4 Challenges for Carbon Cycle Observations

This section discusses the approach to developing a capability for systematic, long-term observations of the terrestrial and atmospheric components of the global carbon cycle. The guiding considerations are: establishing long-term goal and vision; building on existing capabilities; ensuring continuation of existing essential activities, then seeking enhancements; and ensuring close interactions between research and observation specialists. In developing the list of key challenges, the following assumptions have been made:

1 Current satellite missions will achieve or exceed their planned lifetime

2 Presently approved satellite missions will be successfully put into operation

3 Co-operation between research and observation programs and organisations is an effective way to establish systematic, long-term observing capabilities

4 Existing efforts aimed at improving the spatial resolution of meteorological and hydrological information will continue, especially with respect to precipitation and near-surface temperature fields with a high spatial resolution (needed by ecosystem models). This includes liquid and solid precipitation (including satellite missions such as TRMM, Aqua) and numerical weather prediction (NWP) or NWP reanalysis projects. These issues are addressed in satellite observation programs for weather and climate forecasting purposes (by WMO and national meteorological agencies) and by research programs such as WCRP.

5 Only ongoing, repeated observation requirements are of concern to this report. Thus, baseline data products (e.g., soil type maps) are not considered here (refer to Cihlar, Denning and Gosz (2000)).

The strategy for the carbon cycle observation is based on a combination of methods. To understand the importance of the various observations specified in sections 4.2 and 4.3, the observation approach is briefly described below.

4.1 The Concept

The policy community requires information on spatial and temporal patterns of terrestrial CO₂ flux at high resolution and over large areas (section 2.). These requirements imply the use of ecosystem process models linked to spatial measurements that are available everywhere, such as from satellites (section 6.1). Satellite data can also provide up-to-date information frequently, in relation to the rate of change of the variables of interest. In this 'bottom-up' strategy, local land processes are scaled-up in space and time using satellite imagery and other spatial data. The primary limitation of this approach is the difficulty of conclusively establishing the accuracy and reliability of the scaled-up estimates. In addition, the success of the approach depends on the availability of reliable models that represent all the important processes affecting CO₂ exchange with the atmosphere, including the impact of various land use measures. Such models are not yet available for all processes, although inventory- based methods and conversion tables currently provide an acceptable approach (IPCC, 1996). - An alternative and complementary method is to analyse the carbon budget of the atmosphere from a mass balance point of view ('top down'). Such an analysis is predicated on the availability of atmospheric concentration data and other inputs, and it can be carried out in several spatial and temporal domains. Both approaches have been used in various studies.

Used synergistically, the 'top down' and 'bottom up' approaches take advantage of their strengths to compensate for their respective weaknesses (Cihlar, Denning and Gosz, 2000). This may be achieved through atmospheric inversion methods (e.g., Ciais et al., 1995) or through a multiple constraint approach. In the latter, the various data sources are employed to constrain the process parameters in a biosphere model, so that the model predictions are consistent with all available observations (Figure 1). The essence of the approach is to constrain the model parameters to optimal values using inversion theory, and thus to infer the complete space-time distribution of carbon stores and fluxes. In practice, the model predicts the observed variables at locations where measurements (surface- or satellite-based) are available, and then finds the parameter values that minimise the overall difference with the measurements. Although more complex, the multiple-constraint approach offers the possibility of employing multiple types of data with very different spatial, temporal and process resolutions. The predictions of a multiple-constraint approach are subject to verification by confronting the model with a wide range of data sources representing various spatial scales. Failure to accommodate all data streams simultaneously with a common parameter set enables finding model errors, thus preserving the scientific integrity of the approach. It should be noted that the availability of diverse observations is essential to avoid answers that meet the acceptance criteria but may be incorrect due to insufficient data.

After combining the bottom up and top down techniques, the spatial distribution of carbon sources and sinks is produced with high spatial and temporal resolutions, and of the best quality possible with the available observations and understanding of the carbon cycle. The spatial and temporal resolutions will then be constrained by the resolutions of the input satellite (and other) data. This approach is also fully compatible with reporting needs for small areas (such as may be required for the Kyoto Protocol), although additional data may be necessary to fully meet the reporting requirements depending on the case.

4.2 The Long-Term Continuity and Consistency Challenge

This section describes the most important and urgent issues that IGOS Partners face in establishing systematic, long-term observations of the terrestrial and atmospheric components of the global carbon cycle. The discussion presents only the key observation requirements that are needed on a sustained, long-term basis; a complete list of observation requirements is available elsewhere (Cihlar, Denning and Gosz, 2000). Following a brief discussion of each item below, the key issues are identified. The 'issues' represent items that need to be addressed now in order to ensure continuity of existing observations and output products, or they are essential improvements that must be made to balance the effectiveness of the individual observing system elements. In all cases, the observing technology exists, and in most it is deployed at least partly.

This section contains only the specific areas that need attention in implementing an initial observing system. For a more complete description of the initial system, refer to section 6. supported by section 3.2 and Appendix 2.

A. Terrestrial

Information about the land characteristics is required to determine the spatial and temporal distribution of terrestrial carbon sources and sinks. This information is needed at a range of spatial and temporal scales, and is used to run biospheric models and to determine the accuracy of the model outputs. The major types of information needed are described in section 4.2.1 - 4.2.7.

4.2.1 Land Cover and Land Use

A primary characterisation of land cover, and monitoring its changes, is fundamental for carbon cycle observation and assessment, but also to nearly every aspect of land management. Likewise, observations of land use and land use change are fundamental to an understanding of the human impacts on the biosphere. Because of the need to span a range of spatial and temporal scales, land cover and use are ideally suited to satellite-based observation with various sensors. Many years of experience have clearly demonstrated the capability of data from moderate- and fine- resolution earth observation satellites to be classified into a wide range of products that have been highly valuable in providing a synoptic view of the earth, leading toward more sustainable use of earth resources. Global scale land cover classifications have been produced using AVHRR data (available since 1983).

