POINT OBSERVATIONS AND DATA PRODUCTS

Terrestrial measurements
Atmospheric measurements

IN SITU POINT DATA ARE DERIVED FROM MEASUREMENTS MADE AT SPECIFIC LOCATIONS (SITES). "POINTS" CAN VARY IN SIZE DEPENDING ON THE LAND EXTENT OF THE ORIGINAL MEASUREMENTS. EVEN SO, SCALING OF POINT DATA TO AN AREAL EXTENT APPROPRIATE FOR MODEL DEVELOPMENT, TESTING, OR VALIDATION NEEDS TO BE CARRIED OUT. Scaling may be as simple as a few quality measurements made in several identified land cover strata in a 3 × 3 km area, with mean values for measured variables mapped to the appropriate strata.

Alternatively, scaling may involve a geostatistically designed sampling scheme over which a complete set of critical variables are measured using "best-practice" measurement protocols. The measurements might then be integrated with remote sensing and process models to develop a finely gridded set of estimates for the site. These estimates are checked against independent measurement data before a comparison is made to coarsely gridded estimates of carbon-related variables.

In this section, current ecosystem and atmospheric networks of measurement sites are reviewed against variables that are important for model development, testing, and validation of model-derived outputs at local to global scales. The distribution of networks which need to be considered in carbon modelling (such as those involved in bioclimatic, spatial and temporal analysis) are also reviewed. The assessment helps identify coverage gaps within existing networks. Recommendations are provided for achieving a more complete network coverage in the relatively near future.

Terrestrial measurements

Status

Established ecological research networks provide a wealth of in situ measurements, but acquiring and harmonizing adequate ground measurements data for TCO is a major constraint. It is rare that complete sets of measurements are collected for a broad range of field sites and/or over long time. However, climate and other environmental drivers provide a way to bring together diverse collections of data by exploiting relationships in 'environmental space'. Terrestrial carbon dynamics are controlled or influenced by several environmental dimensions. Important axes of variation include (in no specific order and with topic overlap).

Direct:

• climate;
• disturbance and its history;
• land use and its history;
• sensitivity to climate change;
• vegetation functional type and 'successional stage';
• terrain complexity;
• maximum photosynthetic capacity.

• spatial grain and extent;
• temporal grain and extent;
• biogeography, latitude and longitude.

In order to develop global scaled-up data of the terrestrial carbon cycle, empirical relationships can be combined with gridded products derived from remote sensing imagery to predict global patterns of components of the carbon cycle, thus yielding estimates that are approximate but independent of those based on process models. An example of synthesizing in situ data and combining it with remote sensing products are the World Resources Institute's (WRI) estimates of carbon stored in above-ground live vegetation (PAGE, 2000).

WRI estimates are based on those developed by Olson et al. (1983), combined with the Global Land Cover Characteristics Database (GLCCDB, 1998). There is also a need to review and update Olson et al. estimates of carbon storage in vegetation as this has, in some cases, been superseded by more recent analyses at the national and regional levels. The main advantage of the WRI methodology is that the estimates are globally consistent.

GTOS tiers

Although there are vast collections of in situ data, the measurements have been acquired for a variety of reasons unrelated to TCO objectives. In fulfilling TCO in situ data needs, data should be considered for its use in (i) defining stocks and fluxes of carbon components and (ii) providing process-level understanding in synthesize activities described above. The Global Hierarchical Observing Strategy (GHOST) provides a conceptual framework to bring together carbon flux measurements that can be used in TCO.

The core of the GHOST strategy is a permanent hierarchical observing system for the world's key managed and natural ecosystems. This scheme is based on a data sampling strategy involving large-scale studies of the Earth's major environmental gradients; agricultural and ecological research centres; field stations; and a gridded series of some 10,000 sampling sites. GHOST consists of five tiers (Table 2). For tiers 1 to 3, much of the required infrastructure is already in place but there is a need to upgrade some of the existing ecosystem monitoring sites and fill significant gaps in spatial coverage, especially in developing countries. Table 3 lists categories of measurements to be made at tier 1-3 stations. Ideally, existing stations should be strongly encouraged to at least report identified priority 1 and 2 measurements (Table 3) which require minimal resources, labour and time. Tiers 3 and 4 require increasing amounts of national resources but should be reported if possible (also see Scurlock et al., 1999).

