7. IMPLEMENTATION

7.1 The Challenge of Implementation

Turning the Plan into an operational routine collection of observations is a challenging task. The following overall approach has been adopted. First, summarize the present situation with respect to each of the main areas of interest (biosphere, hydrosphere and cryosphere); second, identify which are the crucial observations which must be continued; third, identify enhancements that are required to satisfy the needs outlined in the previous chapters. An additional issue which will need more effort is to identify the international bodies who would play a coordinating role for each of the observations and how they will interact with national bodies to ensure the observations are collected. In the area of atmospheric observations collected primarily for operational weather forecasting, there are mechanisms organized through the WWW by WMO to achieve this goal. Comparable mechanisms need to be developed for many of the other observations described within this report.

This chapter deals with two principal issues. The first is the global implementation of the tier sampling scheme discussed in Chapter 5. Special attention is given to tier 4 whose implementation poses unique challenges. It is anticipated that the overall implementation will be a gradual process, with a full-up global system not being in place for some years. Therefore, the initial period is considered in more detail in the second part. The initial system must start with programmes and facilities now in place or in the process of implementation. Fortunately, there are many initiatives that are directly relevant to the objectives of GCOS/GTOS and can make very important contributions to the initial observing system. This section reviews some of the existing systems for the monitoring of the biosphere, hydrosphere, and cryosphere.

7.2 Implementation of the Tier Sampling Scheme

The tier structure described in Chapter 5 is a classification system to aid implementation, not a rigid formula for implementation. This enables maximum use to be made of existing data sets, sites and facilities. All the tiers are necessary, but not all the variables need to be measured at all tiers. For example, few hydrospheric variables are measured at tier 4 because the characteristic time scale of hydrological processes tends to be too short to benefit from infrequent sampling. The key tiers for early implementation are 2, 3 and 5, because they are served by existing structures.

Guidelines for site selection

The objective of site selection is to take maximum advantage of existing facilities, including biosphere reserves while ensuring an appropriate global distribution of the measurement sites. It is essential that the sites be carefully chosen so as to meet the objectives of GTOS and GCOS. The different objectives associated with the tiers 1-5 (see also Table 5.1) impose different criteria on site selection. However, all GCOS/GTOS sites have to satisfy the conditions of representatives and sufficiency. This section outlines basic guidelines for identifying sites at the individual tiers.

Tier 1: These major, intensive experimental sites should be located with a primary emphasis on spatial diversity of regional ecosystems, land-use patterns, the availability of regional process integrators like catchments, and feasibility. Capturing the range of the major biome types is a critical priority, but the location within biomes will be opportunistic. Several sites can be drawn from the sites of past or present large-scale experiments, including perhaps BOREAS (boreal forest - Canada), Abracos (ARCS) (tropical evergreen forest - Brazil), Harvard Forest (temperate mixed forest - USA), Oasis (dry land agriculture - Australia), and HAPEX. The actual sites will consist of core areas of 100 km² or less, plus a surrounding region of 10⁴-10⁶ km². It is critical to select tier 1 sites so that they include a range of tier 2, 3 and 4 sites.

Although all tier 1 data and research findings are important to GCOS/GTOS, special attention needs to be given to long-term measurements. Tier 1 sites are large experimental areas and various adjustments are required before they can become part of a long-term monitoring programme. The long-term measurements will be a subset of those made during the initial experimental period but the transition from intensive field campaigns to continuous monitoring will require careful planning.

The responsibility for tier 1 sites during the initial period rests with research programmes such as IGBP, GEWEX, etc. Once the initial programme is completed and adjustments are made to institute a long-term measurement programme, the responsibility for the monitoring programme is likely to rest with a national agency. In most cases, tier 1 sites will be converted to tier 2 or 3 for the long-term monitoring. The responsibility of GCOS/GTOS should be to assure that the data are available and that there are links to the data. Lessons learned at these sites should be incorporated into monitoring programmes where appropriate.

Tier 2: The 100 or so tier 2 centres should be chosen to encompass the major climatic zones, ecosystem types, and land management practices. Ideally, tier 2 sites should be located near the centre of the range of environmental conditions (though not necessarily near the centre of the geographical range) of the system which they are representing. The actual locations will depend more on existing infrastructure and feasibility than on strict spatial guidelines, but the need to capture a broad range of ecosystem types may require developing some new sites.

Since the proposed measurement scheme emphasizes a mix of point measurements and spatial studies, the spatial context of each site will be an important consideration. The best spatial context will be large enough regions of relatively homogeneous ecosystems to allow careful assessments against remote sensing, but enough diversity to allow access to a broad range of ecosystem structure and dynamics.

