A13A-0082 1340h
Design of an Atmospheric Observing Strategy for California's Carbon Cycle
We report the design of an atmospheric observing strategy to enable inverse estimation and attribution of CO2 (and other green house gases) exchange between the land surface and the atmosphere in California. Using models of the natural carbon cycle, fossil fuel emissions, and regional meteorology, we predict CO2 concentration "signals" that would be measured with current and proposed observing platforms. The CO2 sources can be broadly categorized as (1) strong and fairly constant positive fluxes from metropolitan areas (e.g., Los Angeles and San Francisco), (2) weaker but widely distributed positive and negative fluxes from diurnally and seasonally varying ecosystem exchange, and (3) weak but broadly distributed fossil emissions from the Central Valley. Maps of predicted CO2 signals for tower-based stations largely mirror the flux maps over land, while aircraft and column integrated signals (as would be observed with sounding instruments) reflect time-averaged fluxes from much larger areas. Predicted concentration signals away from urban areas are dependent on season, due primarily to variations in the natural carbon cycle, and secondarily to variations in fossil emissions and meteorology. Of significance for determining inflow boundary conditions, large plumes of CO2 enriched and depleted air (1-5 ppm) are predicted to advect from the Northwest to positions more than 300 km west of the California coast. This suggests that data from ocean-based stations must be incorporated into inversions using explicit estimations of regional terrestrial influence. We will also discuss ongoing work to: 1) identify a set of candidate observation stations for California, 2) estimate the influence of specific land surface elements to the measured signals, 3) predict atmospheric signals obtained from proposed strategies to increase ecosystem carbon sequestration.
A13A-0083 1340h
Evaluating Model-Observation CO$_2$ Sampling Strategies: Implications for the Strength of the Latitudinal CO$_2$ Gradient.
In CO$_2$ source inversion studies, the combination of model-simulated and observed atmospheric CO$_2$ concentrations suggest that the Northern Hemisphere terrestrial biosphere is sequestering approximately 1-2 PgC/yr. However, in these studies the model simulated CO$_2$ concentrations are generally averaged over all times of day and meteorological conditions whereas most of the observations are made during daylight hours and subject to meteorological restrictions, e.g., wind speed and direction criteria. We investigated the effect of different sampling strategies on modeled CO$_2$ distributions in two different global chemical transport models driven with different meteorological inputs. In our analysis we used diurnally varying CO$_2$ fluxes from the terrestrial biosphere and focused on the stations used in the Transcom intercomparison. We found that annual mean CO$_2$ concentrations simulated in the models were sensitive to the time of day sampling and wind speed and direction. Failing to account for the diurnal cycle of CO$_2$ when sampling atmospheric models leads to an overestimate of CO$_2$ levels at a number of continental and coastal stations. In atmospheric inversions, this bias could lead to an overestimation of the size of the Northern Hemisphere carbon sink. As more observations in non-remote locations are incorporated into model-observation comparisons, extra care will be necessary to sample model simulations in the same manner that the observations were sampled.
A13A-0084 1340h
Measurements of N$_{2}$O and SF$_{6}$ for use in Inverse Model Studies
NOAA CMDL makes measurements of atmospheric nitrous oxide (N$_{2}$O) and sulfur hexafluoride (SF$_{6}$) mole fractions in discrete air samples from $\sim$60 surface sites and $\sim$10 vertical profile sites for use in inverse modeling. Measurements are made by gas chromatography with electron capture detection (ECD) relative to standard scales developed in CMDL's Halocarbons and Other Atmospheric Trace Species (HATS) group. Uncertainties in the standard scales (95% confidence limits) are 0.8 nmol mol$^{-1}$ for N$_{2}$O and 0.05 pmol mol$^{-1}$ for SF$_{6}$. For N$_{2}$O, the ECD response is characterized monthly by a second-order polynomial with a suite of 6 secondary standards covering the range 242-343 nmol mol$^{-1}$. An instrument response function prepared from measurements of the secondary standards relative to a reference cylinder is used to quantify samples. SF$_{6}$ in air samples is quantified with the reference by assuming a linear response with zero intercept in samples that are free of SF$_{6}$. Reproducibility of the measurements, based on agreement between two samples collected in series, is $\sim$0.4 nmol mol$^{-1}$ for N$_{2}$O and $\sim$0.04 pmol mol$^{-1}$ for SF$_{6}$. So far, all measurements have been made on a single analytical system. The measurements impose important constraints on the budgets of N$_{2}$O and SF$_{6}$ based on observed trends, changes in trends, and spatial gradients. These parameters can be used to determine source/sink imbalances, changes in emission rates over time, and the distribution of emissions. In addition, Peters et al. [JGR, in press, 2004] have used the SF$_{6}$ measurements to evaluate transport in the two-way nested chemistry transport model, TM5. They found that TM5 captures vertical gradients and synoptic variability in SF$_{6}$ well at both MBL and continental sampling sites, but it over-estimates the meridional gradient by $\sim$19%. The SF$_{6}$ data also suggest that global SF$_{6}$ emissions increased by $\sim$20% during late-2002.
