B53B-1168
The Use of a Mesoscale Climate Model to Validate the Nocturnal Carbon Flux over a Forested Site
The Savannah River National Laboratory is initiating a comprehensive carbon dioxide monitoring and modeling program in collaboration with the University of Georgia and the Brookhaven National Laboratory. One of the primary goals is to study the dynamics of carbon dioxide in the stable nocturnal boundary layer (NBL) over a forested area of the Savannah River Site in southwest South Carolina. In the nocturnal boundary layer (NBL), eddy flux correlation is less effective in determining the release of CO2 due to respiration. Theoretically, however, the flux can be inferred by measuring the build up of CO2 in the stable layer throughout the night. This method of monitoring the flux will be validated and studied in more detail with both observations and the results of a high-resolution regional climate model. The experiment will involve two phases. First, an artificial tracer will be released into the forest boundary layer and observed through an array of sensors and at a flux tower. The event will be simulated with the RAMS climate model run at very high resolution. Ideally, the tracer will remain trapped within the stable layer and accumulate at rates which will allow us to infer the release rate, and this should compare well to the actual release rate. If an unknown mechanism allows the tracer to escape, the model simulation would be used to reveal it. In the second phase, carbon fluxes will be measured overnight through accumulation in the overlying layer. The RAMS model will be coupled with the SiB carbon model to simulate the nocturnal cycle of carbon dynamics, and this will be compared to the data collected during the night. As with the tracer study, the NBL method of flux measurement will be validated against the model. The RAMS-SiB coupled model has been run over the SRS at high-resolution to simulate the NBL, and results from simulations of both phases of the project will be presented.
B53B-1169
Cross-agency data collaboration for the North American Carbon Program
The central objective of the North American Carbon Program (NACP), a core element of the US Climate Change Science Program, is to quantify the sources and sinks of CO2, CO, and CH4 in North America and adjacent ocean regions. The NACP consists of a wide range of investigators at universities and federal research centers. Although many of these investigators have worked together in the past, many have had few prior interactions and may not know of similar work within knowledge domains, much less across the diversity of environments and scientific approaches in the Program. Thus, coordinating interactions and sharing data are major challenges in conducting NACP. The NACP website provides metadata describing each core and affiliated NACP project. These project-oriented metadata, or "project profiles", can store information such as a project's title, leaders, participants, an abstract, keywords, funding agencies, associated intensive field campaigns, expected data products, and URLs to data centers, datasets, and associated metadata. Examples of data products could include biometric inventories, flux tower estimates, remote sensing land cover products, tools, services, and model inputs / outputs. The NACP project profiles can provide the scientific and social context of each dataset and are an important means of communicating within the NACP and to the larger carbon cycle science community. http://www.nacarbon.org
B53B-1170
A cross-scale remote sensing approach to estimate tree cover and aboveground biomass in pinyon-juniper woodlands of the Colorado Plateau, USA
Vegetation dominated by pinyon pines and junipers (pinyon-juniper [P-J] woodlands) is one of the largest vegetation types in the North America. P-J woodlands maintain the highest level of woody biomass compared to other major dryland ecosystems. However, distributions of tree cover and biomass in the P-J woodlands of the Colorado Plateau have not been well studied. Here we developed a synoptic remote sensing approach to scale up pinyon pine and juniper cover and biomass field observations from plot to regional levels using fractional photosynthetic vegetation cover (PV) derived from airborne imaging spectroscopy and Landsat satellite data. Our results demonstrated strong correlations (p < 0.001) between field and airborne tree canopy cover estimates (r2 = 0.92), and between airborne and satellite canopy cover estimates (r2 = 0.61). Field data also indicated that P-J aboveground biomass can be estimated from canopy cover using a unified allometric equation (r2 = 0.69, p < 0.001). Using these multi-scale, cover-biomass relationships, we developed high-resolution, regional-scale maps of P-J cover and biomass for the western Colorado Plateau. The mean (± standard deviation) P-J cover was 27.4 (± 9.9)%, and the mean aboveground woody carbon (C) converted from biomass was 5.2 (± 2.0)MgC/ha. Combining our data with the southwest Regional Gap Analysis Program vegetation map, we estimated that total contemporary woody C storage for the entire Colorado Plateau P-J woodlands (113,600 km2) is 59 TgC. Our results facilitate further investigation of the processes controlling carbon stocks and fluxes across this large region, which forms a key component of the North American Carbon Program (NACP).
B53B-1171
A bottom-up geostatistical approach for quantifying landcover in Asian desert ecosystems and implications for global and climate models: a case study in Afghanistan utilizing a unique hyperspectral dataset
Political tensions, rough terrain, and remoteness have lead to a gap in the ecological understanding cold, mountainous deserts of Asia. Remote sensing is a time- and cost-efficient way to understand the spatial distribution and temporal dynamics of plant and snow cover in these regions. Here, a unique high-resolution hyperspectral dataset from Afghanistan is employed to classify ground cover at high resolution. The hyperspectral data was taken using a CASI-1500 Visible Near InfraRed (VNIR) spectrometer. The instrument was run in a mode with 1518 crosstrack pixels and 72 spectral bands between 380 and 1050 nm. The GSD was controlled by the altitude above ground level and aircraft speed, which varied resulting in GSD between 4 and 6 meters. Geolocation was provided by a CMIGITS II and the resulting accuracy will be better than 40 m. Atmospheric conditions were challenging and proper atmospheric compensation of the data remains a challenge. A bottom- up geostatistical approach for quantifying the coverage of vegetation and snow will be applied to establish the practical limits of coarse resolution MODIS data for classifying vegetation and snow cover, a scale suitable for monitoring large regions and for modeling. A Multiple Endmember Linear Spectral Mixture Algorithm (MESMA) will be applied to classify land cover. Semivariograms at the multispectral (30 m) and coarse resolution scale (1 km) will be compared with simulated variograms using hysperspectral data. Patches of vegetation and snow cover used for spatial comparison will be identified in the image and characterized using object-oriented image analysis software. The relative amount of cover will be determined using block-kriging and compared between scenes with statistical tests. Insight gained from this analysis can be applied to improve existing data products and can be applied for carbon budget and climate change models.
