subject is reviewed by the white paper by community to improve model physics, Balmaseda et al. so it will not be discussed especially those associated with the bias in here. As part of the CLIVAR/GODAE the mean state. global ocean reanalysis evaluation efforts, many assimilation groups in the US, The estimation of decadal and longer-term Europe, and Japan have participated in an variability remains a challenge due to the effort to compare a suite of derived lack of observations on these time scales in diagnostic quantities among different the ocean and for the forcing fields and the products and with observations. Among insufficient understanding of the errors other goals, the ensemble analysis helps associated with these observations. This is identify the minimum accuracy of compounded by the limitation in model observation that can distinguish the physics. Sustained observations of the products or to constrain the estimation ocean and its forcing are therefore critical effectively. The white paper by Stammer to the improvement of decadal and longer- et al. is related to the intercomparison of term ocean state estimation. various estimation products. Additional feedbacks of state estimation to Many of the state estimation products have observational requirements are addressed resolutions that are too coarse to represent by the white paper of Heimbach et al. mesoscale eddies. As these eddies affect the climate through their interaction with 5. CHALLENGES the larger scales, it is imperative that ocean state estimation efforts move towards Despite significant advances in ocean state eddy-permitting resolutions, to more fully estimation, many challenges remain. The utilize the existing observations that estimates of model and data errors are capture eddy variability (e.g., the multi- fundamental to the accuracy of the altimeter system), and to develop the estimation products. Therefore, the ocean capability to synthesize future observations state estimation community needs to work such as those from the Surface Water closely with the observationalists to obtain Ocean Topography (SWOT) mission. robust estimates of data errors (including Some of the state estimation products biases), an important issue that is often left already achieved eddy-permitting to the hands of assimilation groups. A resolutions (e.g., SODA at 0.25° and close collaboration with the modeling ECCO2 at 18 km). In Europe, several community is also needed to better groups (MERCATOR, INGV, and understand model errors. The University of Reading) are working in a

7 coordinated fashion under the framework studies over a wide range of topics in of the MyOcean project physical oceanography and other (www.myocean.eu.org) to produce 3 sets disciplines. These studies provide of ocean reanalyses at 0.25° resolution in important feedback to the requirement and 2010. Assimilation efforts that currently design of the observing systems. The focus on mesoscale ocean nowcasting estimation systems need further (e.g., HYCOM, Chassignet et al. 2009; improvement through a better MERCATOR, Bahurel et al. 2009) are understanding and quantification of expected to produce high-resolution ocean model, data, and forcing errors, improved state estimation products eventually that model physics and resolution, and the could be used for climate applications. inclusion of other components of the climate system as part of the estimation. Computational resources remain a critical Despite these challenges, ocean state issue for estimation efforts that are based estimation remains a pivotal approach to on ensemble or adjoint methods because understanding the climate system, and will they limit the ensemble size and model be even more so in the future as we aim to resolution that one can afford. Finally, the quantify the feedbacks in the system and coupled nature of the climate system investigate variabilities on longer time prompts for a coupled approach for state scales. estimation that includes different components of the climate system (such as 7. REFERENCES the ocean, atmosphere, land, cryosphere, and biogeochemistry) in order to properly 1. Balmaseda, M.A., G.C. Smith, K. account for the potential feedback among Haines, et al. 2007: Historical different components. Currently, coupled reconstruction of the Atlantic ocean-atmosphere, ocean-ice, and ocean Meridional Overturning Circulation physics-biogeochemistry state estimations from the ECMWF operational ocean are still in their infancy. Examples of reanalysis. Geophys. Res. Lett., 34, emerging efforts include (NOAA) GFDL’s L23615, doi:10.1029/2007GL031645. use of ensemble Kalman filter (Zhang et al. 2007) and (Japan) K-7’s use of adjoint 2. Bahurel P., et al., 2006: MERCATOR method (Sugiura et al. 2008) to perform OCEAN Global to regional ocean estimation using coupled ocean- monitoring and forecasting. In “Ocean atmosphere models. Coupled estimation Weather Forecasting: An Integrated efforts are expected to pick up momentum View of Oceanography”, E.P. in the coming decade. Chassignet and J. Verron (Eds.), Springer, 381-395. 6. SUMMARY 3. Bryden, H.L., H.R. Longworth, and Aided by the development of global ocean S.A. Cunningham, 2005: Slowing of observing systems, significant the Atlantic meridional overturning accomplishments have been achieved in circulation at 25 degrees N. Nature, global ocean state estimation efforts that 438, 655-657. are aiming towards climate applications. A suite of global ocean state estimation 4. Cabanes, C., T. Lee, and L.-L. Fu, products have been produced to describe 2008: Mechanisms of Interannual the time evolving three-dimensional ocean Variations of the Meridional circulation. There have been an increasing Overturning Circulation of the North number of applications of these products Atlantic Ocean. J. Phys. Oceanogr., 38, for oceanographic and climate-related 467-480.

