Argo: the Challenge of Continuing 10 Years of Progress

GODAE Special Issue Feature

By Dean Roemmich and the Argo Steering Team Argo The Challenge of Continuing 10 Years of Progress

Abstract. In only 10 years, the Argo Program has grown from an idea into a functioning global observing system for the subsurface ocean. More than 3000 Argo floats now cover the world’s ocean. With these instruments operating on 10-day cycles, the array provides 9000 temperature/salinity/depth profiles every month that are quickly available via the Global Telecommunications System and the Internet. Argo is recognized as a major advance for oceanography, and a success for Argo’s parent programs, the Global Ocean Data Assimilation Experiment and Climate Variability and Predictability, and for the Global Earth Observation System of Systems. The value of Argo data in ocean data assimilation (ODA) and other applications is being demonstrated, and will grow as the data set is extended in time and as experience in using the data set leads to new applications. The spatial coverage and quality of the Argo data set are improving, with consideration being given to sampling under seasonal ice at higher latitudes, in additional marginal seas, and to greater depths. Argo data products of value in ODA modeling are under development, and Argo data are being tested to confirm their consistency with related satellite and in situ data. Maintenance of the Argo Program for the next decade and longer is needed for a broad range of climate and oceanographic research and for many operational applications in ocean state estimation and prediction.

26 Oceanography Vol.22, No.3 60°E 120°E 180° 120°W 60°W 0° 80°N 400 300 60°N 200 100 40°N 75 50 20°N 25 20 0° 15 20°S 10 5 40°S 4 3 60°S 2 1

10000 9000 Profiles/month 8000 Floats/month 7000 6000

Count 5000 4000 3000 2000 1000 0 JFMAMJJASONDJFMAMJ JASONDJFMAMJ JASONDJFMAMJ JASONDJFMAMJ JASONDJFM 2004 2005 2006 2007 2008 2009 Figure 1. (Top) Map of the number of good Argo profiles obtained in each 1° x 1° box during the period January 2004 to March 2009. (Bottom) Number of floats per month (red) providing good data and number of good profiles per month (green).

Introduction uneven quality, with a mixture of instru- combination of the array’s high-quality Data assimilating models of the histor- ment types and problems due to system- temperature and salinity sensors and ical ocean for the period 1950–2000 are atic errors (e.g., Wijffels et al., 2008). its comprehensive data management limited to using subsurface data sets that The Argo Program presents a unique system produces climate-quality data, were “opportunistic” in nature rather opportunity to correct many of these with new techniques being developed than purposefully designed for regular shortcomings in order to obtain more to identify and minimize systematic global coverage, and are consequently continuous, consistent, and accurate errors. With the global array providing deficient in some respects. The historical sampling of the present-day and future 9,000 temperature/salinity profiles per data were collected mostly for regional states of the ocean. Profiling float month (Figure 1), Argo has far surpassed objectives and were restricted to the technology removes the constraint its historical precursors in data coverage tracks of research vessels and commer- of needing to have a ship present at and accuracy. cial ships. These limitations ensured the time of measurement, making it Here, we summarize plans for sparseness and inhomogeneity in spatial possible to obtain high-quality data enhancing Argo’s value in the coming and temporal data distribution, even in anywhere at any time. Argo is designed years, with attention to ocean data the relatively well-sampled Northern to observe large-scale (seasonal and assimilation (ODA) applications. The Hemisphere oceans. Few measurements longer, thousand kilometer and larger) next section describes plans that include were collected south of 30°S. In addition subsurface ocean variability glob- improvements to data coverage and to sparseness, the historical data are of ally (Roemmich et al., 1999). The to data quality, followed by a section