B53B-1172
Assessing the Influence of Fossil Fuel Emissions on CO2 Flux Measurements Above a Suburban Ecosystem Using Continuous Traffic Data
Cities are a major source of CO2, the most important anthropogenic greenhouse gas. Land use within cities is highly heterogeneous and a significant area can be occupied by vegetated surfaces where CO2 is taken up by photosynthesis and released by ecosystem respiration. Recent remote sensing and modeling studies have estimated that turfgrass lawns cover a surface area of perhaps 163,800~km2 in the continental United States with significant CO2 exchange (Milesi et al. 2005). However, direct measurements of land-atmosphere fluxes above lawn ecosystems have been difficult due to the typically small dimensions of lawns and the heterogeneity of land uses that surround them in an urbanized landscape. We made 2~years of continuous CO2 exchange measurements using a mobile eddy covariance tower over a <1~ha lawn, which was within the footprint of the KUOM 170-m tall flux tower in a suburban residential neighborhood of Minneapolis-St. Paul, Minnesota. A satellite-derived land-cover map was analyzed to assess the characteristic patch dimensions of lawns in the region in comparison to the selected mobile tower site. An important consequence of urban landscape heterogeneity is that CO2 fluxes measured above vegetated patches may be influenced by fossil fuel CO2 sources nearby. In this poster, we use high time resolution, continuously monitored traffic data from the roads surrounding the flux site to quantify the influence of fossil fuel emissions on the net ecosystem CO2 exchange measurements. Using a flux source area model, we assess the relative influence of fossil fuel emissions in relation to variations of wind field, atmospheric stability, and temporal patterns of traffic volume. The results will be useful for validating emissions models and for scaling up the CO2 flux from vegetation in developed land. This study is a contribution to the Mid-Continent Intensive Field Campaign of the North American Carbon Program (NACP).
B53B-1173
Objective refinements to a diagnostic terrestrial biosphere model using satellite data: North America carbon and water cycle simulations
We established a framework for objective improvement of a diagnostic terrestrial biosphere model (Terrestrial Observation and Prediction System; TOPS) using satellite-derived products including snow cover, evapotranspiration (ET), and gross primary productivity (GPP). Based on the TOPS model structure, we established an objective improvement process by first optimizing snow submodel, then soil water submodel, and finally gross primary production submodel. We used MODIS snow cover products (MOD10A2) for snow submodel improvements, Support Vector Machine (SVM) based ET estimation for soil water submodel improvements, and SVM-based GPP estimation for GPP model improvements as satellite-derived products. Snow submodel refinement has shown an improvement on snow dynamics, streamflow, ET, and GPP over high latitude areas. Soil water cycle submodel refinement has shown an improvement on seasonal ET and GPP variations for seasonally-dry regions. GPP submodel refinement has shown an improvement on seasonal and annual GPP over the vegetated regions. Our analysis shows that the objective improvement of terrestrial ecosystem model is an effective way for improving model performance. Carbon and water cycle simulation over North America were greatly improved as a result of the model improvements. http://ecocast.arc.nasa.gov
B53B-1174
Development of Regional CO2 Inversion for Terrestrial Sources/Sinks Using Genetic Algorithm
Global CO2 inversion is a well-developed technique for deducing a reasonable surface CO2 flux combination that is consistent with atmospheric CO2 observation. It provides a global wide distribution of CO2 sources/sinks, however, the obtained CO2 source/sink information is coarse for the use of local level policy making against global warming. Thus, finer spatial resolution information of source/sink is preferable. When CO2 inversion technique is applied to regional study, it is more difficult to relate surface flux to concentration properly because of difficulty in modeling of regional transport. And, especially in regional study, the concentration variation occurs in shorter time scale, which is usually removed as "noise" should be utilized because it may contain spatiotemporal information of regionally distributed sources/sinks. In this study, a regional CO2 inversion that employs Genetic Algorithm (GA) for the inverse method is proposed. It enables us to utilize the concentration variations as signals to detect regionally distributed CO2 sources and sinks. Our framework consists of atmospheric CO2 observation, modeling of atmospheric transport and an optimization technique. In practical, unknown parameters corresponding to fluxes are optimized by matching the modeled concentration with the observed concentration. The flux estimates are calculated using the optimized parameters. The hourly atmospheric CO2 concentration was measured using an infra-red gas analyzer at an observation point located in the target domain and the modeled concentration at the observation point was calculated using our regional transport model driven by hourly meteorological fields from MM5. Genetic Algorithm (GA) is a universal optimization technique that searches within the solution space globally while avoiding local optima. It generates candidate solutions (i.e. combinations of parameters) that result in smaller misfit between the observed and modeled concentration. A surface flux estimate of August 2005 over a region of 126 km x 126 km was obtained using 10-day complete CO2 dataset (August 18-27; 240 hour). The study area is located in the centre of the main island in Japan and we assumed it contained four flux classes including open water, urban area, cropland and forest according to land use data. The surface fluxes of vegetation were prescribed using the modeled short wave radiation while the fluxes of urban and open water were assumed to be constant. Since the obtained flux estimates in our framework have variability, several estimation runs are required. For ensuring our flux estimation, we studied required number of estimation runs using reduced version of our transport model. In reduced model, the spatial resolution was degraded and the time period for assimilating the observation was shortened for the simplification and reduction of computational time of our estimation. In addition to the above, the footprint of the observation was studied.
B53B-1175
Analyzing the Impact of High-Resolution Fossil Fuel Emissions on Atmospheric CO2 Concentrations
Using atmospheric tracer transport models, inverse modelers can estimate the strengths and spatial distribution of carbon sources and sinks; however, fossil fuel CO2 emissions must be accurately estimated to isolate and quantify biospheric and oceanic fluxes. To help achieve unbiased estimates of carbon exchange, high-resolution fossil fuel emissions estimates (10 km spatial resolution at timescales of one hour) are being produced for the United States [Gurney et al., 2006]. To evaluate these emissions estimates and to investigate the impact of high- resolution emissions on the resulting atmospheric CO2 and CO concentration fields, we will use the coupled biosphere-atmosphere model SiB3-RAMS. Simulations across large urban airsheds will be evaluated against both CO2 and CO observations.
