III. WOCE OBSERVATIONS, DATA ANALYSIS AND DATA PRODUCTS

 Continued analysis of observations using traditional methods as well as development of assimilation approaches is required. These methods range from the analysis of single sections and current meter records, to consideration of a wide variety of data sets over a full ocean basin or the globe, to box inverses of hydrographic sections and other more statistically-based analyses of one or more data sets. This mode of analysis provides the benchmarks and data-based understanding of ocean processes at present. These data-based efforts merge naturally into the formal data assimilation directions described in section V, and are also a necessary adjunct to prognostic model development for which they are used for evaluation.

 During the AIMS phase, intensive analysis of the many data sets collected as part of WOCE will accelerate. Analysis of model results should be incorporated, both to aid in interpretation of the data sets, and also to facilitate testing and hence improvement of the models. These analyses are being facilitated by the WOCE regional workshops, which provide target times and venues for discussion of WOCE results for the Pacific, Atlantic, Indian, Southern and global ocean, and by the publications which result from the workshops. The modelling and data assimilation workshop will provide additional focus on comparison of data and models.

 Central to the success of WOCE data analysis, model evaluation, and data assimilation are the Data Assembly Centres (DACs) and Special Analysis Centres (SACs) and the voluntary provision of WOCE data to them by WOCE Principal Investigators (PIs).

 Timely data provision by PIs involved in the field programme is crucial to the achievement of WOCE goals and individuals are urged to comply. (PIs are reminded that data submitted to DACs only comes into the public domain after the originator has so authorised.) Agencies are urged to continue to support DACs and SACs , whose functions and progress will be assessed regularly by appropriate WOCE committees.

 III.1. Field observations

 The intensive WOCE field programme will be completed at the end of 1997: hence no large sea-going projects will be initiated as part of WOCE after this time. However float and drifter data from deployments in the 1990s will continue to be acquired to contribute to the required five-year average; some current meter arrays will be awaiting recovery. The repeat hydrography, XBT, sea level and surface flux programmes which are an integral part of WOCE's assessment of variability have already begun to be and will have to be continued under funding outside WOCE, although it is anticipated that change in the organisational structures which support these data streams will only be evolutionary.

 A detailed and continuously updated summary of WOCE field programmes and data status is currently available through the WOCE Data Information Unit (DIU) (http://www.cms.udel.edu). The committee recommends continuation of support for these activities.

 III.2. Data analysis

 It is expected that WOCE analysed data sets will be used to:

 (1) achieve the descriptive objectives of WOCE (transport estimates, water mass formation rates, description of variability, etc.),

 (2) provide parameters for ocean models such as mixing rates,

 (3) initialise, force and evaluate general circulation models, and

 (4) combine with models to provide more dynamically consistent descriptions of the ocean (state estimation or data assimilation).

 Much of this work is initially heavily based on data analysis using simple models of the ocean. All analyses incorporate a model of the ocean; even those which are based most closely on the data alone rely on assumptions about the statistics of the assumed underlying processes. (A simple model which is incorporated in most hydrographic data analysis is geostrophy.) "Data-based analyses" are not analyses of numerical model output, and they are not data assimilation. However, comparisons with numerical model results are an increasingly important part of the analysis process, for instance for analysis of tracer distributions.

 Priorities for primarily data-based analyses include completion of the Atlantic process experiment analyses, description of the large-scale circulation, estimates of ocean heat and freshwater transports and their divergences, estimates of water mass formation and ventilation rates, and a description of ocean variability as measured in the WOCE period.

 The process experiments were mentioned briefly in II.1.2. They include the Tracer Release Experiments (TREs), which made direct measurements of mixing rates, the subduction experiment, which began to observe the processes and rates by which water connected from the surface to the interior ocean, the Deep Basin Experiment (DBE), which observed the deep flow in detail in the isolated deep Brazil Basin, and the North Atlantic subpolar gyre experiment, which is taking place in 1996-1998, to observe processes in a region of vigorous water mass transformation. All of these process experiments were based primarily on observations. Results from the earliest are already digested and published, while the most recent observations in the North Atlantic have yet to be completed. The TRE and DBE both provide estimates of mixing rates which are useful for models. The DBE is designed to provide a view of upwelling processes as well, which are central to dynamical understanding and very difficult to quantify. The subduction and subpolar gyre experiments provide views of larger-scale upper ocean processes which are central to how the ocean responds to the atmosphere. In these regions, specific processes in models can be tested, and can also be used to help understand the observations.

 Efficient access to the full WOCE data sets is required for the many WOCE investigators who are pursuing data analysis, model evaluation, and data assimilation studies. In the next sections are described the various WOCE data streams and the products that are expected from them.