Oceanography September 2009 27 describing gridded Argo data prod- scale variability. The design was based latitudes is being resolved adequately. ucts, how they differ from historical on statistics from satellite altimetric Argo should be sustained in order to counterparts, and their effectiveness height and from earlier subsurface ocean increase the value of its present five-year in resolving large-scale variability for data sets (Roemmich et al., 1999). The global time series, but it can evolve for comparison and evaluation of ODA prescribed 3° x 3° x 10-day spacing of the greater efficiency and effectiveness. models. The final section addresses the array between 60°S and 60°N decreases Beyond the design of the Argo need to examine the consistency of Argo the distance between instruments with array, recent developments in profiling and in situ and satellite-derived surface increasing latitude, but not as steeply as float technology create opportunities data sets. Consistency is the key issue for the statistics of variability indicate to be for extending Argo’s core objectives integrating global observations through appropriate (e.g., Stammer, 1997). The (Roemmich et al., 2009). Some floats ODA models. present design is a compromise made are now active in the seasonally ice- for accurate mapping of tropical climate covered zones poleward of 60°. What The Evolution of Argo variability (see section below on “The should be Argo’s sampling plan for the New Argo Domains, Sensors, and Argo-Era Global Ocean”) while taking a high-latitude ocean, and how many Sampling Enhancements more exploratory approach to the high- additional floats are needed there? Glider Argo is a broad-scale array, designed latitude ocean. Using five years of global technology (Davis et al., 2002) makes to accumulate about 100 profiles per Argo data accumulated from 2004–2008, systematic sampling of ocean boundary season in every 10° square of ocean. The and simultaneous altimetric height data, currents a possibility. What are the array is not eddy-resolving, but eddy the Argo design is being revisited. An global requirements for high-resolution noise is reduced by averaging over many important question is whether interan- sampling in the boundary currents profiles in a region to estimate large- nual variability at middle and high and marginal seas, to complement the broad-scale Argo array? New sensors for biological and geochemical parameters, Dean Roemmich ([email protected]) is Professor, Scripps Institution of Oceanography, for wind and rainfall, and for better University of California, San Diego (UCSD), La Jolla, CA, USA. Members of the Argo sampling of temperature and salinity Steering Team are (alphabetically): Mathieu Belbéoch (Argo Information Centre, France), structure in the ocean’s surface layer Howard Freeland (Institute of Ocean Sciences, Canada), Sylvia Garzoli (National could all increase Argo’s value, but also Oceanic and Atmospheric Administration, Atlantic Oceanographic and Meteorological its cost. The addition of oxygen sensors Laboratory, USA), W. John Gould (National Oceanography Centre, Southampton [NOCS], to Argo floats holds high promise for UK), Fiona Grant (Irish Marine Institute, Ireland), Mark Ignaszewski (Fleet Numerical addressing global carbon cycle issues Meteorology and Oceanography Center, USA), Brian King (NOCS, UK), Birgit Klein (Riser and Johnson, 2008), and over (Bundesamt für Seeschifffahrt und Hydrographie, German),Pierre-Yves Le Traon (Institut 100 Argo floats are presently equipped français de recherche pour l’exploitation de la me [Ifremer], France), Kjell Arne Mork with oxygen sensors. Prototype floats (Institute of Marine Research, Norway), W. Brechner Owens (Woods Hole Oceanographic carrying many other new sensors have Institution, USA), Sylvie Pouliquen (Ifremer, France), Muthalagu Ravichandran (Indian been deployed. Present float designs are National Centre for Ocean Information Services, India), Stephen Riser (University not capable of operating below 2000 m, of Washington, USA), Andreas Sterl (Royal Netherlands Meteorological Institute, but such measurements may be required Netherlands), Toshio Suga (Japan Agency for Marine-Earth Science and Technology, Japan), because decadal climate signals are Moon-Sik Suk (Korea Ocean Research & Development Institute, Korea), Philip Sutton known to extend into the deepest layers. (National Institute of Water & Atmospheric Research, New Zealand), Virginie Thierry Deep sampling will require develop- (Ifremer, France), Pedro J. Vélez-Belchí (Instituto Español de Oceanografía, Spain), mental work on both floats and sensors. Susan Wijffels (Commonwealth Scientific and Industrial Research Organisation, Marine Careful planning is needed to determine and Atmospheric Research, Australia), and Jianping Xu (Second Institute of Oceanography/ effective strategies for deployment of any State Oceanic Administration, China). extensions to the Argo array.