Data from ATSR/ERS-2 (1996) and VEGETATION (1998) have been used successfully to map land cover at continental scales and detailed plans are in place for their use globally. AVHRR data have also been employed to estimate vegetation cover characteristics which are important to full carbon accounting, including leaf type (broadleaf vs. needleleaf), leaf longevity (evergreen vs. deciduous) and canopy cover. The improved data now available from MODIS, MISR, MERIS and AATSR are expected to yield improvements in classification accuracy, detail and characterisation. In the longer term, the planned NPP, ADEOS-II/GLI, GCOM-B1/SGLI and NPOESS missions will provide follow-on medium resolution calibrated optical data. The spatial resolution will improve from the recent ~1000 m to ~ 300 m but the continuation of such spatial resolution capability is not ensured, nor has its necessity been conclusively established (see section 4.3.1/land cover). In the microwave domain, satellite SARs on Japanese (JERS, ALOS), ESA (ERS) and Canadian (Radarsat) platforms provide land cover information in perennially clouded areas.

At regional to landscape scales, Landsat (since 1973), SPOT (1986) and similar sensors provide land cover with the necessary spatial detail. The spatial coverage of high resolution mapping applications is rapidly growing, and the production of continental and global maps is envisaged to be generated in the near future (GOFC Design Team, 1999). Regarding classification schemes, an agreement has been reached on the highest level classes of land classification schemes. Initiatives such as the FAO's Land Cover Classification System offer a hierarchical approach which allows easy upscaling of detailed land cover classifications from national to global. Similarly, a strategy has evolved to produce intermediate products that can be easily converted into a land cover classification scheme by a particular user (GOFC Design Team, 1999). Much experience has also been gained on the use of coarse resolution land cover classifications for stratification supporting statistical samples of high resolution images subsequently used to quantify rates of change.

Land use is a critical carbon cycle parameter. The knowledge of present and historical (decades to centuries) land use is essential to properly represent carbon exchanges in models. Present land use can be obtained by combining satellite-derived land cover and in situ observations such as statistical reports. Ancillary evidence from satellite images can provide supporting evidence as well; for example, the appearance of logging roads provides a strong indication of particular forms of land use. FAO has estimated that given sufficient resources, a global land use map at the 1:1million scale could be produced in 3-5 years. Historical land use may be derived from existing land use records (e.g., Ramankutty and Foley, 1992). The process of convergence on land cover products is advanced but not completed, and while the observations required for land use products are well understood, the classification approach is less advanced.

The continuity and consistency issues are:

1 [sat]^[7] Continuity of calibrated, fine resolution optical data from both fixed-view (Landsat type) and pointable (SPOT-type) sensors needs to be ensured. Similarly, continuity of complementary SAR data needs to be assured. These observations should be accompanied by a consensus strategy for global satellite data acquisition at high resolution (see also Rosenqvist et al., 1999).

2 [sat] Institutional arrangements need to be made for product generation and quality control, possibly through the GOFC project, and the development of highly automated methods and products for land cover classification should be encouraged.

3 [sat] Archived satellite data (both coarse and fine resolution, optical and microwave) are a very important and often unique source of information on land cover as well as other attributes (fire, seasonal growth cycle). In many cases, these data have not been processed, or have been processed using algorithms that are presently obsolete. Therefore, reprocessing of satellite data to prepare time-stamped or time series products should be systematically addressed. In the context of the Kyoto Protocol, the state of land cover in 1990 and changes since that time has special significance.

4.2.2 Above and Below-Ground Biomass

Estimates of above- and below- ground biomass provide fundamental information on the size and changes of the terrestrial carbon pool as land use and associated land management practices change. Observations of biomass components have a rich history associated with human needs for food, fibre, lumber, and other biomass products. However, there are large gaps in these data in terms of (i) inclusion of above and below ground components, (ii) spatial and temporal consistency, and (iii) completeness in both spatial and temporal dimensions. Carbon cycle science and consequent policy decisions now depend on partial observations from research studies, from inventories focused on commercial interests such as forest inventory or crop yield surveys, and broader surveys or compilations, e.g. country-level statistics assembled by FAO.

Most of the detailed in situ data are not readily available or are only available as highly aggregated summaries. National forest inventories, available for recent decades in many temperate and boreal countries, provide a potentially rich data source but their use requires careful analysis and interpretation. At the present, remote sensing provides high resolution global coverage of land cover and cover changes (section 4.2.1) which are also relevant to the estimation of above ground biomass, but there is no satellite-based capability to estimate biomass directly. Future satellite measurements, e.g. from lidar (ESSP VCL, ICESat GLAS), will provide information on forest height and structure, thus allowing detailed and robust biomass estimates globally. However, the below ground component cannot be obtained from satellite platforms; it requires an increased density of in situ observations and improved scaling algorithms (section 4.3.1).

The existing inventory data, though limited in scope and completeness, have an important role to play. In case of forests, with careful interpretation these inventories can provide estimates of the total forest sink or source, rates of deforestation and regrowth, and losses to disturbance and harvesting. They yield direct estimates of carbon changes associated with forest area or age structure but only partial information on effects of growth rate due to climate change, N deposition etc. Since many countries have conducted forest inventories since the 1930s or even earlier, the inventories provide an important link between current and past effects (see section 4.3.1).