 III.3. Data assembly and products

 Each WOCE data stream has been carefully considered with respect to data collection, data assembly, distribution, and archiving. The WOCE Data Products Committee (DPC) is charged with setting and implementing policies for each data stream. The WOCE Hydrographic Programme Planning Committee (WHP-PC) additionally oversees the hydrographic programme. At least one Data Assembly Centre (DAC) and a final archive site (World Data Centre A) exists for each type of data. For two types of data - hydrography and surface fluxes - there are also Special Analysis Centres (SAC) which perform additional functions. A listing of the DAC and SAC sites, a summary of their holdings and their web and email addresses can be found in the WOCE Data Guide (1997), (also at http://www.soc.soton.ac.uk/OTHERS/woceipo/dguide97/).

 Each data stream is listed at the WOCE DIU website, with links to Data Assembly Centres and to individual sites which can provide data or information on-line.

 In the following subsections, the major WOCE data streams are discussed with an emphasis on data use. Recommendations are made for continuation, expansion, or changes in the approach to the data sets with the goal of making them as easily available as possible to WOCE users. In addition to data, metadata (information about the data and their quality), and mechanisms for identifying, developing and distributing those products which are deemed useful will be considered.

 Widespread use of the world wide web and ftp sites has revolutionised data distribution and availability in the past few years. Data providers (DACs, SACs and archives) should keep abreast of changes in computing and data distribution and storage and stay at the forefront of new approaches. Software provided publicly for data handling and products should be platform-independent.

 III.3.1. General recommendations for datasets and data products

 The primary functions of Data Assembly Centres (DACs) are to assemble data sets, subject them to quality control, and ensure distribution. Providing for easy access to the data sets and error estimates is an essential function of the WOCE data system and on-line access is recommended where the volume of data permits it.

A second function of the DACs, Data Information Unit (DIU), and Special Analysis Centres (SACs) is to provide information (metadata) about the datasets. This information must include origin, quality information and, where resources permit, references to journal publications and technical reports. Metadata must be as accessible as the data sets.

 A third function of the centres is to create, compile and/or track data products which have community value. The DPC working with the DACs and SACs should determine which basic products are necessary and evaluate products which are being made available by both the centres and individuals.

 A proposal to create CD-ROMS of data, metadata and data products, and a common software to access and connect the various data types known as the WOCE Data Resource is described in section III.4

 A series of WOCE "climatologies" will be produced during the WOCE AIMS period . Here we define a "climatology" as a data-based product which includes gridded fields for use in models but may optionally contain the data set itself and methods for customising the products. Such methods (software) must be platform independent. The first set of climatologies will be data-based, not include significant modelling, and will consist of internally consistent basic data sets, gridded products and derived products. Climatologies following later will include gridded fields produced through various assimilation techniques. Gridding standards have been set in place for satellite data and should be adapted where suitable for WOCE in situ data. The list below refers only to the data-based climatologies.

 We anticipate that during WOCE AIMS the following climatologies will emerge: (1) internally consistent datasets, (2) gridded fields of properties, variances and covariances from all data types with error estimates and software to produce alternate versions where appropriate, (3) eddy kinetic energy distributions, and (4) velocities with errors and time scales, from all instruments measuring velocity directly. The gridded fields would be of hydrography (3- D), velocity from floats and drifters hence at two levels (2- D), globally-gridded surface fluxes (2- D plus time), regionally-gridded XBTs (4- D) and time series and repeat hydrography (1- D or 2- D plus time), and globally gridded altimetry (2- D plus time). In some of these analyses it will be necessary to include historical data in order to produce complete fields, although results will be based as much as possible on WOCE and WOCE-era data. The climatologies should include documentation of errors, a bibliography, a compilation of WOCE results for heat and freshwater transports and water mass formation rates. Compilations of other similar derived values might be added: for instance analyses of the three-dimensional velocity field, or isopycnal properties.

 These data products will not result from any one group or individual but will arise from the efforts of organisations (DACs/SACs), individual investigators and investigator groups.

 Although climatologies of WOCE data, error fields and their scientific interpretation will first appear in the scientific literature, the climatologies themselves should be linked to and made available through the DACs/SACs. CD- ROMs of the analyses should also be made widely available, a task which Journals may undertake themselves, but whose implementation will be advised and monitored by the WOCE Publications Committee.

 The nature and use of each data type and the data products expected therefrom are elaborated in the following sections (III.3.2-III.3.10). Model output is not considered in this chapter except as part of the WOCE Data Resource (III.4)

 III.3.2. WOCE Hydrographic Programme data

 Data sets (Level 1). WOCE hydrographic data include CTD and discrete bottle data from one-time and repeated sections and time series stations. Reporting and quality control procedures for all bottle data and the 1 to 2 decibar CTD data were set in place early in the programme and are handled by the DAC, which is the WOCE Hydrographic Programme Office (WHPO). The WHPO will continue this work, including distribution of the data sets to the Special Analysis Centre (SAC) in Hamburg. Detailed recommendations concerning data flow and quality control are the responsibility of the WHP Planning Committee (WHP-PC) and the Data Products Committee (DPC).

 Other temperature and salinity data sets are collected as part of other WOCE data streams - upper ocean thermal/salinity (XBT/XCTD), current meter, surface drifter and profiling float observations. The upper ocean thermal/salinity DACs have a primary function in assembling the XBT/XCTD observations. Temperature/salinity measurements from the three other sources are assembled and archived by their own DACs.