28 Oceanography Vol.22, No.3 Improving Argo Implementation the entire pre-Argo history of oceanog- biofouling or other causes. This drift can Evolving the Argo Program not only raphy (Figure 3), the program is not yet be corrected through careful statistical means reviewing its design and objec- achieving its ambitious sampling objec- comparison of sequences of float salinity tives, but also improving its imple- tive there (Figure 2). values to nearby high-quality profiles mentation with respect to the original Another key objective in Argo imple- (Wong et al., 2003), which might consist objectives. Although more than half of mentation is to minimize systematic of either shipboard conductivity-temper- active Argo floats are presently south of errors in the data stream (see also ature-depth (CTD) data or nearby float the equator, the Southern Hemisphere Le Traon et al., 2009, this issue). Two data. Second, pressure offsets have been is under-sampled relative to the original such systematic errors in Argo data have identified in some floats (e.g., Willis program design (Figure 2). Indeed, over been identified and to a large extent et al., 2007; Uchida and Imawaki, 2008), the Southern Hemisphere, the present corrected. First, over a period of years, resulting from pressure sensor drift Argo array is more than 500 floats short slow drift in conductivity measure- and from errors in float software. Such of its designed number. This shortfall ments occurs in some floats, due to systematic errors, some of which were in the number of floats, in spite of Argo having achieved 3000 active instruments, is due to a combination of factors. Many floats are deployed in marginal seas or poleward of 60°, and although these are of value, they were not considered in Argo’s original design. Other instruments are not producing good profile data due to technical failures, which is being corrected through deployment of improved instruments. The coverage shortfall requires increased attention to Southern Hemisphere deployments by Argo national programs. Occasional ship visits to the remote regions of the ocean are essential for maintaining Argo. Although Argo is producing more Figure 2. Blue bars show the number of Argo floats per 10° of latitude providing good profile data south of 30°S during a single profile data as of March 2009, excluding those in marginal seas. Red bars show Argo’s design austral winter than were produced in requirement for 3° x 3° open ocean sampling.

35°S

45°S

55°S

65°S

75°S 0° 100°E 160°E 60°W Figure 3. Location of all 4,093 temperature/salinity stations to depths of at least 1000 m during austral winter (July/August/September), south of 30°S from 1950–2000 (red dots; source: World Ocean Database), compared to 6,291 Argo station locations (black dots) from July/August/September 2008.