A13A-0085 1340h
Estimating Uncertainty in Atmospheric Trace Gas Measurements for use in Carbon Cycle Inverse Model Studies
The NOAA CMDL Carbon Cycle Greenhouse Gases (CCGG) Group operates an extensive observational network to monitor the abundance of atmospheric trace gases important to understanding the global carbon cycle. Continuous and discrete measurements of these gases (CO$_{2}$, CH$_{4}$, CO, H$_{2}$, N$_{2}$O, SF$_{6}$, and the stable isotopes of CO$_{2}$ and CH$_{4}$) from surface sites, towers, aircraft, and voluntary observing ships constitute the most extensive set of atmospheric greenhouse gas observations that are internally consistent with respect to calibration and methodology. CCGG data are frequently used either by themselves or combined with similar observations made by other measurement laboratories to constrain emission estimates derived from carbon cycle inverse models. At present, model transport may be the dominant source of error in these estimates but uncertainty in the observational data also contributes to the overall uncertainty. Including estimates of measurement uncertainty with the observational data may improve the ability to quantify errors in model transport due to topography, circulation, and resolution. Measurement errors can be introduced during the collection, storage, and analysis of atmospheric air samples. Errors may also be introduced when relative measurements are linked to an absolute scale. Systematic errors are possible when data from two different labs are combined. All potential sources of error must be examined when estimating measurement error. CCGG measurement uncertainty is monitored using a variety of methods including 1) routine analysis of samples filled with air of known composition; 2) comparison of measurements from samples collected in pairs; 3) comparison of measurements made using independent methods; 4) frequent re-calibration of working reference gases; 5) regular calibration of the internal reference scale with an absolute scale; 6) ongoing inter-laboratory comparisons; and 7) periodic inter-laboratory comparisons. Each of these methods provides information useful in estimating uncertainty in CCGG measurements for use in inverse model studies.
A13A-0086 1340h
Optimal estimation of regional N$_{2}$O emissions using a three-dimensional global model
In this study, we use the MATCH (Model of Atmospheric Transport and Chemistry) model and Kalman filtering techniques to optimally estimate N2O emissions from seven source regions around the globe. The MATCH model was used with NCEP assimilated winds at T62 resolution (192 longitude by 94 latitude surface grid, and 28 vertical levels) from July 1st 1996 to December 31st 2000. The average concentrations of N$_{2}$O in the lowest four layers of the model were then compared with the monthly mean observations from six national/global networks (AGAGE, CMDL (HATS), CMDL (CCGG), CSIRO, CSIR and NIES), at 48 surface sites. A 12-month-running-mean smoother was applied to both the model results and the observations, due to the fact that the model was not able to reproduce the very small observed seasonal variations. The Kalman filter was then used to solve for the time-averaged regional emissions of N$_{2}$O for January 1st 1997 to June 30th 2000. The inversions assume that the model stratospheric destruction rates, which lead to a global N$_{2}$O lifetime of 130 years, are correct. It also assumes normalized emission spatial distributions from each region based on previous studies. We conclude that the global N$_{2}$O emission flux is about 16.2 TgN/yr, with ${34.9\pm1.7%}$ from South America and Africa, ${34.6\pm1.5%}$ from South Asia, ${13.9\pm1.5%}$ from China/Japan/South East Asia, ${8.0\pm1.9%}$ from all oceans, ${6.4\pm1.1%}$ from North America and North and West Asia, ${2.6\pm0.4%}$ from Europe, and ${0.9\pm0.7%}$ from New Zealand and Australia. The errors here include the measurement standard deviation, calibration differences among the six groups, grid volume/measurement site mis-match errors estimated from the model, and a procedure to account approximately for the modeling errors.