Tier 2 sites should be located in the major natural land cover (evergreen needleleaf forests, evergreen broadleaf forests, deciduous needleleaf forests, deciduous broadleaf forests, mixed trees and shrubs, closed shrublands, open shrublands, woody savannahs, savannahs, grasslands and permanent wetlands); aquatic systems (lakes, rivers, estuaries); and cryospheric systems (permafrost, ice sheets, ice caps and glaciers).

Tier 3: These are the sites that are most congruent with the existing networks of agricultural and ecological research stations in China, Europe, the USA, and other countries; experimental watersheds; and cryosphere sites. The requirement for permanent staffing and frequent measurements necessitates locating the sites where there is reasonable access, funding, and interest. There is no requirement for spatial representatives at this tier. Although some sites will be needed in under-represented areas, the number of new sites should be a small fraction of the total. Since this tier will provide a primary link with remote sensing observations, selection criteria will include emphasis on reasonable spatial homogeneity over a few kilometres, but this emphasis should not preclude the selection of sites in mountainous zones or in regions with heterogeneous land use or disturbance. Tier 3 sites are intended to sample the range of variation present in the system which they represent. This means that some of them will be close to the average of the various environmental factors which make up the environmental range of the system, while others will be closer to the extremes, and perhaps even at the ecotone of transition to a different system.

Tier 3 sites should be located in natural (see tier 2 above) as well as in managed (12 most important food and fibre crops and agricultural management systems (rice, wheat, maize, potato, sorghum and millet, cassava, sugar cane, extensively-grazed livestock, vegetables, tropical fruits, temperate fruits, and cotton)) ecosystems. The important aquatic systems are rivers, lakes and estuaries; and the cryospheric systems are permafrost, ice sheets, ice caps and glaciers. There are many potential tier 3 sites, but they are not optimally distributed. As a result, some ecosystem types may have more potential tier 3 sites than are needed for GCOS/GTOS purposes, particularly if a rough global balance is to be preserved. Other ecosystem types may have too few sites, or none at all and thus GCOS/GTOS will need to work with funding organizations to enhance and balance the network.

Tier 4: For these points, spatial representativeness is the highest priority. Because the measurement site is small and access is infrequent, they can be located wherever necessary to ensure representativeness. With the order of 10 000 sites globally, the question on optimum sampling design will require a careful consideration (see section on implementation of tier 4 below). A few of these points may be established research stations, but all should be subjected to the prevailing local land management. The land management should not be altered during the monitoring period. It may also be appropriate to continuously add sites so that impacts of the GCOS/GTOS designation can be estimated.

The locations of tier 4 sites should be based on statistical considerations. It is impractical to prescribe one statistical design for all countries. Hence, individual participating nations would be responsible for locating the sites, and may choose either a systematic or a stratified-random approach (or both, for different variables or different systems). This latter approach requires an a priori location specification, but permits rejection of the site and resampling out of the same population if the site is inaccessible or if sampling at the site would compromise national interests.

The main problem with tier 4 is that it is new, and cannot be assembled from pre-existing systems. This makes it apparently expensive and technically untested. The question must be addressed whether tier 4 can be dispensed with, greatly reduced, or phased in without undermining the validity of the entire plan. A detailed discussion of this issue is provided in the sections below.

Tier 5: For the most part, these continuous fields are monitored from space. The frequency of measurement varies depending on the parameter from sub-daily to once in several years. Satellite observations are area averages (for areas < 10² m to >10⁷ m, depending on the sensor and variable), while ground observations are point values. Some variables require surface observations, even for tier 5. These are mostly compilations of existing data sets which took many years or decades to produce, e.g., soil maps, topographic maps, etc. It is virtually certain that all future tier 5 data sets will be obtained through satellite observations.

The implementation of tier 5 requires international collaboration, both in the space and ground components, to produce the required data sets. The preparation of data products from satellite measurements must be based on a long-term programme of data acquisition, archiving, product generation, and quality control. Discussions are now underway in CEOS to set up such a system. For data sets compiled from ground observations (e.g., soils) significant efforts are needed to bring together data from many countries and convert these into homogeneous global data products. International organizations (e.g., the UN) and research programmes (IGBP, WCRP) play very important roles in this regard.

It is important to note that the tier structure is a classification system to aid implementation, not a rigid formula for implementation. All the tiers are necessary, but not all the variables are represented at all tiers. For example, the hydrosphere will have few variables at tier 4, principally because the characteristic time scale of hydrological processes tends to be too short to benefit from infrequent sampling. The key tiers for early implementation are 2, 3 and 5, because they are served by existing structures.