8 climate and postglacial rebound 5. Capotondi A., A. Wittenberg , S. effects, Science, 298, 1975-1977. Masina, 2006. Spatial and temporal structure of ENSO in 12. Du, Y, T.-D. Qu, and G. Meyers, 2008: 20th century coupled simulations. Interannual variability of sea surface Ocean Modelling, 15, (3-4), 274-298 temperature off Java and Sumatra in a (http://dx.doi.org/10.1016/j.ocemod.20 global GCM. J. Clim., 21, 11, 2451- 06.02.004). 2465.

6. Carton, J.A., B.S. Giese, and S.A. 13. Dutkiewicz, S., Follows, M., Marshall, Grodsky, 2005: Sea level rise and the J and W. Gregg, 2001: Interannual warming of the oceans in the Simple variability in phytoplankton Ocean Data Assimilation (SODA) abundances in the North Atlantic. ocean reanalysis. J. Geophys. Res., Deep-Sea Res. II, 48, 2323-2344. 110, C09006, doi:10.1029/2004JC002817. 14. Dutkiewicz, S., M.J. Follows, P. Heimbach and J. Marshall, 2006: 7. Carton, J.A. and A. Santorelli, 2009: Controls on ocean productivity and air- Global Decadal Upper-Ocean Heat sea carbon flux: An adjoint model Content as Viewed in Nine Analyses. sensitivity study. Geophys. Res. Lett., J. Clim., 21, 6015-6035. 33, L02603, doi:10.1029/2005GL024987. 8. Chao BF, Au AY, Boy JP, et al., 2003: Time-variable gravity signal of an 15. Fukumori, I., T. Lee, B. Cheng, and D. anomalous redistribution of water mass Menemenlis, 2004: The origin, in the extratropic Pacific during 1998- pathway, and destination of NINO3 2002, Geochem Geophys. Geosys. 4, water estimated by a simulated passive Art. No. 1096. tracer and its adjoint. J. Phys. Oceanogr., 34, 582-604. 9. Chassignet, E.P., H.E. Hurlburt, E.J. Metzger, O.M. Smedstad, J. 16. Fukumori, I., D. Menemenlis, T. Lee, Cummings, G.R. Halliwell, R. Bleck, 2007: A near-uniform basin-wide sea R. Baraille, A.J. Wallcraft, C. Lozano, level fluctuation of the Mediterranean H.L. Tolman, A. Srinivasan, S. Sea. J. Phys. Oceanogr., 37, 338-358. Hankin, P. Cornillon, R. Weisberg, A. Barth, R. He, F. Werner, and J. Wilkin, 17. Gemmell, A. G. Smith, K. Haines and 2009: U.S. GODAE: Global ocean J. Blower (2009) Validation of ocean prediction with the HYbrid Coordinate model syntheses against hydrography Ocean Model (HYCOM). using a new web application J.Op. Oceanography, 22(2), 48-59. Oceangr., submitted.

