B42C-01 INVITED
Bridging across Spatial and Temporal Scales in North American Carbon Dioxide Flux Estimation through Geostatistical Analysis of Scale-Dependent Relationships Between Carbon Flux and Auxiliary Environmental Data
The first overriding question of the NACP Implementation Plan deals with diagnosis of North American carbon fluxes. One of the critical aspects for addressing this question focuses on "scaling issues inherent in applying data sets over areas that were not measured," including "scaling, in time and space, of the intensive flux measurements to the larger landscape" [Denning et al. 2005]. This need highlights the importance of understanding the scale-dependence of parameters controlling carbon flux variability, and developing methods for using data collected at multiple scales to infer carbon fluxes. Geostatistical kriging and inverse modeling tools offer a unique opportunity to identify, characterize, and quantify relationships between observed or inferred fluxes and a suite of auxiliary environmental variables. This presentation will describe recent work aimed at improving understanding of variables and processes controlling CO2 flux variability at various spatial and temporal scales, by using remote-sensing, in situ, and atmospheric data. The applied method is based on a geostatistical dual kriging analysis of flux variability applicable at AmeriFlux sites, and a geostatistical inverse modeling approach to flux estimation for the North American continent. This approach does not prescribe the effect of auxiliary data on flux variability, and does not involve an explicit biospheric model. Instead, these relationships are estimated based on the information content of available data. Overall, the approach can be used to identify parameters that explain observed or inferred flux variability at various spatial and temporal scales, by analyzing data collected at different scales within a single, consistent, and statistically rigorous framework. http://www.umich.edu/~amichala/
B42C-02 INVITED
Nested Global Inversion for the Carbon Flux Distribution in Canada and USA from 1994 to 2003
Based on TransCom inverse modeling for 22 global regions, we developed a nested global inversion system for estimating carbon fluxes of 30 regions in North America (2 of the 22 regions are divided into 30). Irregular boundaries of these 30 regions are delineated based on ecosystem types and provincial/state borders. Synthesis Bayesian inversion is conducted in monthly steps using CO2 concentration measurements at 88 coastal and continental stations of the globe for the 1994-2003 period (NOAA GlobalView database). Responses of these stations to carbon fluxes from the 50 regions are simulated using the transport model of National Institute for Environmental Studies of Japan and reanalysis wind fields of the National Centers for Environmental Prediction (NCEP). Terrestrial carbon flux fields modeled using BEPS and Biome-BGC driven by NCEP reanalysis meteorological data are used as two different a priori to constrain the inversion. The inversion (top- down) results are compared with remote sensing-based ecosystem modeling (bottom-up) results in Canada's forests and wetlands. There is a broad consistency in the spatial pattern of the carbon source and sink distributions obtained using these two independent methods. Both sets of results also indicate that Canada's forests and wetlands are carbon sinks in 1994-2003, but the top-down method produces consistently larger sinks than the bottom-up results. Reasons for this discrepancy may lie in both methods, and several issues are identified for further investigation.
B42C-03
A Pre-NACP Assessment of the North American Terrestrial Carbon Sink
A pre-North American Carbon Program (pre-NACP) evaluation of North American carbon sources and sinks was generated as part of the U.S. Climate Change Science Program's Synthesis and Assessment Product 2.2 The First State of the Carbon Cycle Report (SOCCR): North American Carbon Budget and Implications for the Global Carbon Cycle. The report found that North America is currently a net source of carbon dioxide to the atmosphere, contributing to the global buildup of greenhouse gases in the atmosphere and associated changes in the earth's climate. North America's fossil-fuel emissions in 2003 (1856 million metric tons of carbon ±10% with 95% certainty) were 27% of global emissions. Approximately 85% of those emissions were from the United States, 9% from Canada and 6% from Mexico. The conversion of fossil fuels to energy (primarily electricity) is the single largest contributor, accounting for approximately 42% of North American fossil emissions in 2003. Transportation is the second largest, accounting for 31% of total emissions. Circa 2003, growing vegetation in North America removed approximately 500 million tons of carbon per year (± 50%) from the atmosphere and stored it as plant material and soil organic matter. This land sink is equivalent to approximately 27% of the fossil-fuel emissions from North America. The imbalance between the fossil-fuel source and the sink on land (a source:sink ratio of nearly 4:1) is a net release to the atmosphere of 1350 million metric tons of carbon per year (± 25%). Approximately 50% of North America's terrestrial sink is due to the regrowth of forests in the United States on former agricultural land that was last cultivated decades ago, and on timber land recovering from harvest. The second largest sink (approximately 25% of the North American sink) is associated with woody encroachment into grasslands and shrublands but is highly uncertain, known to no better than ± 100%, and may actually represent a small source. Other sinks are relatively small (individually 1-10% of the continental sink) and not well quantified with uncertainties of ± 100% or more. The future of the North American terrestrial sink is also highly uncertain. The contribution of forest regrowth is expected to decline as the maturing forests grow more slowly and take up less carbon dioxide from the atmosphere. But, this expectation is confounded by uncertainty in how regrowing forests and other sinks will respond to changes in climate and elevated atmospheric carbon dioxide concentrations. http://cdiac.ornl.gov/SOCCR
B42C-04
Modeling Carbon Cycles for the Western United States using NASA Satellite Products: Focus on Climate and Land Use Change Impacts