Oceanography September 2009 29 not anticipated, highlight the need for The Argo array is not yet “complete” (Curry et al., 2003; Wong et al., 1999; rapid identification and prompt correc- with respect to its original design and Boyer et al., 2005). Second, the Argo-era tion of hardware errors or software flaws. objectives. The highest priority for ocean is better sampled than the histor- A promising technique for detecting Argo’s international partnership is to ical ocean, especially in the Southern systematic errors is comparison of implement further improvements in Hemisphere (Figure 3), leading to lower satellite altimetric height with Argo data coverage and quality to meet these estimation errors. For the first time, it is steric height from sequences of profiles requirements. At the same time as possible to construct mean temperature and flagging large differences for more the Argo Program is being improved and salinity fields for the ocean over careful examination (Guinehut et al., and maintained for its original goals, a specific time period. Historical data 2009). Another technique is the use of extensions to the array should be climatologies are created by blending climatological data for flagging statistical introduced carefully to increase Argo’s regional data collections from different outlier profiles and instruments, a capa- long-term value. eras and, as a consequence, the end bility that is being improved (Gaillard products are weighted toward different et al., in press) as Argo-era data supple- The Argo-Era years in different regions. The sparseness ment earlier data sets with more appro- Global Ocean of historical data and their spatial and priate mean and variability statistics. Argo and Historical Data Sets temporal inhomogeneity make it difficult Because few Argo floats can be recov- A key step in demonstrating the value of to assign error bounds to these clima- ered for sensor recalibration, the final Argo is to show how well it represents tologies (e.g., Roemmich and Sutton, quality of Argo data will depend on the the present-day ocean, including the 1998). Finally, care taken to minimize existence and availability of high-quality mean, annual-cycle, and large-scale systematic errors in the Argo data set shipboard CTD data. Shipboard CTD variability. In modeling applications, leads to Argo-only climatologies that are transects are useful not only to detect data climatologies are used as initial not contaminated by mixing of different systematic errors in Argo float data but states for predictive models, or as mean instrument types. also in joint analyses with Argo data states with known variance, to limit Figure 4 shows an example of the for better description of interannual to unrealistic model variability and trends. differences between Argo (Roemmich multidecadal signals in the ocean. For Climatologies based on Argo data are and Gilson, 2009) and World Ocean example, recent work of Argo Steering more realistic representations of the Atlas 2001 (WOA01) historical data Team member Pedro Vélez-Belchí and modern ocean than historical data climatology (Conkright et al., 2002). The figure shows mean steric height from Argo, 0/2000 dbar, averaged over the period 2004–2008, during which the Argo data set has been used to tackle the Argo array had global coverage. The a wide variety of basic research problems, and difference between this Argo-era mean and WOA01, which is based on data data quality exceeds original expectations. collected over more than 50 years, is especially notable south of 30°S. There, “ the zonal mean difference is 5 dyn cm colleagues compares Argo data in the climatologies for several reasons. First, between 40° and 50°S, and differences North Atlantic with nearby CTD data the ocean has changed substantially are 10 dyn cm or more in some areas. collected by the UK’s 2001–2007 Rapid in the past several decades, becoming” Main causes of the Argo-minus-WOA01 Climate Change program, showing warmer overall (e.g. Domingues et al., differences are decadal change and similar multidecadal changes in these 2008; Levitus et al., 2005, 2009), and mapping errors due to sparseness in the data sets relative to earlier transects exhibiting significant regional changes historical data set. For modeling the along 24.5°N. in temperature/salinity characteristics present-day ocean, Argo climatologies

30 Oceanography Vol.22, No.3 30 25 20 15 40°N 10 8 6 5 4 2 0° 0 –2 –4 –5 –6 –8 –10 40°S –15 –20 –25 –30 60°E 180° 60°W Figure 4. Contours indicate the steric height of the sea surface from Argo data, 0/2000 dbar (dyn cm), 2004–2008 mean. Color shading indicates the difference in steric height, Argo-minus-WOA01 (World Ocean Atlas 2001).

6.0 will replace the historical data products to provide initialization and background 5.0 states that are consistent with the era that is represented. 4.0 Effectiveness of Argo Sampling The effectiveness of Argo in resolving 3.0 large-scale ocean variability can be RMS (cm) tested in a variety of ways (Roemmich and Gilson, 2009), including statistical 2.0 measures that complement model-based Observing System Evaluation activities 1.0 (Oke et al., this issue). One estimate uses satellite altimetric height as a proxy for steric height. On large spatial scales, 0.0 60°S 40°S 20°S 0° 20°N 40°N 60°N steric height and total sea surface height Figure 5. Zonally averaged large-scale (10° x 10° x 3-month) nonseasonal sea are very similar. By subsampling altim- surface height signal (blue) and Argo sampling noise (red) for 15 years of etric height fields at the locations of Argo sustained Argo sampling at the 2007 level, estimated from satellite altimetry profiles, interpolating the subsampled (see text). The black line shows how the apparent signal is reduced in a shorter (four-year) record. Results adapted from Roemmich and Gilson (2009) data, and then comparing to the full altimetric height data set, both the large- scale signal and noise of Argo steric height fields can be estimated. Sampling One such experiment is illustrated maintained. The 15-year gridded altim- experiments have been carried out to test in Figure 5. Here, the goal is to esti- etric height record (Ducet et al., 2000) the impact of Argo’s increasing coverage mate Argo’s ability to detect large-scale from 1993–2007 is subsampled each between 2004 and 2007, and to test variability over 15 years of sustained year at the location and year-day of the its ability to resolve signals of varying sampling by assuming that Argo’s 2007 Argo data set. The subsampled spatial and temporal scales. spatial coverage in the year 2007 is anomalies from the 15-year mean and