10. Cunningham, S.A., T. Kanzow, and D. 18. Gross, R.S., I. Fukumori, and D. Rayner, et al., 2007: Temporal Menemenlis., 2005: Atmospheric and variability of the Atlantic meridional Oceanic Excitation of Decadal-Scale overturning circulation at 26.5 degrees Earth Orientation Variations. J. N. Science, 317, 935-938. Geophys. Res., 110, B09405, doi:10.1029/2004JB003565. 11. Dickey, J. O., S. L. Marcus, O. de Viron, and I. Fukumori, 2002. Recent 19. Halkides, D., and T. Lee, 2009: Earth oblateness variations: Unraveling Mechanisms controlling seasonal-to-

9 interannual mixed-layer temperature 27. Masina S., P. Di Pietro, A. Navarra A., variability in the southeastern tropical 2004. Interannual-to-decadal Indian Ocean. J. Geophys. Res., 114, variability of the North Atlantic from C02012, doi:10.1029/2008JC004949. an ocean data assimilation system. Climate Dynamics, 23, 531-546. DOI: 20. Kim, S.-B., T. Lee, and I. Fukumori, 10.1007/s00382-004-0453-6. 2004: The 1997-99 abrupt change of the upper ocean temperature in the 28. Masuda, S., T. Awaji, N. Sugiura, T. northcentral Pacific. Geophys. Res. Toyoda, Y. Ishikawa, K. Horiuchi, Lett., 31, L22304, 2006: Interannual variability of doi:10.1029/2004GL021142. temperature inversions in the subarctic North Pacific, Geophys. Res. Lett., 33, 21. Kim, S.-B., T. Lee, I. Fukumori, 2007: doi:10.1029/2006GL027865. Mechanisms controlling the interannual variation of mixed layer 29. McKinley, G. M.J. Follows and J. temperature averaged over the NINO3 Marshall, 2000: Interannual variability region. J. Clim., 20, 3822-3843. of oxygen fluxes in the North Atlantic. Geo. Res. Letters., 27, 2933-2936. 22. Köhl, A., D. Stammer, and B. Cornuelle, 2007: Interannual to 30. McKinley, G., M.J. Follows and J. decadal changes in the ECCO global Marshall, 2004: Mechanisms of air-sea synthesis. J. Phy. Oceanogr., 37, 313- CO2 flux variability in the Equatorial 337. Pacific and north Atlantic: Global Biogeochemical Cycles, vol 23. Köhl, A., and D. Stammer, 2008a: 18,doi:10.1029/2003GB002179. Decadal sea level changes in the 50- year GECCO ocean synthesis. J. Clim., 31. McPhaden, M.J., 2002: Mixed layer 21, 1866-1890. temperature balance on intraseasonal timescales in the equatorial Pacific 24. Köhl, A. and D. Stammer, 2008b: Ocean. J. Clim., 15, 2632-2647. Variability of the meridional overturning in the North Atlantic from 32. Oke, P. R., and A. Schiller, 2007: A 50-year GECCO state estimation. J. model-based assessment and design of Phys. Oceanogr., 38, 1913 -1930. a tropical Indian Ocean mooring array. J. Clim., 20, 3269-3283. 25. Lee, T., and I. Fukumori, 2003: Interannual to decadal variation of 33. Ponte, R.M., D. Stammer and C. tropical-subtropical exchange in the Wunsch, 2001: Improving ocean Pacific Ocean: boundary versus angular momentum estimates using a interior pycnocline transports. J. model constrained by data. Geophys. Climate. 16, 4022-4042. Res. Lett., 28, 1775-1778.

26. Lee, T., and M.J. McPhaden 2008: 34. Rabe, B., F.A. Schott, and A. Köhl, Decadal phase change in large-scale 2008: Mean circulation and variability sea level and winds in the Indo-Pacific of the tropical Atlantic during 1952- region at the end of the 20th century. 2001 in the GECCO assimilation Geophys. Res. Lett., 35, L01605, fields. J. Phy. Oceanpgr., 38, 177-192. doi:10.1029/2007GL032419. 35. Romanova, V., A. Köhl, D. Stammer, C. Klepp, S. Bakan et al., 2009:

