Ocean Observations for Climate

Facilitating Access to Global Observing Systems Data and Information

An Ocean Observing System for Climate

Specifications for an Ocean Observing System for Climate were developed by the Ocean Observing System Development Panel and were reported in 1995. Subsequently these requirements were further developed by the Ocean Observations Panel for Climate and other associated activities. The results, extracted from GOOS Document No. 66 "Report Global Physical Ocean Observations For GOOS/GCO - An Action Plan for Existing Bodies and Mechanisms" are reproduced here for convenience.

See references below for more information.

Requirements

The requirements are presented by field, first noting the desired characteristics of the processed signal (output) for different applications. The sampling is presented in terms of a strategy and a set of "benchmark" accuracies. The benchmark accuracy is a standard against which measurement accuracies can be compared. Measurements which fall well below the benchmark may not be useful at all, or may require improved technique and/or quality management. Measurements with accuracy far greater than the benchmark may have reduced cost-effectiveness. Where appropriate specific implications of remote sensing are noted. Alternative sources of information and perceived trends in the requirements are noted where appropriate.

Information on this page:

Sea Surface Temperature

Characteristics of the processed signal

For NWP which supplies stress, heat flux estimates: 0.2-0.5°C on 100 km squares with 3 day resolution. (Note that regional systems and severe weather prediction seek 10-20 km resolution daily (Annex III), and that these are becoming increasingly important for coastal applications (e.g., hurricane forecasts) and some climate applications).
For ENSO prediction and verification: 0.2-0.3°C at 200 km x 30-100 km scales every 5 days in the tropics. The bias requirement is more severe in the convective regions, less severe in the central to eastern Pacific. Meridional resolution has a high premium attached to it.
For climate change detection: 0.1 °C on 2-500 km squares monthly.
Mesoscale and coastal oceanography/GODAE: 0.2°C (relative) 10 km scales daily. Quality and bias is less of an issue, but gradients and features are more important.

The diurnal cycle is a potential source of error for most of these signals.

Sampling strategy and benchmark accuracies

Use geostationary and polar orbiting satellite data for spatial resolution and to reduce geophysical noise in climate signals (Annex II Table B)
Use in situ data for calibration and to produce blended products with optimized bias reduction.
The requirement for remotely sensed SST is 10 km resolution and 3-6 hour sampling, the latter to reduce aliasing error, with 0.1-0.3°C relative error. The temporal sampling implies increased utilization of geostationary platforms. The NWP and mesoscale applications are the dominant determinants of resolution; climate for the error.
The sampling for in situ observations is controlled by the need to remove bias from the satellite product, mainly for climate change applications, but also in the event of unexpected aerosol interference. The best estimate remains at 0.1°C on 500 km squares on weekly time scales and O(25) samples with accuracy better than 0.5°C. ENSO requires an adjustment in the tropics as suggested by the scales mentioned above.

Indirect sources of information

Virtually none. None of the operational analysis systems use model predictions or assimilation to great effect. It remains a field that is far easier to observe than model. It should be remembered that remotely measured SST is indirectly inferred from radiative measure. There is also no unique definition of SST.

Trends

CLIVAR and GEWEX may require resolution of the diurnal cycle and improved accuracy of products in the tropics (0.1°C). There remain some issues concerning the use of bulk, near-surface and skin temperatures in climate applications. This is likely best addressed through greater use of mixed layer models. Applications requiring accurate high-latitude SSTs might also become more important; satellite sampling is poor in some regions and so in situ programmes become more important. (See also the final report of the OOPC/AOPC Workshop on Global Sea Surface Temperature Data Sets, IRI, LDEO, Columbia University, USA, 2-4 November 1998.)

Sea Surface Salinity

Characteristics desired of the processed signal

While sea surface salinity (SSS) products remain largely in the research community, the OOSDP expressed a strong desire for improved monitoring of SSS.

At high-latitudes, surface salinity is known to be critical for decadal and longer time-scale variations associated with deep over-turning and the hydrological cycle (the rms annual variation is ~ 0.2; while interannual is typically ~ 0.2-0.5). These are relatively large-scale signals.
In the tropics, and in particular in the western Pacific and Indonesian Seas, and in upwelling zones salinity is also believed to be important ( rms variation ~ 0.3). A product on 250 km squares and monthly resolution with 0.1-0.2 accuracy would be satisfactory for most applications.