Oceanography September 2009 31 annual cycle are objectively interpolated, compare with multi-data-set model drifter data pairs located within 60-km and then both full and subsampled results and with data-withholding “scaled-distance” of one another during anomaly grids are smoothed with a model experiments. the period 2004–2008. Here, “scaled- 10° x 10° x 3-month running mean. After distance” includes a time difference term, smoothing, the temporal RMS signal is Argo and Ocean Surface with 1 day equivalent to 10 km. Figure 6 estimated from the full data set, and the Data Sets shows the means and standard devia- RMS noise is estimated from the full- Argo is the dominant subsurface data tions of Argo-minus-drifter temperature minus-subsampled differences. Figure 5 set for the present-day ocean, but ODA as a function of distance, sorted into shows the zonal means of the RMS models assimilate sea surface data sets 5-km bins. The number of nearby pairs signal and the RMS noise. As expected, as well, including sea surface height, sea increases from 214 in the 0–5 km bin the signal-to-noise ratio is highest in surface temperature, and air-sea fluxes to 2,987 pairs in the 55–60 km bin. The the tropics due to enhanced signal and of heat, water, and momentum. It is mean differences are small, 0.02°C and reduced noise. Figure 5 also illustrates important to examine the consistency of less, and not statistically significant. how large-scale variability grows with the these data sets with Argo and with one Significant stratification was found duration of the data set as a longer-term another where they have complementary between the 1-m and 5-m measure- mean is estimated and removed and or overlapping information content. The ments only in a small subset of low-wind decadal variability starts to be observed. ODA models allow for random data daytime conditions, so Argo data are This method is one of several being used errors, but problems may arise when a good approximation of bulk SST at to assess Argo’s errors and its effective- systematic errors create inconsistencies most times. The comparison (Figure 6) ness (Roemmich and Gilson, 2009), between data types. suggests that Argo data may be valuable including mapping and comparison for SST estimation on a global basis. of Argo subsets, comparison to other Sea Surface Temperature observing system elements, and formal Sea surface temperature (SST) is esti- Sea Surface Height optimal interpolation error estimates. mated from satellite measurements, The relationship of satellite-derived sea In addition to the interpolated Argo- using sparse ocean surface drifters surface height (SSH) and steric height only data set (Roemmich and Gilson, (at ~ 1-m depth) and other in situ variability is central to Argo. A global 2009) used in Figure 4 for purposes of SST measurements for bias correc- study of SSH and steric height varia- illustration, groups around the world are tion (e.g., Reynolds et al., 2002). Argo tions during 1993–2003 (Guinehut et al., developing similar products. To promote profiles, which collect their shallowest 2006) revealed high correlation between their dissemination and usefulness, the data at around 5 m, are not used in most the two with some systematic differences Argo Steering Team is identifying global SST products at present. Questions due to barotropic ocean forcing. The Argo analyses that are available for include the magnitude of stratifica- consistency of global SSH variability distribution (http://www.argo.ucsd.edu/ tion between the depth of drifter SST in 2003–2007 with the component AcGridded_data.html). There is much measurements and the shallowest Argo changes in Argo steric height and ocean to be gained from comparing techniques data and whether Argo floats, which mass (from the Gravity Recovery And and results of different analyses as are more plentiful than surface drifters, Climate Experiment [GRACE] satellite well as from developing products for are useful for SST estimation. Similarly, mission) has been examined by Willis different applications. It is essential to Argo’s potential usefulness in combi- et al. (2008), Cazenave et al. (2009), and provide ODA applications not only with nation with a scheduled sea surface Leuliette and Miller (2009). Although accurate data sets and best estimates of salinity satellite mission is of interest. the annual cycle in globally averaged the error. While ODA models continue To investigate the issue, the surface SSH was consistent with the sum of to develop, it is also critical to provide drifter and Argo data sets were searched steric and mass-related components, individual data sets that are extensive to identify nearby pairs of measure- differing conclusions were reached enough for statistical interpolation, to ments. There were 21,100 Argo profile/ regarding the four-year increase in

