SEARCH

SEARCH BY CITATION

Keywords:

  • carbon dioxide;
  • carbon monoxide;
  • emissions;
  • fossil fuel;
  • radiocarbon

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Results and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] Flask samples from two sites in East Asia, Tae-Ahn Peninsula, Korea (TAP), and Shangdianzi, China (SDZ), were measured for trace gases including CO2, CO and fossil fuel CO2 (CO2ff, derived from Δ14CO2observations). The five-year TAP record shows high CO2ff when local air comes from the Korean Peninsula. Most samples, however, reflect air masses from Northeastern China with lower CO2ff. Our small set of SDZ samples from winter 2009/2010 have strongly elevated CO2ff. Biospheric CO2 contributes substantially to total CO2variability at both sites, even in winter when non-fossil CO2 sources (including photosynthesis, respiration, biomass burning and biofuel use) contribute 20–30% of the total CO2 enhancement. Carbon monoxide (CO) correlates strongly with CO2ff. The SDZ and TAP far-field (China influenced) samples have CO: CO2ff ratios (RCO:CO2ff) of 47 ± 2 and 44 ± 3 ppb/ppm respectively, consistent with recent bottom-up inventory estimates and other observational studies. Locally influenced TAP samples fall into two distinct data sets, ascribed to air sourced from South Korea and North Korea. The South Korea samples have low RCO:CO2ffof 13 ± 3 ppb/ppm, slightly higher than bottom-up inventories, but consistent with emission ratios for other developed nations. We compare our CO2ff observations with modeled CO2ff using the FLEXPART Lagrangian particle dispersion model convolved with a bottom-up CO2ff emission inventories. The modeled annual mean CO2ff mole fractions are consistent with our observations when the model inventory includes the reported 63% increase in Chinese emissions from 2004 to 2010, whereas a model version which holds Chinese emissions flat is unable to replicate the observations.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Results and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] Emissions of greenhouse and polluting gases in East Asia are of increasing importance, as countries in this rapidly developing region increase their emissions. China is now believed to be the world's largest emitter of carbon dioxide from fossil fuels (CO2ff), contributing more than 20% of global total CO2ff emissions [Gregg et al., 2008; Boden et al., 2010; BP, 2010]. Emissions of other anthropogenic trace gases from East Asia are also large and growing rapidly, and are more poorly constrained than CO2ff. Carbon monoxide (CO) is of particular interest as it contributes to air pollution and is hazardous to human health. CO is produced during incomplete combustion, and rapid economic development and increasing fossil fuel use in East Asia, particularly China, would suggest an increase in CO emissions. However, concurrent improvements in combustion efficiency, replacement of older power plants, and a change from inefficient biomass combustion to fossil fuel use act to decrease CO emissions. The balance of these competing effects is not yet clear.

[3] Bottom-up inventories tabulate anthropogenic CO emissions by combining fuel use statistics with “emission factors,” the amount of CO emitted per unit of fuel burned. It can be difficult to accurately quantify fuel use, and there may also be biases in emission factors. For CO, emission factors in China vary from 0.1 mmol CO per mole of CO2 (hereafter referred to as ppb/ppm or parts per billion of CO per parts per million of CO2) for fuel oil and diesel combustion, to 100 ppb/ppm for coal used in kilns [Streets et al., 2006] and open biomass burning [Andreae and Merlet, 2001]. Bottom-up inventories typically do not account for other sources of variability such as CO production by oxidation of natural and anthropogenic volatile organic compounds, and destruction in the atmosphere by reaction with hydroxyl radical, with an atmospheric lifetime that varies seasonally between one month and more than one year [Sander et al., 2006; Spivakovsky et al., 2000]. Therefore, uncertainties in inventory-based CO emission estimates may be large. For example, the suite of published CO inventories for China vary by a factor of two [European Commission, 2009; Kopacz et al., 2009; Zhang et al., 2009; Tanimoto et al., 2008; Ohara et al., 2007; Yumimoto and Uno, 2006; Allen et al., 2004; Wang et al., 2004; Tan et al., 2004; Streets et al., 2003; Carmichael et al., 2003; Heald et al., 2003].