GCOS/GTOS must simultaneously have a top-down and bottom-up approach to site selection. The bottom-up approach can be implemented by publicising the existence and benefits of GCOS/GTOS, and inviting sites (and existing networks of sites) to apply for membership. The GTOS would then screen the applicants for suitability. The guidelines are spelled out in the GTOS Planning Group Report (ICSU, et al., 1996). Briefly, they are:

• Each participating country should have at least one per biome;

• They should be capable of collecting, documenting and making available the appropriate data;

• Sites in under-represented environmental systems have priority over already-represented systems;

• Reasonable permanence is required.

All else being equal:

• Sites where research is also carried out are preferred;

• Long-established sites are preferred;

• Existing sites are preferred over sites which need to be established;

• National support for the site is preferred to dependence on external funding;

• Accessible, practical sites are preferred.

The bottom-up approach would not fill gaps where no or insufficient sites exist, or where the sites are not aware of the needs of GCOS/GTOS. A complementary top-down approach is therefore required to fill gaps in ecosystem types where no sites have been identified. If insufficient candidates are found through reviews of existing sites, an active process may be required to upgrade or establish sites in that type.

The above described tier observing system can be built largely out of existing national and international observation systems, research centres and stations, with modest additions of stations and sites where representation is inadequate, and a major effort towards methodological consistency and data management. An exception is tier 4 which, for the most part, cannot be assembled from pre-existing systems.

Role of tier 4

Tier 4 serves two primary purposes, only the first of which is strictly speaking a monitoring function, but both of which are necessary for an operational system: repeated measurements, for purposes of change detection; and one-time measurements, for purposes assessing the state of the system and model parameterisation.

Tier 4 has the unique role of providing accurate and spatially-resolved data on variables which at present cannot be remotely sensed. An example is the soil carbon content which is important because it is the largest biospheric carbon pool, is subject to change, and influences other factors such as the soil water holding capacity. It cannot be observed from space. Tier 4 can also act as the calibration and validation points for indirect remotely-sensed variables. Tier 3 can also provide these data, but may have insufficient sites to calibrate or validate all the ecosystem classes. It is important to separate the unique role from the additional roles, because they have different sampling requirements.

Table 7.1 summarizes the variables that are unlikely to be available with useful accuracy and resolution in the foreseeable future without tier 4 and those whose accuracy will likely be seriously affected. Furthermore, the GTOS also has several important non-climatic requirements from tier 4, including land use inputs, disturbance regime and soil chemistry (pH, nitrate, phosphate, bases, acidity, e.g.). It is thus evident that functions met by tier 4 are a critically important component of the global observing system. The benefits of enriching the land surface data go beyond the needs of GCOS and GTOS, into issues such as natural resource management at a regional or local scale. Tier 4 can deliver national-level information and national involvement in the global programmes.

Table 7.1 - Impacts of excluding tier 4 from the sampling hierarchy.

Sampling design for tier 4

Various options can be envisioned for the location of tier 4 sites: systematic (e.g., gridded), stratified random and targets of opportunity, among others. Each scheme has implications in terms of cost, political feasibility and statistical analysis. Table 7.2 provides an analysis of the financial, political and statistical sensitivities of the issue.

Table 7.2 - Implications of different tier 4 sampling patterns.

Tier 4 need not have a single sampling scheme for all variables and all places. For instance, where the purpose of the observation is calibration and validation of a remotely-sensed variable, the target-of-opportunity approach is acceptable. For statistical change detection of a variable not indirectly measurable, the scheme must be stratified or systematic. A systematic scheme in one country remains compatible with a stratified scheme in another, if they are designed to the same accuracy specifications.

Implementation strategy for tier 4

Among all the tiers, tier 4 is the least established at the present with only a few countries having monitoring programmes at this level. The implementation strategy is therefore particularly important in view of the new resources required. Some possibilities are considered here.

While there are costs associated with analysing and storing any collected data, reducing the number of variables in tier 4 is not a useful strategy, since the costs involved in sampling have largely to do with accessing the sites, not the time spent at an individual site. There may even be a case for increasing the at-site data collection in order to be able to share the sampling effort with a wider range of clients. For example, could GTOS geo-referenced socio-economic data be collected this way? The main cost-reducing options are thus fewer sample sites, use of targets of opportunity, and phased implementation.