32 Oceanography Vol.22, No.3 1.00 to seasonal displacement of the zonal oceanic boundary current fronts at those 0.80 latitudes. Comparison of air-sea fluxes of freshwater with oceanic freshwater Standard Deviation storage is more problematic. Patterns 0.60 of freshwater storage are spatially more complex than heat storage, and esti- 0.40 mates of evaporation and precipitation are subject to large errors. A challenge 0.20 for ODA models is to exploit Argo’s global measurements of salinity for Mean Difference

Mean and Std Dev of T-difference Mean and Std Dev of improved estimation of variability in the 0.00 hydrological cycle.

–0.20 Discussion 0. 10. 20. 30. 40. 50. 60. A key goal of the Argo Program is to Scaled Distance (km) provide a global data set of value for Figure 6. Argo-minus-surface-drifter mean and standard deviation of temperature difference (°C) as a function of “scaled distance” (see text), in 5-km bins, for assimilation by ODA models that is also 21,100 nearby pairs of observations. extensive enough to enable evaluation of the results of those models. The Argo array now includes about 3,000 instruments providing 9,000 globally distrib- SSH (by about 12 mm) in relation to reproduce the annual cycles in data sets uted temperature and salinity profiles its components. A longer time series with overlapping information content monthly from the sea surface to mid- is needed to be more definitive. These and can successfully rationalize differ- ocean depth. Five years of Argo data, and other examples illustrate the strong ences such as those seen in Figure 7. including 400,000 profiles, have been need to close the ocean’s mass and heat collected since sparse global coverage was budgets with careful measurements Air-Sea Fluxes achieved in early 2004, comprising stable of all components over an extended Air-sea exchanges of heat and fresh- estimates of the mean and annual cycle period of time. water on seasonal time scales are nearly for this period. All data are freely avail- As an example of the close relation- balanced by oceanic storage (Gill and able, with about 90% of profiles accessible ship between SSH and steric height, Niiler, 1973), with seasonal advection at two Global Data Assembly Centers Figure 7 shows the zonally averaged being a small residual term in ocean within 24 hours of float surfacing. Argo’s annual cycle of both quantities, using interiors. Roemmich and Gilson (2009) ground-breaking open access data policy Archiving, Validation, and Interpretation found good agreement seasonally, on is central to the value of the program and of Satellite Oceanographic data (Aviso) hemispheric and global scales, between to building its international partnership. SSH (Ducet et al., 2000) and steric height the Southampton Oceanography Centre/ The Argo data set has been used to tackle from Roemmich and Gilson (2009). In National Oceanography Centre historical a wide variety of basic research problems, spite of the high similarity in the annual data climatology of air-sea fluxes (Josey and data quality exceeds original expecta- variations of SSH and steric height, there et al., 1998) and Argo-derived ocean heat tions. The Argo Program has made rapid are also significant differences between storage. Regionally, maximum seasonal progress in the decade since its planning them, for example, in the amplitudes amplitudes in heat storage at 40°N and began. Further increases in float numbers at about 10°N and 35°S. A good test 35°S exceeded the amplitudes of air-sea and improved coverage in the Southern for ODA models is to see whether they flux by about 25 W -2m , possibly due Hemisphere, better ability to identify and

Oceanography September 2009 33 Figure 7. The annual cycle of zonally averaged steric height (0/2000 dbar) from Argo data (left panel), 2004 –2007, is compared to that from altimetric height (right panel, Aviso product) from the same period.