Synthesis of carbon cycle components

Several major synthesis studies have compiled carbon cycle component measurements. These studies provide the process-level understanding that can be used for the synthesis activities described above. The status of these in situ data compilations is summarized by Olson and Cook in Appendix 5.

Table 2. The Global Hierarchical Observing Strategy (GHOST, www.fao.org/gtos/gtospub/pub13.html)

Tier	~ sample numbers	Sample area (km2)	Example variables
1. Large area experiments and gradient studies	10	1,000	Boundary layer gas exchange, biome shifts
2. Long-term research centres	100	10	Energy, water, carbon and nutrient cycling
3. Field stations	1,000	1	Crop yield, ecosystem productivity, land use
4. Periodic, unstaffed sample sites	10,000	0.01-0.1	Land cover, soil state
5. Frequent low-resolution remote sensing	100,000,000	0.001-1	Leaf area dynamics, land cover

Table 3. Measurement priorities for desirable carbon cycle attributes to be collected at study sites, network sites and ground stations

Priority	Parameters to be measured
1. Basic site characterization	· latitude/longitude/elevation using Global Positioning System (GPS) · details and position of nearest weather station with long-term monthly precipitation and temperature considered representative of the study site · vegetation type at site and surrounding vegetation · slope, surrounding topography · basic description of soil type and texture
2. A little more resources needed	· estimate of total standing crop, preferably with separated above-ground live and dead matter (report the peak value if this varies greatly during the year) · soil coring to determine soil layers, structure and texture of different layers, total soil carbon · simple components of NPP/carbon cycling, e.g. monthly tree litter fall for 24 months
3. Partial characterization of carbon cycling	· non-woody/foliar biomass increment, preferably estimated monthly · woody biomass increment, generally estimated yearly · below-ground standing crop, preferably with above-ground live and dead matter separated
4. More difficult and complete measurements	· carbon flux components (eddy correlation, chambers,etc.) · estimates of coarse and fine below-ground increment/productivity · other components of NPP, e.g. herbivory, fire losses, mortality, large woody litter fall (over several years), volatile organic carbon, root exudates · other ecosystem/stand characterization, e.g. stand age, tree density, age classes, successional status, management history

Ideally, all TCO/IGBP/GCTE measurement stations should be strongly encouraged to report at least at 1 and 2 levels which require minimal resources, labour and time. Levels 3 and 4 require increasing amounts of resources but should be reported if possible (see also Scurlock et al., 1999).

Networks

Existing observing networks provide a set of sites at which either controls on carbon dynamics (i.e. productivity and respiration) are studied or measurements of a single/few components are made. The most valuable networks are those that obtain all the critical measurements necessary to understand trace gas fluxes and related key ecosystem processes, thus permitting the development and testing of models as well as validation of model-derived output products. These networks have developed regionally and are associated with FLUXNET (www.daac.ornl.gov/FLUXNET/).

They also include sites established through intensive field campaign studies such as the International Satellite Land Surface Climatology Project (ISLSCP) First Field Experiment (FIFE); the Boreal Ecosystems - Atmosphere Study (BOREAS); the Southern African Regional Science Initiative (SAFARI); and the Large Scale Biosphere-Atmosphere Experiment in Amazonia (LBA). Although the latter networks provide a wealth of in situ and remote sensing data, the campaigns are typically short (one to several years in duration). Other networks focus on specific aspects of carbon cycle.

The International Long Term Ecological Research (ILTER; www.ilternet.edu/) network is an example of a system for comprehensive studies of ecosystem productivity. In contrast, the Biosphere-Atmosphere Stable Isotope Network (BASIN) network focuses on measurements of stable isotope concentrations. Other longer-term networks are summarized by Olson and Cook (Appendix 5).