Fewer samples

There are two issues which need to be distinguished, accuracy and spatial resolution. The debate on tier 4 initially concentrated on the number of sites needed to validate a land cover product with a given number of classes, to a given level of precision and confidence for each class (i.e., discrete variables). However, most of the variables which depend on tier 4 are continuous values, not categories. The determination of an adequate sample size for these is more complicated, since it requires a knowledge of their statistical distribution, which is largely lacking, and fluctuates from variable to variable. For a completely unbiased and efficient sampling scheme (stratified random, for instance) and a normally-distributed variable with a coefficient of variation of 30%, an accuracy of ± 10% with 95% confidence would need 36 samples, ± 5% would need 144 samples, and ± 0.5% would need 14 400 samples. In practice, most of the variables in question are log normally distributed and then only once they have been stratified, so the sample number needs to be multiplied by the number of strata if each is to meet the accuracy criteria, or by some area- or value-weighted number for a given global accuracy. If only a global estimate is needed, the sample number is greatly reduced; perhaps by 75% (to allow for increased coefficient of variation as disparate classes are lumped). If regionalised estimates are needed, the requirement goes up in rough proportion to the number of regions.

Although a rigorous analysis has not been performed, it seems that the original estimate of the order of 5 000-10 000 sites for tier 4 remains valid. However, where the tier 4 sites are simply required to calibrate or validate a remote-sensing algorithm, the required sample numbers are much smaller, and the sample location requirements are much less rigorous. Typically, more than 30 samples each are required for calibration and validation of a continuous, linear model if the errors are normally distributed, the model is reasonably predictive (accounting for > 75% of the variance), and the sample points cover the full range of variation. If the number of points needed can be reduced to 500-1 000, then tier 4 can be substituted by tier 3, but all the problems associated with a biased sample scheme remain. Such reduction would also not provide adequate information on variables listed in Table 7.1.

Targets of opportunity

By piggy-backing on other activities, the costs of sampling can in theory be reduced to the marginal costs of the additional effort needed to collect the GCOS/GTOS data. An example is the Soil and Terrain Data (SOTER) Project which aims to improve global soil data products. If the vegetation component were slightly enhanced and geo-location specifications were tightened to ± 10 m, many GCOS/GTOS requirements for surface climate-related observations would be met.

There are two main drawbacks with using targets of opportunity: the sample locations are likely to be biased, biasing the values in an unknown way; and the chances of being able to revisit the point at the same low cost are small. This approach could be useful for calibration and validation of indirect algorithms and for one-time parameterisations, but is not suitable for change detection except in the sense of archiving a current state, which future generations may find useful. At a minimum, the data system should make provision for the recording of tier 4-type data from activities outside of GCOS/GTOS, and should actively pursue their acquisition from the original collectors.

Initial implementation

All of the tier 4 variables have relatively slow rates of change, which allows them to be infrequently collected. Thus only 10-20% of the target sample needs to be collected in a given year. Alternatively, or additionally, each tier 3 station could be tasked with collecting one or two tier 4 data points per year, in an a priori determined location. This approach has the added advantage of a closer link between the tiers at the regional level.

Recommendations for implementation of tier 4 sites

Despite the significant logistical challenges associated with implementing tier 4, this tier should remain as a key part of the hierarchical global sampling scheme. A specific action plan could have the following components:

• Publish a brochure explaining and publicising the hierarchical system;

• Publish and distribute a methods handbook for tier 4 data and actively encourage the placement of target-of-opportunity tier 4 data in public domain databases;

• Phase in the implementation of tier 4 by making it a tier 3 responsibility. Each tier 3 site would appoint one or more dedicated tier 4 observers, who will simultaneously do in-field training of tier 3 personnel and collect tier 4 data.

7.3 Management of the Tier Sampling Scheme

The TOPC has concluded that the tier sampling scheme should be a joint responsibility of both GCOS and GTOS, but that a single observing system should take prime responsibility for the overall management. The general management structure outlined in the GTOS Planning Group Report (part 1, page 15) (ICSU, et al, 1996) is appropriate, but in its initial implementation it should focus on establishing an effective secretariat for this purpose and developing strong links to national implementing agencies and networks.

The steps in an initial operating strategy for managing the system should be:

• Make the Terrestrial Ecosystem Monitoring Sites (TEMS) database more comprehensive by asking regional and national experts to check and populate it;

• Charge the GTOS secretariat, in conjunction with discipline and system experts, to develop a "methods manual" (see comment below), reporting procedures and a training programme. Training exercises should be in-field (not central), and should double as data-collection exercises;

• Establish the communication and distributed database functions of GCOS/GTOS so that they are ready to handle the data flow;

• Begin a dialogue with existing site networks (either international or national) to produce a draft of workable data exchange rules and procedures and to refine the potential site database;

• Conduct an initial tier 3 selection according to the procedures described above;

• Identify under-sampled regions and ecosystem types;

• Hold an international meeting of site and network managers, plus representatives of organizations to establish or upgrade sites in under-sampled regions. Its purpose would be to establish and ratify the rules and methods of data collection, data sharing and quality control and to establish agreement on their participation in a network.