Depending on the resolution of the Kyoto Protocol reporting requirements, there will likely be a need for repeated measures of biomass/carbon density with high degree of accuracy for small land parcels. Traditional forest survey methods may meet this need, but satellite sensors such as profiling lidars are expected to offer an important additional and consistent information.

Regarding below-ground biomass, the FAO/UNESCO soil map of the world (based on soil surveys carried out during 1960s; FAO, 1995) remains the only global inventory of soil information to date. Coarse resolution soil carbon density information has also been derived using this map^[8]. Several regional updates of the global map have been undertaken using the SOTER approach (FAO, 1993). These updates contain georeferenced analyzed soil profile information with quantitative soil characteristics, including soil carbon. Under preparation are SOTER products for Southern Africa and for Western Europe; the global update is expected to be completed by 2006, subject to resources being available for West Africa and Southeast Asia segments. In addition, ISRIC and FAO have compiled >4000 georeferenced soil profiles that include carbon data and have been used to provide the best estimates to use for soil properties including soil carbon for each mapping unit (Batjes et al, 1997). National holdings of georeferenced soil profile information are of variable quantity and quality, and some are difficult to access given stringent copyright rules (Nachtergaele, 2000).

The continuity and consistency issues are:

1 [sat] Above ground: ensure ongoing availability of canopy structure measurements from satellite sensors, beyond the planned ESSP VCL mission. The best current prospects are lidars and advanced SARs (see also section 4.3.1).

2 [ins]^[9] Below ground: increased density of in situ observations (i) by improving or adding observations within existing networks; (ii) by significantly expanding the soil profile databases available through SOTER and similar programs; and (iii) through more efficient use of national inventories, in combination with land cover derived from satellite data (refer also to section 4.3.1).

4.2.3 Seasonal Growth Cycle

Seasonal growth characteristics such as leaf area, growing season duration, and timing of growth (onset and senescence) provide strong constraints on carbon sequestration (Cramer and Field, 1999). Many ecological/carbon models require leaf area index (LAI) and, more recently, information on vertical and horizontal leaf distribution to account for different light use efficiencies of sunlit and shaded leaves. LAI is typically obtained from passive optical measurements, using empirical or model-based estimates calibrated by ground measurements. Satellite-based LAI products have been generated globally as well as for specific regions from AVHRR data, and they are planned to be produced globally by Terra. A co-ordinated LAI validation program has been initiated under CEOS WGCV. - Simple indices from passive remote sensing (e.g. NDVI), while giving reasonable first order estimates tend to saturate at LAI»4 depending on canopy type, structure, and leaf state. Adjustments have been developed but intelligent methods are required to extend the maximum estimable range of LAI. Multi-angular measurements from sensors such as MISR, POLDER and EPIC offer strong potential in this respect even though the latter two are limited by their spatial resolution (~10 km). Although further development needs to be undertaken (section 4.3.1), the evidence of the importance of multi-angle measurements (Knyazikhin et al., 1998) is sufficiently strong to justify the requirement for their continuity in support of LAI products.

Depending on the biome, the growing season is limited by temperature (temperate and boreal) or moisture (tropical). In addition, the onset of both green-up and senescence varies with species within plant assemblages. Satellite observations have shown the ability to detect green-up and senescence in ecosystems. While such observations are valuable, reliable determination of growth season duration, whose inter-annual variation is on the order of days, requires calibration and cross-comparison of methods from ground-based meteorology and space. For temperature-limited ecosystems, it has been shown (Frolking et al., 1999) that SAR measurements are a sensitive indicator of freeze-thaw transitions.

The continuity and consistency issues are:

LAI:

1 [sat] Ensuring continuity of moderate resolution optical sensor measurements: same as for land cover (section 4.2.1).

2 [sat] Ensuring continuity of multi-angular sensor measurements of the MISR type.

3 [sat] Ensuring agency commitments to generating global LAI products beyond the MODIS/Terra period.

GROWING SEASON DURATION:

1 [sat] Ensuring continuity of moderate resolution optical sensor measurements: same as for land cover (section 4.2.1)

2 [dev]^[10] Performing calibration and cross-comparison of products from ground-based meteorology and satellite-based, including sensitivity analysis of the space-based estimates over long periods and an assimilation strategy employing data from multiple satellites (AVHRR, VEGETATION, MODIS, ATSR).

4.2.4 Fires

In many regions of the world, fire causes the strongest disturbance of vegetation. World-wide information about fire is necessary to calculate net carbon sink, and fire may be a major cause of the large observed inter-annual variations in carbon emissions from ecosystems. Fire is also a very important factor influencing ecosystem succession and land use. Information about the fire timing, areal extent, intensity, and trace gas emissions has many societal implications beyond the carbon cycle (Ahern et al., 2000). Among these are health implications of major fire events; the effects on timber and range resources; and the risk to lives, valuable economic infrastructure, and private property.

The status and needs for global biomass burning were determined in a 1999 GOFC workshop (Ahern et al., 2000). Large fires in forests and grasslands, which are most important for the carbon cycle, can be detected using satellite-based thermal and optical sensors. Clouds prevent detection of a significant fraction of fires, but statistical corrections can be applied to obtain reliable estimates of the location and timing of fires. Recent work with SAR shows considerable promise for burnt area detection. The area burned in large fires can be mapped with good accuracy using satellite data, notably ATSR on ERS-1 and -2, VEGETATION on SPOT-4, and MODIS on Terra. Prototype burned area products from these sensors will be produced by ESA/ESRIN, JRC and NASA and the development of community consensus algorithms is underway. The continuity of optical data is assured through the NPP, ADEOS-II, GCOM and NPOESS missions. However, the planned sensors do not have the preferred spatial resolution (~200 m, Ahern et al., 2000). Further assessment of this spatial resolution requirement is needed (see also section 4.3.1/land cover). Smaller fires can be mapped, when needed, using fine resolution sensors such as Landsat and SPOT; these sensors also can also provide more accurate information on the spatial distribution of fires and for more accurate area estimates through double sampling approaches. Historical information on area burned is important to estimating secular changes in fire frequency. The most promising approach is through re-processing of the AVHRR archive (1982-2000).