correct systematic errors, and greater the Argo’s design and objectives. cycle rate, and other mission param- uniformity in production and release Low-latitude interannual variability is eters could increase Argo’s value in of delayed-mode data are all required well resolved in the present data set, many applications (Gary Brassington, to achieve the core objectives of the while additional floats are needed at Australian Bureau of Meteorology program. The broad Argo user commu- southern latitudes. Continuing advances Research Centre, pers. comm., 2009). nity is needed to demonstrate the high in profiling float, ocean glider, and In each case, for new objectives, value of the array, and the international sensor capabilities raise new challenges energy and other added costs need to Argo partnership must prove its ability for expansion of Argo’s activities. Deeper be weighed against the benefits, and to maintain the array for a decade and profiling and sampling of seasonal ice new resources are needed to cover any beyond. A caution is that the provision of zones, marginal seas, and boundary new costs. An important challenge for highest-quality Argo data is a continuing currents could all extend Argo’s limits. Argo is to expand its constituency by process, and users should ensure the data Inclusion of new sensors could add demonstrating the value of the data set is appropriate for their applications. important geochemical and biological set in a growing number of applica- As the Argo era of quasi-uniform, dimensions. Operational control of the tions while maintaining the high data high-quality global sampling lengthens, Argo array using two-way communica- quality and spatial coverage needed for it is important to review and improve tion systems to change profile depth, Argo’s core objectives.