[4] Atmospheric observations can provide an independent method (“top-down”) of evaluating emissions. Regional CO enhancements (ΔCO) can be determined relative to background mole fractions, and measurements of the regional depletion of14C in atmospheric CO2 (expressed as Δ14CO2) can be used to obtain enhancements in the CO2ff mole fraction [e.g., Levin et al., 2003; Turnbull et al., 2006]. The ratio of ΔCO to CO2ff (RCO:CO2ff) in the region can then be determined. As the uncertainty in the CO2ff emission flux reported in inventories (3–30% at the annual, national scale [Boden et al., 2010; Gregg et al., 2008]) is likely small relative to the uncertainty in the inventory-based CO emission flux estimate, the regional CO emissions can be evaluated from RCO:CO2ff. This ratio method does not depend strongly on atmospheric transport, as both species are advected in the same way, as long as they are approximately co-located at the spatial scale influencing our observations, and the transport time is short enough that significant CO removal (e.g., by OH) has not occurred.Turnbull et al. [2011] used this method to show that U.S. EPA CO inventories for Sacramento, California, are too high by a factor of two. Vogel et al. [2010] and Van der Laan et al. [2010]used the same top-down method to evaluate diurnal and seasonal variability in CO emissions in two different European regions.

[5] In addition to assessing RCO:CO2ff and CO emission inventories, measurements of CO2ff from Δ14CO2 observations can be used to quantify the contributions of CO2ff and biological CO2 (CO2bio) exchange to total CO2 variability. It is sometimes assumed that CO2ff dominates CO2 variability in urban regions, whereas CO2ff measurements can directly evaluate this assumption [Turnbull et al., 2011; J. B. Miller et al., Linking emissions of fossil fuel CO2 and other anthropogenic tracers using atmospheric 14CO2, submitted to Journal of Geophysical Research, 2011].

[6] Finally, although the inventory-based CO2ff emission flux is more certain than that of other species, the uncertainty is still significant, particularly for East Asia. Annual CO2ff emissions from China are reported to have increased by 6–10% per year from 2004 to 2009 [Boden et al., 2010], yet the uncertainty in total Chinese emissions is ±15–20% [Gregg et al., 2008]. CO2ff observations have the potential to directly constrain the emission flux and trends in that emission flux, if atmospheric transport can be adequately described.

[7] Here we use flask samples from two sites in East Asia to determine CO2ff, RCO:CO2ff and CO2bio, and examine CO emissions. Δ14CO2 measurements in the flask samples are used to determine the recently added fossil fuel CO2 (CO2ff) mole fraction in each sample. Combining this information with CO2 and CO measurements from the same flasks, we determine the contribution of recently added (or removed) biospheric CO2 (CO2bio) in the same samples, and the emission ratio (RCO:CO2ff) of ΔCO to CO2ff. Our observed RCO:CO2ffvalues are then compared with inventory estimates and other observational data sets. We use the FLEXPART Lagrangian particle dispersion model (LPDM) to aid in determining the source regions and sources of variability in the observed gases and emission ratios. We combine the FLEXPART back-trajectories with bottom-up CO2ff inventories to obtain a modeled prediction of CO2ff at our sites, and compare this with observed CO2ff.

2. Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Results and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

2.1. Sampling Sites

[8] Flask samples were collected approximately weekly at two sites in East Asia (Figure 1), Tae-Ahn Peninsula, Republic of Korea (TAP, 36.73°N, 126.13°E, 20 masl), and Shangdianzi, People's Republic of China (SDZ, 40.65°N, 117.12°E, 292 masl). TAP is located on a small Peninsula on the western coast of Korea. Samples are collected during mid-afternoon from a small cliff above the Yellow Sea, and only collected when there is an onshore wind. The local area is rural, with a small village and rice cultivation within one kilometer. The site is about 100 km southwest of the Seoul National Capital Area (population 25,000,000). SDZ is located near a small village about 100 km northeast of Beijing (population 18,000,000), and is influenced by strong pollution events from Beijing and surrounding urban areas during southwesterly wind flow [Lin et al., 2008; Zhang et al., 2010].

image

Figure 1. Map of measurement sites (TAP, red; SDZ, blue). Lines and white labels indicate the TAP source regions identified from the model (see text for details). Pressurized water nuclear reactors are shown as white triangles, all other nuclear facilities are shown as yellow triangles Reactor locations and types are from IAEA [2006].