A network of receiving and processing facilities, the World Fire Web led by the Joint Research Centre of the European Commission, is nearing completion^[11], and will provide global coverage of active fires as well as burned areas by the end of 2000. However, this is a demonstration project and there is no assurance that this capability will be maintained into the foreseeable future. An important global outlet for fire information is the Global Fire Monitoring Centre, operated for the United Nations by the University of Freiburg^[12]. It collects fire information world-wide and provides it in a consistent format, as well as providing critical analyses to help national and international agencies develop more effective approaches to fire management.

The continuity and consistency issues are:

1 [sat] Making commitments to the continuity of fire products generation (WFW, GOFC) and to reprocessing the archived AVHRR data to obtain a global fire history (as an input to estimating the current C fluxes).

2 [sat] Ensuring long-term operation of the World Fire Web project.

4.2.5 Solar Radiation

Global solar radiation (shortwave, SW) and its photosynthetically active radiation (PAR) component are major drivers of surface processes such as photosynthesis and evapotranspiration. Global long term monitoring is required for providing inputs to terrestrial photosynthesis models (Cramer and Field, 1999), and for a variety of agrometeorological applications. For carbon uptake modelling, daily PAR estimates are needed at a resolution of 50 km (minimum) to 10 km (preferred). Several methods have been developed to derive SW from satellite radiance measurements (Charlock and Alberta, 1996). Global data set of monthly averages of SW have produced for climate purposes using ERBE data by the NASA Langley Research Centre for the GEWEX Surface Radiation Budget Project (SRB^[13]); Version 2 (global, 1^o resolution, 3-hourly, 1983-95) will be completed in 2001. The products generated so far are the result of various research programs. Estimates of PAR averages at weekly and monthly time scales can be derived as a constant fraction of SW. The limitations of these products are coarse spatial and temporal resolution, and their availability for a limited period only (to 1995).

The continuity and consistency issues are:

1 [sat] Ensuring commitment to the development and routine production of daily to monthly SW products for the period beyond 1995 and ongoing for the future, at the highest feasible spatial resolution.

4.2.6 Surface-Atmosphere Fluxes

Ecosystem flux measurements are a critical element of a terrestrial carbon observing system. The data provide essential input to process studies, the development and testing of models, and to upscaling from sites to regions (Cihlar, Denning and Gosz, 2000). Fluxes of carbon, water and energy are continuously measured with sub-hourly time steps at 140 stations world-wide (30 of which have data for >3 years) encompassing a range of terrestrial ecosystems and climate. Data are collected in regional networks (CarboEuroflux - Europe, Ameriflux - North America, LBA - South America, Asiaflux - Japan, Thailand, Ozflux- Australia) and analyzed as synthesis products within the framework of FLUXNET. The current network design provides useful data for model and remote sensing products validation at the scale of 1km as well as insights on biome-specific responses to environmental factors and their temporal and spatial variability (Valentini et al., 2000).

The continuity and consistency issues are:

1 [ins] Maintaining the existing flux measurement programs for at least 10 years at a site. These measurements are essential to capture the seasonal and inter-annual variability.

2 [ins] Expanding the current network (i) in underrepresented regions (especially Africa and Asia), and (ii) in ecosystems undergoing major disturbances or highly dynamic responses (complex and highly disturbed landscapes; after fire regeneration, logging, degraded lands, grasslands, savannas; across gradients of succession, stand age, and land-use intensity).

3 [ins] Operating a selected set of long-term 'ideal' stations for monitoring carbon, water and energy fluxes on representative biomes where disturbances, and direct human impacts are minimal, to serve as anchor stations for understanding climate variability.

4 [ins] Improving international co-ordination with on-line data transfer to a centralised facility, responsible for data quality check/assurance, data assimilation, and synthesis products.

4.2.7 Ecosystem Productivity

Ecosystem productivity measures quantify carbon uptake by terrestrial ecosystems. Net primary productivity (NPP) is the net biomass increase through photosynthesis, while net ecosystem productivity (NEP) refers to net carbon exchange with the atmosphere after accounting for soil respiration and organic matter decomposition. Carbon remaining in the ecosystem is obtained as the difference between NEP and losses due to fires, insects, harvest and other disturbances. Together, these measures meet the needs of various clients, including those interested in climate change and in resource management (Cihlar, Denning and Gosz, 2000). Ecosystem productivity quantities are determined using satellite-derived products (section 4.2.1 to 4.2.5); soil and meteorological data bases; and biogeochemical models that mimic the ecosystem processes involving carbon uptake and transformations within the ecosystem as well as the exchange with the atmosphere. Surface-atmosphere fluxes (section 4.2.6) and associated site observations are essential for the development and validation of the ecosystem models. To date, regional or ecosystem productivity products using AVHRR data have been generated by various groups at global and regional levels, using both top down and/or bottom up approaches^[14]. Refinements of products and methodologies for both approaches are subject of intensive research, stimulated in part by the increased capabilities of the new satellite sensors. Core algorithms can be validated locally, but the behaviour of the models beyond the scale of eddy covariance towers is not well constrained a priori. Some measure of overall mass-balance constraint may be possible through airborne campaign sampling of trace gas concentrations in conjunction with mesoscale atmospheric transport modelling (Stephens et al, 2000).

The continuity and consistency issues are:

1 [sat] [ins] Ensuring commitments to providing the required data inputs (sections 4.2.1 - 4.2.6, 4.2.8) and output products.