A13A-0087 1340h
Global Inverse Modeling of Aerosols Using the Adjoint Method
The feasibility of using the adjoint method to provide spatially and temporally optimized inventories of aerosol and aerosol precursor emissions is explored for the first time. The global chemical transport model GEOS-CHEM is used to simulate hourly total concentrations of the ammonium-sulfate-nitrate particulate system for one month at 4$^\circ$ $\times$ 5$^\circ$ resolution. An adjoint model is developed which explicitly treats aerosol thermodynamics and transport. The gradients from the adjoint model are used to minimize the discrepancy between model output and simulated observations. Starting with assumed scalar parameters representing monthly emissions over a few large areas, the aerosol inventories are optimized to have increased temporal and spatial resolution in order to improve not only globally averaged model performance, but also agreement between predicted and observed time varying aerosol concentrations at specific surface sites. The extent to which the success of this method depends upon the quality and frequency of the observations is investigated, and development of an increasingly detailed adjoint model (including further physical processes such as tropospheric chemistry, wet deposition, etc.) is explored.
A13A-0088 1340h
Inverting for Emissions of Ozone Precursors Using the Adjoint of a CTM
In order to optimise the emissions of ozone precursors (CO, NOx, hydrocarbons) in the IMAGES global chemical transport model, we apply the adjoint technique. Misfits between modelled and measured concentrations are quantified by introducing the cost function and looking for a solution that corresponds to its minimum. The minimum of the cost function is calculated via an iterative procedure that makes use of the adjoint model operator, that is, the gradient of the cost function with respect to a set of control parameters to be optimised. The advantage of the adjoint model technique compared to other inversion methods is that no linear response of the calculated concentrations to changes in the emissions is assumed. Furthermore, the emissions of several chemical compounds can be varied and optimised simultaneously and the chemical feedbacks existing between different chemical compounds can be explicitly taken into account. These features are very important for compounds like CO and NOx which have common emission sources (like biomass burning), and are strongly inter-related through the chemistry of the OH radical. In the present study, the control parameters to be optimised are the annual emissions of CO, NOx and a few NMVOCs (ethane, propane and acetone) over large regions and for different broad categories. Making use of an emission inventory based on EDGAR 3 and GEIA, we present the results for emission optimisations performed using different combinations of the following observational datasets considered for the same year: ground-based measurements of CO and NMVOCs, MOPITT-derived CO columns, GOME-derived NO2 tropospheric columns, and aircraft measurements of several NMVOCs. Finally, the a posteriori concentrations are compared to independent observations provided by aircraft campaigns.
A13A-0089 1340h
Inverse Lagrangian Modelling of Methane Emissions Using Cabauw Tall Tower Concentration Gradients of 2000-2004
At the tall tower of Cabauw (The Netherlands) we measured from June 2000 to May 2004 high precision, half-hourly, vertical concentration profiles of ambient methane. Measurement heights were 20, 60, 120 and 200 m AGL. The COMET Lagrangian transport model was adopted to use for the same time period hourly 96-hour backward trajectories. Mixing layer depth was determined from ECMWF model data using the Critical Richardson number approach. Forward calculated methane concentrations show a high correlation with the measured vertically averaged concentrations of the mixing layer (r$^{2}$=0.70). Using `uncertainty' trajectory data, an estimate can be made of the quality of the trajectories at individual hours. Using criteria for the error related to the uncertainty in the trajectory path the selected data show model-measurement correlations with values for r$^{2}$ of 0.93. Several optimization methods have been tested on the inversion of the COMET derived Source-Receptor Matrix (SRM) and will be discussed. The robustness and Monte Carlo simulated error estimates of the calculated emission rates as a function of important parameters as model error, length of the concentration time series, background concentration data and source area configuration are also considered. The potential and requirements for application of tall tower measurements of other important greenhouse gases like N$_{2}$O, SF$_{6}$ and HFC's for emission verification by inverse modeling are also shown.