Sampling strategy and benchmark accuracies

One sample per 200 km square every 10 days with an accuracy of 0.1 is the benchmark [the signal to noise ratio is typically not favourable]. The tropical western Pacific and Indian Oceans, and high latitudes are the highest priorities.

Indirect sources of information

Precipitation estimates provide some useful indirect estimates of SSS. In theory, a combination of altimetry and ocean temperature should also be useful for inferring SSS, but this has yet to be demonstrated in practice.

Trends

Retrievals using passive microwave L-band with an accuracy of 0.1-0.2 over 200 km are sought. ESA's SMOS mission is aimed at such accuracy.

There remains the possibility of remotely-sensed SSS, at the threshold level listed in Annex II, Table B. The need for improved salinity networks has been a theme in CLIVAR and in the OOPC, principally because of the significant interest in the tropics and the interest in decadal-to- centennial variations at high latitudes.

Surface Wind Vectors

Characteristics desired of the processed signal

Estimates come from NWP, from direct analyses of wind data (e.g., the Florida State University (FSU) product) and from products generated directly from remote sensing. Re-analysis products are also popular in the research community.

For ENSO applications: 5% in direction and 0.5 m/s in speed estimates are required at 5° longitude and 2° latitude horizontal scales monthly. For longer periods the accuracy requirements are slightly weaker, but a global resolution of 2° x 2° is desirable (such products are not used directly for detecting climate change but for driving models studying climate change).
Many mesoscale, coastal, and some climate applications seek much finer temporal and spatial resolution. Research applications also have demanding requirements.

Sampling strategy and benchmark accuracies

The OOSDP did not give a specific sampling rate, citing the many different applications as one of the mitigating circumstances. The following is a guide:

2° x 2° resolution at an accuracy of 0.5-1.0 m/s in the components every 1-2 days is the benchmark for climate applications;
Daily 50 km resolution at an accuracy of 1-2 m/s daily is the benchmark for mesoscale/GODAE and coastal applications.

Indirect sources of information

Clearly NWP and forecasts based upon previous data are an important source of indirect information, as are the other contemporary atmospheric and ocean surface data (e.g., cloud drift winds; mean sea level pressure (MSLP)). Atmospheric assimilation systems continue to have problems ingesting surface wind data, so direct estimates are essential, particularly in the tropics (e.g., TAO).

Trends

ADEOS/NSCAT showed that estimates of around 2 m/s accuracy every 2 days could be obtained, at resolution of around 50 km. If such an instrument is flying operationally, then the role of in situ data would be more like that of in situ SST data for SST estimates. That is, providing ground truth for bias correction. The reanalysis projects have yielded improved products, which are popular, but which have short-comings with respect to quality and resolution. The demand for higher resolution, particular for cyclones and hurricanes, is growing. There is consensus that at least one operational double swath scatterometer is justified, and an emerging case for two to eliminate aliasing of these high-frequency variations into climate signals.

Surface Flux of Heat, Water

Characteristics desired of the processed signal

For surface heat flux: 10 W/m2 accuracy over 2° latitude by 5° longitude by monthly bins.
For precipitation: 5 cm/month over 2° latitude by 5° longitude by monthly bins.

Sampling strategy and benchmark accuracies

Use flux estimates from NWP/reanalysis projects and adopt the sampling requirements of WWW.
Use direct calculations based on surface marine data, both satellite and ocean based (e.g. FSU, SOC) with O(50) observations of the main parameters (wind, air temperature, humidity, MSLP, SST) per bin. Specific high priority actions include:
- Improved SST, air temperature, humidity, MSLP, precipitation and absolute wind velocity on selected VOS;
- Shortwave and longwave radiometers on selected VOS;
- Satellite-based estimates of radiation and precipitation; and
- A number of flux buoys to provide high-quality verification.