 Level 1 processing of CTD data requires application of laboratory calibration data, vertical averaging of bottle/CTD data at scales not exceeding 1-2 decibar, correction using bottle values collected simultaneously, and quality control. These functions are carried out by the data originators, with subsequent assembly and quality control by the WHPO.

 No provision has been made for retention of the original CTD data of much higher sampling rate (24 Hz with variable fall rate). Scientific uses of these data are in studies of fine-structure and in assimilation work where noise estimates are required. It is recommended that where the data is of good quality, these original data be retained so that they can be made available in response to special requests.

 Almost all WHP data sets include discrete salinity sampling. WHP data sets from the one-time survey also include oxygen, nutrients, chlorofluorocarbons, tritium, helium- 3, carbon- 14, and several CO₂ parameters. Each data type carries its own set of processing and quality control procedures, which are carried out by the originators and overseen by the DAC, which has produced operations manuals for some data types. The primary issue for Level 1 discrete bottle data is assembly and quality control of these multi-originator data sets.

 Data uses. Uses of these data include observationally-based approaches to determination of circulation, transports and time scales, through sections, maps and time series. Combination of results from the WHP hydrographic data and other datasets, such as the upper ocean thermal/salinity data, floats, current meter estimates and satellite altimetry as well as historical hydrographic data is expected. A multitude of methods is and will be used. These range from traditional approaches based solely on the data sets, to inverse modelling, to data assimilation. Another use of the data is for model evaluation. An important use of the WHP data in combination with historical data will be in three-dimensional climatologies to be incorporated in models and the determination of ocean climate change.

 Data products (Level 2) will be useful for varied purposes. Printed atlases of the WHP one-time sections and an on-line database for all WHP data, with interactive programmes and displays for property plots and sections, are envisioned. Original data, bottle data interpolated to standard depths, and gridded sections would be available. Isopycnal maps incorporating WHP data and historical data should be made available. Three-dimensional gridded fields will be especially useful; some versions of these would be suitable for prognostic models and some simply for easy visualisation of the data.

 Additional Level 2 products which would be useful for data assimilation include estimates of covariances induced by smaller scale (10-10⁴ m) processes.

 Level 2 products may also take the form of software operating with Level 1 data. This will provide greater flexibility of use. Software production can be decentralised, with some products coming from the DAC/SAC and some from individual investigators and co-ordinated by the DAC and SAC.

 Hydrographic and tracer atlases should be produced as outlined above. It is also recommended that the WHP DAC/SAC along with its primary concern for WOCE data should assemble recent high quality historical data . These data will supplement the WOCE dataset and will also allow evaluation of ocean climate variability and change in support of GOAL 2 of the programme.

 III.3.3. Drifter and float data

 Data sets (Level 1). The WOCE drifter data set includes positions, times, temperature and drogue depth and status; on some drifters atmospheric pressure and salinity are also measured. The WOCE float data set includes positions, times, float pressure; for some records there are temperature and salinity profiles. The drifter data are assembled by the NOAA AOML Drifting Buoy Data Assembly Centre and archived by the Marine Environmental Data Service (MEDS) in Canada. The float data are assembled and archived by the WOCE subsurface float DAC in Woods Hole.

 Data uses and Level 2 data products. For the WOCE synthesis the most general uses of pseudo-Lagrangian data are expected to be:

 1. Determination of the large-scale, multi-year average circulation patterns near 15 m and 1000 m depth over much of the globe and at other selected levels in certain regions like the Brazil Basin (2000 and 4000 m), the Subduction Experiment region of the eastern subtropical North Atlantic (200-400 m), and the subpolar North Atlantic (600 and 1500 m);

 2. Determination of the regional patterns of single-particle eddy statistics like eddy kinetic energy, eddy polarisation, Lagrangian timescale and the Taylor single-particle eddy diffusivity K_h(t); and

 3. Description of the patterns of low-frequency large-scale current variations in particular regions (e.g., surface circulation changes in the tropical Pacific associated with ENSO).

 WOCE data products should be designed to convey these general purpose descriptions in forms that permit combining or comparing them with other observed fields or with dynamical models and in ways that they can be used inside assimilation studies.

 In addition to these general uses the WOCE Lagrangian observations will support a number of specific research studies such as (a) investigations of path variations of selected currents receiving repeated deployments of current followers; (b) "stirring" descriptions from multi-particle dispersion where current-followers were deployed in initially coherent arrays; (c) studies of the relationship between surface-current fluctuations and sea-level variability observed by altimeters; and (d) the nature of the near-surface velocity response to wind. The requirements for these studies are unique enough, and the data sets are small enough, that their data requirements should be met directly from the archived data.

 In addition to their use as current-followers, WOCE drifters are being used to observe sea surface temperature, surface salinity and atmospheric surface pressure while some PALACE floats report temperature and salinity profiles. These data should be retained in the basic instrumental data archives and should be merged with similar observations made in other ways. It is recommended that drifter SST observations should be merged with VOS and research vessel reports and that all PALACE profiles should be merged into the XBT and XCTD archives.