The overall collection of sites at present is not fully representative of either ecosystems observed or climatic zones. There is therefore a need for networks to increase the number of sites that measure and contribute key variables significant to bioclimatic, temporal, and spatial dimensions of the carbon dynamics variability. Except for the Amazon and some Asian networks, the emphasis has so far been mostly on temperate and boreal forest ecosystems. Extensive but under-represented ecosystems are particularly agricultural areas, savannas, and wetlands. In addition, several other biome gaps are evident and include:

• areas undergoing rapid shrub/forest encroachment are under represented but are important, as they are a likely growing sink of carbon;
• wetlands which are almost totally absent from current networks but are critical for estimating methane flux;
• the (fairly productive) dry tropical forests which are under-represented in existing networks;
• woodland and poorly vegetated/desert environments which appear to be poorly represented. Although these latter systems are not particularly important for present and short-term carbon dynamics, they are often the most heavily utilized/disturbed systems and may therefore be of importance over longer time scales;
• tundra, which is important due to the relatively large amounts of carbon stored in its soils that could be released into the atmosphere due to increases in temperature.

Site inventories

Terrestrial Ecosystem Monitoring Sites (TEMS, www.fao.org/gtos/tems) database is an international directory of sites and networks that carry out long-term terrestrial monitoring and research activities. The TEMS database provides information on the "who, what, and where" in long-term terrestrial monitoring that can be useful to both the scientific community and policy-makers. In addition to this database, a useful list of sites and networks was compiled in 1998 as part of the Land Surface Processes Interactions Mission (LSPIM, http://imk-isys.fzk.de/isys-public/datasets/geo/Land-use-exp/sites-tables-1.htm/and www.estec.esa.nl/vrwww/LSPIM/).

The review carried out at the workshop provides only a partial assessment of the degree to which existing site networks fulfil the needs of TCO. A more definitive assessment needs to be conducted to finalize gaps in coverage, as a step in facilitating future improvements of the networks.

Gap, issues and considerations

The main gaps and issues regarding terrestrial measurements are:

1. Many estimates of forest stocks and productivity data at global to sub-national levels from point samples are unavailable or inadequate. Issues include:

• different inventory methods used with varying levels of accuracy;
• accurate data on stock changes (due to harvest, fires, other disturbances) not available, particularly at sub-national levels;
• limited access to the original plot-level measurements;
• exact coordinates of plot data not generally released due to confidentiality concerns, and this greatly limits the value of the data for carbon modelling and budget studies;
• not all the biomass is measured (usually only above-ground components with the focus on commercial tree species and merchantable volumes, coarse woody debris and litter production is not usually included), so how this data can be used for total flux estimates is not clear;
• inventories which usually have a strong statistical design to estimate volumes and growth for large areas of forest do not provide sufficient quantitative information at the local level.

2. Measurements of soil carbon have high level of uncertainty due to the natural heterogeneity of soils at local scales and at different soil depths. Soil profile measurements are available at national or global levels (in SOTER) and a method for the spatial distribution of soil carbon has been developed by IGBP (see Nachtergaele, page 91 and Global Soils Data Task, 2000). However, the quality of soil carbon inventories is limited by the lack of soil carbon measurements and soil depth data.

3. Below-ground coarse and fine root biomass and root turnover rates are poorly known, especially their distribution over large areas. The measurements are scarce, especially for tropical deciduous and boreal deciduous needleleaf forests such as larch, and the measurement methods are inconsistent.

4. Flux tower measurements need to be carefully processed to ensure their usefulness. Issues include (FAO, 2002b) representation of night time fluxes, approaches to filling measurement gaps (including those of complex terrains), footprint characterization, quantification of soil respiration, closure of site energy balance, site biometric measurements and the possibility of high precision estimates of CO₂ concentrations. Given the central role that flux towers play in the TCO in situ measurement structure, such critical problems need ongoing close attention.

5. Water fluxes, nitrogen and phosphorus need closer and more systematic analysis. As they are needed for soil databases and process models incorporating these aspects.

6. In models there is a need for incorporating genetic and bacterial controls on soil carbon flux. Additional measurements are needed at targeted sites with improved methods of upscaling the data.

7. A more definitive assessment of the gaps in the coverage of in situ observations is needed so that the location of future observation sites and networks can be identified.

8. General guidelines are needed on sampling design, measurements, and upscaling to a minimum area of 3 × 3 km that meet the needs of TCO. Examples from similar programmes are available (e.g. the BigFoot field guide; Campbell et al., 1999; Gower et al., 1999). When selecting ad ditional sites, those with flux towers should be given highest priority since they provide the most important basic infrastructure and measurements around which to develop additional measurement sets.