2 [ins] [dev] Expanding the in situ observation networks and obtaining data to improve the quality of satellite-derived products and the performance of biogeochemical models.

B. Atmospheric

The trend in the global mean concentration measured for the last 40 years is one of the most basic observations characterising global change, and reflects the integral of all source and sink processes at the surface. Concurrent changes in CO, CH₄, O₂/N₂, and the stable isotopic ratios d¹³C and d¹⁸O in CO₂ as well as other trace gases provide important information about source/sink mechanisms. Nearly all of the these data are collected by flask sampling in remote areas to capture variations in clean, background air far from local sources and sinks. In the past decade, the networks of sampling sites maintained by various governments have reached a spatial coverage sufficient to make inferences about sources and sinks on continental or ocean-basin scales in addition to the global-scale inferences that have long been made on the basis of trends, e.g. at Mauna Loa and the South Pole. This is accomplished by tracer transport inversion using wind and climate data collected by the weather forecasting infrastructure and numerical models of the transport of trace gases^[15]. Inverse calculations from atmospheric data can provide valuable mass-balance constraints to the integral of surface exchanges estimated from "bottom-up" inventory and model or satellite-driven aggregation methods. This also requires accounting for the anthropogenic combustion source of CO₂, which is available from econometric inventories (e.g., Andres et al., 1996).

The current network provides only gross constraints on very large spatial scales. Network optimisation studies have shown that adding additional observations, especially over the continents, would reduce flux uncertainties substantially (Rayner et al, 1996; Gloor et al, 2000). Airborne sampling and continuous monitoring have been shown to be particularly effective to achieve this aim. Augmenting the existing network with only stations in the remote marine boundary layer would require many more additional stations to achieve a similar reduction in uncertainty.

4.2.8 Long-term Continuity

The current sampling system is operated by research funds from many agencies in over a dozen countries. Individual sites are thus subject to elimination due to unstable funding in many quarters, and issues of methodological consistency and interlaboratory calibration and standards make the use of the data difficult at the global scale. These difficulties are being addressed by an international voluntary effort to co-ordinate and ensure consistency among the many observing networks ( GLOBAL-VIEW-CO2, 2000^[16]). This program facilitates round-robin intercomparison of standard air among several laboratories, adjustments for calibration offsets among participating laboratories, filling and smoothing of the data record, and unified access to the data record across the many national programs.

1 [ins] The issue is commitment to ensure the long-term continuity and stability of the atmospheric sampling program.

4.2.9 Calibration and Accuracy

All measurements of atmospheric CO₂ are made relative to standards derived from the WMO primary standard linked to fundamental physical constants by high-precision manometric techniques maintained by the US National Oceanic and Atmospheric Administration. It is crucial to maintain these standards, generate high-quality secondary standards, and make them available to measurement laboratories in the many countries that participate in he observing network. This is expensive and difficult to achieve, and as long as this activity is funded by research funds there will be sacrifices made to support other worthwhile activities. Past difficulties in propagating the standard to many laboratories, due in part to high gas consumption requiring an extended hierarchy of standards and to difficulties in ensuring uniform quality control over many laboratories, have limited the overall effectiveness of the current networks. An immediate challenge therefore is to significantly improve the ability for all laboratories to link to the fundamental calibration scales. As part of its Global Atmospheric Watch (GAW) program, WMO CO₂ measurement expert forums have identified 0.05 to 0.1 ppm as a target "network" precision for merging data from different laboratories in order to determine regional fluxes (Francey, 1997). More than 20 years of international effort has so far failed to consistently achieve this target. The effective spatial coverage with the current measurement sites is compromised by these calibration errors, as is the consistency of sufficiently long time-series for source attribution studies. These problems are even more serious for CO₂ tracers such as d¹³C and O₂/N₂. In all of these measurements, achieving high precision depends on repetitious alternating analysis of sample and reference air in a controlled environment and with a sufficient sample size.

Recent WMO (mainly for CO₂) and International Atomic Energy Agency (mainly d¹³C) measurement expert forums have unanimously endorsed a global intercomparison strategy for greenhouse gas measurements (Francey, 1997). Called GLOBALHUBS, its first priority is to establish much more frequent, comprehensive and transparent comparisons between many more measurement laboratories than currently possible, and to incorporate this information into GLOBALVIEW for CO₂ and other species. Four globally distributed HUB laboratories are proposed which maintain and/or develop suites of standards (linked to primary standards), but also maintain intensive HUB intercomparisons using a variety of techniques, e.g. a highly precise low flow rate CO₂ system (see section 4.3.2). The GLOBALHUBS program 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.

[ins] The issue is to ensure timely intercalibration of laboratories to the primary WMO standards with a goal of reducing interlaboratory differences in measured CO₂ concentrations to £ 0.1 ppm.

4.2.10 Expansion of the Observing Network

Estimation of sources and sinks at continental or ocean-basin scales by atmospheric transport inversions is strongly data-limited at the present time. This is known because (a) trial data inversions and network optimisation studies have shown that adding a few vertical profiles over continental interiors could dramatically reduce flux uncertainty (Rayner et al, 1996; Gloor et al, 2000); and (b) flux uncertainty due to sparse data is of the same order as the differences arising from very different transport models in estimating sources and sinks^[17]. There are many opportunities to add substantial data constraints to the mass-balance of atmospheric CO₂ over regional scales (section 4.3.2). To do this effectively, additional network optimisation studies must be undertaken that seek to quantify the reduction in uncertainty for proposed sampling sites before they are deployed. Such studies should also address alternative sampling methods, particularly airborne sampling, continuous monitoring, and remotely sensed concentration measurements.