 Products for the general uses above are all statistical in the sense that measurements from different times and places are combined together to achieve representativeness or statistical reliability. These results can be displayed as gridded values. Unfortunately, the best way to average depends on the data themselves (since we want to combine statistically similar observations rather than smoothing away spatial structure of these statistics) and on the specific application or desired trade-off between resolution and statistical reliability. In a similar way, the method of preparing gridded values may employ various corrections or hypotheses about the field. For example, windage estimates might be applied to drifter velocities, float velocities might be "corrected" to a common depth using mean shear estimated from gridded hydrography data, or observations of either type might be objectively mapped using estimated time and space scales and/or hypothesised dynamics such as minimal divergence or near conservation of potential vorticity.

 Because the techniques for gridding Lagrangian observations vary and the differences between them are important, it is not enough to simply present results gridded by one scheme. The optimum solution is the creation of a range of products, and of well documented programmes and data to generate such products. These should be supplied by the main data producers or by individuals or committed centres, lodged with DACs and distributed to the broader community.

 III.3.4. Upper ocean thermal and salinity data

 Data sets (Level 1). The upper ocean thermal data set is divided into two differing streams by their spatial and temporal resolution and by their expected uses: (1) broad-scale with relatively low along track resolution (greater than 1 degree), and (2) high density (high along track resolution) XBT data consisting of ocean spanning boundary-to-boundary transects with horizontal spacing between probes of 10-40 km.

 Broad-scale data, usually collected by ship's crew, are passed in real time via the Global Telecommunication System (GTS) to the Marine Environmental Data Services (MEDS) of Canada. These data are quality controlled with documented procedures and passed on to US NODC where they are added to the Continuously Managed Database (CMD). NODC is the centre for processing the delayed mode stream and merging and eliminating duplicates between the real-time and delayed mode streams. Collectively the activity is known as the Global Temperature and Salinity Project (GTSP).

 The US NODC also acquires delayed mode high vertical resolution XBT data, passes them through GTSP quality control, and merges them into the CMD. The delayed mode high resolution XBTs replace real time, low resolution XBTs when duplication occurs. All types of data (Level 1) are available from MEDS, US NODC (on line), and participating GTSP data centres.

 High density section data are usually collected by an onboard scientist or technician, with probe spacing designed to resolve narrow features such as eddies, fronts, and boundary currents. Data flow is essentially equivalent to that for broadscale data (a real-time data set at 1 probe per degree via the GTS plus a complete dataset to NODC in delayed mode).

 Profiles from PALACE-type floats are included in the GTSP data collection, beginning in October 1995 for real-time data and some time later for delayed mode profiles.

 Data uses. The principal objective of broadscale XBT sampling is to determine the variability in ocean heat content. The main objective of the high density XBT sections is to determine the variability in geostrophic velocity and the associated transports of mass and heat.

 In addition to individual scientists, the main users for the Level 1 products from the broad-scale data are meteorological centres engaged in climate assessment and prediction activities (NCEP/NOAA, ECMWF, JMA (Japan), and UKMO). Real-time analyses (nowcasting) for tropical Pacific and Atlantic are produced at NOAA/NCEP (USA), BMRC (Australia), UKMO (UK) and JMA (Japan). These analyses are produced using statistical interpolation technique based on the level data alone, or using ocean model based data assimilation system to combine information from model, forcing fields (surface fluxes) and the observations. Retrospective analyses for the Pacific and Atlantic have also been produced at NOAA/NCEP which can take advantage of delayed mode data and improved data assimilation schemes. In addition, both NCEP and ECMWF use real-time analyses to initialise coupled ocean-atmosphere model for El Niño prediction.

 The level 1 data are also used by research groups world wide to facilitate development of ocean data assimilation techniques towards the WOCE data assimilation objective.

 Users of high density section data are likely to be: (1) seeking comparison/constraining data for modelling studies, (2) carrying out heat budget studies, (3) looking at subsurface structure in relation to satellite altimeter surface height, (4) addressing temporal variability/representativeness with respect to WHP transects.

 Data products (Level 2). Periodically, high vertical resolution, broadscale data are sent to the UOT science centres (AOML, CSIRO, and Scripps) where additional quality control is applied to these data, and gridded fields such as SST and mixed layer depth are produced. These products are available through these centres, some on line.

 In addition Scripps Institution of Oceanography's Upper Ocean Thermal DAC is working with investigators to develop a World Wide Web page for display and dissemination of high density section data as transects.

 Outside of the UOT science centres other groups and individuals are generating valuable products from this data stream. Links to these investigators should be sought and displayed by the science centres.

 III.3.5. Sea level data

 Data set (Level 1). Hourly, daily and monthly sea level time series covering at least part of the TOPEX/POSEIDON data period are available from over 100 stations, accessible via Internet anonymous ftp from the Hawaii sea level DAC in a short delay mode (1 month).