9. The role of biodiversity in ecosystem functioning is a subject of current research (e.g. www.abdn.ac.uk/ecosystem/intro.htm). It appears that dominant plant species have a strong role on ecosystem processes, including carbon uptake. Thus the distribution of the dominant species will be a valuable input to ecosystem models simulating carbon uptake. Such information is available in national forest inventories but is not for other types of ecosystem.

10. An important cross-cutting issue is land-cover classification. What is an appropriate land cover classification scheme within which other schemes would nest? Many countries have developed their own national and regional land cover classifications that are not optimal for carbon modelling. Continuous field classification schemes (e.g. DeFries et al., 2000) may be an effective way to deal with this issue in some models.

11. Existing in situ data should be registered in a system/data search engine such as Mercury (http://mercury.ornl.gov/; see http://mercury. ornl.gov/ornldaac/). Such a registration process would involve the identification of important data sets and obtaining the owners' agreement to make data available for TCO purposes (Section 6). Although the Internet and the expanding regional campaigns have both contributed to increased data sharing, accessing existing in situ data with adequate documentation (Cook et al., 2001) remains a challenge. GTOS has developed a set of operating principles to cover the sharing of data that should be considered for use by TCO (GTOS, 1998). One of the key ways to encourage data access is to provide recognition to the data provider, including options to "publish" data in peer-reviewed, citable journals, i.e. Ecological Society of America Data Archives (Olson et al., 1999).

Priority needs and recommendations

The following recommendations were made for terrestrial measurements:

1. Use TEMS and similar databases, develop inventories of the sites within the main networks, analyse the site coverage within a multi-dimensional environmental framework, and produce guidelines for locating future sites.

2. Conduct a thorough assessment of TCO relevant data available from sites within the major existing networks. From the evaluation, identify gaps in coverage that need to be filled.

3. Examine ways to produce (data fusion/assimilation) data products needed by TCO which reduce estimate errors and incorporate methods for filling gaps.

4. Adopt or develop guidelines for sampling design, measurements, and scaling from a point to a minimum area of 3 × 3 km.

5. Register existing in situ data and sources useful for TCO in a system with search capabilities and acquire or establish links to key data.

Atmospheric measurements

Status

An overview of the global measurement networks for CO₂ is provided at www.cmdl.noaa.gov/ccgg/ globalview/index.html. GLOBALVIEW-CO₂ presently involves 17 organizations from 13 countries. An internally consistent, 21-year global time series has been compiled so far. Actual data from most of the individual laboratories are available via the World Meteorological Organization (WMO) World Data Centre for Greenhouse Gases (WDCGG), Tokyo (Japan) and the Carbon Dioxide Information Analysis Center (CDIAC), Oak Ridge (US). Most of the data comes from flask samples collected from marine boundary layer air, typically at weekly intervals. Many laboratories in addition to CO₂ also analyse the flask air samples for significant greenhouse gas species (e.g. CH₄, N₂O) or provide information on carbon exchange (isotopes, CO, etc.).

Issues and considerations

The current main atmospheric observation issues are addressed in the previous TCO reports (FAO, 2002a and FAO, 2002b). The fundamental concept under-pinning the monitoring of long-lived trace gases is the ability of atmospheric mixing to provide large-scale integration of surface fluxes, and to accurately reflect the sum of global exchanges. The key issues address the challenge of improving the specificity of atmospheric methods in identifying the globally significant processes. In this report, the main aspects are highlighted and the areas that need to be futher developed identified.

Long-term continuity

Long-term continuity of observations is necessary for a variety of reasons, both scientific and policy. An important factor is the interannual and multi-decadal variability in global CO₂ growth rates. These variations are significant and are related to the variability in climate per se rather than direct anthropogenic forcing. Consequently, climate variability underpins a requirement for globally distributed atmospheric sampling over decadal or longer time scales.