B53B-1176
Reconstruction Of Air-Sea Fluxes And Meridional Transport Rates Of Anthropogenic Carbon With An Ensemble Kalman Filter Data Assimilation
Regional air-sea fluxes and meridional transport of anthropogenic carbon are inferred by assimilating anthropogenic carbon concentrations within the ocean from different data-based reconstructions. An inverse, Ensemble Kalman Filter method with the Bern3D ocean model is applied. The Bern3D model (Müller et al., 2006) is a computationally-efficient, 3-dimensional coarse resolution ocean model. The Ensemble Kalman Filter (Evenson, 2003) is suited for the assimilation of spatially and temporally varying data into a range of models, for model tuning or for model initialization. Regional fluxes through the air-sea interface and meridional transport rates in the ocean are determined by minimizing deviations between the distributions of anthropogenic carbon from the GLODAP database (Key et al., 2004) and from the Bern3D ocean model in the Ensemble Kalman Filtering optimzation. The resulting anthropogenic carbon fluxes are in agreement with those from another ocean inversion study using the same GLODAP data (Mikaloff Fletcher et al., 2006). Transport uncertainties are addressed by utilizing different configuration of the Bern3D model. The inferred transport uncertainties are comparable in magnitude to the uncertainties obtained by Mikaloff Fletcher et al. The fields of anthropogenic carbon reconstructed with six different reconstruction methods: CFC-shortcut (Thomas et al., 2001), C-star (Gruber et al. 1996), IPSL (Lo Monaco et al., 2005), PHI-CT (Vazquez Rodriguez et al, submitted), TrOCA (Touratier et al., 2004), and TTD (Waugh et al., 2006) from four sections in the Atlantic are assimilated individually to investigate the influence of data uncertainties on the inferred fluxes. Deviations in the inferred fluxes from the different reconstruction methods are comparable or even larger than uncertainties arising from model transport uncertainties. For example, anthropogenic carbon uptake is more than twice as large for the IPSL reconstruction than for the PHI-CT reconstruction in the subpolar and tropical Atlantic.
B53B-1177
Modeling and Synthesis Support for the North American Carbon Program
The Modeling and Synthesis Thematic Data Center (MAST-DC) supports the North American Carbon Program by providing data products and data management services needed for modeling and synthesis activities. The overall objective of MAST-DC is to provide advanced data management support to NACP investigators doing modeling and synthesis, thereby freeing those investigators from having to perform data management functions. MAST-DC has compiled a number of data products for North America, including sub-pixel land-water content, daily meteorological data, and soil, land cover, and elevation data. In addition, we have developed an internet-based WebGIS system that enables users to browse, query, display, subset, and download spatial data using a standard web browser. For the mid-continent intensive, MAST-DC is working with a group of data assimilation modelers to generate a consistent set of meteorological data to drive bottom-up models. http://nacp.ornl.gov/mast-dc/
B53B-1178
Is Convectively Driven air-sea CO2 Exchange Important When Evaluating Regional and Global Estimates?
Present estimates of air-sea carbon dioxide (CO2) transfer during low wind speeds and strong solar heating underestimate the net amount of gas exchanged, as they fail to consider the response of the near surface ocean to diurnal temperature variability and subsequent buoyancy-driven convective overturning. We present results that aim to assess the significance of this missing term for regional and global calculations of CO2 exchange. A two equation k-ε turbulence closure model served as a 1-d test-bed for evaluating a modified version of the National Oceanic and Atmospheric Administration/Coupled-Ocean Atmospheric Response Experiment (NOAA/COARE) air-sea gas transfer parameterization. The improved parameterization includes a new term based on a water-side convective velocity scale (w*w), to better represent convective gas transfer. Meteorological data from the PIRATA mooring located at 10°S10°W in the Tropical Atlantic was used, in conjunction with cloud cover estimates from Meteosat-7, to calculate fluxes of longwave, latent and sensible heat along with a heat budget and temperature profiles during February 2002. Results from twin model experiments, representing idealistic and realistic conditions showed a 20% enhancement to the instantaneous gas transfer velocity, and when integrated over a week; CO2 exchange was increased by approximately 3%. In addition we present results from simulations utilizing the UK Met Office FOAM-HadOCC system (Forecasting Ocean Assimilation Model with Hadley Centre Ocean Carbon Cycle Model biogeochemistry). We aim to highlight the spatial and temporal variability of enhanced low wind speed CO2 exchange and ultimately assess its global importance.
B53B-1179
Temporally smoothed and gap-filled MODIS fPAR data to improve satellite-derived estimates of gross vegetation productivity
Satellite-derived measurements of terrestrial productivity such as the MOD17 8-day gross primary production (GPP) product generated at the University of Montana (UMT) enable a current and spatially comprehensive understanding of the global carbon cycle. These data are limited by the quality of upstream datasets including the MOD15 fPAR (fraction of photosynthetically active radiation) product. Problems associated with this dataset pertain to poor quality or missing data due to sub-optimal atmospheric conditions or sensor malfunction. The most recent MOD17 product (collection 4.8) utilizes fPAR data (UMT fPAR) that has undergone a temporal filling process based on simple linear interpolation. This process generally acts to increase fPAR, as unreliable data is replaced by linear interpolation of the nearest reliable values. We present an alternate smoothed and spatially gap-filled fPAR product derived using a modified version of the TIMESAT software. This approach provides a weighting mechanism based on the MODIS quality assessment layers maximizing the use of high-quality retrievals to fit an annual curve. The curve is used to replace missing or poor-quality observations. If large gaps exist in the time-series, TIMESAT does not fit a curve and a separate spatial gap-filling procedure is used. Processing has currently only been conducted for North, Central and the northern part of South America. Initial comparisons between four years (2002 – 2005) of the UTM and TIMESAT fPAR reveals that both exhibit similar spatial and temporal patterns reducing the noise that is present in the original MOD15 dataset. In regions with extensive periods of missing data due to high cloud cover, UMT fPAR values show greater deviations from the original high-quality MOD15 fPAR values compared to the TIMESAT-derived fPAR. We utilize the TIMESAT fPAR in the MOD17 GPP algorithm to test the sensitivity of the product to the TIMESAT smoothing algorithm. In addition, we compare seasonal and annual estimates of GPP derived from flux measurements at 30 field sites spanning representative vegetation types across North America.