Download figure to PowerPoint

2.2. Trace Gas and Isotope Measurements

[9] Flask samples have been collected at TAP and analyzed at NOAA/ESRL for greenhouse gases since 1990 [Conway et al., 2011; Kim et al., 2008]; measurements of 14CO2 were added beginning in October 2004. The China Meteorological Administration (CMA) began flask sampling at SDZ in August 2006. In addition to a pair of flasks collected and analyzed by CMA, since September 2009, a second pair of flasks has been collected simultaneously and analyzed at NOAA/ESRL. We examine only those samples from the two sites in which Δ14CO2 measurements have been made, from October 2004 to March 2010 for TAP, and October 2009–May 2010 for SDZ.

[10] Flask samples are always collected in pairs, and each flask is measured at NOAA/ESRL for mole fractions of CO2, CH4, CO, H2, N2O and SF6 [Conway et al., 2011], and at the University of Colorado for stable isotopes of CO2 [Vaughn et al., 2004]. For 14C, the remaining CO2was extracted from both flasks of the pair and combined into a single Fe-C target at the University of Colorado for Δ14CO2 measurement by accelerator mass spectrometry at the University of California Irvine [Lehman et al., 2011; Turnbull et al., 2007]. In some cases, a second pair of flasks was collected at TAP, immediately following the first flask pair. When this occurred, the CO2 mole fraction in each flask from the second pair was measured, and if the CO2 mole fractions from both flask pairs agreed within the reported uncertainties (±0.1 ppm), all four flasks were combined for a single Δ14CO2 measurement. At SDZ, four samples were highly polluted with CO mole fractions above 500 ppb, outside the calibration range of the CO instrument, resulting in larger errors for those samples [Conway et al., 2011]. The standard deviation of individual Δ14CO2 measurements is typically 1.8‰ [Turnbull et al., 2007]. All measurements are available for download at ftp://ftp.cmdl.noaa.gov/ccg/co2c14/flask/event/.

2.3. Background Values

[11] For this analysis, we are primarily interested in enhancements (or reductions) in each trace gas species as the air masses contact the surface over East Asia. We therefore determine mole fraction enhancements for each species relative to a representative background value, which may be seasonally varying. We derive our background values from marine boundary layer observations from multiple sites in the NOAA/ESRL Carbon Cycle Cooperative Global Air Sampling Network. The background value for a given sampling time and trace gas species is obtained by interpolating in both space and time from a global marine boundary layer map [Masarie and Tans, 1995]. We considered alternative background choices, such as measurements from sites upwind of TAP at Mt Waliguan, China (WLG, 36.29°N, 100.90°E, 3,810 masl) [Zhang et al., 2011], or Ulaan Uul, Mongolia (UUM, 44.45°N, 111.10°E, 914 masl) for this purpose, but in both cases, these sites sometimes exhibit mole fractions higher than those at TAP, suggesting occasional local influences, and/or that these sites are not always upwind of the TAP site. High altitude clean air sites such as Mauna Loa, Hawaii (MLO, 19.54°N, 155.58°W, 3,402 masl) or Niwot Ridge, Colorado (NWR, 40.05°N, 105.58°W, 3,526 masl) are also possible choices, but we find that these sites do not adequately characterize the seasonal cycles at the surface, particularly for CO. However, we found that using any of these sites as background did not substantially change our results, since the enhancements in most species at TAP and SDZ are typically large relative to the differences caused by choice of background.

[12] For Δ14CO2, we have insufficient observations to develop such a global marine boundary layer reference, and instead we interpolate for each sampling date from a curve fit to a filtered data set of NWR Δ14CO2 values [Turnbull et al., 2007] (and extended to May 2010 for this work) using the fitting procedure of Thoning et al. [1989]. NWR appears to be a reasonable choice of background Δ14CO2 for Northern Hemisphere midlatitudes [Turnbull et al., 2009, 2007], and the NWR Δ14CO2 values are similar to those measured at WLG and UUM during the winter of 2004/2005 (Figure 2). The NWR values are also similar to those from the high altitude site at Jungfraujoch, Switzerland which has also been used for Δ14CO2 background [Levin et al., 2007; I. Levin, personal communication, 2010].

image

Figure 2. Time series of observed Δ14CO2 at NWR, TAP, SDZ, UUM and WLG. Measurement uncertainties are omitted for clarity, and are typically 1.8‰.