The current sampling system is operated through research funds provided by many agencies in over a dozen countries. Individual sites are thus subject to uncertainty or elimination due to unstable funding. Issues of methodological consistency and calibration make the use of the data difficult at the global scale. Furthermore, in a research environment it often becomes difficult to justify the continued measurement of observations that have become routine. Adequate resources for quality assurance, data archiving and distribution are more reliable as they are provided by institutes with specific responsibilities for maintaining long-term records, however research links may often be neglected. The challenge is to maintain active interactions between providers and users of the data.

TCO needs to contribute to the long-term continuity, quality, and stability of a global atmospheric sampling programme. Technically, this could be achieved by maintaining and expanding the current networks, while seeking ongoing support and funding through agencies with observational mandates. Methodological developments (instrumentation and data assimilation in models) will also be an important addition to the configuration of future systematic, ongoing measurements. Long-term commitments by national institutes continue to be a crucial factor in implementing recommendations from the atmospheric observation community.

Calibration and accuracy

The WMO expert group has recommended a target precision of 0.05 ppm, 0.10 ppm for the Southern and Northern Hemispheres, respectively to allow the merging of CO₂ data from different sampling networks (WMO, 1993, 1981). CO₂ measurements are used to determine CO₂ differences in space or time that in turn provide air-surface fluxes of CO₂. Generally, spatial or short-term differences can be monitored more precisely than individual measurements as they can be linked to a primary standard.

Three round-robin laboratory calibrations, involving circulations of high-pressure cylinders of air through ~20 laboratories have been conducted since 1990 by the National Oceanic and Atmospheric Administration (NOAA) under the auspices of WMO. Although there have been improvements, at best only 50 percent of cylinder air measurements achieve the target precision for merging data, and there is little consistency from one circulation to the next.

Stricter adherence to WMO guidelines, by maintaining regular calibration of high-pressure cylinder air in the WMO-endorsed central calibration laboratory, would make a significant impact. However, it is now clear that this is not a complete answer. For example, within advanced laboratories there are examples of systematic differences between methods greater than the target precisions for inter-laboratory data merging; in less advanced laboratories the effectiveness and cost of maintaining and propagating the link to primary standards has proven to be a limiting factor.

Recent WMO (mainly for CO₂) and International Atomic Energy Agency (mainly for d¹³C) expert forums have unanimously endorsed a global intercomparison strategy for greenhouse gas measurements called GLOBALHUBS (Global Quality Control for Long-Lived Trace Gas Measurements, Francey et al., 2001). The first priority of GLOBALHUBS is to establish more frequent, comprehensive, and transparent comparisons between measurement laboratories, and to incorporate this information into GLOBALVIEW (Cooperative Atmospheric Data Integration Project for CO₂).

Four globally distributed HUB laboratories are proposed which will maintain and/or develop suites of standards (linked to primary standards), and maintain intensive HUB comparisons using a variety of techniques, such as a highly precise low flow rate CO₂ system. The GLOBALHUBS programme requires initial set-up funding to obtain the needed hardware, to maintain or strengthen the links to primary standards, and to fund significant communications and processing development efforts.

In summary, the calibration issue consists of the identification and removal of systematic measurement errors within and across sampling networks allowing data to be accurately merged.

Expansion of the observing network

Estimation of carbon sources and sinks at continental or ocean-basin scales by atmospheric transport inversions is presently strongly data-limited (FAO, 2002a). There are many opportunities to add substantial data constraints to the mass-balance of atmospheric CO₂ over regional scales, including:

Continental lower troposphere profiling by aircraft

A recognized way to improve the scale linkage between fixed surface sites and the atmospheric transport parameterizations is by improving vertical profiling of trace gases through the boundary layer and above, with flasks or continuous analysers. The vertical profile of CO₂ concentration from the surface to the middle troposphere provides the connection between surface processes and larger scale (ultimately global) atmospheric CO₂ concentrations. Airborne sampling provides valuable information on carbon fluxes at the intermediate (regional to continental) scales. This helps bridge the gap between the flux estimates at local/site (eddy correlation) and global (inferred from inversion models) scales. The feasibility and effectiveness of obtaining such measurements from aircraft platforms has been demonstrated in long-term monitoring programmes (e.g. Australian, Japanese and French national reports in WMO, 1997), and in regional campaigns including the CO₂ Budget and Rectification Airborne study (COBRA), Terrestrial Carbon Observing System - Siberia (TCOS-Siberia) and LBA (e.g. Nakazawa et al., 1997 and Gerbig et al., 2000).