B53B-1180
New Coupled Vegetation-Carbon Model Used Inversely for Reconstructing Historical Terrestrial Carbon Storage from Pollen Data
A long-standing issue exists between data concerning the discrepancy of paleocarbon storage reconstructions since the Last Glacial Maximum by means of pollen, carbon isotope, and general circulation model (GCM) analysis. In this study, a new estimate of past biospheric carbon stocks is reported using a new paleocarbon model (PCM), which is defined as a physiological process vegetation model (BIOME4) coupled to a process- based biospheric carbon model (DEMETER). The PCM was constrained to fit pollen data to obtain realistic estimates. It was estimated that the probability distribution of climatic parameters, as simulated by BIOME4, was compatible with pollen data while DEMETER successfully simulated the carbon storage values with the corresponding outputs of BIOME4. The carbon model was validated with observable global vegetation biomass and soil carbon, and the inversion scheme was tested against 1491 surface pollen spectra sample sites procured in Africa and Eurasia. Results showed that this method can successfully simulate most biomes at selected pollen sites, and that the coefficient of determination (R2) calculated between the observed and reconstructed modern climates vary from 0.70 to 0.96. Comparisons between the simulated biome-average terrestrial carbon variables with the available observations also indicated a consensus: R2 variability of 0.92 for vegetation carbon density and 0.81 for soil carbon density. Results demonstrate the reliability and feasibility of this paleoclimate reconstruction method and its efficiency in reconstructing historical terrestrial carbon storage.
B53B-1181
Empirically Modeling Carbon Fluxes over the Northern Great Plains Grasslands
Grasslands cover nearly one-fifth of the global terrestrial surface and store most of their carbon below ground. The grassland ecosystem in the Great Plains occupies over 1.5 million km2 of land area and is the primary resource for livestock production in North America. However, the contributions of grasslands to local and regional carbon budgets remain uncertain due to the lack of carbon flux data for the expansive grassland ecosystems under various managements, land uses, and climate variability. A quantitative understanding of carbon fluxes across these systems is essential for developing regional, national, and global carbon budgets and providing guidance to policy makers and managers when substantial conversion to biofuels are implemented. Additionally, these estimates will provide insights into how the grassland ecosystem will respond to future climate and what systems are sustainable and offer net carbon sinks. This knowledge base and decisions support tools are needed for developing land management strategies for the region under a variety of environmental conditions and land use options. In the past, we used a remote sensing-based piecewise regression (PWR) model to estimate the grassland carbon fluxes in the northern Great Plains using the 1-km SPOT VEGETATION normalized difference vegetation index (NDVI) data. We estimated the carbon fluxes through integrated spatial databases and remotely sensed extrapolations of flux tower data to regional scales. The PWR model was applied to derive an empirical relationship between environmental variables and tower-based measurements. The PWR equations were then applied through time and space to estimate carbon fluxes across the study area at 1-km resolution. We now improve this modeling approach by 1) using Moderate Resolution Imaging Spectroradiometer (MODIS) data with higher temporal, spatial, and spectral resolutions (8-day, 500-m, and 7-band) as input; 2) incorporating the actual vegetation evapotranspiration data derived from the VegET model, which takes into account soil moisture and land surface phenology; 3) adding an additional flux tower from Brookings, SD, and additional years at other flux towers to the training data sets; and 4) considering the lag response of vegetation production to precipitation. We modeled and mapped 8-day and 500-m carbon fluxes for the years 2000–2006 in the northern Great Plains grasslands. These maps were then used to assess the regional and temporal trends of carbon fluxes in this region, identify carbon sink and source areas, and determine important transitions and environmental drivers of carbon sinks/sources. Cross-validation at sites showed that the improved model increases the estimation accuracies and reflects the variations in water stress that may not be monitored by vegetation indices alone because of the lag-response of vegetation indices to water deficits.
B53B-1182
Retrieving high resolution CO2 fluxes in the South West of France by direct and inverse methods
The CarboEurope Regional Experiment (CERES) is designed to test the coherence of top-down and bottom-up estimates of carbon flux in a limited domain in southwestern France. Here we compare estimates from a biosphere model (ISBA-A-GS), an atmospheric inversion using the same model as a prior estimate and a network of flux towers. We focus on a four-day intensive period in which continuous measurements at two towers were augmented by aircraft measurements. ISBA-A-gs was used to estimate hourly CO2 fluxes at 8km resolution. Based on the model prior, we inverted CO2 fluxes on a limited domain of 700 by 700 km using the mesoscale atmospheric model MesoNH coupled with the Lagrangian particle model LPDM, using boundary conditions from the global transport model LMDz. Both tower and aircraft data were used for the flux inversion, and the model performance was evaluated by an ensemble of simulations. The prior flux uncertainty was estimated by comparing flux tower observation with the biosphere model results at hourly time step. The consistency of the top-down and bottom-up estimates is studied by comparing the structure of the correction in the atmospheric inversion with the differences between ISBA-A-GS and the flux observations.