[ins] The issue is adding sites for flask sampling, based on network optimisation studies.

4.2.11 Improved measurements of atmospheric composition

Improved vertical profiling of trace gases through the boundary layer and above, with flasks or continuous analysers, has long been recognised as one way to improve the scale linkage between fixed surface sites and the transport parameterisations. This is a key strategy incorporated into major continental carbon budget initiatives such as CARBOEUROPE and the proposed CARBON-AMERICA plan (Tans et al., 1996). The rapid development of mesoscale transport models, an integral part of the continental studies, also has the potential to improve the linkage in the case of the background observations, but has not been routinely exploited. Mesoscale atmospheric transport modelling in conjunction with continuous, high-precision trace gas measurements has the potential to be particularly useful over continental regions.

[ins] The issue is continuation of vertical profile measurements begun as part of continental-scale carbon budget experiments.

4.2.12 Availability of Transport and Emissions Data

Atmospheric tracer transport inversions require detailed information about winds, turbulence, and convective transport by clouds at high spatial and temporal resolution. These data are typically available from operational forecast centres, but frequent changes in forecast models and the high costs of obtaining these data have precluded consistent and accurate real-time analysis. Trace gas transport by unresolved vertical motions (e.g., in thunderstorms) is an important control on concentrations, yet these transports are generally not archived by forecast centres and are therefore unavailable for inverse modellers. As more concentration data become available at higher spatial and temporal resolution in the future (section 4.3.2), these data will become much more important for analysis of sources and sinks at regional scales.

To provide a useful constraint on process-based models of surface CO₂ exchange from atmospheric data requires detailed knowledge of combustion sources. These are currently known reported on a national annual basis, and have been distributed by population and seasonal cycles in the research literature (e.g., Andres et al, 1996). Regional mass-balance constraints will require more detailed emission estimates by location (city) and date. These may not be available everywhere, but will be extremely valuable for campaign-style scaling experiments. In addition, the measurements generally have low spatial density, thus necessitating use of proxy data (see also section 4.3.2).

The continuity and consistency issues are:

1 [ins] Nations making available location- and time- specific fossil fuel emission data. Currently nations aggregate their data to the national and annual scale before reporting it.

2 [ins] Archival and distribution of subgrid-scale vertical mass fluxes from operational weather analysis centres to facilitate atmospheric inverse modelling of sources and sinks.

4.3 The Knowledge Challenge^[18]

This section contains a list of issues that are considered important to improving the carbon cycle observations, beyond the initial observing system. They include observations, observing approaches, and methodologies for making or using the resulting measurements. Most of the issues are in the purview of various IGOS Partners, although this is not a condition for including the items.

4.3.1 Terrestrial

LAND COVER AND LAND USE

1 [dev] Effective methods are needed to map the global distribution and temporal variability of wetland cover types (Sahagian and Melack, 1998), possibly in conjunction with soil moisture monitoring (see below). This information is necessary to estimate carbon fluxes for wetlands.

2 [dev] Global land use products should be developed through an integrated in situ / satellite approach that uses a harmonised classification system and provides historical and current land use with the highest feasible spatial detail.

3 [dev] Consensus on the following methodological aspects is necessary: a common hierarchical land cover classification scheme, a common terminology for land use and protocol for data collection, and an agreement on minimum mapping units for land cover and land use at various scales.

4 [dev] Examine the trade-offs between spatial resolution and information obtained from medium resolution (200 m to 1000 m) satellite optical data for global land cover, fires and biophysical parameter extraction.

BIOMASS

1 [dev] New soil carbon estimation techniques need to be developed, using primarily a combination of in situ and modelling strategies. Given the costs and other limitations of current methods, new techniques are the main potential means for increasing the amount of actual observations. Direct satellite-based methods would be very desirable but no viable approaches are known at the present. However, satellite-based biophysical parameter products (section 4.2) may be helpful in a modelling approach. In addition to the new estimation techniques, new surveys of soil carbon also need to be promoted, and should include observations of the factors that regulate its distribution and dynamics so that effective models may be developed. New pedotransfer functions need to be developed for soil data from Eastern Europe to deal with variations in soil data caused by differences in soil analytical techniques. The existing field measurements should be fully used to update the 1:5 million soil map of the world; this is now underway through the combined work of FAO, ISRIC, UNEP and IUSS.

2 [sat] Satellite-based methods offer great promise in quantifying above ground biomass and its changes at high spatial resolution. Lidar, bidirectional reflectance measurements, and radar interferometric techniques (Rosenqvist et al., 1999) are the most promising sensing methods. The satellite technology needs to be developed, beyond the current or planned sensors (VCL/ESSP, single-frequency SARs, and experimental multiangle optical sensors such as MISR/Terra and (lidar) and EPIC/TRIANA).

3 [ins] [dev] Effective techniques need to be developed to utilise forest and agricultural inventory data in quantifying biomass and its changes. In particular, this includes scaling algorithms for above and belowground biomass for all major ecosystems (i.e., beyond those of economic interest), and should be conducted in conjunction with improved satellite observations. Allometry based on the fractal properties of plants holds promise to develop the relationships rapidly and cost-effectively.

SEASONAL GROWTH CYCLE

1 [sat] The use of new satellite sensing techniques (multi-angle optical, lidar) needs to be further investigated in combination with current data types. Important questions include improving LAI estimation accuracy at high LAI values and obtaining information on the spatial distribution of sunlit and shaded leaves within the canopy. The planned POLDER/ADEOS-II and EPIC/TRIANA will contribute in this respect, but their spatial resolution is limited. There is no planned follow-on to MISR/Terra.