34 Oceanography Vol.22, No.3 Acknowledgements Ducet, N., P.-Y. Le Traon, and G. Reverdin. 2000. and M. Ignaszewski. 2009. The Argo Program: Global high resolution mapping of ocean circula- Observing the global oceans with profiling floats. The Argo data used here were collected tion from TOPEX/Poseidon and ERS-1 and -2. Oceanography 22(2):34–43. and are made freely available by the Journal of Geophysical Research 105:19,477–19,498. Stammer, D. 1997. Global characteristics of ocean International Argo Program and by the Gaillard, F., E. Autret, V. Thierry, and P. Galaup. In variability estimated from regional TOPEX/ press. Quality control of large Argo data sets, POSEIDON altimeter measurements. Journal of national programs that contribute to it. Journal of Atmospheric and Oceanic Technology. Physical Oceanography 27:1,743–1,769. The efforts of many international part- Gill, A., and P. Niiler. 1973. The theory of the Uchida, H., and S. Imawaki. 2008. Estimation of seasonal variability in the ocean. Deep-Sea the sea level trend south of Japan by combining ners in planning and implementing the Research 20:141–177. satellite altimeter data with in situ hydrographic Argo array are gratefully acknowledged. Guinehut, S., P.Y. Le Traon, and G. Larnicol. 2006. data. Journal of Geophysical Research 113, C09035, What can we learn from global altimetry/ doi:10.1029/2008JC004796. Preparation of this manuscript was hydrography comparisons? Geophysical Research Wijffels, S., J. Willis, C. Domingues, P. Barker, N. supported by US Argo and US GODAE Letters 33, L10604, doi:10.1029/2005GL025551. White, A. Gronell, K. Ridgway, and J. Church. through NOAA Grant NA17RJ1231 Guinehut, S., C. Coatanoan, A.-L. Dhomps, P.-Y. 2008. Changing expendable bathythermograph Le Traon, and G. Larnicol. 2009. On the use of fall-rates and their impact on estimates of (SIO–JIMO). Graphics were produced satellite altimeter data in Argo quality control. thermosteric sea level rise. Journal of Climate using Ferret software, a product of Journal of Atmospheric and Oceanic Technology 21:5,657–5,672, doi:10.1175/2008JCLI2290.1. 26(2):395–402, doi:10.1175/2008JTECHO648.1. Willis, J.K., J.M. Lyman, G.C. Johnson, and J. Gilson. NOAA’s Pacific Marine Environmental Josey, S.A., E.C. Kent, and P.K. Taylor. 1998. The 2007. Correction to “Recent cooling of the upper Laboratory. The Aviso altimeter prod- Southampton Oceanography Centre (SOC) Ocean- ocean.” Geophysical Research Letters 34, L16601, ucts were produced by the CLS Space Atmosphere Heat, Momentum and Freshwater Flux doi:10.1029/2007GL030323. Atlas. Southampton Oceanography Centre Report Willis, J.K., D.P. Chambers, and R.S. Nerem. Oceanography Division as part of the No. 6, 30 pp. 2008. Assessing the globally averaged sea level Environment and Climate EU ENACT Leuliette, E.W., and L. Miller. 2009. Closing the budget on seasonal to interannual timescales. sea level rise budget with altimetry, Argo, and Journal of Geophyiscal Research 113, C06015, project (EVK2-CT2001-00117) and with GRACE. Geophysical Research Letters 36, L04608, doi:10.1029/2007JC004517. support from the Centre Nationales doi:10.1029/2008GL036010. Wong, A., N. Bindoff, and J. Church. 1999. Large-scale Levitus, S., J.I. Antonov, T.P. Boyer. 2005. Warming of freshening of intermediate waters in the Pacific and d’Études Spatiale (CNES). Argo/drifter the world ocean, 1955–2003. Geophysical Research Indian oceans. Nature 400:440–443. comparisons (Figure 6) were provided Letters 32, L02604, doi:10.1029/2004GL021592. Wong, A.P.S., G.C. Johnson, and W.B. Owens. 2003. by M. Scanderbeg. Levitus, S., J.I. Antonov, T.P. Boyer, R.A. Locarnini, Delayed-mode calibration of autonomous CTD H.E. Garcia, and A.V. Mishonov. 2009. Global profiling float salinity data by theta-S climatology. ocean heat content 1955–2008 in light of Journal of Atmospheric and Oceanic Technology References recently revealed instrumentation problems. 20:308–318. Boyer, T.P., J.I. Antonov, S. Levitus, and R. Locarnini. Geophysical Research Letters 36, L07608, 2005. Linear trends of salinity for the world ocean, doi:10.1029/2008GL037155. 1955–1998. Geophysical Research Letters 32, Reynolds, R.W., N.A. Rayner, T.M. Smith, D.C. L01604, doi:1029/2004GL021791. Stokes, and W. Wang. 2002. An improved in situ Cazenave, A., K. Dominh, and S. Guinehut. 2009. Sea and satellite SST analysis for climate. Journal of level budget over 2003–2008: A reevaluation from Climate 15:1,609–1,625. GRACE space gravimetry, satellite altimetry and Riser, S.C., and K.S. Johnson. 2008. Net produc- Argo. Global and Planetary Change 65(1–2):83–88. tion of oxygen in the subtropical ocean. Nature Conkright, M.E., R.A. Locarnini, H.E. Garcia, 451:323–326, doi:10.1038/nature06441. T.D. O’Brien, T.P. Boyer, C. Stephens, and J.I. Roemmich, D., and P. Sutton. 1998. The mean and Antonov. 2002. World Ocean Atlas 2001: Objective variability of ocean circulation past northern New Analyses, Data Statistics, and Figures, CD-ROM Zealand: Determining the representativeness of Documentation. National Oceanographic Data hydrographic climatologies. Journal of Geophysical Center, Silver Spring, MD, 17 pp. Research 103:13,041–13,054. Curry, R., B. Dickson, and I. Yushaev. 2003. A change Roemmich, D., and the Argo Science Team. 1999. in the freshwater balance of the Atlantic Ocean On the Design and Implementation of Argo: An over the past four decades. Nature 426:826–829. Initial Plan for a Global Array of Profiling Floats. Davis, R.E., C.E. Eriksen, and C.P. Jones. 2002. International CLIVAR Project Office Report 21, Autonomous buoyancy-driven underwater gliders. GODAE Report 5. GODAE International Project Pp. 37–58 in The Technology and Applications of Office, Melbourne, Australia, 32 pp. Autonomous Underwater Vehicles. G. Griffiths, ed., Roemmich, D., and J. Gilson. 2009. The 2004–2008 Taylor and Francis, London. mean and annual cycle of temperature, salinity and Domingues, C.M., J.A. Church, N.J. White, steric height in the global ocean from the Argo P. Gleckler, S.E. Wijffels, P.M. Barker, and J.R. Program. Progress in Oceanography doi:10.1016/j. Dunn. 2008. Improved ocean-warming estimates: pocean.2009.03.004. Implications for climate models and sea-level rise. Roemmich, D., G.C. Johnson, S. Riser, R. Davis, Nature 453:1,090–1,093. J. Gilson, W.B. Owens, S. Garzoli, C. Schmid,

Oceanography September 2009 35