B53B-1183
Seeing the Forest Through the Trees: Investigating Signal to Noise Problems in Regional Atmospheric Inversions
Estimation of regional carbon fluxes from sparse atmospheric data by transport inversion is complicated by high- frequency variations in surface fluxes in both space and time. We assume that a forward coupled model of the vegetated land surface and atmosphere adequately captures most of the high-frequency variations (SiB-RAMS) as a `preprocessor` of input data from remote sensing and large-scale weather. We then use continuous CO2 observations and backward-in-time Lagrangian particle modeling to estimate persistent multiplicative biases in photosynthesis and ecosystem respiration, constraining the temporal pattern of these fluxes with the forward model. With a sparse network of continuous observing sites in North America, the inverse problem is still badly underconstrained for flux biases on the model grid scale. Previous studies have reduced the dimensionality of this problem by using large `regions` such as biomes or ecoregions, or by seeking a smooth solution in space. This could introduce substantial bias in the solution because the actual flux biases are likely to be quite heterogeneous. We have evaluated the degree to which carbon flux over large regions (500 to 1500 km) can be recovered when the true spatial pattern is not smooth. We performed ensembles of inversions for a 4-month case study in May- August, 2004 over North America with synthetic mid-day CO2 observations from a network of 8 towers. A smooth regional field of model biases was superposed with ensembles of various degrees of grid-scale `noise,` and these were then used to create synthetic concentration data. The pseudodata were then inverted to estimate gridded values of the biases, which were then combined with time-varying model fluxes to create regional maps of sources and sinks. We found that the degree to which corrections in regional fluxes are possible will depend on the relative amount of variance in the regional vs grid scales, but that the system is quite successful in estimating regional monthly fluxes even when the regional scale constitutes a smaller percentage of the overall variance.
B53B-1184
Assimilation of AmeriFlux Data in Two Terrestrial Carbon Cycle Models: Are Ecosystem- Scale Parameters Temporally and Spatially Coherent?
The growing volume of eddy covariance observations presents an opportunity to test and calibrate large-scale terrestrial biosphere models with ecosystem-scale carbon and energy flux observations. Such models inherently assume that many parameters are coherent across space and time for a given plant functional type. We test these assumptions by assimilating CO2 and H2O fluxes to optimize parameters at several broadleaf deciduous forest eddy covariance sites in the Ameriflux network each with multiyear records. Two models are used to examine if the results are model dependent: Top-down Representation of Flora and Fauna Including Dynamics (TRIFFID) and the Local Terrestrial Ecosystem Carbon model (LoTEC). A genetic algorithm (global optimization method) is used to avoid misconvergence to local minima in the objective function. Models are forced with gap- filled meterological data from the AmeriFlux sites. Observed hourly net ecosystem exchange (NEE), gross primary productivity (GPP) and latent heat flux are used as data constraints. Optimized parameters are subsets of model parameters that most strongly control photosynthesis, transpiration, heterotrophic and autotrophic respiration, and phenology. Initial model carbon pools are also estimated as parameters to avoid long equilibrium runs. Separate optimizations are performed for each site-year to obtain estimates of the variance for each model parameter. We then calculate the variance of each model parameter at each site and across all site- years. The spatial autocorrelation of each parameter is also calculated, but is highly uncertain due to the limited number of sites. A joint optimization is performed using all site-years to obtain a single set of parameters for each model to predict carbon fluxes across all site-years. We evaluate the ability of the models for out-of-sample prediction using cross-validation.
B53B-1185
Remote Sensing of Giant Kelp Forest Cover and Biomass Using SPOT Multispectral Imagery
We present a method for the remote assessment giant kelp ( Macrocyctis pyrifera) canopy cover and carbon biomass using SPOT multispectral imagery is presented for the kelp forests near Santa Barbara, California. After atmospheric correction by the dark pixel method, kelp-covered pixels are identified by their high near-infrared and low green spectral signals using a principal components approach. Diver measurements of frond density and total kelp biomass are available at three kelp forest sites. Comparisons of these field observations with satellite determinations of normalized difference vegetation index (NDVI) show high correlation with canopy carbon biomass measurements (r2 = 0.63) providing a satellite based method for determining giant kelp areal biomass. The SPOT satellite determinations of cover and biomass are validated on both pixel and bed scales using monthly visual estimates of biomass, available aerial photographic surveys of kelp cover, and diver mapping of kelp bed extent and biomass. Bi-monthly SPOT imagery for these reef systems start in 2004 and continue through 2007. Comparisons are also made with field estimates of kelp net primary production (NPP), the flux of photosynthetically available radiation (PAR) at the sea floor and available data on sea surface temperature, horizontal currents, surface gravity wave action, terrestrial runoff, and ambient chlorophyll concentrations. Our goal is to observe intraseasonal to interannual changes in kelp cover and biomass and to attribute the processes driving the time/space variations in kelp canopy growth, disturbance and colonization.
B53B-1186
A Landscape Investigation of Carbon Dioxide Flux Drivers Using a Geostatistical Regression at Various Temporal Scales
Methods used to quantify regional sources and sinks of atmospheric carbon dioxide (CO2) typically involve the modeling of biophysical and ecological processes using relationships derived from mechanistic studies performed at smaller spatial scales. However, estimates from biospheric models coupled with atmospheric transport models have difficulty reproducing measured atmospheric CO2 concentrations. Although there may be several reasons for these inconsistencies, they point towards a need to study the relationships between critical biophysical and ecological processes and CO2 flux at the spatio-temporal resolution of biospheric models. In addition, the most influential flux drivers identified at diurnal scales may differ from those at larger timeframes. As such, this research aims to investigate parameters driving landscape scale CO2 flux at various temporal scales. The study focuses on a mixed northern hardwood forest site (~1km2) at the University of Michigan Biological Station (UMBS), where eddy-covariance net ecosystem exchange (NEE) measurements from an AmeriFlux tower have been collected since 1999 together with other site-specific environmental datasets such as air temperature, friction velocity, and leaf area index (LAI). This study employs a geostatistical regression at monthly, daily and hourly time scales which, unlike classical statistical techniques, has the ability to account for temporal correlation in the NEE residuals. The regression analysis characterizes the relationships between NEE and the available environmental datasets by estimating regression coefficients which provide insight into the process-based mechanisms controlling biospheric CO2 flux variability at the three examined temporal scales. Preliminary results show that at the monthly scale, LAI and the fraction of canopy intercepted photosynthetically active radiation (fPAR) have the strongest associations with monthly averaged NEE, with LAI being associated with a sink and fPAR being associated with a source of atmospheric CO2. These findings at the landscape scale are consistent with previous results from a global geostatistical inversion study [Gourdji et al, in prep] that suggest that, at a monthly temporal resolution, LAI is better able to explain the strong seasonality expected for photosynthesis while the weaker seasonal cycle of fPAR captures variability expected for total ecosystem respiration. At the daily time resolution, during the early growing season, LAI and fPAR show similar contributions as those at the monthly resolution. Forest biophysical theory supported by mechanistic studies argues that fPAR is a better proxy for photosynthesis than LAI, which as a partial measure of above ground biomass, is more indicative of autotrophic and heterotrophic respiration. However, results from this study and the global work suggest that representations of these vegetative parameters may have a different spatio-temporal correlation to flux at aggregated scales than those inferred at smaller resolutions. Overall, results indicate that it is the combination of the LAI and fPAR datasets that is able captures the largest part of the photosynthesis and respiration signal at the monthly and daily resolutions during the growing season.