 Quality controlled hourly sea level data are available in delayed mode delivery at BODC. WOCE requirements state that the elevation must be accurate to 1 cm, the atmospheric pressure to 1 mb and the timing to 2 minutes. BODC provides also an inventory of available data, details of the gauge location locations, maps, summary statistics, plots of low frequency variations.

 Several bodies already exist who are co-ordinating and promoting the use of this type of data: GLOSS group of experts, SWT of TOPEX/POSEIDON, and IOC sea level sub-group. Meetings as part of planned workshops and conferences will be an efficient complement.

 Data uses. The WOCE user community concerned with this type of data is the one addressing the following scientific problems.

 * Use of pairs of gauges across straits or choke points to estimate the variability of the surface geostrophic currents (in conjunction with altimetric data).

 * Use of sea level measurements for correction of satellite altimetry, especially for removing the drift of the satellite measurements (in conjunction with geodetically located tide gauges).

 * Use of the archive of the long term data set of sea level collected by GLOSS (partly supplied now by the WOCE sea level network) to allow estimates of long term trends and assess the representativeness of the WOCE observation period within the long term records.

 * Use of tide gauge data to constrain or evaluate numerical models. This has been of particular benefit for the evaluation of the ocean tide models developed for altimetric tide corrections.

 The fast and delayed mode sea level delivery centres provide an excellent and important service. Because they have been under threat in the past, the need for their continued existence is stressed, particularly because of their use with satellite altimetry.

 III.3.6. Current meter data

 Data sets (Level 1). These consist of velocity, time, and position at each current meter. Additional data might include temperature, salinity and pressure records at the current meters or from separate instruments. The data are assembled by the DAC at Oregon State University and distributed on-line through their website. If data arrive at the DAC corrected for mooring motion, this is noted in the file header; otherwise no attempt is made to correct for motion as the requisite data (mooring line size, tension, flotation, design) are not provided to the DAC.

 Data uses. Current meter data in WOCE were collected primarily in narrow passages or strong currents, with the purpose of computing velocities, transports and fluxes of defined currents or throughflows and their variability. These provide valuable information for model assessment, for property flux calculations, and for comparison with geostrophic, ADCP and altimeter-derived velocities where they are coincident.

 Historical current meter records are also of great value to WOCE. Dickson (1990) has compiled estimates of variance for a large number of records from around the world, and the DAC at Oregon contains a very substantial body of detailed historical data which is available on line.

 Data products (Level 2). The DAC currently provides spectra to standard scales and time series plots of both the filtered and unfiltered data; it is likely in the near future to include movies from reasonably coherent arrays. Level 2 data products also include the means, variances, and time scales of velocity and other measured variables from each current meter record, with software to produce alternate versions. Other derived quantities (e.g., higher order statistics, tidal constituents) can be made generally available as well. These would be produced by some combination of effort from the DAC and from individual investigators. As with other types of data, the DAC may accept proposals from individuals for products, and will also evaluate and link to products made available by individuals as their resources allow.

 Recognising the significance of flux measurements derived from WOCE current meter arrays, it is recommended that the DAC prominently displays all references to publications that mention them and, where possible, provides links to reports of such (flux) measurements.

 III.3.7. Air-sea flux data

 Data sets (Level 1). Observations of surface winds, air and sea temperatures, humidity, atmospheric pressure, precipitation, and numerous other variables (clouds, radiation) are recorded by research vessels during WOCE cruises and by moored buoys during WOCE-related projects (e.g., Subduction Experiment). Frequency of observations, specific observed variables (e.g., some platforms have radiometers installed to provide short and long-wave radiation components), and quality of observations varies widely by platform. The WOCE Surface Meteorology DAC at Florida State University (FSU) is collecting these data and associated metadata, performing well-documented quality-control procedures, and distributing the resulting files to the WOCE community. From these data, air-sea flux values for momentum, turbulent heat, and moisture can be estimated using bulk formulae.

 The FSU SAC for surface fluxes is producing near-global fields of atmospheric variables and fluxes as well as uncertainties. Current products include surface wind products for the Indian and tropical Pacific basins, and historical meteorological and flux fields for the Indian Ocean. Similar fields are being developed for other basins. All data products are offered in netCDF and ASCII formats and are of a size easily manageable by most investigators. These files are available via WWW and FTP servers at the FSU DAC/SAC.

 Data uses. The community of Level 1 (DAC) product users includes scientists incorporating these data into atmospheric prediction models (e.g., NCEP, ECMWF, JMA, UKMO) for production of surface synoptic fields. Note however that many research vessels are not equipped to provide near-real time data reporting to GTS nor does much of these data appear in COADS. Data from a number of cruises in regions with very limited historical observations are of particular value due to their dense sampling of a large number of relevant variables. These data should be incorporated into historical archives. Additional users are groups involved in mixed-layer and process-oriented investigations, and those analysing geochemical fluxes, all of whom rely on accurate surface condition data. Lastly, other potential users are those involved in global evaluation of remotely sensed data such as NSCAT (winds), TRMM (precipitation), and AVHRR (SST) using data from automated recording systems aboard numerous research vessels.