10 Surface freshwater fluxes from 2004: Estimating Air-Sea Fluxes of GECCO and HOAPS. Submitted. Heat, Freshwater and Momentum Through Global Ocean Data 36. Schoenefeldt, R., and F. A. Schott, Assimilation. J. Geophys. Res., 109, 2006: Decadal variability of the Indian C05023, doi:10.1029/2003JC002082. Ocean cross-equatorial exchange in SODA. Geophys. Res. Lett., 33, 43. Stammer, D. 2005: Adjusting internal L08602, doi:10.1029/2006GL025891. model errors through ocean state estimation. J. Phy. Oceanogr., 35, 37. Schott, F.A., W.-Q. Wang, and D. 1143-1153. Stammer, 2007: Variability of Pacific subtropical cells in the 50-year ECCO 44. Sugiura, N., T. Awaji, S. Masuda, et al. assimilation. Geophys. Res. Lett., 34, 2008: Development of a four- L05604, doi:10.1029/2006GL028478. dimensional variational coupled data assimilation system for enhanced 38. Schott, F.A., L. Stramma, W. Wang , analysis and prediction of seasonal to et al., 2008: Pacific Subtropical Cell interannual climate variations. J. variability in the SODA 2.0.2/3 Geophys. Res., 113, C10017, assimilation. Geophys. Res. Lett., 35, doi:10.1029/2008JC004741. L10607, doi:10.1029/2008GL033757. 45. Toyoda, T., T. Awaji, S. Masuda, N. 39. Stammer, D., C. Wunsch, I. Fukumori, Sugiura, H. Igarashi, T. Mochizuki, and J. Marshall, 2002: State estimation and Y. Ishikawa, 2009: Interannual improves prospects for ocean research. variability of North Pacific eastern EOS, Transactions, American subtropical mode water formation in Geophysical Union, Vol. 83, Nr. 27, p. the 1990s derived from a 4- 289, 294--295. dimensional variational ocean data assimilation experiment. J. Geophys. 40. Stammer, D., C. Wunsch, R. Giering, Res., submitted. C. Eckert, P. Heimbach, J. Marotzke, A. Adcroft, C.N. Hill, and J. Marshall, 46. Volkov, D. L., T. Lee, and L.-L. Fu, 2002: The global ocean circulation 2008: Eddy-induced meridional heat during 1992 –1997, estimated from transport in the ocean, Geophys. Res. ocean observations and a general Lett., doi:10.1029/2008GL035490. circulation model. J. Geophys. Res., 107(C9), 3118, 47. Wang, W.-M and M.J. McPhaden, doi:10.1029/2001JC000888. 2000: The surface-layer heat balance in the equatorial Pacific Ocean. Part II: 41. Stammer, D., C. Wunsch, and R. Interannual variability. J. Phy. Giering, et al. 2003: Volume, heat, and Oceanogr., 30, 2989-3008. freshwater transports of the global ocean circulation 1993-2000, estimated 48. Wang, O., I. Fukumori, T. Lee, and B. from a general circulation model Cheng, 2004: On the cause of eastern constrained by World Ocean equatorial Pacific Ocean T-S variations Circulation Experiment (WOCE) data. associated with El Nino. Geophys. J. Geophys. Res., 108, 3007, Res. Lett., 31, L15310, doi:10.1029/2001JC001115. doi:10.1029/2004GL02472.

42. Stammer, D., K. Ueyoshi, A. Köhl, 49. Wunsch, C. and P. Heimbach, 2006: W.B. Large, S. Josey, and C. Wunsch, Estimated decadal changes in the

11 North Atlantic meridional overturning circulation and heat flux 1993-2004. J. 51. Zhang, S., M.J. Harrison, A. Rosati et Phys. Oceanogr., 36, 11, 2012-2024. al, 2007: System Design and Evaluation of Coupled Ensemble Data 50. Wunsch, C., R.M. Ponte, P. Heimbach, Assimilation for Global Oceanic 2007: Decadal trends in sea level Studies. Mon. Wea. Rev., 135, 3541- patterns: 1993-2004. J. Clim., 20, 3564. 5889-5911.

