The North American Carbon Program Plan (NACP)
Chapter 2: Major Elements of the North American Carbon Program Plan (NACP)
Major program elements are noted briefly here, and elaborated on further in the following sections:
The following sections survey these elements in greater detail.
|Atmospheric Sampling Programs|
Sources and sinks of CO2, CH4, and CO impart their signatures on the distribution of atmospheric concentrations, under the influence of atmospheric transport processes (e.g., advection, turbulent mixing, cloud venting). Thus the spatial and temporal distributions of CO2, CH4, and CO in the atmosphere provide spatially resolved information on surface-atmosphere fluxes over time, once atmospheric transport is accounted for.
This section discusses atmospheric observations to characterize the distributions of CO2, CH4, and CO that will serve as input for the analysis framework. There are two complementary components:
|(a) Network of Long-Term Atmospheric Observations|
Guiding Principles for the Network. Long-term measurements of concentrations of CO2, CH4, and CO are currently made at a network of dispersed surface sites, each intended to represent air over a large region. Thus, most sites take samples at remote oceanic or desert locations, avoiding the influence of strong and variable sources and sinks on the continents. The purpose is to provide an accurate measure of global and hemispheric concentrations, including seasonal variations and long-term trends. Attempts to use these global data sets to infer net sources and sinks over North America have been limited by the absence of data from the interior of continents and near-shore environments.
Unfortunately, suitable data cannot be obtained by simply adding a few continental measurements to the existing network. Continental stations at ground level observe highly variable concentrations, reflecting the influence of proximate sources and sinks on the composition of the planetary boundary layer (PBL). This difficulty is only partially mitigated by making observations from tall towers, because even the tallest does not span the PBL on most days.
Tans et al. (1996) considered the characteristics of a network suitable for determining regionally resolved fluxes of the key gases. The requirements include vertical soundings through the PBL and into the middle troposphere using small aircraft, plus continuous measurements of fluxes and concentrations from both tall and short towers. Most aircraft soundings should be made with light planes equipped with simple, portable instruments to measure CO2, CH4, and CO; 2-4 stations in the network should have capable small jet aircraft with more sophisticated instrumentation and much greater range and ceiling. Corollary developments include specification of the actual vegetation cover and its physiological state and determination of the atmosphere-ocean fluxes in adjacent ocean basins.
Observations and models indicate that the correlation lengths for concentrations of CO2, CH4, and CO are typically about 1,000 km, similar to the size of weather systems. Thus, to determine an overall budget for North America we require a “ring” of stations along the coasts and through northern Canada, and stations in the interior at about 1,000 km spacing. These scales are suitable for measuring the influence on atmospheric CO2 of regionally distributed processes, such as regrowth of forests in the eastern United States, agriculture in the Midwest, or woody encroachment in the Southwest. At least 20 sites would be needed across the conterminous United States, with lower density in deserts. The number needed in Canada and Mexico will depend on the goals and objectives established by those countries. A total of 30 sites for North America are anticipated.
The spatial density and the accuracy of the measurements must be adequate to distinguish signals due to uptake/release from other sources of variability. Consider uptake of CO2 due to woody encroachment as an example. Pacala et al. (2001) estimated a notably high number, 0.12 GtC/yr, primarily in the Southwest. If spread out over an area the size of Texas, the annual mean decrease of CO2 in the column would be 0.11 ppm/day. The signal would typically be concentrated in the boundary layer and lower free troposphere, and flushed out in about two days. The associated depletion in atmospheric CO2 over 1,000 km could be 0.6 ppm in the lowest 3 km, comparable to the CO2 from fossil fuels, which has to be carefully accounted for using other trace gases, especially CO. Our plan thus addresses the critical question of the capability of the measurements to define net fluxes from large-scale processes, such as woody encroachment, in the large-scale intensive experiments. The atmospheric “integral” would get increasingly difficult to use at scales of less than 1,000 km.
Vertical profiles should be obtained at a frequency of every other day, to avoid undersampling on the time scales for passage of weather systems, a major cause of variance of trace gas fluxes and concentrations. One could argue that two flights should be conducted each flight day to account for diurnal variations, but this is not yet certain. Perhaps co-located ground stations or towers could substitute. This question is a principal target of the large-scale intensives described below.
Measurements will have to be continuous in flight to allow sampling in controlled air space and to provide suitably accurate resolution of the PBL. A critical need for the program is to develop robust, easy-to-operate instruments for routine use in measuring CO2, CO, and CH4. Better instrumentation for CO2, CH4, and CO is needed for deployment on aircraft, at ground stations, on ships, and on tall towers. Likely more than 100 such instruments will be needed. Oceanographers and many others could put them to great use. Their development should not be delayed, otherwise the whole program will be impeded. Important contributions could be made also by total-column observations made using high-resolution upward-looking Fourier transform spectrometers, which can potentially determine very accurately changes over time in the vertical column amounts of CO2, CH4, CO, and other gases.
Fair-weather bias attends measurements made aboard aircraft, and observations are influenced by the fact that atmospheric variance, especially for CO2, peaks close to the ground. Thus the PBL needs to be sampled as densely as possible. A network of tall towers, smaller AmeriFlux towers, and near-shore buoys should observe continuously. Tall towers provide robust statistical information on the covariance of the target gases under all weather conditions (Bakwin et al., 1998). Flux measurements, as in the AmeriFlux network, provide crucial biophysical information on the relationships between ecosystem types, flight days versus non-flight days, and uptake or emission of key carbon gases (Baldocchi et al., 2001) and on the covariance of target gases under all weather conditions (Potosnak et al., 1999). Hence towers allow adjustments for systematic errors associated with fair weather bias. Offshore aircraft profiles will be paired with continuous measurements from permanently moored buoys.
Flask samples provide quality control, especially for merging data from different stations, and allow measurement of other tracers and isotopes. Flasks should be deployed on a subset of flights, and at ground stations, for example, a set of 12 flasks could be obtained at each aircraft and 2 at each ground site at 10-day intervals.
Phased Implementation of the Network. The observing system is designed to infer the magnitudes of sources and sinks of CO2, CH4, and CO from observed concentration differences, requiring analysis of regions sufficiently large that the impact of important postulated mechanisms for carbon storage or loss can be measured. We seek to discern the effects of current and historic land use and land management-for instance, an increase of organic matter in agricultural soils, regrowth of forests in the East, or woody encroachment in the Southwest. The effects of regional environmental anomalies, such as drought, cloudiness, air pollution, or changes in the length of the growing season, should be resolved. The design of the initial version of the observing system ought not to rely on atmospheric transport models to fill in sparse measurements.
An observing network of this kind has not been implemented previously. The design therefore includes phased development of new observations, with improved models to be applied continuously to such problems as optimizing network design and selecting sites. Early stages will demonstrate proof-of-concept by focusing on areas with already existing information on sources and sinks (e.g., for croplands), coordinated with intensive field studies to test instruments and develop infrastructure, personnel, and diagnostic models. Later stages will use knowledge from the early phases to refine the network design.