A13A-0090 1340h
TransCom 3 Atmospheric Carbon Inversion Intercomparison: Comparing the Long-term Means and Biogeochemical Interpretation of the Interannual Carbon Exchange
This talk will interpret results from the TransCom 3 control interannual time dependent inversion. First, the long-term mean carbon exchange will be compared across the three different TransCom 3 inversion levels: the annual mean, seasonal, and interannual control experiments. We will highlight the agreement among these experiments in spite of the differing degrees of freedom, and the differing CO2 observing networks employed. Comparison will be made to independent decadal estimates of land and ocean carbon uptake with the sensitivity to different CO2 networks noted. We will also interpret the model mean interannual carbon fluxes as they relate to key indices of climate variability. In particular, correlation to the El Nino/Southern Oscillation index will be made suggesting a propagation carbon flux anomalies from the tropics to the extra tropics following the peak of the ENSO warm phase in the tropical Pacific ocean. These correlations will be explained via anomalies in temperature and precipitation from NCEP reanalysis. Acknowledgement goes to all the TransCom 3 modelers for supplying their model output.
http://transcom.colostate.edu
A13A-0091 1340h
Diagnosing Atmospheric Inverse Calculations: What Exactly Are We Learning About?
It is well-known that our current ability to quantify the fluxes of long-lived atmospheric tracers, such as $CO_{2}$ is limited by the sparse coverage of the global observation network used to estimate fluxes. The global growth rates of $CO_{2}$ and the distribution of fluxes between the Northern and Southern Hemispheres is well-constrained by observations, however, the longitudinal distribution of fluxes and especially the partitioning of fluxes between the continents and oceans is much more uncertain. In this study we use numerical diagnostics, such as the resolution kernal, in conjunction with covariance estimates to diagnose resolved features of flux estimates. This work suggests that the partitioning of fluxes between the global oceans and the terrestrial biosphere can indeed be constrained by the current observational network. Zonal average land and ocean fluxes may also be resolved in general. On the other hand, many of the 22 TransCom III source regions are not resolved by the current observation network. The resolution of some regions improves during the growing season for the terrestrial biosphere, suggesting that more certain flux estimates may be obtained for certain seasons. The size of the observational network has increased substantially over the past two decades. Using 29 observation sites to represent the network around 1985, and 120 sites to represent the network around 2000, we find that resolution of carbon fluxes increases for some regions, but has remained disappointing low for most of the TransCom III source regions.
A13A-0092 1340h
CO2 and CO simulations and their source signature indicated by CO/CO2
Three years (2000-2002) atmospheric CO2 and CO fields are simulated by a Chemistry Transport Model driven by the assimilated meteorological fields from GEOS\_4. The simulated CO2 and CO are evaluated by measurements from surface (CMDL), satellite (MOPITT/CO), and aircraft. The model-observation comparisons indicate reasonable agreement in both source and remote regions, and in the lower and upper troposphere. The simulation also captures the seasonality of CO2 and CO variations. The ratios of CO/CO2 are analyzed over different representative regions to identify the source signature, since the anthropogenic CO comes from the same combustion processes as CO2. This work enables us to improve satellite inversion estimates of CO2 sources and sinks by simultaneously using satellite CO measurement.