Indirect sources of information

There are no direct methods for measuring the net heat and water surface fluxes, though there are methods for measuring some components. NWP takes advantage of many indirect (non-ocean) sources of information. Ocean budget techniques (e.g. TOGA COARE) have proved quite effective for estimating net heat flux; a similar technique can be employed for net water flux based on salinity (water) budgets. Ocean models with assimilated ocean temperature data can also be used to infer surface fluxes.

Trends

As noted above, there is increasing emphasis on the oceanic water budget, so at-sea measurements of precipitation (e.g., from TAO, VOS) are becoming increasingly important. Several methods are available based on satellite data (e.g. TRMM), and high-quality in situ data are needed for algorithm development and calibration. NWP prediction estimates are still plagued by large uncertainties and systematic bias, particularly in those components influenced by cloud cover. Ocean models are extremely sensitive to bias errors, so the sampling strategy must endeavour to provide as much ground truth as possible. This strategy then places a high premium on data quality, and hence on improving the quality of in situ data streams.

Sea Level

The OOSDP report discussed long-term trends and ocean variability needs, but was not specific with respect to the in situ gauges or altimetry. The OOPC, CLIVAR and NOAA, convened a workshop to refine these requirements, in conjunction with GLOSS and its Implementation Plan (1997).

Characteristics desired of the processed signal

For climate change: annual global sea-level change on large space scales (~ 500 km), with accuracy of around 1-2 mm a year.
For estimates of sea surface topography anomalies (for ENSO and ocean variability studies): for 10-30 day periods an accuracy of 2-5 cm and a spatial resolution of:
- 500 km zonal x 100 km meridional in the tropics;
- 2° x 2° elsewhere.
For estimates of mesoscale variability: on a 25-100 km square with an accuracy of 2-10 cm every 5 days (see also Table B, Annex II).
For ocean circulation (estimates of absolute sea level): on a 200 km scale and 2-5cm accuracy (dependent on a gravity mission).

Sampling strategy and benchmark accuracies

Long-term trends require a dual strategy.
The preferred observing strategy comprises:
- altimetry for global sampling, at approximately 10 day intervals;
- approximately 30 in situ gauges for removing temporal altimeter drift;
- additional gauges at the margins of the altimeter (e.g., continental coasts and high latitudes); and
- a program of geodetic positioning.
An alternative observing system, proposed due to the lack of guaranteed availability of altimetric data and due to the lack of experience and confidence in the application of altimetry to measuring long-term trends, would comprise
- a globally distributed network of in situ measurements, with similar effect to the GLOSS Long Term Trends (LTT) set of tide gauges; and
- a program of geodetic positioning.
For large-scale variability, sites for in situ measurements are limited. The TOGA network should be maintained (at higher priority than assigned in OOSDP, 1995), with increased focus on the tropical western Pacific and Indonesian Throughflow, and in the western boundary current regions. The GLOSS Implementation Plan and OOPC/CLIVAR Sea Level Workshop (GCOS,1998) detail priority stations for monitoring large-scale variability. TOPEX/Poseidon (T/P)-class altimetry with 100-200 km resolution and ~2 cm accuracy is also highly recommended. Altimetry, in general, is now rated far more highly than it was at the time of OOSDP (1995).
Mesoscale variability is only accessible with multiple altimeters, at least one being T/P class. The optimal sampling is at a 25 km scale and an accuracy of 2-4 cm every 7 days.

Indirect sources of information

For long-term trends there are no viable alternatives, though acoustic thermometry may offer some sort of alternative measure. For ENSO monitoring and prediction, there is redundancy between wind, SST, sea level and subsurface temperature; sea level has the advantage of a history stretching back into the 1970's, and the fact that it measures the joint effect of thermal and haline variations. For large-scale variability in general, thermal data offer similar types of information. However their complementarity would seem a more powerful attribute, with sea level measuring the vertically integrated variability, and temperature profiles measuring vertical structure. There is no alternative for mesoscale variability.

Trends

For ENSO prediction, sea level is enjoying a revival, courtesy of TOPEX/Poseidon and improved methods for assimilating sea level information. There is more confidence in altimetry for long-terms trends (c.f. OOSDP 1995). For the mesoscale, the number and type of altimeters required still remains open (see notes in Table B, Annex II). The gravity missions GRACE and GOCE (OOPC, 1998) will provide an opportunity to exploit absolute measures of sea level.