An initial conceptual plan for the phased implementation of the network is outlined in Appendix 1. Components include the following:
Definition of an Observing Station. The ideal observing station has four components, not all of which will be realized at every site.
|(b) Intensive Atmospheric Field Campaigns|
Intensive field programs (intensive operation periods, IOPs) are part of the phased implementation of new continental observations. Currently we have only rough estimates of the spatial and temporal sampling density needed to resolve seasonal and annual budgets for CO2, CH4, and CO, based on limited field data (e.g., the recent CO2 Budget and Rectification Airborne Study, COBRA 2000; see example data below) and model runs. The IOPs will combine deployments of research aircraft, biophysical studies, development of models and analysis tools, and remote sensing to address critical subsets of research questions required for the program. The enriched sets of data will help to determine how well data from the long-term stations (element 1, page 13) represent ambient distributions, and to evaluate the accuracy of tracer budgets computed from the data of the long-term network. Enhanced data sets from the IOPs will also help develop and test models described in element 3 and help to constrain bottom-up scaling approaches driven by biophysical data (elements 2 and 3, page 13).
Large-scale measurements in all NACP phases will link up with airborne observations from coordinated atmospheric missions such as NASA's Global Tropospheric Experiment, and proposed NCAR and NOAA experiments. Appendix 1 illustrates possible configurations and mission profiles envisioned for the intensive program of measurements.
|Land: Measurements and Models of Terrestrial Carbon Fluxes|
Large-scale carbon fluxes among land ecosystems, the atmosphere, and the ocean reflect the responses of diverse ecosystems to climate, soils, natural disturbances, and direct and indirect human perturbations, including air pollution, elevated atmospheric CO2, and land use and management. The size and distribution of carbon stocks are roughly known for the major ecosystem types in the United States. However, knowledge of changes in stocks, and the mechanisms that cause such change, is far from complete.
Changes in soil carbon control net ecosystem carbon flux in many non-forested ecosystems such as grasslands, croplands, and wetlands, which comprise two-thirds of the land surface in North America. In forests and woodlands, changes in C stocks can be significant in live or dead biomass, or in soils. Carbon stocks in soils and vegetation may respond differently to environmental changes. Carbon fluxes are especially poorly documented for important ecosystems with limited commercial exploitation. For example, increases in woody vegetation in grasslands and savannas, resulting from long-term suppression of fire, appear to be a significant carbon sink (Pacala et al., 2001), but few data are available.
Peatlands are another poorly known but potentially important ecosystem, covering 12% of the land in North America. Net primary production is low, but stocks of soil carbon are huge (Harden et al., 1992), with approximately 455 PgC (about 60% of the C in the atmosphere) stored within 1 meter of the surface. Peatlands and wetlands are also major sources of CH4 (e.g., Crill et al., 2000). Peatlands are vulnerable to small changes in climate. Changes in temperature, evaporation, precipitation, or hydrology can quickly change a peatland from a small sink for atmospheric CO2 and a source of CH4, to a strong source of CO2 and a small sink for CH4.
Over 70% of CH4 emissions are anthropogenic, dominated by biogenic sources (e.g., landfills, domestic sewage, rice agriculture, ruminants, animal waste), with a smaller amount associated with fossil energy. Agriculture accounts for about 50% of human-related CH4 sources globally, about 30% in the United States. Agricultural sources of methane include concentrated (e.g., feedlot) and diffuse (non-point source) emissions, and are sensitive to production practices such as applications of water, fertilizers, and manures. Determination of agricultural methane emissions is needed to quantify the North American and global carbon budgets.
Natural wetlands account for more than 20% of the global CH4 source, largely northern peatlands and tropical wetlands. CH4 exchange from these environments is intimately linked to hydrology, system productivity, and carbon accumulation and balance. At the regional scale, CH4 emissions for many landscapes in North America are dominated by natural sources (termites, wetlands, lakes and coastal waters). Different mixes of anthropogenic and natural sources and sinks determine the net fluxes in different regions. For example, in New England, northern peatland sources dominate CH4 emissions in Maine, but landfills and energy use dominate in Massachusetts and the south.
Historical legacies of natural disturbances and past land use and land management appear to play a large role in the long-term carbon balance of North America. Forest and agricultural inventories, historical data on forest harvesting and clearing, and agriculture can all provide clues to historical factors that regulate carbon fluxes. Currently incomplete historical records and limited process understanding generate substantial uncertainty in the fraction of the current carbon sink in U.S. forests that can be attributed to the trajectory of land use and land management (e.g., Pacala et al., 2001). Natural disturbances, including severe storms, fire, and insect outbreaks, appear to be important also (Kurz and Apps, 1999) and may be increasingly important in the future.
Objectives for Enhanced Measurements and Models on Land
The goals for NACP ecosystem measurements and models are to quantify, and reduce uncertainty in, spatial patterns and mechanisms accounting for changes in carbon stocks and CH4 release and uptake. The effort is intended to make a critical contribution to integrated analysis of the North American carbon balance.
Principal associated research objectives for NACP terrestrial studies include the following:
The long-term observational strategy should include several elements:
Except for CO2 flux towers, current land surface observations do not explicitly measure or monitor changes in C stocks (or fluxes). The data therefore lack critical features, including lack of complete ecosystem C measurements (particularly below-ground C), gaps in spatial coverage, inconsistent procedures relative to time and location, and lack of sufficient temporal resolution (re-measurement intervals as long as 15 years in important areas). The NACP plan will address this issue by including quantification of carbon stocks and fluxes as a key objective for ecosystem measurements at all scales.
Advances in modeling, such as newly emerging dynamic global vegetation models (DGVMs) and high-resolution biophysical models, may play important roles in integrating land and atmospheric data. The key will be to link spatial and temporal scales up to continental (or larger) and decadal (or longer) scales. A major goal of the NACP is to develop new frameworks for cross-scale connections, and to link biophysical studies of the carbon cycle with socioeconomic models that address the needs of policy makers and land managers.
Implementation of Biophysical Measurements: A Hierarchical Approach
Diverse data on land ecosystems, collected at a range of temporal and spatial scales, are required to determine regional- and continental-scale carbon exchange. The data must be integrated within a comprehensive framework that includes statistical scaling techniques as well as process models. Sources of data will include reconstructions of land use and land management history, past and ongoing resource inventories, satellite remote sensing, as well as process studies, including ecosystem gas exchange with eddy covariance.
Land use records and inventory measurements have low temporal resolution (5-10 years), but sampling density can be very high, with more than 100,000 sites in some networks. Data from eddy covariance sites provide information with very high temporal resolution (30 minutes) at a few (about 40) sites. Satellite data cover the whole landscape at frequent intervals, but the data are only indirectly related to carbon fluxes. Remote sensing with the highest temporal resolution, but moderate spatial resolution (e.g., from MODIS), enables extension of process-based simulations, while high spatial resolution but lower temporal resolution data (e.g., LandSat) are best to define land use and management.