In the context of TCO, the main issues are related to the broader implementation involving both instrumentation and feasible sampling programmes. From an operational perspective, accurate measurements are difficult and expensive to sustain. However, some programmes are establishing similar capabilities. For example, commercial aircraft flights are being used for atmospheric measurements at ~11 km altitude. However, this programme does not give useful profile data since most airports are in industrialized/urban environments.

Recently, a pilot project has been underway in France to develop a prototype CO₂ and carbon monoxide (CO) measurement system for installation on ~5 regional aircraft in association with the MOZIAC programme (Measurement of OZone on Airbus In-service AirCraft, http://sedac.ciesin.org/ozone/WMO/france.html/). Although aircraft profiling continues to be a vital component of regional campaigns and research programmes, its larger-scale operational implementation is developing more slowly. It should also be noted that aircraft observations can only be made under favourable flying conditions and thus suffer from 'fair weather bias'. Table 4 indicates current options for improvements.

Direct CO₂ concentration measurements using tall towers

CO₂ concentrations are being measured at an increasing number of tall (a few hundred metres) towers. Such measurements can provide, among other benefits: continuous information on the carbon budget for the lowest part of the troposphere; continuous measurements and thus a long-term context into which the airborne measurements can be situated; measurements under less-than-ideal flying conditions, providing an estimate of the 'fair-weather bias' in aircraft measurements (Gerbig et al., 2000).

The practical feasibility of measurements at tall towers has been demonstrated (e.g. Bakwin et al., 1998) and more towers are being built. Tower measurements are strongly affected by transport in horizontal advection which integrates the signal caused by the upwind sources/sinks. Thus, methods are needed to account for surface inhomogeneity and transport in coupling observations with a transport model, driven by actual winds. These may also use the Lagrangian experimental framework (i.e. observation platform sampling nominally the same air parcel).

Calibration of flux tower measurements to obtain accurate CO₂ concentrations

There are ~140 flux towers currently registered around the world. If these measurement sites could be enhanced to measure CO₂ concentrations with precision and calibration consistent with global monitoring, the network density would be significantly increased, particularly in continental areas that are not well represented in the current networks (FAO, 2002b).

This would have the advantage of adding important measurements at a low cost; the disadvantage is that reliability of the extrapolated vertical CO₂ profile depends to some degree on the local, near-surface meteorological conditions and the biome activity which are not well represented in current regional or global models. However, continuous monitoring offers considerable opportunity to select and/or average data, and is therefore an important option for increasing the density of the atmospheric observation network in the terrestrial environment.

Increased frequency of measurements in the global monitoring network

The prevalence of flask sampling at background sites means that the site atmosphere is usually sampled for a few minutes for each sample. Normally 2-8 samples are used to calculate a monthly average. Recent breakthroughs in CO₂ measurement technology (e.g. Da Costa and Steele, 1997 and Steele, private communication) have achieved effectively unattended, continuous, high-precision (~0.01 ppm) CO₂ monitoring at a remote location for periods of > 5 months. Such instruments are now under development for possible wider usage.

Increased range of gas species measured

The source of the carbon found in the atmosphere can be identified through analysis. Stable carbon isotopes and O₂/N₂ mainly indicate terrestrial (rather than oceanic) exchange; ¹⁴C in CO₂ is used to identify carbon from the combustion of fossil fuels; and CO, Methane (CH₄) and Hydrogen (H₂) can also be exploited as a proxy for combustion in certain areas. There is a potential to use ¹⁸O in CO₂ and VOCs with a range of lifetimes to better localize and differentiate the large-scale influences of photosynthetic vs. respiration fluxes in terrestrial biomes.

Detailed knowledge of combustion sources is necessary to provide a constraint on process-based models of surface CO₂ exchange from atmospheric data. These emissions are currently reported on a national and annual basis. The spatial distribution needed in inversion models has been achieved by using population as proxies within each country. The needed temporal variation of the combustion sources has so far been determined only for the United States. However, reliable regional mass-balance constraints will require more detailed emission estimates by location (city) and date.