B53B-1187
Ensemble modeling of carbon and water fluxes for NACP
We began the process of conducting a VEMAP-style modeling exercise over North America in support of NACP. Public-domain versions of the following ecosystem models: BGC, TOPS, SIMCYCLE, BEAMS, CASA, PnET, LPJ will be used in the proposed experiment to evaluate uncertainties associated with model logic. A standard set of model inputs a la VEMAP is prepared using the TOPS modeling framework. The input data consists of satellite- derived LAI/FPAR(MODIS, AVHRR), climate data (station networks and gridded data), soils data (ECOCLIMAP), terrain data (SRTM) and other data over the North American region. Upon the projects completion, all inputs and ensemble outputs will be made available to the community as is for further analysis.
B53B-1188
Evaluation of the LPJ-DGVM for modelling tropical ecosystem processes in the Amazon Basin
Dynamic global vegetation models (DGVM) provide spatially continuous information for carbon and water fluxes and vegetation distribution. In tropical regions, DGVMs are necessary to resolve large-scale questions related to carbon cycling and disturbance and for evaluating field-based data from eddy covariance or biomass measurements. We evaluate the sensitivity of these parameters in the LPJ-DVM by comparing biomass, plant functional type distribution (PFT), and carbon fluxes against field and remote sensing observations. Based on literature review, we adjusted the range of rooting distributions, sapwood to heartwood turnover rates, and the development of phenology (leaf longevity and response to drought) within LPJ. We found that rooting distributions were most sensitive in the southern and eastern Amazon where water is more frequently limiting. Carbon allocation and phenology were most sensitive in the wetter regions of the Amazon and strongly influenced the competitiveness and biogeography of PFTs. The optimal distribution of parameters is likely more variable than the number of PFTs in use; this suggests that dynamic sub-modules rather than fitted parameters may be required for determining rooting and phenology patterns.
B53B-1189
Upscaling Carbon Fluxes from Stand-Level Towers to the Footprint of a Very Tall Tower in a Heterogeneous Landscape
Eddy covariance fluxes measured from a very tall tower (~400 m, WLEF near Park Falls, WI) have indicated that the heterogeneous landscape surrounding the WLEF tower has been a weak CO2 source. Carbon fluxes derived from stand-level towers, however, have indicated that mature hardwood forest, willow-alder wetland, and old- growth mixed forest were CO2 sinks. We hypothesized that potential CO2 sources from wetlands and clearcuts could be prevalent within WLEF footprint, thus making it a weak CO2 source. We tested this hypothesis by deploying two roving eddy covariance systems to measure carbon fluxes in two clearcuts (~ 1 and 4 years old in 2005) and two wetlands (ericaceous bog and sedge fen) during the 2005-06 growing seasons. We modeled ecosystem respiration (Re) with night-time NEE (net ecosystem exchange for CO2) versus temperature relationships. Daytime gross ecosystem production (GEP), calculated as Re minus observed NEE, was modeled as a function of photosynthetically active radiation. We observed a wide range of carbon fluxes (e.g., NEE ranged from -548 to 152 g C m-2) for the entire growing season (150 days in May to October 2005) among the six stand-level ecosystems. Measurements from stand-level towers are used to upscale carbon fluxes to the landscape level, i.e., footprint around the very tall tower (WLEF). Remotely sensed data at various spatial resolutions (e.g., MODIS and QuickBird) are used to classify forest ecosystems. Upscaling of carbon fluxes is achieved through integration of flux tower data with forest classification. We expect that carbon fluxes measured from clearcuts and wetlands will substantially improve our ability to upscale local observations to the region.
B53B-1190
Modeling of Local and Regional Soil Carbon Balance under Various Agricultural Managements
We employed the Daily CENTURY (DayCent) soil organic matter model in a number of sensitivity simulations, in an attempt to quantify the effects of various factors on soil carbon (SC) levels. The site chosen for the simulations was Ames, Iowa, USA (42.03°N, 93.63°W), between 1894 and 2005. A control simulation was carried out using realistic assumptions of farming practices (tillage type, rotation pattern, fertilization, cultivar choice) and actual observed weather data (temperature and precipitation). The observed increase in atmospheric CO2 concentration during this time period was accounted for in the simulations. The sensitivity simulations varied one aspect of the simulation for the entire period. When compared against the control run, the sensitivity runs quantify the response of the modeled ecosystem to changes in the various aspects. Results illustrated that SC levels are generally higher without the use of tillage, or when the soil is disturbed only minimally. Application of nitrogen fertilizer to corn was found to maintain a higher SC level than unfertilized corn. A prescribed cooled climate run, as well as a dry run, produced higher SC levels than the control. We expanded the study above while attempts were made to simulate ecosystem dynamics over parts of the Midwestern United States. A large number of sites in two meteorological networks - the Automated Surface Observing System (ASOS) and the National Weather Service Cooperative Observer Program (Co-Op) - were chosen from the available stations in Illinois, Iowa and Missouri. Soil profile and land-use characteristics for each station were determined from gridded terrain data used in a mesoscale meteorological model; using these in conjunction with the weather observations, DayCent was run at each site, and total seasonal net primary production (NPP) was spatially interpolated over the region. For each of the years 2000 through 2006, the simulated NPP fields were compared against satellite observations of total annual NPP reported by the NASA- TERRA program, thus providing some validation tests for the model performance. Initial results suggest that refinements are needed in the application of the two-dimensional DayCent system. Accounting for such refinements, results of NPP and SC sensitivity to the previous aspects, over entire Midwest region, will be presented.