Extensive observations are critical because the land surface is so diverse. Given the clear importance of vegetation, climate, soils, natural disturbance, and land use history, every plot can potentially have a unique carbon balance. Without the comprehensive coverage provided by remote-sensing data and inventories, it is exceedingly difficult to constrain the carbon balance of a large region. But interpreting the extensive data is impossible without calibration and understanding from intensive studies at selected sites. Observations at intensive study sites complement large-scale data in at least five ways. First, they provide a context for testing hypotheses about mechanisms. Second, they serve as a test bed for model development and testing. Third, they function as test sites for evaluating extensive methods, including methods based on both inventories and remote sensing. Fourth, they provide data that are unique in terms of both processes covered and accuracy for comprehensive analysis with data fusion. Finally, intensive sites can serve as test sites or controls for experiments aimed at enhancing carbon storage.
Since intensive studies are impractical at more than a few hundred sites, the NACP will use a hierarchical approach to land research, integrating four networks spanning intensive to comprehensive. This hierarchy will:
The large-scale land monitoring programs will adopt a sampling strategy with four tiers distinguished by spatial averaging, resampling frequency, and type of observations. Some data elements are identically defined and collected at each tier, providing direct links among tiers, while other variables may be unique to one or a subset of tiers. The combination of remote sensing, extensive inventories, medium-intensity sampling, and intensive observations at selected sites together comprises a powerful, flexible, and potentially efficient data collection system. The sample tiers must, however, be linked statistically so that inferences about the entire population within cover classes can be made. The observation system should have the capability to closely integrate with atmospheric monitoring, but should also stand alone to provide independent estimates of C fluxes for validation and as a contribution to ecosystem science.
Multi-tiered sampling and analysis has previously been implemented for land inventories and, more recently, for linking new remote sensors with field data. The first tier of the NACP involves comprehensive measurements with remote sensing at continental scale. Middle tiers include (tier 2) existing, densely sampled, extensive land inventories composed of a large number of sample plots, and (tier 3) a proposed new set of approximately 1,000 medium-intensity plots with process monitoring at medium-intensity sites, also selected to represent typical conditions across the landscape. The fourth tier includes the existing and potentially new intensive observation sites where the most detailed observation are made, such as at LTER and AmeriFlux sites.
Table 1 illustrates the multi-tier concept with a listing of a few of the variables likely to be central to the land observation system.
Table 1. Multi-tier concept with likely variables central to land observation system
NACP implementation envisions that a complete and well-defined set of variables will be phased in systematically over three to five years. During implementation, appropriate estimators will be defined through modeling and analytical studies, and recommendations made for enhancing observations to produce an efficient, continuing multi-tier network optimized for estimating C flux at multiple temporal and spatial scales. An efficient way to integrate across scales is not apparent for some critical variables such as soil CO2 and methane flux. Preliminary studies and pilot implementation tests will be undertaken to develop a strategy for these variables.
The following summary shows how the hierarchy of networks will link observations and understanding across space and time, carefully stratified to ensure appropriate coverage of climate, soils, vegetation, disturbance, and land use history. The intent is to link networks with coordinated observations and analysis, implement strong QA/QC at all levels, quantify estimation errors, and provide appropriate inputs for data fusion through diagnostic models and analysis.
Tier 1: Comprehensive Measurements with Remote Sensing. Current U.S. land inventory systems use a combination of high-altitude aerial photography and Landsat Thematic Mapper (TM) data to sample the largest scale and to detect change. Several sensors with coarse spatial resolution but high temporal resolution have been used to drive terrestrial biogeochemistry models that estimate carbon stocks and fluxes, typically with limited information on land use history. Products from the 20-year record of NOAA AVHRR (advanced very-high-resolution radiometer) data provide a picture of recent temporal dynamics. SeaWiFS provides a vegetation index of higher quality but for a shorter record (Behrenfeld et al., 2001). The NASA MODIS sensor, launched in 1999, should provide data of very high quality over the coming years, enabling a number of new biogeochemical models (Running et al., 1995).
Additional sensors in the planning stage will provide new information, including aspects of canopy chemistry and structure and soil moisture. The Vegetation Canopy LIDAR (VCL) is a promising sensor. VCL data may provide estimates of above-ground biomass, an important quantity in the carbon budget that has been beyond the scope of most remote-sensing analysis. Enhanced utilization of remote-sensor data is also needed. Data on land use and cover (from LandSat) need to be better integrated with the data from AVHRR, SeaWiFS, and other satellites with high temporal resolution, and remote-sensing data need to be better informed by historical, atmospheric, and weather data. The range and quality of remote-sensing data products need improvement.
Several specific needs have been identified to provide data to the NACP on carbon stocks at tier 1, using remote-sensing products and in situ data: (1) timely systematic and routine processing of satellite data from the North American continent into land cover and land cover change products, covering both natural and human disturbances; (2) integration of satellite observations with in situ measurements of carbon stocks and existing inventories; (3) augmentation of satellite and in situ estimates of carbon stocks with airborne and surface measurements; and (4) development of appropriate estimation models.
Additional details about remote-sensing contributions to the NACP are included in Appendix 4a.
Tier 2: Dense Sampling with Inventory Techniques. Current large-scale land inventories conducted by USDA, the Forest Inventory Analysis (FIA) and National Resources Inventory (NRI), employ multi-tier sampling strategies using remote-sensing and ground measurements. These continuous inventories provide baseline information about land cover, management intensity, productivity, and disturbance that can be used to estimate carbon stock changes over 5 to 10 year periods. Very high sampling intensity allows detailed description of some of the causes of observed carbon stock changes, such as the effects of vegetation growth, mortality, and harvesting. Historical data are available to trace land use and management history.
Current land inventories are limited in several critical ways: incomplete coverage of regions and ecosystem types; lack of complete ecosystem C measurements; limited temporal resolution; and lack of easily available and usable historical data. There is little or no coverage of some “reserved” areas: lightly sampled areas of the Intermountain West, the Pacific Coast, Alaska, urban, and suburban areas; and large areas of public non-forest land (mostly grazing land in the West). Large areas of Canada and Mexico have been sparsely sampled. Enhancements to ongoing inventories are projected to fill some of these gaps, especially in forests, but others remain.
Carbon pools that are poorly quantified in existing inventories include stumps, live and dead roots, mineral soil, litter, and coarse woody debris. Land inventories are generally designed to provide a “rolling average” estimate with a temporal resolution of 5 to 10 years, sufficient for some applications, incompatible with the temporal resolution needed by the NACP.
New designs for forest inventories address temporal resolution using successive sample “panels” to approximate continuous sampling. Each panel is re-sampled with a period of 5 to 10 years. Supplemental data with higher time resolution are merged with the inventory data to estimate the major causes of variations in C flux-productivity, mortality, harvest, and land use change-using advanced statistical techniques to estimate annual changes in C stocks. Sources of supplemental data include flux towers (productivity and trace gas dynamics), aerial and satellite disturbance surveys (land use change, damage and mortality), and timber and agricultural product surveys (harvest quantities).