A13A-0093 1340h
Estimations of Regional CO2 Fluxes - Analysis of Concentration Data From the Ring of Towers in Northern Wisconsin
Inverse studies of CO2 mixing ratio are traditionally conducted at coarse spatial and temporal resolution. This limits our ability to evaluate efforts to upscale chamber and stand level CO2 flux measurements to regional scales, where coherent climate and ecosystem mechanisms govern the carbon cycle. We present an effort to implement atmospheric budget or inversion methodology on a regional scale. A first step towards this end is the evaluation of a network of six relatively inexpensive CO2 mixing ratio measurement systems deployed on towers in northern Wisconsin as part of the Chequamegon Ecosystem-Atmosphere Study (ChEAS). Five systems were distributed on a circle of roughly 150-km radius, while one system is centrally located. All measurements were taken at a height of 76 m. The systems used LiCor-820 infrared CO2 analyzers and were calibrated every two hours using four samples known to within +-0.1 ppm CO2. Field tests prior to deployment in which the six systems sampled the same air indicate agreement of the systems to better than 0.3 ppm from the mean. The six systems were fielded from April to August 2004. Several frontal passage events were observed, and the progression of the front is evidenced in the CO2 concentrations measured by the network. Pollution events were also observed and trajectory analysis was used to examine the source of the air. Results from the 2004 deployment are presented.
A13A-0094 1340h
Applying Atmospheric Measurements to Constrain Parameters of Terrestrial Source Models
Quantitative inversions of atmospheric measurements have been widely applied to constrain atmospheric budgets of a range of trace gases. Experiments of this type have revealed persistent discrepancies between 'bottom-up' and 'top-down' estimates of source magnitudes. The most common atmospheric inversion uses the absolute magnitude as the sole parameter for each source, and returns the optimal value of that parameter. In order for atmospheric measurements to be useful for improving 'bottom-up' models of terrestrial sources, information about other properties of the sources must be extracted. As the density and quality of atmospheric trace gas measurements improve, examination of higher-order properties of trace gas sources should become possible. Our model of boreal forest fire emissions is parameterized to permit flexible examination of the key uncertainties in this source. Using output from this model together with the UM CTM, we examined the sensitivity of CO concentration measurements made by the MOPITT instrument to various uncertainties in the boreal source: geographic distribution of burned area, fire type (crown fires vs. surface fires), and fuel consumption in above-ground and ground-layer fuels. Our results indicate that carefully designed inversion experiments have the potential to help constrain not only the absolute magnitudes of terrestrial sources, but also the key uncertainties associated with 'bottom-up' estimates of those sources.
A13A-0095 1340h
Influence of Varying Biospheric Parameters on CO$_{2}$ Seasonal Cycles in a Forward Transport Model
An atmospheric tracer inversion is only as good as the forward transport model. Reducing uncertainty and bias in final flux and mixing ratio estimates of CO$_{2}$ begins with having a model that accurately depicts the state and trace gas evolution over time. One such indicator of model validity is the estimate of seasonally varying annual cycles of CO$_{2}$ in the atmosphere. This research focuses on the differences effected in precision with measurements by choice of input parameters such as fossil fuel estimates, ocean carbon flux estimates, biomass burning datasets, and changes in sampling level for specific geographic locations. The seasonality mismatches are also compared with new SF6 measurements. Also compared is the satellite-based seasonal cycle measurements of CO$_{2}$ from the SCIAMACHY near-infrared instrument aboard the ENVISAT platform, and the observed seasonal cycles as seen in global surface measurements. For these comparisons, the forward atmospheric transport models TM3 and TM5 are used in global and regional simulations.
A13A-0096 1340h
Will Satellite CO$_{2}$ Measurements Experience a Clear-Sky Bias?