Satellite remote sensing and vegetation-soil predictions from the NASA-CASA model were used in this study to estimate the past and future carbon balance for ecosystems in the western United States. We report on spatially detailed (< 10 km resolution) terrestrial carbon budgets for ecosystems of the Rocky Mountain and Pacific regions of the county. Although net primary production (NPP) was estimated to increase on a western region-wide basis during the 1990s, the total terrestrial sink in all western U.S. ecosystems did not exceed 0.01 Pg C per year between 1982 and 1997 and continues to recover slowly from drought impacts up to 2006. Forested mountain areas of the Cascades, the Sierra Nevada Range, the northern California Coast Range, and the southern Rockies were estimated as the most consistent ecosystem carbon sinks areas within the region. Future climate scenarios imply that major ecosystem carbon losses in the western United States will be experienced in all but the most isolated forest areas of the high mountain zones. http://geo.arc.nasa.gov/sge/casa/
B42C-05
Initial Results From an Integrated Terrestrial Carbon Model for North America: Constraining Process Models with Experiments and Measurements
To develop finer temporal and spatial resolution terrestrial carbon cycle simulation models must incorporate understanding from experiments, be constrained and validated by observations, and provide robust methods for their extrapolation. The Integrated Terrestrial Carbon Model (ITCM) develops an operational framework in which different aspects of terrestrial carbon processes are integrated to estimate and mechanistically explain current carbon sources and sinks and forecast their future behavior and influence on atmospheric CO2 concentration and climate. Preliminary runs for North America with a one-degree spatial grid and hourly temporal resolution show features that have important implications for the NACP. North America simulations show strong and distinct patterns of CO2 dynamics based on weather patterns associated with ENSO. The cause of this result can be explained by examining the relative response of photosynthesis and heterotrophic soil respiration. Simulations indicate that the large increase in NEP from 1991 to 1992 may be attributed to a reduced rate of decomposition in New England and Great Lakes regions. These preliminary results offer hypotheses to be examined and tested with model-data assimilation at site and regional scales. These model-based bottom-up analyses also need to be confronted with top-down inferences based on fine spatial and temporal resolution atmospheric CO2 concentration measurements.
B42C-06
Applications of Lagrangian Particle Transport Modeling in the Top-Down Regional CO2 Studies
Atmospheric transport plays a critical role in top-down studies where observations from towers and/or aircraft are inverted to estimate net sources and sinks of CO2 for the study area over short periods of time. Lagrangian particle dispersion models are well suited for this modeling task since they 1) can be easily linked to any regional scale meteorological model, 2) can be run both forward or backward in time (in an adjoint mode), 3) can accurately resolve any CO2 observational system without limits of gridded transport models, and 4) can be applied to different spatial scales even across grids or domains of meteorological models. In the modeling framework developed at CSU, the Lagrangian Particle Dispersion Model is linked to SiB-RAMS: Regional Atmospheric Modeling System combined with Simple Biosphere model. For our North America studies the SiB-RAMS domain extends over the entire continental US with nested grids centered in the mesoscale area of interest. The CO2 lateral boundary conditions are provided by a global transport model - PCTM (Parameterized Chemistry and Transport Model). Influence functions derived from the LPDM output allow us to quantify each CO2 data point (e.g., concentration at a specific sampling time and tower) in terms of contributions from different sources: 1) surface fluxes, 2) inflow fluxes across domain boundaries and 2) initial CO2 concentration in the domain at the beginning of the analysis period. The surface contributions can be furher quantified by a physical process (respiration, assimilation or fossil fuel emission) and/or land cover type. Therefore, the influence function approach is very useful for interpretation of CO2 observations and source apportionment, designing tower network and, finally, deriving source-receptor information for the inverse studies. We are going to review our modeling efforts based on the SiB-RAMS/ LPDM and the influence function approach to the meso- regional scales from a few tens to several thousands of kilometers: 1) ~10km - quantifying both CO2 concentration and flux measurements from real and hypothetical towers in the Tapajos River region in the Amazon using very high resolution SiB-RAMS simulations, 2) ~300km - estimation of mesoscale CO2 fluxes using the summer 2004 observations from the "ring of towers" in northern Wisconsin, 3) ~600km - extension of the previous work to a larger domain of the second "ring of tower" run in summer 2007 within the NACP's Midcontinental Intensive Study (preliminary pseudo data inversion experiments), 4) ~5000km - deriving influence functions and transport characteristics for the US continental scale CO2 inversions.
B42C-07
Carbon Fluxes over North America during 2001-2005: A Comparison between Land- and Atmosphere-based Approaches
Carbon fluxes inferred from atmospheric inversion experiments are generally conditioned on prior knowledge of biospheric fluxes, yet the influence of the latter on the inversion results is rarely assessed. As an initial step to address this question, this study compares two independent datasets: CarbonTracker NEE (net ecosystem exchange) and MODIS GPP (gross primary production) over North America between 2001 and 2005. CarbonTracker NEE (Peters et al. 2007) is a newly released product based on inversion of atmospheric measurements and MODIS GPP (Zhao et al. 2005) is based on forward modeling and satellite observations. Annual mean MODIS GPP and CarbonTracker NEE for the continent are about 13.8 (PgC/yr) and -0.65 (PgC/yr, negative sign indicates a carbon sink), respectively, consistent with previous studies; their monthly anomalies are correlated significantly ( r=-0.57, p<0.01), and also appear to be correlated with variations of ENSO (El Nino- Southern Oscillation) indices during the period, suggesting the impact of large-scale climate variability on regional ecosystems. Despite the overall agreement, however, the spatial patterns of the two datasets show remarkable differences. In particular, while MODIS shows the most substantial GPP over forest ecosystems, CarbonTracker suggests that a large proportion (30%) of the estimated carbon sink is located in the agricultural regions of the Midwest, which also contribute importantly to the interannual variability of NEE. Regression analyses suggest that the particular spatial pattern of CarbonTracker is influenced by the biospheric fluxes used in the inversion algorithm. Therefore, ensemble experiments with different prior biosphere fluxes may be necessary to verify the robustness of the spatial distribution of carbon sinks and sources estimated by atmospheric inversions.