Some of the existing sites for intensive studies (e.g., LTER and Ameriflux) are not monitored with inventory techniques, inhibiting the extension of results of intensive studies using inventories. This is especially true for terms like soil CO2 flux and CH4 flux, which are beyond the scope of traditional inventories. New direct measurements of C fluxes will be needed during intensive field programs, and in the new medium-intensity sampling network (tier 3).
Tier 3: Process Monitoring at Medium-Intensity Sites. To take full advantage of comprehensive remote-sensing data (tier 1) and extensive inventories (tier 2) will require a set of sites of intermediate intensity and number. It is not practical to deploy intensive sites in all of the nation’s ecosystem types and across the full range of land use histories. Yet, process data from this entire range is crucial for robust analysis. A new network of approximately 1,000 medium-intensity sites is planned to provide appropriate coverage.
Measurements at these sites should address the major processes in the carbon balance, including net primary productivity (NPP), leaf area index, leaf nutrients, soil respiration, litterfall, dynamics of coarse woody debris, and CH4 flux. Depending on technology development, it may be practical to measure C balance with eddy flux at many of these sites. It will be important to obtain accurate records of land use, including past history as well as current management. Environmental conditions, including soil moisture and solar radiation, should be measured at each site to facilitate assessment with ecosystem models.
The new tier 3 sites will serve as invaluable links between the approximately 100 intensive sites and the much larger number of inventory sites. Because tier 3 sites will directly measure components of carbon balance, they will be a centerpiece for testing models developed at the intensive sites. Direct measures of components of the carbon balance will also be critical for setting appropriate conversion factors for the quantities that need to be estimated in inventories.
Tier 4. Mechanistic and Process Studies at Intensive Measurement Sites. Intensive sites provide direct estimates of C flux and C stock changes across a range of temporal scales. In addition, research at these sites will include detailed studies on the mechanisms controlling the fluxes. Data from intensive sites will be critical for developing and testing models, for interpreting large-scale patterns, and for constraining data fusion models.
Many of the approximately 100 intensive sites that will be needed are already in place. Net ecosystem CO2 exchange is presently measured at more than 30 sites in North America that are part of the AmeriFlux network. Additional sites are planned in the Flux-Canada network. Only a few sites are currently making the complete range of measurements that will be needed for the NACP. Enhancements will be necessary to insure that all of the flux sites measure the full range of controlling variables and that they also deploy all of the measurements used in tier 1, 2 and 3 sites.
Flux tower sites. Net ecosystem CO2 exchange is presently measured at more than 30 sites in North America in the AmeriFlux network (supported largely by DOE through the Terrestrial Carbon program and the National Institute for Global Environmental Change). Additional sites are planned in the Flux-Canada network. Summed over the course of a month, season, or year, data from these sites provide direct measures of ecosystem CO2 source or sink strengths. In contrast to the network of tall towers described under element 1, most flux towers are “small” (<60 m) and provide information specific to one ecosystem type.
Data from flux sites help test physiological models of C exchange and are critical to relating fluxes and remote-sensing data. Companion physiological and ecological measurements enable partitioning carbon fluxes into plant and soil components and reveal mechanisms responsible for these fluxes. At some sites, biomass-based estimates of C storage have validated C budgets from direct flux data (e.g., Curtis et al., 2002; Barford et al., 2001). Data from the flux sites have been applied in ecology, weather forecasting, and climate studies, especially for sites with several years of data to quantify interannual flux variations.
Important enhancements are required for this network in the NACP:
At least one-third of North America is topographically too complex for eddy flux measurements, and gaps will have to be filled using remote sensing, The Tier 2 intermediate sites, and modeling techniques (see “data assimilation”, below).
FACE, LTER, and agricultural study sites. The National Science Foundation’s Long-Term Ecological Research (LTER) sites can contribute to understanding terrestrial C budgets. Some include unique measurements such as dissolved organic C and particulate C losses. Existing long-term agricultural experiments provide another major resource. The CASMGS (Consortium for Agricultural Soils Mitigation of Greenhouse Gases) program involves a number of USDA and University Experiment Station long-term sites, focused on soil management decision-making issues. Free Air Carbon Exchange (FACE) experiments examine stimulation of ecosystems by elevated CO2. Enhanced instrumentation on a number of these sites, adding tower- and chamber-based measurements of CO2 and CH4 fluxes, isotopic measurements to support understanding, and modeling the controls on carbon cycling, are envisioned as parts of the NACP.
New intensive observation sites. In addition to enhancement of current sites, the new intensive sites will be needed to sample under-represented ecosystems and patterns of land use and current management. Additional sites will be needed in actively managed cropland, pasture, and arid ecosystems, as well as in Mexico and northern Canada.
The tier 4 sites will need strong coordination to function with a high level of reliability and quality control. In addition, key sites should be networked to deliver data in near real time, facilitating the prospects for real-time analysis. A more formal structure than exists at present, with defined site selection, QA/QC, review, and analysis procedures, will be necessary for NACP. At least one-third of North America is topographically too complex for eddy flux. Developing methods for obtaining tier 4 data from these regions is a key priority that needs to be addressed at the start of the NACP.
Tier 4 measurements of terrestrial sources of methane. Atmospheric measurements of CH4 should be complemented by surface observations at representative sites to enable optimal evaluation of source/sinks, and to quantitatively resolve the major elements that produce the net flux. Identification of sources and long time series of data are required to quantify causes of variability and to resolve emission processes at interannual or longer time scales.
Between 1 and 12 million hectares of forest and other vegetation burn annually in North America, resulting in emission of 40-200 Tg of carbon as CO2, CH4, CO, and other trace gases. These emissions represent a potentially major source of error in the analysis of atmospheric measurements, and they may make significant contributions to annual budgets for North America. Large fires in Mexico, the coterminous U.S., Canada, Alaska, and even Russia produce emission plumes in the study region with significant enhancements of concentration. Significant research is now focused on developing methods to collect the required data and to model emissions from biomass burning, but considerable uncertainties need to be addressed in the near term.
Data Assimilation, Analysis and Models
The Data Assimilation Challenge. This section outlines the development of the soil and plant components of an integrated data assimilation framework for the C cycle, built on closely coupled, data-driven models for the atmosphere, soils and plants. The atmospheric components, and the links between biophysical and atmospheric components, are discussed elsewhere in this document.
The challenge is to define a vegetation-soil-biogeochemistry modeling framework that can interface optimally with input data from diverse sources. In principle, a model in this framework could be as complex as nature itself, with countless parameters, but we must develop a diagnostic model whose parameters can be constrained by observations, focusing on quantitatively defining those processes that regulate the key emergent properties of the ecosystem (fluxes, stocks, structure) on the relevant time scales (hours, years, decades). Models within this framework will be used, to analyze data from the NACP using conventional methods and full-system assimilation methods, and will function prognostically when linked to climate models for the future.