Sea Ice

Desired characteristics of processed signal and available techniques (for climate)

Although sea-ice is a basic component of the climate system, systems to observe sea-ice properties are limited. The limited OOSDP recommendations reflect this situation.

Sea ice extent: daily 10-30 km resolution is attainable using passive microwave sensors and meets the requirement for large-scale observations at seasonal to interannual time scales but serious problems remain in their interpretation. Sea ice regions vary greatly in character and there is difficulty in establishing algorithms to describe sea ice extent and concentration in the presence of snow, melt water, thin ice, etc. Synthetic aperture radar (SAR) where feasible provides finer accuracy. In situ techniques are largely insignificant for large-scale monitoring.
Sea ice concentration: 2-5% in sea ice concentration, measured daily, provides a target for microwave sensors at the same spatial scales as for sea ice extent but the same interpretation problems exist.
Sea ice drift: Measurement of drift as opportunities arise, using buoys and pattern-tracking from remote sensors (SAR, AVHRR).
Sea ice thickness: 2-500 km scale mapping of ice thickness on monthly time scales with accuracy O(0.2m), using upward-looking sonars and other devices. Sea ice thickness and volume are an important climate variable but are the most difficult to obtain on the large scale.

Other comments

Operational sea ice systems are more advanced in the Northern Hemisphere than in the Antarctic. Work in the Antarctic is largely driven by climate concerns. In the Arctic operational real-time prediction of sea ice is also a major issue. For decadal-to-centennial variability, sea ice extent, concentration and volume are required. Surface salinity and sea-ice export estimates are complementary. For models to be useful for sea ice prediction (on short time scales), good wind data are essential.

There are extensive services for the provision of real-time sea-ice date in the vicinity of the Arctic. In some cases, observational programmes have been going for over 50 years.

At this time, GOOS has not fully considered just how these activities should be dealt with. For JCOMM and the several activities that were being covered by CMM, it is clear sea-ice needs to be considered more fully in future versions of this action plan. In the meantime, the requirements set down by WMO will be used as a guide.

Surface Waves

Like real-time sea-ice monitoring and prediction, the requirements for surface wave/sea state analysis and forecasting have not been considered in detail by GOOS - Kamen and Smith (1998) examined some of the issues related to present forecasting systems but did not examine the requirements in detail. Within WMO, wind waves have been the province of CMM and there has been an active sub-group on wave modelling and forecasting. It is the published requirements of this programme that have been added to Table B of Annex II.

A paper has been solicited for the OceanObs99 conference to develop an agreed set of requirements for wind waves. In broad terms, we can expect wind wave requirements.

Significant wave height at 100-250 km and 6-12 hour with accuracy 0.5m.
In situ (wave ride buoy) measurements at several locations, preferably in deep water, to verify remote measurements and operational models. These data should be circulated on R/T.
A wind-wave verification scheme whereby in situ data are assembled and made available to operation agencies.

Surface Carbon Flux

For the most part, these measurements remain within the research community. But the technology exists to use VOS and drifters to collect pCO₂ in situ measurements, and satellite ocean colour provides effective proxy data for pCO₂.

Sampling strategy and benchmark accuracies

Seek pCO₂ and total CO₂ measurements with an accuracy of ñ2-3 æatm and ñ2 æmol respectively.

In situ sampling is not expected to reach threshold rates, so simply aim for enhanced VOS, mooring and drifter measurements, piggy-backing wherever possible on existing operational systems. Ancillary SST and atmospheric data are important.

Aim for continuing global satellite ocean colour measurements, at 25-100 km resolution and daily coverage, with 2-10% accuracy.

Development and validation of satisfactory remote sensing algorithms is important.

Time-series stations are playing a key role in research and the Ocean Climate Time-Series Workshop (Baltimore, MD, USA, March 1997) co-sponsored by GOOS, GCOS, WCRP and JGOFS (GOOS Report No. 33, GCOS Report No. 41) saw an important role in the future for such Time- Series.

Comments

Some non-biological applications (e.g. tropical ocean modeling) are using ocean colour to estimate opacity. Independently of any non-physical applications, this suggests that there is a good case for adding ocean colour to the list of needed remote sensing techniques.