B53B-1191
Ensemble Calibration and Sensitivity Tests of a Photosynthesis Model Using an Optimization Algorithm
Carbon exchange between the atmosphere and terrestrial ecosystem is a key component affecting climate changes. Except for some measurements from spotty sites, the CO2 exchange is simulated by numerical ecosystem models. These models contain large uncertainties in estimating CO2 exchange owing to incorporating a number of empirical parameters on different scales. This study applied a global optimization algorithm and ensemble approach to a surface CO2 flux scheme to 1) identify sensitive photosynthetic and respirational parameters, and 2) optimize the sensitive parameters and improve the model skills. The photosynthetic and respirational parameters of corn (C4 species) and soybean (C3 species) in NCAR land surface model (LSM) are calibrated against observations from AmeriFlux site at Boundville, IL during 1999 and 2000 growing seasons. Results showed that the most sensitive parameters are maximum carboxylation rate at 25oC and its temperature sensitivity parameter (Vcmax25 and avc), quantum efficiency at 25oC (Qe2), temperature sensitivity parameter for maintenance respiration (arm), and temperature sensitivity parameter for microbial respiration (amr). After adopting calibrated parameter values, simulated seasonal averaged CO2 fluxes were improved for both the C4 and the C3 crops (relative bias reduced from 0.09 to -0.02 for the C4 case and from 0.28 to -0.01 for the C3 case). An updated scheme incorporating new parameters and a revised flux- integration treatment is also proposed.
B53B-1192
Implementation of a Boundary Layer Heat Flux Parameterization Into the Regional Atmospheric Modeling System (RAMS) and its Effects on Regional Carbon Budgets
Estimates of regional carbon budgets depend upon the depth of the planetary boundary layer. Overshooting thermals that entrain free tropospheric air down into the boundary layer modify the characteristics and depth of the lower layer through the insertion of energy and mass. This alters the surface energy budget by changing the bowen ratio and thereby altering the vegetative response and the surface boundary conditions. Although overshooting thermals are important in the physical world, their effects are neglected in most regional models. We have included a parameterization to include the effects of boundary layer entrainment into a coupled ecosystem-atmosphere model (SiB-RAMS). The parameterization is based on a downward heat flux at the top of the boundary layer that is proportional to the heat flux at the surface. Results with the parameterization show that the boundary layer simulated is deeper, warmer, and drier than when the parameterization is turned off. These results alter the vegetative stress factors thereby changing the carbon flux from the surface. The combination of this and the deeper boundary layer change the concentration of carbon dioxide in the boundary layer.
B53B-1193
Assessment of Regional Differences in Frontal Atmospheric CO2 Variations Across North America, Europe, South America, and Africa Using a Continuous Global Surface Observation Network
A whole new suite of continuous CO2 observations across the globe has opened up opportunities for understanding the carbon cycle at local and regional scales, through both observational analysis and top-down CO2 inversion techniques. CO2 surface observations contain high frequency variations due to local biology and atmospheric transport. Frontal passage events, common in mid-latitudes, allow for air mass exchanges to occur, which can sometimes cause abrupt CO2 variations on the order of a typical continental seasonal cycle to occur. Since these spikes, which often occur under the presence of frontal cloud cover, are not necessarily observed under the current flask network or very likely satellite missions due to cloud contamination, they should not be ignored with respect to data assimilation. In this poster, CO2 patterns established during frontal events in Northern Hemisphere mid-latitudes are analyzed using surface observations from over 20 towers across the globe and a global chemistry transport model (PCTM, and other CTM's&pused in TRANSCOM), and compared to other frontal variations across mid-latitudes. We find that systematic patterns exist and that they appear to be controlled regionally by upstream influences. We also find an important difference between mid-latitudes and tropical regions in that in mid-latitudes, air mass separation by fronts contributes significantly to day-to-day variations, while tropical variations are controlled more by transport by trade winds and convective depletion of the PBL.
B53B-1194
Regional and Local Carbon Flux Information from a Continuous Atmospheric CO2 Network in the Rocky Mountains
We will present preliminary carbon flux estimates from the Regional Atmospheric Continuous CO2 Network in the Rocky Mountains (Rocky RACCOON). In order to improve our understanding of regional carbon fluxes in the Rocky Mountain West, we have developed and deployed autonomous, inexpensive, and robust CO2 analyzers (AIRCOA) at five sites throughout Colorado and Utah, and plan additional deployments on the Navajo Reservation, Arizona in September 2007 and atop Mount Kenya, Africa in November 2007. We have used a one- dimensional CO2 budget equation, following Bakwin et al. (2004), to estimate regional monthly-mean fluxes from our continuous CO2 concentrations. These comparisons between our measurements and estimates of free- tropospheric background concentrations reveal regional-scale CO2 flux signals that are generally consistent with one another across the Rocky RACCOON sites. We will compare the timing and magnitude of these estimates with expectations from local-scale eddy-correlation flux measurements and bottom-up ecosystem models. We will also interpret the differences in monthly-mean flux signals between our sites in terms of their varying upwind areas of influence and inferred regional variations in CO2 fluxes. Our measurements will be included in future CarbonTracker assimilation runs and other planned model-data fusion efforts. However, questions still exist concerning the ability of these models to accurately represent the various influences on CO2 concentrations in continental boundary layers, and at mountaintop sites in particular. We will present an analysis of the diurnal cycles in CO2 concentration and CO2 variability at our sites, and compare these to various model estimates. Several of our sites near major population centers reflect the influence of industrial CO2 sources in afternoon upslope flows, with CO2 concentration increasing and variable in the mid to late afternoon. Other more remote sites show more consistent and decreasing CO2 concentrations throughout the afternoon. These measurements provide insight as to when and under what conditions mountaintop CO2 signals are regionally representative, as well as first-order constraints on boundary-layer heights and flux rates for use in evaluating model fidelity. Because of coarse representation of topography and boundary-layer mixing biases, forward model CO2 diurnal cycles can be 180 degrees out of phase with respect to assimilated mountaintop CO2 observations if care is not taken in the choice of model level used.