Features must include improved estimates of carbon stocks in soils and in natural and managed vegetation, and accounting for prior disturbance, nutrient limitations and inputs, pollution, extreme meteorological events, chronic and acute stress and herbivory, and invasive species. Considerable effort will be needed to design this new class of model, including observations and manipulations to test the models over long and short time scales.
Input to biophysical models. Historical land use, exposure to air pollution and deposition, severe weather, insect outbreaks, and management are key factors explaining current observed terrestrial CO2 fluxes and associated ecosystem structure, including age class distribution, soil fertility, and species composition. Historical data in a spatially explicit database are essential inputs for models. A project to develop a prototype of such a database has begun, with special attention to historical information about selected intensive study areas. It will be a dynamic database, updated using remote sensing and enhanced inventories outlined in the previous section. The goal is an accurate, high-resolution, time-varying map of land cover and land use in North America, with the following data:
These data are the foundation of biophysical/ biogeochemical models.
Structure of biophysical/biogeochemical models. Biogeochemical models simulating fluxes of mass (CO2, H2O, and CH4) and energy, productivity, respiration, and effects of disturbance will be driven by data from three sources: (1) the dynamic map of land cover/land use/inventories establishes the slowly varying state variables of the ecosystem (stock, vegetation, soils) at each grid point; (2) remote-sensing and in situ biophysical data provide current values of a subset of transient state variables of the system (soil moisture, phenology, recent events such as drought, wind, ice); and (3) atmospheric data (from satellites and assimilated meteorological products) provide drivers for ecosystem processes (sunlight, evaporation, temperature, precipitation). It is very advantageous to maintain consistency in the conceptual framework and driving data across the diverse spatial and temporal scales applicable to biophysical or biogeochemical models.
Land surface biophysical models will ideally be driven at time steps of about 1 hr to provide temporally and spatially complete estimates of surface CO2 and CH4 flux comparable with aircraft data. The models must first accurately compute energy and water balances for the major vegetation and climatic regimes on the continent and in the coastal ocean. The model then must compute hourly CO2 and CH4 inputs and outputs, that is, photosynthetic uptake and autotrophic and heterotrophic production of CO2 and CH4. Nutrient cycling and other long-term factors (e.g., vegetation structure, soil organic carbon, and permafrost) must be dynamically treated.
The NACP requires realistic simulation of fluxes for time scales shorter than a day using an “ecological data assimilation” approach. Each iteration updates from the previous time step using ongoing satellite data and in situ observations (mainly from AmeriFlux sites). AmeriFlux and satellite data will have to reach a central location promptly. A pilot study is currently underway in which MODIS data are obtained weekly for evaluation. There is clearly much work to be done, though, to obtain daily or hourly downloads, and the uniformity and reliability of the AmeriFlux data products require improvement. Many remote-sensing data products are currently available, or expected by 2004, including the following at 1 km resolution: snow cover, albedo, surface evaporation resistance (for energy partitioning, LAI and FPAR, or Fraction of absorbed Photosynthetically Active Radiation), GPP (Gross Primary Product, for defining regional gradients and phenology), fire area coverage and plume dispersion, and surface moisture/wetlands delineation/drainage class.
Models will need to deliver outputs of hourly values for surface CO2 flux, GPP, NPP, (Net Ecosystem Exchange (NEE), and fluxes of CO2, CH4 and latent heat, respiration components, water and energy balances (hourly and daily), albedo, roughness, and so forth at about 10 x 10 km resolution, to function optimally with aircraft and meteorological data.
At least one-third of North America is topographically too complex for eddy flux towers or low-altitude aircraft measurements. These areas are generally forested with significant potential for sources or sinks of carbon. A multi-step procedure will be needed to derive NEE values for these regions, lacking flux measurements. Remote-sensing data will provide land cover and weekly GPP (MODIS). Daily surface meteorology from the National Weather Service can be extrapolated using topo-climatology principles, such as elevational lapse rates and aspect to map the surface microclimate. Then, photosynthesis, autotrophic and heterotrophic respiration can be computed. Model-based estimates of NEE in the mountains may best be tested using gauged watersheds to estimate hydrologic fluxes, and biomass inventories to validate carbon fluxes. This procedure is not as direct a test of surface NEE as are flux data, but can provide complete and consistent NEE estimates for mountain areas suitable for deriving regional and continental flux data from NACP atmospheric measurements.
Complete Carbon Accounting. Carbon may be transported in or out of an analysis region through erosion/sedimentation or product harvest (crops, timber). Hence, atmospheric estimates must be carefully matched with complementary accounting for fluxes on the land. Also, deposition or mobilization of inorganic C in soil as carbonates can play a significant role, especially in arid and semi-arid soils, and must be considered in C transfers between land, atmosphere, and oceans.
Data Management. Many of the required data streams exist today, but are not produced consistently at the time/space resolution needed, and the data are not assembled into an integrated set for data fusion. Because of the diversity of data and multiple temporal and spatial scales, it will be a significant challenge to make these data available for data assimilation activities and for public use. Hence, enabling activities are needed in this area.
|Oceans: Measurements and Models of Marine Carbon Fluxes|
The Ocean and the North American Carbon Cycle
The oceans absorb half of the 4 to 5 Pg C sequestered annually from the atmosphere. The location and year-to-year variations of marine carbon uptake are uncertain. Carbon exchange between the atmosphere and ocean is caused by a variety of physical effects and ecosystem responses to weather, climate, and land-ocean interactions. Large-scale climate shifts (e.g., El Niño, the Pacific Decadal Oscillation and North Atlantic Oscillation) cause variability in regional fluxes and distributions of CO2 in waters surrounding North America, with potentially important implications for efforts to measure net carbon fluxes for the continent.
Ocean programs generally focus on open ocean processes, thus missing the CO2 exchange along the ocean margins that can affect the CO2 content of air entering or leaving North America. Coastal upwelling and biological production rates are high in these regions, which also receive large carbon fluxes from rivers. In addition, a large fraction of the ocean's surface waters may acquire the chemical and biological characteristics that control net CO2 exchange via margin processes. Thus, the influence of nearshore processes may extend beyond the geographic boundaries of ocean margins. The North American Carbon Program therefore requires marine observations and diagnostic models focused on understanding the role of coastal systems on adjacent ocean basins and on atmospheric CO2 distributions.
Climatic perturbations may affect the coastal ocean quite differently than the open ocean. For example, when sea surface temperatures in the open ocean increase in an ENSO (El Niño-Southern Oscillation) event, CO2 evasion may increase due to higher surface water pCO2. But ENSO events can sharply decrease upwelling in coastal regions, reducing CO2 out-gassing. Meteorological and long-term (land use) influences on runoff have dramatic impacts on nutrient inputs and the export flux of carbon in coastal margins. Runoff changes resulting from ENSO events may also change the partitioning of material discharged to the ocean, and impact fluxes of DOC, POC, and DIC to the ocean. Because of the sensitivity to changes in winds, river runoff, and anthropogenic inputs of nutrients, the CO2 fluxes in nearshore waters will likely respond strongly to climate change.