 Data products (Level 2). These include gridded fields of surface meteorology variables such as winds, temperatures, humidity and fluxes of heat, momentum, and moisture, that are commonly utilised as surface boundary conditions in ocean models. Experience has shown that surface fields from atmospheric GCM prediction models can be quantitatively evaluated through comparison to other products (e.g., data-based products from the WOCE Flux SAC and others). Results assist in guiding future GCM model development and possible modifications to existing products. Finally, fields from reanalysis projects are attractive options for ocean modelling efforts; recent results demonstrate these flux fields must also be evaluated by comparison with other data-based products. In summary, Level 2 products, including reanalysis products, will be used by major groups involved in the definition of the forcing function of the ocean, and in evaluation studies of the air-sea fluxes.

 Other Level 2 fields (both time-dependent and climatologies) are and will be produced by investigators utilising satellites as well as by climate and weather analysis/prediction centres (NCEP, ECMWF). The critical questions for WOCE PIs (particularly ocean modellers) are - Which products are best for their investigations and how can the products be improved to an accuracy sufficient for modelling studies (section IV.4.2). The SAC should act as a resource for information on flux fields, providing details and links to various flux products suitable for WOCE investigations. Surface flux intercomparison/evaluation efforts are planned by the flux SAC as well as by groups outside of WOCE, such as the SCOR/WCRP Working Group on air-sea fluxes. The results of all such evaluations in light of oceanographic modelling are crucial to WOCE synthesis plans.

 The flux SAC, in conjunction with the SCOR/WCRP working group on air-sea fluxes should carry out flux evaluation studies, should submit suitable products for companion studies, and provide guidance to WOCE modellers as to the most appropriate flux products for their work.

 III.3.8. Satellite altimetry data

 The satellite altimetry data are expected to have major influence and use in WOCE analysis. They are collected from the ERS- 1, ERS- 2 and TOPEX/POSEIDON (T/P) platforms. In addition, the T/P follow-on mission JASON is under preparation. ERS- 1 provides altimeter, scatterometer and ATSR (Along Track Scanning Radiometer) data from the start of the mission in 1991 and ERS- 2 is essentially a duplicate. The T/P altimetric mission has collected sea-surface topographic observations with an unprecedented accuracy, extending the range of applications of those data and giving rise to new findings about the ocean circulation.

 ERS data are distributed through the European Space Agency (ESA) with some reservation in the dissemination policy: users must be investigators approved by ESA, or at least their co-workers. TOPEX/POSEIDON data are distributed in the scientific community through the GDR (Geophysical Data Records), which include the altimeter sea level height measurements and associated corrections based on algorithms evaluated by the T/P Science Working Team. NASA and CNES are processing the GDR for each agency's own altimeter system. The delay between data reception by space agencies and distribution into the science community is generally no longer than 30-45 days. T/P is the first ocean research mission that delivers high-quality data on a near-real time basis.

 Substantial progress has been achieved during 1996 in the correction algorithms. For altimetry, ocean tidal errors were the main concern until 1995. Improved hydrodynamic models have recently been developed for TOPEX/POSEIDON applications, reaching a global accuracy of ~3 cm. In addition, a combined set of ERS-T/P data has been produced by the French agency CNES to correct for the orbit error of ERS- 1 satellite and to provide a refined spatial coverage of the sea-surface. Sea level measurements from tide gauges have established satellite altimeter drift and these corrections must continue for the lifetime of these and follow on missions.

 Because of both their global coverage and time-series nature, virtually all forms of oceanographic data analysis benefit from satellite measurements. The most exciting growth-areas are occurring in the assimilation of altimetric data into primitive-equation models of the ocean circulation. Methodological developments are currently underway, and their application still need further research, but already novel descriptions of seasonal and interannual variations are emerging.

 The excellence of the satellite programmes is acknowledged, with special mention of the efficient distribution of T/P products by the responsible agencies. Because of the very large amount of data represented by the ensemble of altimetric observations, only groups having sufficient resources can reprocess and use the whole data set as and when improved corrections will be available from Agencies. Several groups now routinely update altimetric data with the most recent corrections, and routinely share their data sets with other investigators. Such an approach should be continued.

 It is also recommended that responsible agencies (CNES, ESA, NASA) be encouraged to continue their search for excellence for altimetric observations and obtain support to continuously provide the best global altimetric data set to the community. Software produced by the agencies and various research groups to handle level 2 data would be of great service in improving data access.

 III.3.9. ADCP data

 Data sets (Level 1). Acoustic Doppler current profiler (ADCP) data are collected in WOCE in two modes: with ship-mounted ADCPs, sampling velocity along the cruise track in the upper 300 metres, and with instruments lowered to the bottom from the ship (Lowered ADCP or LADCP), sampling velocity through the full water column. Data sets include all velocity components, position and time.