Detailed characterization of fossil and anthropogenic emissions, are required for multiple species (e.g. source isotopic or stoichiometric ratios for fuels and for terrestrial ecosystem sites) if these are to be used as additional inversion constraints. Specific information is required on the timing and location of the emissions, including their measurement uncertainties and, where gridded/generalized data are provided, the horizontal resolution and possible spatial and temporal covariances of the uncertainties (e.g. by reporting both detailed systematic and random uncertainties of the source estimates).

More spatially and temporally detailed emission data products should also be prepared based on statistical reports within countries. Such products may be possible by working with international agencies to encourage communication of cross-disciplinary requirements and limitations between sectors of the carbon cycle community; change the international reporting requirements; solicit access to more detailed data already collected but not processed, and/or encourage new and better measurements.

Network optimization studies

Additional studies on network optimization must be undertaken to quantify the reduction in uncertainty for proposed sampling sites and strategies. Such studies should address alternative sampling methods, particularly airborne sampling, continuous monitoring, and remotely sensed concentration measurements. This should result in a strategy for evolving the observation networks in relation to the evolution of measurement techniques and data assimilation models. Some studies have been reported in the literature, and the TransCom project is now engaged in such an effort through an intercomparison of model results.

Table 4. The status of the different options for improving atmospheric trace gas monitoring

Issue/Option		Category
Calibration and accuracy		I, S
Global network calibration		I, S
Calibration of flux tower CO2 concentration measurements leading to network expansion		I, S
Expansion of the observing network
Continental lower troposphere profiling by aircraft		I, S
Tall towers
Installation of towers in continental biomes		I, S, L
Continuous measurements		I, S, L
Network design studies		R, S, L
Multitracer approach
CO, 13C, 14C, 18O, O2/N2, etc.		R, S, L
Long-term continuity of observations
Consolidation of satellite sensors for CO2		R
Assimilation of satellite measurements of CO2		R
Other greenhouse gases		R
Requirements from "bottom-up" community
Horizontal transport at the surface, particularly		I, R, S
	* carbon trade data products	I, R, S
	* carbon transport by rivers	I, R, S
	* erosion/burial	I, R, S
CO2 pre-cursor transport (CO, VOC, CH4, etc.)		R
Improved anthropogenic emissions data		R, S
Improved prior surface flux estimates, including uncertainties and their covariances		R
Requirements from atmospheric modelling community
Development of data assimilation techniques (similar to weather forecast systems)		R
Improvements in atmospheric transport modelling.		R

I= implementation issue, R= research issue; S= short-term, L= long-term

Recommendations for atmospheric measurements:

1. Existing national programmes should be vigorously promoted by TCO if improvements in the current understanding of the global carbon cycle are to be made.

2. Methods should be developed to identify and remove systematic errors across atmospheric composition measurement networks/laboratories and for on-going monitoring of network intercomparability (e.g. GLOBALHUBS). This would provide improved links to primary standards. Centralized international coordination is preferable. In the absence of centralized funding, community efforts to focus and informally coordinate national/regional programmes can provide significant improvement over the status quo.

3. Current global atmospheric sampling networks should be extended, both spatially (into continental regions using towers and aircraft) and temporally (with continuous analysers). New extensions should be guided by network design studies to ensure maximum effectiveness. Thus, network design studies are therefore a priority and should include an assessment of the operational complementarity of these networks with future satellite sensing of CO₂ missions.

4. TCO should continue to promote increased communication between terrestrial and atmospheric communities (both measurement and modelling). In particular this should occur at the planning stages of major experiments, with the view of providing complementary data and methods that link atmospheric composition changes to appropriately parameterized surface processes.

5. Top-down methods of chemical identification of the major sources of atmospheric composition change need to be improved. TCO should support the multi-species (e.g. isotopes, O₂/N₂, VOCs etc.) monitoring of the global atmosphere, the development of suitable chemical transport models, and the improved characterization of emission ratios of significant exchange processes (e.g. through campaign/process studies).

6. Improved methods are needed (both modelling and measurement) for using with integrated chemical signatures in atmospheric composition to identify the major sources of change.