The Marine Component of the NACP
The oceans component of the NACP is designed to leverage the impressive suite of existing and developing marine programs on the carbon cycle (see Appendix 3 for a summary), to define the role of the oceanic regions bordering North America. The main objective will be to provide information on processes controlling seasonal and interannual air-sea CO2 fluxes within ocean margins and ocean basins adjacent to North America, to define the net effect of the marine system on the CO2 content of air exchanging with the continent. Basin-scale ocean flux balances are also critical for placing the North American continent in the Northern Hemisphere context. Quantification of these processes is necessary to understand and interpret the large-scale regional and continental CO2 flux estimates that will be obtained during the intensive field experiment in 2004-2005. Plans for ocean research to address global issues, summarized in the U.S. Large Scale CO2 Observation Plan (LSCOP, Bender et al., 2001), and other programs (see Appendix 3) will contribute to the NACP.
Goals of the marine component of the NACP are as follows:
The NACP ocean carbon component, summarized in Appendix 3, includes both the open ocean domain and the coastal ocean domain; it will coordinate the efforts of ongoing programs into an integrated observing system, and undertake selected efforts not proceeding in other programs.
Open Ocean Domain. Characterization of the air-sea fluxes in oceanic regions bordering North America is critical for isolating processes related to the study region. A long-term basin-scale observation network of underway and time-series measurements, as laid out in the LSCOP (Bender et al., 2001), will provide the measurements necessary to constrain the boundary conditions in the NACP. The NACP portion of the open ocean domain will include expansion of surface ocean transects across the North Atlantic and North Pacific.
One of the objectives of the open ocean domain component of the NACP is to better characterize the spatial and temporal variability of air-sea fluxes in the North Pacific and North Atlantic. Currently we have a reasonably good understanding of the global-scale sources and sinks of CO2 in the oceans based on the sea-surface pCO2 climatology developed by Takahashi and coworkers (2001). However, there is still very little information on temporal variations of CO2 sources and sinks. The open ocean measurements will improve this constraint for the NACP. They will also help place the NACP results in a more global context by monitoring changes in air-sea CO2 gradients in the remote North Pacific and North Atlantic that correlate with observed seasonal and interannual changes in the net North American uptake.
Coastal Ocean Domain. Continental margins are particularly important for the NACP. Specific objectives of new ocean margins studies are better estimates of air-sea fluxes of CO2 and carbon burial and export to the open ocean, elucidation of factors controlling the efficiency of the solubility and biological pumps in coastal environments, quantification of the influence of margin biogeochemical processes on the chemical composition of open ocean surface waters, and development of coupled physical-biogeochemical models for different types of continental margins. River-dominated margins and coastal upwelling regions merit special attention due to their dominant role in coastal C budgets (see Appendix 3, Figure A3.1).
The NACP coastal program (Appendix 3) will include (1) long-term observations using coastal transects and buoys with autonomous sensors, and (2) intensive process studies. The long-term observations will be coordinated with aircraft profiles and coastal terrestrial study sites to provide the most complete picture possible at these sites. The long-term sites will also be coordinated with the anticipated location of the process studies to better characterize the dominant controls on the observed CO2 signals.
Ocean Carbon Modeling. Tracking changes in organic and inorganic carbon pools in coastal and open oceans requires detailed understanding of ecosystem dynamics, interlinking biogeochemical cycles, and oceanic physical circulation. Accurate determination of air-sea CO2 fluxes requires an understanding of such processes as upwelling, primary production, physical and biological transport, remineralization, and sedimentation. The NACP modeling effort will be designed to assimilate process study information and estimate regional sources and sinks for carbon. Integration of such a wealth of information will be a formidable task, but it is envisioned that developing a cohesive ocean carbon program within the framework of the Carbon Cycle Science Plan will act as a first step in such an effort.
Quantification of coastal and open-ocean carbon fluxes will involve a hierarchical approach, with widely distributed in situ observations, remote sensing, and modeling efforts. The field experiments will provide a foundation for satellite and model-based interpolations of oceanic CO2 fluxes over a broad range of temporal and spatial scales. The modeling program will be designed to assimilate field results and determine regional CO2 sources and sinks, providing the data for oceanic CO2 fluxes required as constraints for determination of continental fluxes.
|Data Analysis and Modeling: Data Fusion for Atmospheric, Land Surface, and Ocean Observations|
Models and data must be tightly integrated to identify uncertainties and strategies for measurement programs (prognostic models), and to analyze the observations to yield quantitative results and understanding (diagnostic models). Models are the link between processes and observations. They provide a quantitative representation of the physical processes (atmospheric, oceanic) and biological processes (ecological) that together regulate surface fluxes (sources and sinks) and the atmospheric responses to surface fluxes. Thus, model results can be compared to observations of the atmosphere, to satellite data, and to large assemblies of data such as the FIA or land cover maps. Models allow evaluation of the contributions of various mechanisms to the regional flux. Ultimately, predictive models of the future behavior of the carbon cycle will be developed, tested, and improved following systematic comparison to data collected by the NACP and related programs.
The atmospheric measurements collected by the NACP will provide a number of independent means for analyzing the budgets of major carbon gases, divided into two general approaches:
Both approaches require that models and data sets for atmospheric transport and for biogeochemical processes be combined and synthesized.
The NACP envisions development and application of a suite of diagnostic models and data assimilation techniques for a range of space and time scales, to obtain improved process understanding and quantitative estimates of carbon flux. A phased approach is planned. Research in data assimilation modeling and numerical experiments to test new concepts occupies the initial phase. Mass-balance and assimilation methods will be employed to estimate regional fluxes during intensive observing periods in the development and testing phase. The required data assimilation system will be built for the installation and operational phase, providing regionally resolved estimates of trace gas fluxes and improved model parameters for prediction of future trends, each with quantified uncertainty.
Most significant sources and sinks of carbon arise from “slow” processes: climate trends, recovery from disturbance and regrowth, the long-term rise in atmospheric CO2, changes in water tables, nitrogen deposition, and the long-term effects of management (fire suppression, tillage, forestry). However, most of the variance in atmosphere concentrations results from “fast” processes: the diurnal cycle, daily weather, and seasonal variability in climate. This mismatch defines the fundamental problem in defining the long-term budgets of the major C gases. Models used for analysis of data, either top-down or bottom-up, must simulate both fast and slow patterns of variability; measurements and analysis tools must be sufficiently comprehensive and accurate to delineate the long-term trends against the “noise” of fast processes. This is the challenge of the NACP.