For each variable, standardization should be primarily achieved by specifying the quantity measured, necessary accuracy, measurement frequency, and spatial resolution. Standard methods can be encouraged by publishing manuals and providing training, but these are not obligatory. Full descriptions of the methods used at the site should be provided to the Secretariat at the time that a site joins the network, and updated when necessary. Variables with a significant bias due to the method used should be subject to expert review and cross-calibration exercises.

The first and most important line of quality control is the point of data collection. Data items not accompanied by time, exact location (± 100 m or less) and method should be rejected. On entering a GCOS/GTOS data system, data should, as a minimum, be passed through a "smart" filter to detect gross errors and inconsistencies, and then be checked for reasonableness by a discipline expert.

7.4 Initial Operational System for the Biosphere

It is anticipated that the terrestrial/ecosystem observations for climate will be further developed and implemented jointly with the GTOS, GEMS, World Hydrological Cycle Observing System (WHYCOS) and other activities as appropriate. The following activities will be closely coordinated with projects of IGBP and WCRP, since these programmes are currently very active in these areas. Probably the largest single task in implementing the IOS will be to gain agreement on the data management system and to have individual sites begin to openly share their data.

Existing components

The UNEP GEMS monitoring networks, GEMS-Air and GEMS-Water, working with UN agencies and national and international institutions, collect global pollution data and publish technical assessments on a range of environmental issues. The Man and the Biosphere Programme (MAB) biosphere reserve programme also provides valuable observations in selected sites around the world. Except for these two programmes, there are no global programmes that coordinate observations of global terrestrial systems. Many individual countries, however, have established ecological monitoring and/or long-term research sites. There are over 2 000 such sites located throughout the world. The majority are in the Northern Hemisphere. Even though there is a general lack of sites in the Southern Hemisphere, there are some notable exceptions primarily in Australia, Brazil, Chile, and New Zealand. Consequently, it is proposed that the initial operational system should start at existing sites, and if resources become available additional sites should be added, first in the Southern Hemisphere.

There is an urgent need to:

• Continue the established data collection programmes including those in biosphere reserves. Assure all existing sites are collecting the minimum set of variables listed in chapter 2, using comparable methods;

• Complete UNEP/FAO initiative on Standardization of Land Use and Land Cover Classification;

• Support existing rainfall and discharge measurement networks and enhance their effectiveness using modern data collection, transmission and dissemination systems (see GCOS-17);

• Continue assessments of land cover and land use change using fine resolution satellite data and provide systematic multi-temporal monitoring of selected sites in a variety of biomes (see GCOS-15). Landsat and SPOT have proven particularly useful to the terrestrial community for a number of measurements including detecting and measuring changes in land cover and land use. It is important that data from these and other similar later missions, such as the Indian Remote Sensing Satellite (IRS) and the Japanese Earth Resources Satellite (JERS), are collected systematically for the Earth's land surface and made readily available to user communities;

• Continue to develop and validate LAI algorithms for additional biomes and produce LAI data sets using best available algorithms;

• Routinely produce improved land-related data fields, including albedo, short-wave and long-wave fluxes Photosynthetically Active Radiation (PAR) and the fraction of PAR absorbed by the plant canopy (FPAR). The AVHRR of NOAA has provided valuable source data to derive these variables at coarse resolutions. It is important that data from these missions continue to be made available;

• Continue the experimental 1 km AVHRR data collection for the complete global land surface and derivation of registered calibrated spectral bands and vegetation indices;

• Continue to develop and validate optimal FPAR algorithms for various biomes, and stimulate production of FPAR data sets;

Table 7.3 - Global distribution of potential sites of relevance for TOPC objectives. In brackets () number of sites which have already shown interest in contributing. (From GTOS Plan. The list is not complete, and although those expressing interest in contributing have accepted certain general principles, e.g., regarding the free and ready access to data, there has been no attempt to spell out the full obligations.)

Enhancements

The following enhancements and augmentations to current systems are of high priority, and given modest increases in funding, could be implemented in the near term:

• Water and gas flux monitoring needs to be integrated into intensive ecological monitoring sites in additional biomes and global regions;

• Improve reliability of existing monitoring stations, especially in developing countries, where some ground-based networks are declining in quality;

• Derive improved global soils information through integration of existing soil maps and detailed profile observations;

• Increase the availability of passive microwave data to monitor microwave emissivity and hence surface characteristics such as grain size, wetness, and temperature;

• Assemble global biomass data sets from existing sources within the next 5 years;

• Topographic data should be derived through the integration of existing sources of data;

• Improve parameterizations of land surface interactions in operational 4-dimensional models including making available improved soil moisture fields; and

• Improve global coverage of high resolution satellite imagery.

Several new sensors are due to be launched in the next five years such as MODIS, the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and the vegetation instrument on SPOT-4 to provide improved data at medium to coarse resolutions (250 m to 1 km). The availability of these data will be vital for many aspects of monitoring the biosphere.