Table 1. Brief summary of ocean state estimation systems.

System Method Data Period Server ECCO- Adjoint Altimetry; scatterometry; 1992-2008 www.ecco-group.org GODAE (MIT- tide gauges; gravity; SST, AER), SSS; T & S profiles from USA XBT, CTD, Argo, TAO & other buoys, elephant seals (SeaOS); Florida Current; RAPID array ECCO1, USA Adjoint Altimetry; scatterometry; 1992-2001 www.ecco-group.org tide gauges; geoid; SST, SSS; T & S profiles from XBT, CTD, ARGO, TAO & other buoys,. G-ECCO, Adjoint Altimetry; scatterometry; 1952-2001 www.ecco-group.org Germany tide gauges; geoid; SST, SSS; T & S profiles from XBT, CTD, ARGO, TAO & other buoys,. ECCO-JPL, Kalman Altimetry, T profiles 1993- www.ecco-group.org or USA filter & from XBT/CDT, Argo, present www.ecco.jpl.nasa.gov/exte RTS TAO. rnal smoother ECCO2, USA Green’s Altimetry, SST, T & S 1992-2008 www.ecco2.org functions profiles from XBT, CTD, Argo, TAO; sea ice data GMAO/NASA, OI, Altimetry, T & S profiles 1993- gmao.gsfc.nasa.gov/research USA ensemble from XBT, CTD, Argo, present /oceanassim/ Kalman TAO filter GFDL/NOAA, Coupled SST, T profiles from 1979-2008 Data1.gfdl.noaa.gov/nomads USA Data XBT, CTD, ARGO, TAO /forms/assimilation.html Assimilati & S profiles from CTD, on ARGO (Ensemble Kalman Filter) GODAS, 3D-VAR SST, T profiles from 1979- www.cpc.ncep.noaa.gov/pro

12 NCEP/NOAA, XBT, CTD, Argo, TAO present ducts/GODAS USA SODA, USA OI Altimetry, Satellite and 1958-2007 www.atmos.umd.edu/~ocea in-situ SST, T & S n/data.html or profiles from MBT, XBT, soda.tamu.edu CTD, Argo and other float data, TAO and other buoys. ORA-S3 3D OI Altimeter (sea level 1959- Graphical: ECMWF, EU with anomalies and global present www.ecmwf.int/products/fo online trends), SST, T & S from recasts/d/charts/ocean/reanal bias XBT, CTD, Argo, TAO ysis correction Data: ensembles.ecmwf.int/thredd s/ocean/ecmwf/catalog.html MERCATOR- 3D-VAR Altimetry, SST, T & S 1960-2006 www.mercator-ocean.fr 2, France from XBT, CTD, Argo, TAO MERCATOR- SEEK Altimetry, SST, T & S 1980- www.mercator-ocean.fr 3, France filter from XBT, CTD, Argo, present TAO CERFACS, 3D-VAR SST, T & S profiles from 1960-2006 http://www.ecmwf.int/resear France EN3 ch/EU_projects/ENSEMBL ES/data/data_dissemination. html INGV, Italy OI SST, T & S profiles from 1958-2006 www.bo.ingv.it/contents/Sci XBT, CTD, Argo, TAO entific- Research/Projects/oceans/en act1.html DePreSys, OI SST, T & S profiles from 1950-2007 http://www.ecmwf.int/resear UK XBT, CTD, Argo, TAO ch/EU_projects/ENSEMBL ES/data/index.html

Reading, UK OI with T & S profiles from EN3 1960-2007 www.resc.reading.ac.uk/god S(T) and Argo at 1° and iva2 1987-2007 at 1/4° K-7, Japan Adjoint Altimetry, SST, T from 1960-2006 www.jamstec.go.jp/frcgc/k- XBT, CTD, Argo, TAO 7-dbase2/ MOVE-G, 3D-VAR Altimetry, SST, T & S 1948-2007 www. mri- Japan from XBT, CTD, Argo, jma.go.jp/ Dep/oc/oc.html TAO