The NACP envisions two approaches to this problem: (1) mass-balance and atmospheric data assimilation methods focus on quantifying fast processes (time scales of hours to months), with slow ecosystem changes deduced from residuals versus time-mean behavior; and (2) fully coupled process models allow direct assimilation of both fast and slow processes. Both approaches require substantial investment in improved capability for numerical simulation of atmospheric transport and for evaluation of the transport properties of the models, which are derived from weather forecasting. This investment is the necessary complement to that recommended for the observations.
|Forward Simulation of Atmospheric Concentrations|
Process models that simulate the carbon balance of terrestrial ecosystems and the ocean mixed layer will be coupled to meteorological models to provide detailed “forecasts” of atmospheric CO2, CH4, CO, and other trace gases. The models must be structured to allow direct, detailed comparison with data, thus to identify shortcomings in the simulation of the carbon cycle and the meteorology. High-frequency sampling of the continental atmosphere in the NACP will provide new information on much finer spatial scales than at present, and on diurnal to synoptic time scales. Hence model-data comparisons must represent fast ecophysiological processes and their interaction with atmospheric transport. Correlations (e.g., diurnal) between carbon fluxes and atmospheric transport (the “rectifier effect”) remain a major source of uncertainty in global inverse modeling studies. Correct representation of these processes in coupled models is necessary for robust interpretation of global data. Observational tests of these models, including correlations among processes at all time scales, represent a central goal of regional atmospheric sampling and analysis in the NACP.
Evaluation of forward simulations of atmospheric CO2, CH4, and CO will make use of both statistical analyses of data from the long-term network and case studies of data collected during intensive observing periods. Statistical studies can often use large-scale transport calculated using winds and sub-grid-scale parameterized mass fluxes produced by Numerical Weather Prediction (NWP) centers (NCEP, ECMWF, NASA/DAO). Global analyses are currently available on grids of 1° x 1° (latitude x longitude) every 6 hours, expected to be 0.5o x 0.5o, or even 0.25° x 0.25° by 2006 (see Appendix 4a on NASA DAO proposed activities). Case studies, especially for intensive operation period (IOP) data from aircraft, may require simulation of small-scale circulation features down to the scale of the PBL, for which mesoscale models will be needed to interpolate to finer resolution at higher temporal frequency (e.g., RAMS, ETA). Accurate trajectories with fine spatial and temporal detail from nested models will also be extremely valuable when using aircraft data from the long-term network to constrain models of carbon and other trace gas fluxes.
Most mesoscale atmospheric models in current use can be run from analyzed meteorological fields at larger scale and are thus already available for use to support field campaigns. Few, if any, are coupled to ecosystem models that predict spatial and temporal variations of photosynthesis and respiration. Intensive field campaigns such as those envisioned for the NACP motivate development of these models, and the results will provide excellent tests. The models will be applied to simulate the coupled interaction of weather, hydrology, and biogeochemistry at scales of 10 km or finer over the whole of North America for periods of up to a year, and at much finer scales for IOPs. These simulations will be evaluated rigorously against data from surface, airborne, and spaceborne platforms, providing new understanding, algorithm development, and code for the coupled global assimilation models envisioned below. These coupled meteorological-biophysical models should develop in parallel with the observational network for the NACP.
It is crucial that the full three-dimensional mass fluxes, including those resulting from parameterized (unresolved) processes like cumulus convection and turbulent entrainment, be archived by the NWP centers for use in global and regional transport calculations, and that techniques be developed to ensure that archived analyzed wind fields conserve mass. These issues have been major stumbling blocks for attempts to derive surface fluxes from measurements of atmospheric tracers. A few analyzed fields (NASA DAO) currently include sub-grid-mass fluxes, but most do not. To represent the interaction of diurnal cycles in convection and turbulence with variations in surface fluxes of carbon, energy, and water, models will have to provide fields every 3 hours or less, instead of the current practice of reporting analyses every 6 hours. Additional vertical resolution may be needed near the surface, to capture both ventilation and PBL top entrainment processes. Many model outputs are regridded from dynamic height variables (e.g., sigma coordinates) to conventional coordinates, requiring exquisite attention to conservation of mass in the derived products. These improvements pose major challenges for global models.
Improved inventories of anthropogenic emissions, with higher spatial resolution and tested algorithms for temporal disaggregation (time of day, day of week, season) are needed for the detailed analyses envisioned here (both forward and inverse) (see Appendix 4b). The heterogeneous distribution of fossil fuel sources, with localized emissions in urban areas, leads to significant variance of atmospheric CO2 concentrations, complicating interpretation of tracer data. CO also has large fossil fuel contributions, and analysis of the CO observations will allow rigorous testing of emission inventories and transport, especially during intensive field studies. Thus, CO serves to distinguish fossil fuel signals from other sources of variance for the major carbon gases, allowing quantification of ecosystem fluxes and other non-fossil influences.
|Inverse Modeling and Data Assimilation|
Spatial and temporal variations in trace gas concentrations contain information about surface fluxes and the processes that produced them, which can be quantitatively extracted using a family of methods collectively referred to as inverse modeling. Fluxes can be estimated at local to regional scale for short periods by direct mass-balance techniques, in which airborne measurements are used to calculate the time rate of change of concentration in an airmass or the horizontal flux divergence of tracer in a control volume of air. Spatial patterns of concentration can be related to time mean fluxes using methods known as synthesis inversion techniques. Variational data assimilation, Kalman filter, and other estimation methods have been proven for weather analysis and forecasting. These methods allow information to be combined from essentially incomparable data, such as NDVI and wind fields, or high-frequency variations in trace gas concentrations and long-term observations. They provide formal mechanisms to merge disparate data streams (e.g., from flux towers, daily aircraft ascents, satellite imagery) to estimate model parameters.
Mass Balance Methods. The most direct approach to estimating carbon fluxes is by balancing mass flows for either a fixed volume of air (Eulerian frame) or a volume of air that follows the atmospheric flow (Lagrangian frame), both of which will be used to prepare for data assimilation at a later stage. Lagrangian approaches are best suited for intensive campaigns, especially in regional experiments, where aircraft flights sample the same airmass at multiple times. This approach has been developed and successfully applied in the COBRA regional experiments (Gerbig et al., 2001, Lin et al., 2001). The method provides a direct and precise measurement of surface fluxes at regional scale, as it eliminates horizontal advection terms present in budgets conducted in a non-airmass-following framework. Eulerian approaches are more readily applied to estimate fluxes over North America from the long-term network, with many observation points distributed discretely across the continent.
Synthesis Inversion. Estimates of surface CO2 fluxes from atmospheric concentrations by “synthesis inversion” are typically completed in two steps. In the first step, forward simulations are carried out for prescribed surface sources and sinks over large regions (e.g., temperate North America in July). In the second (inverse) step, the magnitudes of unknown surface fluxes are estimated by fitting the forward predictions to atmospheric observations. Variations in the prescribed fluxes within the large regions must be specified from prior knowledge (e.g., fossil fuel combustion patterns and satellite vegetation imagery), due to the sparse observations available. An analogous procedure may be carried out using a time-reversed Lagrangian approach, where winds are run backward (an adjoint model) and the influence of ground sources is estimated. These techniques allow the quantitative insertion of data from other sources (e.g., emission inventories, satellite vegetation data) and their uncertainties, and produce final uncertainty bounds along with flux estimates.