Upper Ocean Temperature

In the past, upper ocean thermal networks have largely been the province of research. Making significant parts of these networks operational is one of the key themes of OOPC and remains a high- priority issue.

Characteristics desired of the processed signal

General large scale requirement is for 2-500 km scale bimonthly global maps of the heat content and the first few vertical modes of variability; and monthly climatologies on 1° resolution. An accuracy of ~ 0.5°C is useful.
For ENSO forecasts: 1° latitude and 5° longitude resolution every 10 days and over 500m vertically (mixed layer depth (MLD) and ~5 vertical modes) to an accuracy 0.2-0.5°C.
For mesoscale applications: 25-50 km resolution every 2 days over 500 m with an accuracy of around 0.5°C.
For climate trend, better than 0.1 C/year accuracy.

Sampling strategy and benchmark accuracies

Maintain TOGA/WOCE broad-scale VOS sampling (1 XBT per months with 1.5° latitude and 5° longitude resolution). Priority to lines with established records, of good quality, and in regions of scientific significance (e.g., tropics, particularly outside the domain of TAO, and the TRANSPAC region).
Maintain TOGA Pacific network, in particular TAO (OOSDP did not specify part or all of the present array, but did suggest "close to" 1994 levels). Around 4 samples every 5 days per 2° x 15° bin, with 10-15 m vertical resolution is deemed satisfactory.
Enhanced coverage in the equatorial regions in the vicinity of sharp gradients (e.g. Kuroshio): O(18) sections per year, with 50-100 km resolution.
Boost routine sampling of the polar regions (at broadcast mode levels)
Use of profiling floats to implement a truly global observing system. This is a technology that is developing rapidly and real-time data are now available; sampling strategies have yet to be defined for "operational" use but a float profile per 2-300 km square every 10 days might be a feasible target. Argo, developed under the auspices of GODAE, will become the mechanism for developing a strategy for deploying ~ 3000 floats globally for GODAE in the period 2003-5. As such it will serve as a pilot project for the longer term use of profiling floats in the GOOS/GCOS OOS.

Other sources of information

Clearly altimetry offers complementary data. For the tropics, it is feasible a good model plus SST and wind-forcing may be able to forecast subsurface temperature structure with useful skill. However, at the present time, there is no reason to lessen the requirements outlined above. Several groups are using empirical relationships plus assumptions about the T/S relationship to infer sub-surface structure from altimetry (variously known as synthetic or pseudo XBTs). Acoustic thermometry has good potential, particularly for long-term change and in regional modeling. It seems highly unlikely that an in situ solution will be found for the mesoscale applications. Rather, it is likely a mix of moorings, XBTs and profiling floats may be used to pin-down the global, large-scale thermal structure, and a mix of altimetry, SST and colour used to specify the horizontal structure of the mesoscale field.

Trends

Profiling floats, and in particular the Argo initiative, are arousing a great deal of interest and seem to offer the one real chance for global temperature sampling (VOS are limited in terms of geographic coverage, and moorings are better suited to tropical and boundary regions). A program called PIRATA is testing TAO-like moorings in the tropical Atlantic, and the Japanese TRITON program is testing moorings for mid-latitude climate studies, and for Indian Ocean studies. (See also section on Time Series Stations below.)

Heat and Water Transports and Budgets

The OOSDP recognised that observing changes in the ocean circulation and its inventories of heat, fresh water and carbon would require the use of profiling floats, precision altimetry, knowledge of the surface forcing fields, etc. which are discussed elsewhere in this section. In addition, transocean sections at key latitudes and in regions of watermass formation would be essential. The OOSDP report, which was published at the end of 1994, states that, although repeat hydrography and transocean sections are essential, they lacked some urgency as part of the initial ocean observing system because of the global coverage being provided by WOCE and the expected repeat time of five to ten years. The OOPC has not yet reviewed the question of transocean sections and repeat hydrography given the experience of WOCE.