B53B-1195
Prediction of Continental-Scale Net Ecosystem Carbon Exchange by Combining MODIS and AmeriFlux Data
There is growing interest in scaling up net ecosystem exchange (NEE) measured at eddy covariance flux towers to regional scales. Here we used remote sensing data from the MODIS instrument on board NASA's Terra satellite to extrapolate NEE measured at AmeriFlux sites to the continental scale. We combined MODIS data and NEE measurements from a number of AmeriFlux sites with a variety of vegetation types (e.g., forests, grasslands, shrublands, savannas, and croplands) to develop a predictive NEE model using a regression tree approach. The model was trained using 2000-2003 NEE measurements, and the performance of the model was evaluated using independent data over the period 2004-2006. We found that the model predicted NEE with reasonable accuracy at the continental scale. The R-squared values are 0.50 for all vegetation types combined and 0.72 for deciduous forests. We then applied the model to the conterminous U.S. and predicted NEE for each 500m by 500m cell over the period 2001-2006. Based on the wall-to-wall NEE estimates, we examined the spatial and temporal distributions of annual NEE and interannual variability of annual NEE across the conterminous U.S. over the study period (2001-2006). Our scaling-up approach implicitly considered the effects of climate variability, land use/land cover change, disturbances, extreme climate events, and management practices, and thus our annual NEE estimates represents the net carbon fluxes between the terrestrial biosphere and the atmosphere in the conterminous U.S.
B53B-1196
An Improved Methodology to Estimate Terrestrial Net Primary Productivity by Integrating MODIS-LAI to Ecosystem Model SimCYCLE
Net Primary Productivity (NPP) is the difference between total photosynthesis and total plant respiration in an ecosystem. Estimating terrestrial NPP accurately is important as world's forest plays a vital role in the global carbon budget and overall environmental sustainability. Existing ecosystem models synthesize disparate time/space data into single coherent analysis of terrestrial carbon fluxes by incorporating known parameterizations of different ecosystem processes. However, because of the differences in mechanisms between the model and natural ecosystems, often the simulation of key parameters within the model is not accurate. Leaf area index (LAI) is one such key parameter simulated inside most ecosystem models based on averaged climate and soil conditions, carbon allocation scheme and fixed specific leaf area assumed for each biome. It is the second most important factor to NPP after growth period, and now with advanced modeling techniques, realistic and accurate estimates of LAI can be estimated globally from remote sensing imageries such as MODIS. In a previous study (Scheme I), considerable improvement was achieved in estimating global NPP at 0.5 degree spatial scale by constraining simulated LAI in an ecosystem model SimCYCLE with MODIS-derived LAI(MODIS- LAI). In this study (Scheme II), we used a similar strategy to Scheme I, but employed an improved methodology to estimate global NPP by integrating MODIS-LAI to ecosystem model SimCYCLE. Validation and comparison of results were done using GPPDI (Global Primary Productivity Data Initiative) ground-truth NPP dataset at 0.5 degree spatial scale. With this new integration scheme (Scheme II), estimation accuracy improved considerably (R2: 0.67, RMSE: 1.27 MgC ha-1yr-1, Stdev:3.46 MgC ha-1yr-1), when compared with Scheme I (R2: 0.56, RMSE: 1.88 MgC ha-1yr-1, Stdev:7.52 MgC ha-1yr-1) or with the model-alone estimates without integration of MODIS-LAI (R2: 0.44, RMSE: 2.40 MgC ha-1yr-1, Stdev:8.59 MgC ha-1yr-1). Validation at several locations in the tropics with another dataset also showed the new Scheme II producing better estimates (R2: 0.44) when compared with Scheme I (R2:0.01) or with the model estimates using climate and soil data alone (R2: 0.11).
B53B-1197
Remote Estimation of Gross Primary Production in Crops at Field and Regional Levels
Accurate estimation of spatially distributed CO2 fluxes is of great importance for regional and global studies of carbon balance. We have found that in irrigated and rainfed crops (maize and soybean), GPP is closely related to total crop chlorophyll content. The finding allowed development of a new technique for remote estimation of crop chlorophyll specifically for assessing gross primary production. The technique is based on reflectance in two spectral channels: the near-infrared and either the green or the red-edge. The technique provided accurate estimations of daily GPP in both crops. Validation using independent datasets for irrigated and rainfed maize and soybean documented the robustness of the technique. We report also about applying the developed technique for GPP retrieval from data acquired by both an airborne imaging spectrometer (AISA-Eagle) and Landsat ETM+. The Chlorophyll Index, retrieved from Landsat ETM+ data, was found to be an accurate surrogate measure for daily crop GPP with a root mean square error of GPP prediction of less than 1.58 g C m-2d-1 in a GPP range of 1.88 g C m-2d-1 to 23.1 g C m-2d-1. These results suggest new possibilities for analyzing the spatio-temporal variation of the GPP of crops using not only the extensive archive of Landsat Thematic Mapper imagery acquired since the early 1980s but also the 500-m/pixel data currently being acquired by MODIS.
B53B-1198
Lidar measurements of atmospheric CO2
Improved spatial and temporal coverage of atmospheric CO2 measurements is essential for a better understanding of the regional carbon cycle. Active remote sensing has the capability to make studies at various scales and could ultimately be operated from space. In this context, a 2-µm Heterodyne Doppler and Differential Absorption Lidar (HDIAL) has been developed and operated at IPSL/LMD to monitor both CO2 mixing ratio and velocities in the troposphere. In this paper, we will describe the instrument and field experiment results and comparison with in-situ sensors. In addition we will review the potential for future applications of CO2 lidar in regional-scale carbon cycle studies.