2 [sat] An observing strategy needs to be developed to detect frost-free season duration at high latitudes and in mountainous areas. An experimental algorithm based on SAR imaging has been demonstrated (Frolking et al., 1999). SeaWinds scatterometer on Quickscat provides an opportunity to test the concept.

FIRES AND OTHER DISTURBANCES

1 [sat] Satellite-based observation methods need to be developed to estimate the spatially and temporally varying fire intensity and fire burning efficiency (Ahern et al., 2000). These parameters are essential but presently missing inputs to quantifying carbon emissions to the atmosphere. The potential of data from new sensors of the MODIS type needs to be examined as an initial step.

2 [sat] [dev] Robust methods are needed to detect and quantify partial disturbances in forests such as insect damage and selective harvesting.

3 [sat] New methods are required to map areas in dense forests burned by 'ground fires' (in which the tree canopy is not seriously damaged, thus not detectable with medium resolution optical sensors). Among the current or planned sensors, very high resolution (few metres) optical images or canopy penetrating (imaging) lidars may be helpful, in combination with fire behaviour models.

SOLAR RADIATION

1 [sat] Development of daily PAR products from geostationary and polar orbiting sensors is needed. The aim should be spatial resolution near ~10 km and direct estimation for the PAR spectral region (0.4 - 0.7 micrometres).

SURFACE-ATMOSPHERE FLUXES

1 [ins] [dev] More accurate CO₂ concentration measurements should be made at the flux towers, using better instrumentation and frequent automated calibration to standard gases. An operational goal of 0.2 ppmv accuracy is attainable and would allow these measurements to be integrated into the world-wide observational network of trace gas concentrations (section 4.2). These measurements are an important link to the larger-scale flux estimates using atmospheric inversion.

2 [ins] Flux tower methodologies among the regional networks need to be standardised in several areas: (i) ecological measurements including characterisation of site variables, phenology, soil, and site history; (ii) effective use of high resolution remote sensing measurements to characterise the flux - contributing areas; (iii) algorithms for data treatment (atmospheric corrections, data quality checks and assurance, and gap filling); and (iv) hardware and data communication.

3 [dev] A significant effort is needed to integrate tower-based fluxes and ecological information with larger-scale observations and analyses. A key approach is through the use of local flux and ecological data to constrain parameters in process-based models which are then applied over larger scales using satellite-derived products.

METHANE-RELATED PRODUCTS

1 [sat] [dev] Satellite observation techniques and modelling tools should be developed to estimate methane fluxes from wetlands in all major biomes. This includes effective use of satellite - derived products, especially land cover, surface wetness, and seasonal growth cycle. Experience obtained with CH₄ estimates from MOPITT/Terra as well as HIRDLS/Aura, TES/Aura, and GCOM-A1/SOFIS will be very valuable for the future evolution of satellite - based techniques.

2 [sat] Development of sensors to detect the depth to water table is an important area to pursue. Water table divides the aerobic part of the profile (producing CO₂) from anaerobic (producing CH₄), thus greatly affecting the composition and global warming potential of the gases emitted by the wetland ecosystems.

SOIL MOISTURE

1 [sat] [dev] Intensified research and development should be carried out leading to a satellite-based capability to monitor soil moisture world-wide (i.e., soil water content within the root zone; the depth varies by vegetation type but may be nominally considered to correspond to the 0-1 m layer). The experimental MIRAS/SMOS mission is an important step in this direction. Current efforts of GEWEX/BAHC, the LDAS project^[19], and other related initiatives should be supported. They should be structured to take full advantage of the planned NPOESS products and of the satellite data preceding NPOESS, including AMSR-E/Aqua, AMSR/ADEOS-II and GCOM. Soil moisture availability exhibits a major control on CO₂ uptake by vegetation. In addition to soil moisture, surface wetness (water content at or within a few cm from the surface, including flooded or saturated condition) provides an important constraint on trace gas exchange with the atmosphere, in both herbaceous and forested wetlands. Flooded forests can be satisfactorily mapped using satellite SARs, especially L-band SAR on JERS-1 and ALOS.

CANOPY BIOCHEMISTRY

1 [sat] Experimental programs should be supported to determine the operational feasibility of producing robust estimates of canopy biochemical properties. Leaf nitrogen content is of most interest because of its role in photosynthesis and the close coupling between carbon and nitrogen cycles; it is most important in nitrogen-limited ecosystems such as the boreal forest. Based on the research to date, high resolution imaging spectrometry measurements are the most appropriate satellite observing strategy, and they might take advantage of the correlation between chlorophyll concentration and nitrogen content. The planned EO-1/Hyperion mission should provide a valuable contribution in this respect, as a first-time demonstration of the feasibility. However, significant additional effort will be required in this area.

4.3.2 Atmospheric

ATMOSPHERIC COMPOSITION AND TRANSPORT

1 [dev] Problems of relating point observations at surface sites to the scale of atmospheric transport representations (potentially significant at the mostly-marine GAW sites) become critical over terrestrial regions of strongly heterogeneous exchanges under low wind-speeds. Modelling studies have shown that added atmospheric observations in such continental regimes hold great potential for more robust inverse flux estimation. Interpretation of spatial and temporal patterns in trace gas composition under these conditions will require new field and modelling studies. Field campaigns including sampling from tall towers, tethered balloons, and light aircraft should be conducted in conjunction with ground measurements of fluxes and ecosystem condition, and would be used to test upscaling methods based on remotely sensed imagery. Process models and spatial data would then be coupled to mesoscale transport models to test their predictions against the airborne samples.