Characteristics desired of the processed signal

For the estimates of the variability of meridional heat, fresh water and carbon fluxes, transocean sections are required at key latitudes with station spacing that resolves mesoscale variability, 25-100 km, at specific latitudes and at a repeat time to be determined based on the experience of WOCE.
For the determination of the changing inventories of heat, fresh water and carbon, additional sections with station spacing appropriate to the sales of variability may be required to supplement the transocean sections for transport estimates.
For the measurement of water mass formation, sections are at least annually to observed yearly watermass formation and at a station spacing adequate to sample region.

Sampling strategy and benchmark accuracies

The sampling strategy, desirable accuracies and operational procedures for deep sea hydrographic observations are fully described in the documentation prepared for WOCE implementation and can be seen in WCRP (1988 a, b) and WOCE (1991), WOCE Hydrographic Programme Office (1994).

Trends

Hydrographic sections remain the fundamental tool for observing changes in watermasses and the climatically important meridional ocean transport of heat, fresh water and carbon. The availability of profiling floats measuring T and S, moored profiling instruments, and precision altimetry combined with the increasing power of ocean dynamical models and techniques for assimilating observations could lead to more comprehensive approaches in the future.

Upper Ocean Salinity

Upper ocean salinity remains primarily an experimental field in terms of applications. An exception for the OOSC are the upper ocean segments of the hydrographic data to be obtained from transocean and repeat sections as well as time series stations for which the techniques of obtaining accurate salinity data are well established. Expendable CTDs (XCTDs) on selected VOS lines and perhaps also high density lines, and salinity sensors on some TAO moorings, were recommended by OOSDP.

Characteristics desired of the processed signal

Monthly subsurface profiles with an accuracy of 0.1 on 3° squares would serve most large-scale purposes.

Sampling strategy and benchmark accuracies

A profile per month per 3° square at better than 0.02 accuracy is a benchmark.

Trends

There are suggestions that sub-surface salinity is important for ENSO forecasting and CLIVAR Upper Ocean Panel has given high priority to enhanced sampling.

Again, the profiling floats of Argo would seem to offer the best opportunity for increased global coverage, though there remains some questions about the stability of the salinity sensor. Current plans suggest Argo will deliver in excess of 50,000 profiles of 5 to 2.000m. Studies using a combination of altimetry, sea surface salinity and ocean temperature have shown promise for estimating salinity (Reynolds, pers. comm.). CLIVAR is intent on pursuing a better description of the hydrological cycle which implies greater emphasis on subsurface salinity.

Ocean Currents

The OOSDP (1995) report was vague with respect to the need for velocity measurements, principally because there were few, if any, operational applications. They recommended a minimal array of current meter moorings and VOS acoustic doppler current profilers (ADCPs) for validation of models as well as gathering surface drift data from buoys.

Sampling strategy and benchmark accuracies

At the surface: a global surface drifter program can yield very good surface current estimates. The benchmark is global coverage of one drifter measurement per 600 km square per month, with current- following accuracy of around 2 cm/s which would give estimates of the mean velocity good to 10% of the eddy variability.

For the subsurface: a minimal array for model verification. Accuracies of the order 5 cm/s for monthly averages would be the benchmark for the tropics.

Trends

Several groups are experimenting with surface current estimates derived from altimetry and from SST- pattern following techniques. GODAE will place greater premium on surface velocity data since its short-range forecasting goal includes estimates of the surface currents.

There is considerable interest in the prospects from the gravity missions GOCE and GRACE. GOCE to be launched in the period 2001-3 will provide geoid accuracy of ~1.0 cm on scales of 500km and ~0.1 cm on scales of 1000 km. GRACE to be launched after 2003 will provided geoid accuracy of `2.0 cm on 100 km scales and better than 1.0 cm on 1000 km scales. If successful, these missions would allow the calculation of absolute surface geostrophic currents on smaller scales (down to mesoscale at mid- latitudes) and greater accuracy than presently available, and enhance the already substantial impact of satellite altimetry.