Variational and Kalman Filter Methods. Variational and Kalman filter assimilation methods use estimation theory to optimize a set of parameters to time-varying data, minimizing a cost function whose derivatives relative to observed quantities are known. These techniques form the basis of modern weather analysis and forecasting, and allow maximum flexibility with respect to the temporal and spatial filtering of the observations. Thus, high-time-frequency variations in concentration (e.g., that result from passing synoptic weather disturbances) can add large amounts of information to the flux estimation, rather than being treated as “noise.”
The wealth of observations envisioned from the NACP will provide much tighter constraints on inverse calculations than have been possible previously. Given a complete archive of both resolved and unresolved transport, the generation of the adjoint of a transport model is straightforward. This approach estimates fluxes at the resolution of the gridded meteorological analysis, which are then aggregated to coarser scales according to the degree of data constraint available. Systematic, regionally coherent offsets between forward and inverse results will allow testing key ideas about factors that underlie slowly varying ecological processes, such as forest regrowth and woody encroachment.
The NACP will initiate the development of formal data assimilation methods to use comprehensive observations relevant to CO2, CH4, and CO. Data will cover atmospheric composition from flask collections, tall towers and buoys with continuous samplers, instrumented aircraft with continuous profiles, upward-looking spectrometers measuring column amounts, and satellites with global coverage. Measurements from satellites and at the surface will provide data on the state of soils, vegetation (including biomass), and ocean biota. Surface fluxes will be measured at instrumented towers by eddy covariance. Inventories will define changes in biomass, agricultural productivity, and soil carbon over long time scales. Buoys, moorings and ships at sea will define air-sea fluxes.
Simple process-based descriptions of photosynthesis, ecosystem respiration (or methanogenesis, for CH4), growth, and air-sea gas exchange will be coupled to the atmospheric transport model, and the adjoint of the coupled model developed. A generalized cost function will then be minimized, allowing estimation of key parameters in the carbon process models, rather than area-averaged surface fluxes as for synthesis inversion. The system provides best-fit values for parameters in the underlying biophysical models that describe processes responsible for the fluxes, such as temperature--moisture sensitivity of soil respiration, wind-speed dependence of the air-sea gas exchange coefficient, photosynthetic capacity of forest canopies, and seasonal or annual imbalances in mass flows to longer lived pools of organic matter. In this way, assimilation into global coupled models not only provides time-resolved maps of surface carbon exchange, but also leads to progressive improvement in the predictive capability of the process models. This approach has already been demonstrated using simple atmosphere-land biosphere models (Wang and Barrett, 2001; Rayner et al., 2001), which show quantitative uncertainty reduction for estimates of both regional flux and model parameters.
A fully coupled variational data assimilation system that combines meteorological analyses with carbon cycle process models, simultaneously constrained by meteorological as well as carbon data, is a long-range goal of the NACP. That effort will require participation by one or more operational NWP centers, because substantial computational, data-handling, and human resources are required. Real-time and reanalysis products will both be needed because some observations (e.g., flask sample analyses, satellite data on state of the vegetation) take weeks to obtain.
Assimilation of atmospheric CO2, CH4, and CO data and spaceborne data directly into the analysis of an operational numerical weather forecast model would be an optimal way to perform the data assimilation task envisioned as a long-term goal. Tracer concentrations carried as prognostic variables in the assimilation provide strong constraints from the “memory” in the atmosphere and reduce artificial noise from poorly constrained satellite retrievals. Assimilated concentrations of CO2 and other tracers in operational NWP models would produce fully populated, global gridded data, filling the gaps left by clouds, yet remaining optimally consistent with existing observations. The underlying source fields, derived by the models to be consistent with observations, provide estimates of emission or uptake rates gases.
Coupled meteorological, biogeochemical, and trace gas data assimilation will contribute better weather forecasts and climate models. Comparison of simulated trace gas data with observations will help expose model flaws. Knowledge of atmospheric CO2 concentrations has recently been shown to improve the retrieval of temperature profiles from infrared spectroscopy, reducing forecast initialization error by as much as 1°K over some regions (Engelen et al., 2001).
This program is technically feasible. Implementation could start as early as 2002, when CO2 estimates begin to be available from the Atmospheric Infrared Sounder (AIRS) aboard EOS-Aqua. Indeed, ECMWF has already begun such a development effort, with the aim of real-time assimilation of both atmospheric CO2 and its surface sources and sinks within three years. NASA DAO may pursue a similar effort. New resources would be required, with potentially large scientific return, especially later in the decade when higher quality global satellite products are expected. A workshop involving operational weather centers and carbon modelers is planned to examine performance and resource requirements, and to begin work on an implementation plan.
An important function for early versions of coupled models is analysis of bias in trace gas data from satellite retrievals, requiring substantial field sampling programs and sophisticated methods for comparing in situ and remote-sensing data from the sustained and intensive campaigns (elements 1a and 1b). We will be able to determine bias associated with sensors relying on reflected sunlight, which see only daytime conditions and thus have systematically lower CO2 concentrations over vegetated land than the true mean. Similarly, spaceborne observations of CO2 concentration will be biased to clear sky conditions.
A study to design the NACP long-term observational network is one of the highest priorities for initial model development. Previous atmospheric and biospheric data and the atmospheric reanalysis data (see Appendix 4a) can be used with the new models to assess network designs, intensive campaigns, and other NACP concepts before implementation. A first step in this effort is a modeling study/workshop focusing on network design scheduled for summer 2002.
By the end of the decade, the goal should be an assimilation system that includes a model treating the fluid dynamics and physics of the atmosphere and oceans and the biology and biogeochemistry embedded in each. This coupled model would predict many quantities that are directly observable (including temperatures, winds, and radiances at the top of the atmosphere that result from radiative interactions with vegetation, phytoplankton, and atmospheric trace gases such as CO2 and CO). The system would then minimize a generalized cost function that includes deviations of each of the predicted quantities from the observations, to include observations at the surface, by automated in situ sensors, and from space, enabling near-real-time analysis of the elements of the carbon cycle on land and in the oceans, and of the processes that give rise to sources and sinks.
This new data fusion system will be invaluable for monitoring present conditions and for learning about the coupled Earth system. Most important, it would enable development of falsifiable predictive models for future behavior of the carbon cycle and the climate system. This is a very ambitious program that calls for substantial effort in computational and human resources, requiring significant resources in advance of the actual field observations.
Goals for atmospheric modeling and data assimilation support for the North American Carbon Program are the following developments:
|PDF Version of Report | Report Home Page | Report Table of Contents | Top of Page|