Satellite measurements of CO$_{2}$ can help enhance our understanding of the carbon cycle; however, retrieval of near-surface CO$_{2}$ from satellites will require cloud-free conditions. Using continuous CO$_{2}$ measurements from tower sites in Wisconsin and Harvard Forest, we investigated the systematic bias that could occur from retrieving CO$_{2}$ concentrations only in clear-sky conditions. The clear-sky bias in near-surface CO$_{2}$ at each site was defined by detrending the CO$_{2}$ concentration timeseries, creating a clear-sky subset from mean daytime CO$_{2}$ concentrations by comparing photosynthetically active radiation (PAR) measurements taken at both towers to the top-of-the-atmosphere solar radiation, fitting two harmonics to both the clear-sky subset as well as all the mean daytime CO$_{2}$ measurements, and subtracting the fit for the daytime CO$_{2}$ measurements from the clear-sky fit. At both towers, clear-sky measurements experienced a negative bias relative to the mean of all daytime measurements: Wisconsin had a mean surface bias of -1.1 ppm while Harvard Forest had a mean surface bias of -2.85 ppm, with a larger bias in the winter due to fossil fuel combustion compared to a smaller bias in the summer. We obtained similar results investigating the bias at 1pm rather than using mean daytime measurements. These results indicate that satellite measurements will have a systematic negative bias, underestimating the total column CO$_{2}$ concentration by as much as .6 ppm at some locations and times of the year. Quantitative interpretation of satellite CO$_{2}$ retrievals will require accurate modeling of cloud radiative forcing of ecosystem carbon exchange.
http://biocycle.atmos.colostate.edu
A13A-0097 1340h
Grid-Scale CO$_{2}$ Flux Error Estimates from a Variational Data Assimilation Approach Processing Satellite-Based CO$_{2}$ Measurements
Inversion of atmospheric CO$_{2}$ concentration measurements, using atmospheric transport models, to solve for surface CO$_{2}$ fluxes has proven to be a useful `top-down' check of the carbon fluxes produced by `bottom-up' process models -- the comparison of flux integrals at fairly coarse time and space scales can reveal model mis-tunings, and can even suggest processes missing from the models. For this comparison to be most useful, it should be performed at the finest time/space scales possible (not only to help localize possible model deficiencies, but also to best represent the available fine-scale data, particularly over the continents). The time/space density of carbon observations currently limits the effective resolution at which this comparison may be performed. Should the atmospheric CO$_{2}$ measurements from two satellites scheduled for launch in 2007 prove useful, this resolution will be greatly increased, but then computational limits may become important. To help evaluate the usefulness of the up-coming satellite observations, we have built a variational data assimilation system (4-D Var) that solves for the surface CO$_{2}$ fluxes at resolutions as fine as the horizontal resolution of the underlying global transport model (2.0 x 2.5 degrees) and the model time step (\~15 minutes), using measurements localized up to an equally fine scale. The performance of the method has been assessed using simulated observations (i.e., with an observing system simulation experiment, or `OSSE'). As the size of the flux vector to be estimated is too great to permit a direct estimate, an iterative method is used (in the form of an unconstrained optimization problem). The final estimation accuracy is limited then not only by the number and accuracy of the observations, but also by the number of iterations computationally feasible. With this OSSE set-up, we compute the flux estimation error obtained for measurement sets similar to those planned for the upcoming satellite missions, as well as those from the current flask network augmented with planned flux towers. This error is given as a function of the number of optimization iterations, measurement span considered, and measurement correlation time/length scale assumed. The estimation error obtained by comparing the true and estimated fluxes for the single OSSE case (averaged over time/area to obtain meaningful statistics) is compared to the corresponding error from the leading terms in the covariance matrix built up by the descent method. For those cases in which the estimation errors for the fluxes at the grid scale are too high to be of use, we will discuss at which (coarser) scales the estimates may be believed.
A13A-0098 1340h
Inverse Modeling of Tropospheric CO Using Satellite Measurements
The availability of remotely-sensed CO measurements from the MOPITT instrument provides an opportunity to better characterize the spatial and temporal distribution of CO sources. This is especially true for biomass burning emissions given the recent availability of complementary datasets of fire parameters derived using other satellite measurement products. In this study, we present estimates of time-dependent CO sources by region and sector derived from a global synthesis inversion using MOPITT CO column measurements. We present monthly CO emission estimates for April 2000 to March 2001 and focus, in particular, on the seasonal distribution of large-scale biomass burning emissions. A satellite-constrained fire emission product is used as a starting point in the inversion. Our analyses show the combined utility of MOPITT CO measurements and fire emission product in characterizing the spatial and temporal CO biomass burning emissions. Differences between top-down estimates and satellite-constrained fire emission inventories are identified as a first step towards improving the global fire emission product.