Time Series Stations

Time series stations do not fit neatly into the above field-by-field description for the OOSC. They provide long records with temporal resolution that is short compared with the characteristic dominant variability, as well as co-located measurements of several different variables, sometimes including chemical and biological parameters. These attributes make such data sets powerful and complementary to the data mentioned previously, particularly for physical and phenomenological studies. The Ocean Time Series Workshop (IOC, 1997) discussed the merits of time series as a strategy for both GOOS/GCOS and CLIVAR. The CLIVAR Implementation Plan (WCRP-103, 1998) includes a summary of the attributes of 8 existing time-series and attempts to evaluate their relevance to meeting the goals of CLIVAR. The OOPC has yet to attempt this with regard to the OOS. However, it can be noted that the Time Series Workshop presented a strong case for continuing the long time series at Bravo and station "S". The TAO array also contains several important long records (e.g. at 110 W) which should be maintained. Station "Papa" is to be the subject of sustained study within CLIVAR and may be another potential site for consideration for the OOS. Others may be equally relevant.

Ocean Modeling

As noted at the beginning of this section and in OOSDP (1995), models are essential for the effective and efficient use of observations. Equally, data from the real ocean are essential if a model is to move beyond theory and concept. Ocean data assimilation, or ocean state estimation in the nomenclature of GODAE, is the preferred methodology for merging theoretical knowledge of the ocean (models) with data. Note the data may be ingested through both boundary conditions and adjustments to the sate variables. The development of models is not the purview of JCOMM. However, the end-to-end chain of observation- processing-service inevitably involves models of varying levels of sophistication and so JCOMM must take into consideration the implementation and routine use of models.

Management and oversight

The OOSDP (1995) stressed the importance of scientific involvement in all parts of the data flow, from measurement through to end product. The OOSDP recommended the establishment of an evaluation process, perhaps built around a distributed network of contact points in operational centres, whose prime objective was to ensure that the data gathering, processing and dissemination was consistent with observing system plan. It was important that this evaluation process provided feedback to the sources of the data in regard to quality, timeliness, percentage consumption (that amount of data that were actually ingested), and so on.

The OOSDP all set out several principles for data management:

the information management system will be built as far as is possible and appropriate on existing systems;
the information management system should be "operational" (c.f. experimental) in the sense as that for the observational network;
the information management system should be consistent with the objectives, needs and priorities of the scientific design;
data should be transmitted from instrument platforms to appropriate data centers and made available for further processing as soon after measurement as is feasible and practical;
quality assurance of data and products should receive high priority to maximize the benefit drawn from the often difficult and expensive ocean measurements;
the information management system should be user-oriented to ensure that the needs of users, the ultimate sponsors of the observing system, are served well;
full and open sharing of data and information among the participants and users of the observing system is essential to its successful implementation and operation;
observing system participants should contribute data voluntarily and with minimal delay to data archival centers which in turn should be able to provide information to users effectively free of charge;
the observing system will be most effective if practical international standards are developed for all phases of information management;
information management will be most effective if it is part of the overall monitoring and evaluation process of the system.

References

Nowlin, W.D., Jr, N. Smith, G. Needler, P.K. Taylor, R. Weller, R. Schmitt, L. Merlivat, A. V‚zina, A. Alexiou, M. McPhaden, and M. Wakatsuchi 1996: An ocean observing system for climate. Bull. Amer. Met. Soc., 77, 2243-2273.

Ocean Observations Panel for Climate, 1996. Report of the 1st Meeting of the Ocean Observations Panel for Climate, RSMAS, Miami, March 1996, publ. by the IOC.

Ocean Observations Panel for Climate, 1997. Report of the 2nd Meeting of the Ocean Observations Panel for Climate, Cape Town, South Africa, February 1997, publ. by the IOC.

Ocean Observations Panel for Climate, 1998. Report of the 3rd Meeting of the Ocean Observations Panel for Climate, Grasse, France, April 1998, publ. by the IOC.

The Ocean Observing System Development Panel 1995: The Scientific Design for the Common Module of the Global Ocean Observing System and the Global Climate Observing System: An Ocean Observing System for Climate. Report of the OOSDP, publ U.S. WOCE Office, Texas A&M University, College Station Texas, 285 pp.

Smith, N.R., G.T. Needler and the Ocean Observing System Development Panel, 1995. An ocean observing system for climate: The conceptual design. Climatic Change, 31: 475-494.

IOC, 1997. Ocean Time Series Workshop, Baltimore, March 1997.

WCRP-103, 1998. The CLIVAR Implementation Plan, The World meteorological Organization, WMO/TD No. 869.