7.5 Initial Operational System for the Hydrosphere

While there is not an overall design for an observing system of the hydrosphere for climate purposes, it is clear that there are current activities and there are enhancements to those activities which should be included in the system design.

The number of precipitation, river flow, ground-water and water quality measuring sites exceeds by at least one order of magnitude the number of those represented in "global" systems. Data are collected primarily to be able to respond to local, regional, and national needs, but operational organizations are often unaware or sceptical of their usefulness to a broader community. This puts a major burden on global environmental change interests to convince national bodies of the benefit to the world of sharing their data.

Understanding the impact of climate change on the distribution and abundance of hydrological resources is critical to many socio and economic factors. In this regard the IOS should systematically compile and analyse data for: globally significant freshwater bodies, pristine and/or "stable" ecosystems, research basins used for studying universal-type problems.

The globally significant freshwater bodies should include:

• A set of largest rivers of the world, accounting for at least 50% of total runoff to oceans;

• Selected large lakes;

• Largest world's aquifers;

• Water bodies representative of major anthropogenic influences, selected on a basis of population, economic development, possible water induced international conflicts, etc.

The pristine and/or "stable" ecosystems should be distributed throughout the world reflecting various climatic and physiographic conditions. The stability concept should be rigorously defined.

The following list is preliminary in nature and it is expected to grow as a result of the specific design that is eventually developed. Table 7.4 gives an overview of the existing elements to form the Initial Operational System. It is proposed to build upon these elements to achieve the goals of the IOS.

Existing components

This section presents a description of existing observational systems. A detailed sampling strategy still needs to be developed. A number of existing networks are providing information on the required hydrological variables (Table 7.4). While many existing sites provide useful information for climate purposes, no overall climate strategy which identifies the most critical stations and identifies gaps has yet been developed. The development of such a detailed sampling scheme for climate purposes will be the subject of future TOPC and WHYCOS planning activities.

Table 7.4 - Hydrological variables for the IOS.

All hydrological parameters, except for evapotranspiration and soil moisture, should be observed to provide a global coverage of information. Given the highly variable spatial and temporal nature of evapotranspiration and soil moisture, and given the limitations in the methods for their observation, it was deemed that data for these two variables should be obtained from site-specific observations, that is from tier 1 and tier 2 ecological stations. (See Chapter 5 for details on tiers.)

Currently, the WWW, working with national governments, collects, and makes available, data from over 20 000 stations globally. All available data from national and international archives, whether from meteorological or hydrological stations, that meet GPCC standards, should be accepted for precipitation observations.

Ground-water data are needed from all major aquifers of the world. The total number will probably be of the order of 6 000-7 000 stations globally. Locally, the network density of observation wells will vary, depending not only on aquifer conditions, but on the state of information already available to characterize the aquifer. Thus, aquifers in the "start up" phase of characterization will need more intense monitoring in order to estimate future change.

Discharge data are currently collected on all major rivers of the world, but many of the data are available only at the national level. To study impacts and climate change detection, discharge data should be obtained for several selected rivers that are minimally affected by human intervention. This information - the degree of human intervention - can be stored as metadata in the site characterization, as has been developed by the United States Geological Survey (USGS), (Slack and Landwehr, 1992). To construct global 100 km² grid information for GCM and other climate model analysis and validation, enough discharge data should be collected to satisfy the 10% error constraint of accuracy. Similarly, changes in surface water storage reflect climate impacts and climate variation, although the climate forcing signal can be confounded by human withdrawals. Thus, surface storage information - for lakes, reservoirs and wetlands - needs to be observed only at such sites that are relatively free of human intervention. Finally, biogeochemical transport via sediments needs to be observed at the mouths of large rivers (WMO, 1988).

It should be noted that many hydrological data, specifically discharge data, are gathered by national hydrological services throughout the world, but it has generally not been seen as beneficial to individual countries to share the data to construct global data sets. Archiving the latter will be a formidable task, but a necessary one to accomplish if GTOS and GCOS are to obtain truly global-scale relevant observations. It should be noted that the USGS is currently in the process of making public via the Internet (USGS home page) all of its hydrological data. In particular, the design of the sampling strategy for climate needs to be done in close cooperation with GEWEX, and WHYCOS.

In terms of current capabilities the following recommendations are made:

• The current hydrological observational capability is essential as a component of climate observing system;

• The 4 500 stations that report on precipitation in any given day to the GPCC must be continued;

• The existing network of national water quality stations that provide information on biogeochemical fluxes to the oceans must be continued;

• Coordinated global collection, storage and distribution of hydrological data, including ground water and surface water storage is essential. Existing data need to be compiled, primarily from national and local state agencies;

• Continue WHYCOS in the Mediterranean region.