 Shipboard ADCP: The WOCE ADCP DAC is co-located at the Japan Oceanographic Data Centre (JODC) and at the University of Hawaii (UH). Since 1991, JODC has been the Responsible Oceanographic Data Centre for ADCP measurements. In May 1995, the JODC was designated the primary host of the International WOCE DAC for shipboard ADCP. A secondary hub, hereafter referred to as the co-DAC, was established at the Joint Archive for Shipboard ADCP (JASADCP), a collaboration between the E. Firing ADCP Laboratory at the University of Hawaii and the US National Oceanographic Data Centre (NODC). The JASADCP has been primarily focusing on the US WOCE contribution of ADCP, although it has served as a hub for other multi-national programmes such as TOGA COARE.

 ADCP data along all WHP lines are being assembled and archived at the DAC, in standard formats. Data sets include position, time and velocities at given levels depending on the ADCP instrument, with metadata on the depth range, bin length, number of bins plus other instrument, navigation, and calibration information. At JODC weather and sea conditions are also stored. The JODC data base started as a national (Japanese) collection and currently remains so biased. The co-DAC also started out as a national (US) collection. Both sites were designated WOCE DACs late in the field programme and are now actively striving to acquire the backlog of data. Exchange of data holdings between each site is proceeding and must continue.

 Lowered ADCP: The LADCP provides a complement to the thermal-wind shear estimated from hydrographic data, and is especially useful in regions of strong baroclinic shear such as western boundary currents or in strong ageostrophic flow such as at the equator. As this is a new mode of data collection, standard formats have yet to be proposed and a data archiving system set up. A proposal for archiving this data type is currently under consideration by the DPC.

 Data uses. The velocity values from the shipboard and lowered ADCP are being used to study ageostrophic transport in the surface layer and through the water column, to characterise the internal wave field, as reference velocities for geostrophic profiles, in transport studies, and mixing studies using a Richardson number. For these uses, the basic velocity data (Level 1) as reported by the originators and archived are needed. Level 2 products (maps and statistics) as well as the Level 1 data will be useful for model evaluation and data assimilation, but some sort of error estimate will be necessary for the latter.

 Data products (Level 2). These include maps and vertical sections of subsurface velocity along the cruise tracks. In some regions, for example, the western North Pacific (the Philippine Sea) and the sea around Hawaii, the space and the time coverage of ADCP data is enough to make complete maps of averaged subsurface currents. Statistics of the velocity fields and mixing parameters would also be Level 2 products. The DACs may make vertical sections and, in addition are encouraged to make available useful products provided by individual investigators.

 The ADCP DAC and co-DAC are urged to pursue the acquisition of WOCE underway ADCP data and also to implement the lowered ADCP archive when so authorised by the DPC.

 III.3.10. Bathymetry data

 Data sets (Level 1). Bathymetric data are collected along most WHP lines and while setting moored arrays. From many of the ships, these records consist of short time averages of the uncorrected depth, position and time. Other ships might provide multi-beam swath data for specialised purposes.

 WOCE does not archive bathymetric data per se. Currently the WOCE Hydrographic Programme Office will accept bathymetric data along with hydrographic data, and will forward it to the World Data Centre for Marine Geology and Geophysics (National Geological Data Centre or NGDC). Investigators however are encouraged to submit the data themselves direct to NGDC. At NGDC the data are absorbed into their data base, which is updated annually on a CD-ROM. Extraction of the data requires knowledge of the NGDC identifier, which is available from the WOCE DIU. The number of WOCE data sets currently at NGDC is small in relation to the number of WOCE cruises.

 The WOCE Data Information Unit (DIU) is currently extracting datasets that are on the NGDC CD- ROMs, as their resources allow. This effort is greatly facilitated if investigators tag their datasets as WOCE when they submit them.

 Data uses. WOCE cruises often work in ocean regions that other ships do not reach, where the new bathymetric data are an important supplement to previous datasets. The along-track bathymetry can help to map seamounts, spurs, sills, etc., to supplement sometimes meagre charts. In the case of multi-beam surveys, critical information can be added. Special regions that were surveyed this way in WOCE include the Brazil Basin and the Samoan Passage. These data sets will be retained by the originators, but should also be offered to NGDC. New bathymetric data sets based on a blend of shipboard and altimetric information are being offered by W. H. F. Smith and D. T Sandwell (http://www.ngdc.noaa.gov/mgg/announcements/announce_predict.html). As a data set based on a new concept and under development, it has some shortcomings but is emerging as a valuable tool.

 WOCE uses for the bathymetric data include refinements of transport estimates using the detailed bathymetry, and mapping of deep velocities onto the actual cross-section of the water column, particularly if a large feature is not well resolved by the CTD/bottle station spacing or by the current meter array.

 Data products (Level 2). Bathymetric data (including multi-beam data) should be submitted to NGDC. The WOCE DIU should set up links to NGDC and these specific data sets. It is also recommended that the WOCE data set (CD-ROMs) for the WOCE Conference contain the WOCE bathymetric data and links to the most up-to-date versions of gridded bathymetry .