2 [dev] Continuous in situ CO₂ analysis would be highly advantageous, particularly in continental areas and in conjunction with high-resolution transport modelling because the full range of atmospheric mixing conditions at a site is sampled, permitting proper averaging of data (aided by local area atmospheric transport models). The value of multi-species information from flask sampling is also significantly enhanced when the sampling is in done conjunction with a continuous analyser. The challenge is to achieve much wider deployment of robust remotely-operated continuous CO₂ analysers, with an acceptable trade-off between logistical independence and precision. Deployment can also be extended to moving platforms (ships, aircraft) if potential vibration sensitivity is overcome. An important performance breakthrough in conventional Non-Dispersive Infra Red CO₂ analysis, the most precise technique currently employed in global studies, has recently occurred (Da Costa and Steele, 1997). The remotely-controlled operational system has substantially lower sample size requirements, much higher long term stability, and significantly lower operating costs. However, engineering development is required to miniaturise, ruggedise and decrease costs through mass production techniques. An interim deployment of cheaper low precision continuous CO₂ analysers is worthy of consideration.

3 [ins] The accuracy and resolution of the spatial distribution of CO₂ sources and sinks computed through atmospheric inversions would increase substantially with better characterisation of the transfer and mixing behaviour of the atmosphere. In particular, subgrid-scale vertical mass fluxes are routinely calculated by operational weather forecast models, but are not reported or archived, yet these are crucial for analysing trace gas spatial structure at mesoscales.

TECHNOLOGY DEVELOPMENT FOR TERRESTRIAL ENVIRONMENT

Increased deployment of continuous analysers and flask sampling on platforms which permit characterisation and/or minimise surface layer influences (tall towers, balloon sondes, piloted and pilot-less aircraft) is a key part of current approaches to the problem of CO₂ sampling and integration in the heterogeneous terrestrial environments. Results from these studies will have a significant input to both top-down and bottom-up studies of the carbon cycle. Consequently, improved calibration against primary standards is a challenge for these as well as for background monitoring sites. Several options offer promise:

1 [sat] Remote infrared CO₂ monitoring via satellite is the subject of considerable current interest. Although integrated column CO₂ can be obtained with a relatively low precision (~1ppm), the potential for reducing regional flux uncertainties, particularly in continental regions now poorly determined by surface observations, appears quite high. Given the lead-time for developing the required sensor technology, this area should be pursued as soon as feasible. Among the planned sensors, AIRS/Aqua and IASI/METOP will provide opportunities to develop and test this concept.

2 [dev] Another new technique of considerable promise is Fourier Transform InfraRed spectroscopy which provides relatively low precision but continuous, multi-species monitoring. In terrestrial systems, the potential for chemical identification and monitoring of significant sources is high.

3 [dev] Spatial and temporal coverage will be significantly expanded if a gap can be closed between instrument size and payload limits for the smaller, less expensive sampling options (drone aircraft, balloon sondes); technological developments in this direction should be encouraged.

4 [dev] There is a need to develop techniques for estimating CO₂ emissions from fossil fuels with a higher spatial and temporal resolution than presently feasible. Direct satellite sensing (see above) could make an important contribution in this respect.

4.4 Organisational Challenge

There are four major organisational challenges in establishing a global observing capability for terrestrial carbon:

• Continuity and improvements of satellite and in situ observations. For satellite observations, this can be addressed within IGOS-P as most of the satellite agencies are part of the decision process in CEOS. The situation is more complicated for in situ observations. For atmospheric in situ observations (trace gas concentrations,..) the link is through WMO to national agencies. For terrestrial ecosystems, the links between the national funding agencies and international programs are weak. Internationally, GTOS and its sponsors (particularly FAO, UNEP, WMO) have national points of contact but at the national level, the programs and funding are not typically handled centrally. It is therefore difficult and unwieldy to establish effective links between operational national terrestrial in situ observations and global carbon observation effort. In situ observation networks are gradually being linked (e.g. ecology, hydrology, permafrost) through initiatives such as GT-Net (Appendix 2.) but progress has been slow and will need to be accelerated if the in situ community is to contribute effectively to TCO.

• Transition from research to ongoing operations. This is an issue for most of the elements of the carbon observation theme, except for weather-related observations. A major reason is the relatively recent interest in the global carbon cycle and the associated experimental nature of the observing technologies and programs. However, transition to routine operations is essential if IGOS-P is to achieve long-term, systematic observations that are required to meet the requirements (section 2.). This may best be achieved gradually, by pursuing the transition for individual components of the observation system as they mature. The initial candidates are satellite observations, atmospheric concentration measurements, and some in situ terrestrial observations.

• Supporting research and technology development. Improvements beyond the initial observing system will necessitate vigorous support for development in several areas, including: instrumentation for in situ and satellite observations (section 4.3); in situ networks enhancement and design optimisation studies, in turn requiring the capability to evaluate trade-offs in performance based on various hypothetical improvements in the observations; and models and algorithms that are able to effectively use the improved observations and eventually perform well with reduced observations. Therefore, technology development needs to be carried out in close collaboration with the research community, internationally as well as nationally. This is already recognised by some current programs (e.g., the EU Fifth Framework Programme; ftp://ftp4.cordis.lu/pub/eesd/docs/d_wp_en_199901.pdf).

• Co-ordination. From the above it is evident that the establishment of systematic observations of the terrestrial (and the linked atmospheric) component of the global carbon cycle is a complex undertaking, mainly because the complexity of the carbon cycle intersects with the political and economic structures that have been set up at the international as well as national levels. Effective and efficient arrangements for global carbon cycle observations must therefore rely on several components, most of which have multiple clients and are sponsored for different reasons. Since it is unlikely that this situation will change appreciably in the near future, it is essential that an effective co-ordination mechanism be established by IGOS-P. This mechanism needs to meet the needs of the three observing systems as well as those of international research programs. A possible way forward is outlined in section 6.