Enhancements

A number of enhancements are proposed to provide the observations required for operational characterization of the hydrosphere.

• Develop WHYCOS into a fully operational system. Critical elements of WHYCOS include the following:

  (a) Approximately 1 000 discharge stations worldwide;

  (b) Data collection should be performed according to well-defined international standards;

  (c) Data collection will involve the upgrade of existing facilities and selected installation of new ones.

• Provision of underpinning logistical support for data collection;

• Improved data archiving, especially through use of satellite communication;

• It is proposed that data be collected on the world's 50 largest rivers, and selected others sensitive to climate change. One site should be located near the mouth, though above the extent of saline incursions;

• The GPCC at the Deutscher Wetterdienst presently can use only approximately 4 500 stations that report on any given day via the GTS of the WWW. To provide adequate, continuous global fields of monthly precipitation on a 2.5 degree grid, the estimated number of stations required is 40 000. This enlargement would be achieved by including data from existing non-synoptic stations operated separately by national meteorological and hydro-logical services;

• Additional approaches to obtaining discharge data must be explored to overcome the problems of the past, namely the reluctance of individual countries to submit data to global organizations. Efforts should be made to provide global products of interest to individual countries, so that they can see benefit in return for their contributions;

• It is suggested that a coordinating unit or centre should be established for ground water storage data, possibly located with a national hydrological service;

• Establish within five years, a mechanism by which: (1) remotely-sensed land cover data are routinely converted into surface storage area; (2) rating curves are developed for each relevant site; and (3) protocols are established for archiving information;

• Assure that ground water quantity is measured in the 50 largest aquifers in the world.

7.6 Initial Operational System for the Cryosphere

It is clear that there are current activities and there are enhancements to those activities which should be included in any system that may be designed. However, an overall design for a cryosphere observing system for climate purposes has not been developed. The following list is expected to evolve as a result of the specific design for cryospheric observations.

Existing components

There are a number of current sites which can and will serve as the basis of a global monitoring network.

Snow and snow water equivalent

Daily observations of snowfall and depth of snow on the ground are currently made at first-order synoptic weather stations; most of these are distributed over the GTS of WMO. A much greater number of climatological stations also record snowfall and snow depth and these are held in national archives (as hard copy and digital files). In countries where snow cover is an important water resource, ground surveys of snow depth and water equivalent are made at regular intervals (10-day to monthly); these are conducted by various local, regional and national agencies and may not all be held in national archives.

Snow cover extent and sea ice extent

Operational products have been prepared by NOAA/ National Environmental Satellite Data Information Service (NESDIS) for the Northern Hemisphere snow cover extent on a weekly basis since 1966 and by the U.S. Navy-NOAA-Coast Guard National Ice Centre for sea ice in both polar regions since 1972/73. The basis of these products is visible band data from NOAA Polar Orbiting Satellites (since 1971, AVHRR 1-km data). More recently, daily all-weather passive microwave data have been incorporated into the ice analyses which depict ice extent, concentration categories and ice type. Plans are underway for a similar product blending satellite visible and passive microwave data and station observations. Both sets of charts are distributed by the National Climatic Data Center (NCDC) and NSIDC in Boulder, Colorado, USA.

Glaciers

Glacier monitoring is site-specific concerning the direct signal of mass balance. Regional to global representativeness in space and time must be tested by more numerous observations of glacier length changes. As coordinated by the World Glacier Monitoring Service, mass balance is being measured on about 50 relatively small glaciers, and glacier length changes are observed for about 1 000 glaciers. Glacier lengths are easily measured in the field, from aerial photographs, or with high resolution satellite imagery. About 100 regularly observed glaciers and ice caps constitutes a minimum for assessing the long-term representativeness of the mass balance programmes in 10 major mountain ranges of the world.

Ice sheets

At present, observations are sufficient to constrain the mass balances of the Antarctic and Greenland ice sheets to within 20% of their mass turnover, and to provide some idea of the spatial distribution of mass balance and the interannual fluctuation in surface accumulation and (to a lesser extent) surface ablation. The dynamic behaviour of some drainage basins in West and East Antarctica has been studied in greater detail. This status is insufficient for GCOS. An improvement in mass balance estimates by a factor five is required to determine the ice sheet contribution to the present rise in sea level. A secular trend may be the result of the long-term adjustment of the sheets to the Holocene warming and precipitation patterns, or more local changes in the margins due to variations in the atmosphere, oceans or ice shelves. Because the rate of ice flow is largely determined by basal conditions, changes in the trend are unlikely on scales of less than a century. It is sufficient to sample at decadal intervals the large scale mass balance and changes in extent, flow and conditions at the margins. However, for this trend to be observed within the next two decades, a considerably better description of interannual fluctuations in surface accumulation and ablation is required. In addition, to reduce the uncertainty in predicting the consequences of climate change, a more representative sampling of surface ablation is required, together with a closer understanding of its dependence on regional climate, particularly in Greenland. What is needed is an effort in the next decade to greatly improve our knowledge of the interannual fluctuations in space and time, and to assess our ability to forecast them. This requires a tier-based approach to: (i) determine the detailed surface energy balance and its relation to climate variables; (ii) determine the spatial and temporal variability of surface accumulation and its relation to climate variables at present and in the recent past; and (iii) assess the ability of forecasting models to correctly predict surface precipitation and ablation, and therefore provide proxy measurements of ablation and accumulation.