 III.4. Future directions for WOCE Data Management

 A priority area for work by the WOCE data facilities and WOCE Data Products Committee is to provide better and more comprehensive access to WOCE data of all kinds. One concept that is assisting in this direction is that of a WOCE "Data Resource." The idea is that the seeker of WOCE data should begin to see the highly distributed WOCE data system more as a single entity (despite the fact that the data assembly, quality control, distribution, and documentation functions will remain highly distributed). The distributed system continues to allow the value-adding processes of quality control, collation of metadata and establishment of common data formats for each data type. However, it is not a natural arrangement of the data for the analysis, modeling, data assimilation and interpretation. The development of a 'virtual' WOCE Data Resource that logically combines access to all of the WOCE data streams (including metadata, modeling results, and data products) may better satisfy a range of generic user needs during the AIMS phase. Each data facility in WOCE will willingly participate in creating the Data Resource to better serve the community of researchers seeking access to WOCE data.

The aim of the Data Resource is to provide researchers with an interface through which a wide range of WOCE data and data products may be accessed without concern for the complex internal structure and distributed nature of the WOCE data infrastructure. The WOCE Data Resource idea, if successfully implemented by the facilities participating in WOCE, will allow integrated data products to be developed more easily and transparently from the available data because component data sets will be more readily accessible and compatible.

Some specific illustrations from WOCE to motivate the concept for the AIMS phase are useful. It is common, for example, on a single hydrographic cruise to deploy XBTs, surface drifters, neutrally buoyant profiling floats, an acoustic Doppler current profiler, meteorological sensors and bathymetric sounders in addition to a full and complex suite of water property profilers and samplers. For many good reasons, these data types are processed in different ways, at different WOCE facilities, in different formats, and by different investigators. However, recovering the various data sets from a single cruise can be a challenge under present conditions. An investigator should be able to access this data and meta data by simple search of the WOCE Data Resource for data from a particular cruise.

Another illustration addresses some needs of the data assimilation and modeling community. Because of the relative sparseness of the oceanographic data, assimilation and inverse models need to include all of the available measurements to make quantitative estimates of the largescale circulation and transports of tracers. Even a partial description of the ocean circulation will not result from a single instrument type, but must blend a wide range of data with a model. For example, different aspects of the thermal structure of the ocean are obtained from satellites, PALACE floats, surface drifters, moored measurements, XBTs and CTDs. At some point, assembly of the individual data streams and metadata must occur. The WOCE Data Resource aims to ease such assembly.

The re-assembly of data from a single cruise and blending of data from different instruments that measure the same field are two examples of the requirements of WOCE data users. By design, during the field phase, WOCE researchers and facilities have been devoted to providing the highest quality data for each data stream. However, both quality and access define the value of the data set. Inadequate or difficult access means that the data will never be fully utilised. An essential part of the WOCE AIMS phase, to enhance synthesis, thus is to provide the fullest possible access to WOCE through the development of the WOCE Data Resource concept.

The integrated data resource will also allow gridded products to be developed easily and transparently from the available data. Data products will evolve rapidly as scientific issues are addressed and as synthesis occurs. This evolution necessarily implies that many initial products could become redundant as new data are added to the integrated data set and as new ideas for forming products are developed.

Specifications for the integrated data resource are that it:

 * provides a single interface to the entire WOCE data set, thus improving access for every researcher to the quality-controlled data.

 * provide for searching and selecting of data.

 * maintain all the metadata and quality flags intact. This is critical to maintaining quality and for updating data when corrections become available as a result of ongoing analysis and comparisons.

 Implementation of the integrated WOCE data resource

 The WOCE data resource is to be built upon the already existing data streams. Most individual DACs have established interfaces that allow the selection and search of their own individual data streams. These data streams should now be linked in more systematic way creating one cross-linked searchable data set. This will be accomplished in several phases.

 An initial phase of implementation has already been completed. With the advent of the world wide web, WOCE has been able to create an embryonic "virtual" data resource. All the WOCE data facilities offer data and products via the web and continue to work toward common standards.

 In the next phase of WOCE each Data Assembly Centre has been asked to provide a data set, browse products, and documentation on CD-ROM for the WOCE scientific conference in May 1998. This volume of CD-ROMs will be the first physical manifestation of the WOCE Data Resource (call it phase 1). These products will exist in parallel, on-line, at the DACs and also in CD-ROM form to allow more scope for distribution and richness of content.

 The second phase of implementation of the WOCE Data Resource will be the integration of the data sets, documentation and products from the individual data streams into a single logical 'virtual site' that can be accessed and searched by individual investigators. We propose that this logically linked database be released with software as a suite of CD-ROMs (analogous to phase 1). The task will be facilitated by the current results of the pilot "comprehensive data set" project of the WOCE DIU.

 The third and final phase of the project is incorporation of final "complete" data streams into the WOCE data resource, thereby becoming the legacy of the WOCE field programme.

 To create the WOCE data resource, support is required to combine the heterogeneous data streams into a single apparent 'virtual' on-line data set, a physical manifestation thereof using appropriate media (presently CD-ROM). The emphasis is on developing a good user interface and provision of platform-independent tools to allow easy access and manipulation of the WOCE data and products.

 Return to Table of Contents