A better understanding of the dynamics of the marine West Antarctic Ice Sheet is needed before the impact on ice sheet discharge of atmospheric and oceanographic driven changes at the ice sheet margins can be assessed. This requires detailed geophysical and glaciological investigations of particular drainage basins and ice sheet/ice stream/ice shelf systems, with a view to better understanding the conditions at the bed and the resulting bed shear stress; the pattern of subglacial drainage; the stress distribution between ice streams and shelves and their effect on flow rate; and the surface and bottom interactions of ice shelves with the atmosphere and ocean.

Permafrost

Many permafrost sites are presently being monitored for active layer thickness and permafrost temperature at depth in North America, the Russian Federation, and in some high-altitude areas (Tibet Plateau, Tien Shan Mountains, European Alps). Permafrost data are held by national agencies, the private sector and university researchers. Efforts are presently being made by the IPA to systematically rescue and collect such information.

Lake freeze-up/break-up

There are over 500 stations in both the Northern and Southern Hemispheres that collect data on lake freeze-up and break-up.

There is a demonstrable need to maintain the following components:

• Daily observations of snowfall and depth of snow on the ground at first-order synoptic weather stations;

• Regular ground surveys of snow depth and water equivalent; these are conducted by various local, regional and national agencies;

• Observations of lake freeze-up/break-up dates on about 200 lakes and rivers in northern latitudes and high altitude areas. Expand the network to include lakes in the Southern Hemisphere;

• Use of array(s) of ULS for measuring sea ice thickness in a key area of the Arctic (Fram Strait, Bering Sea);

• Acquisition of sea ice data by operational agencies for both polar regions and adjacent seas;

• Gridded daily brightness temperature and ice concentration products from SSM/I and future passive microwave sensors;

• Measurements of mass balances on about 50 relatively small glaciers and glacier length for about 1 000 glaciers;

• Radar image mapping at 10 metres resolution of the Greenland and Antarctica ice sheets;

• Complete altimetric mapping and image mapping of the Antarctica and Greenland Ice Sheets by satellite or aircraft platforms at decadal intervals with a resolution of:

  10 cm accuracy at 10⁴m² resolution of ice sheet margins.

Enhancements

The following enhancements are strongly recommended:

• Develop routine daily products of snow cover and sea ice extent, snow water equivalent and sea ice concentration from EOS MODIS and planned Multifrequency Passive Microwave Radiometer (MPMR);

• Develop and implement a routine procedure for optimal merging of the various satellite data streams and surface observations. The data products should be distributed through appropriate data centres;

• Complete radar and optical mapping at 100 m resolution of ice sheets, ice caps and glaciers, with a ten-year interval. The ERS-1 SAR, Radarsat SAR and EOS ASTER satellites should be used to provide an initial baseline;

• Digitize and quality control the existing records of snow and snow water over the next 5 years;

• Develop a central archive of all available snow water equivalent data, at daily intervals;

• Assemble a central archive of lake ice records;

• Implement an upward-looking sonar network in the Antarctic corresponding to the moored ULS array in the Arctic. Ensure data are transferred to a designated central archive (NSIDC, for the Arctic, designated by the WMO-WCRP ACSYS) for distribution;

• Rescue and archiving of earlier permafrost borehole data are urgently needed;

• Develop a strategy for internationally coordinated long-term monitoring of permafrost and install deep (> 20 m) boreholes at selected sites in polar and high-altitude areas. This will be the subject of a future TOPC workshop;

• Establish baselines for long-term monitoring using laser altimetry in combination with kinematic Global Positioning System (GPS). Measure surface elevation changes along representative flow lines of large glaciers in Alaska, Canada, Patagonia, and the Himalayas;

• Glacier mass balance programmes should be strengthened in the Southern Hemisphere. Systematic application of remote-sensing products (air photos, high-resolution space imagery) should enable global coverage to be reached with respect to glacier length/area change and inventory data.