Download figure to PowerPoint

2.4. Calculation of CO2ff

[13] The recently added fossil fuel CO2 component was determined for each (combined) flask sample, using the method described by Turnbull et al. [2006, 2009], and briefly summarized here, such that

  • display math

[14] CO2obs and Δobs are the observed CO2 mole fraction and Δ14CO2 value in the same sample. Δff is, by definition, −1000‰ (zero 14C content) and Δbg is the background Δ14CO2 value. The second term of the equation is a typically small correction for the effect of other sources of CO2 which have a Δ14C value differing by a small amount from that of the atmosphere, principally heterotrophic respiration and also including CO2 from biomass burning (since natural fractionation is accounted for in Δ14C [Stuiver and Polach, 1977], processes such as photosynthetic uptake, are ignored). CO2other and Δother are the added mole fraction and average Δ14C value of these other sources. We estimate this correction as −0.2 ± 0.1 ppm during winter, and −0.5 ± 0.2 ppm in summer [Turnbull et al., 2009; Turnbull et al., 2006], hence increasing CO2ff as this correction is subtracted in equation (1). The typical uncertainty is about 1 ppm in CO2ff, including uncertainty in the sample Δ14CO2 value (±1.8‰) choice of background and uncertainty in the bias term.

[15] We do not make any correction to CO2ff for 14C from nuclear activities, since we believe that the majority of our samples are not influenced by nuclear emissions. There are no nuclear power plants in the upwind region for the majority of the TAP samples which see air from the northwest (Figure 1). Samples which received air from the east or southeast, having traveled over Korea and/or Japan, could be influenced by nuclear emissions, but the influence is likely small since three of the four sites in Korea are pressurized water reactors, which produce mainly 14CH4 rather than 14CO2 (shown as white triangles in Figure 1) [Graven and Gruber, 2011; International Atomic Energy Agency (IAEA), 2006]. The North Korean nuclear facility in Yongbyon could potentially contribute to 14C enrichment in a few samples, but the magnitude and variability of 14C emissions from this site are unknown and therefore we cannot reasonably account for them.

2.5. Method of Determining Δ14C Source Values

[16] To determine the isotopic composition of the added CO2, we use a variation on the Keeling plot described by Miller and Tans [2003]. This method takes into account time varying background conditions, such that

  • display math

where Δ indicates the Δ14CO2, and the subscripts obs, bg and s indicate the observed (measured) value, background value and source value, respectively.

[17] The Δs of the source is then determined from the slope of a regression of (CO2obs − CO2bg) versus (ΔobsCO2obs − ΔbgCO2bg). The linear least squares regression technique (used both for this method and for Keeling plots) does tend to bias Δs toward the samples with the largest enhancements in CO2, which for this data set, likely results in a bias toward (local) fossil fuel sources. Δsin this formulation represents a concentration-weighted sum of all the sources, and if the isotopic composition of the various source categories is known, then the fraction from each source category can be determined. The CO2ff fraction (fff) is estimated using Δff = −1000‰, and we assume that all other CO2 sources (such as biospheric respiration) have Δ14C the same as the current atmosphere, of 40‰ (this value has only a small impact on fff. E.g. for Δs = −500‰, a 50‰ change in Δ14C of other sources value equates to a 2% difference in fff). Therefore,

  • display math

2.6. FLEXPART Footprint Method

[18] We use the FLEXPART LPDM to obtain back trajectories, “footprints” and modeled CO2ff mole fractions at TAP and SDZ, following the methods of Stohl et al. [2009], summarized here. To obtain each footprint, FLEXPART is run backward for seven days at 1 × 1° resolution, using meteorological data from the NCEP GFS zero and three hour forecasts. For each sampling time at the “receptor” site, we obtain the emission sensitivity in a surface footprint layer from 0 to 100m above the ground (in units of s m2 kg−1). The emission sensitivity in a given grid cell is proportional to the particle residence time in that grid cell. It is the modeled mass mixing ratio at the receptor site that would be produced by a unit strength (1 kg s−1 m−2) source. The spatial pattern of these emission sensitivities is the footprint.

[19] These footprints are first used to identify the source region for each sample. For TAP, we separate the samples into four broad regions (Figure 1), which are similar to those used by Kim et al. [2008]. The four regions are: oceanic background (OBG) air which has recently passed over the ocean to the south of the Korean Peninsula; local Korean air (LK) transport from the north and east, having passed over parts of North and South Korea and/or Japan; regionally polluted continental north (RPCN) air which has passed over Northeastern China; and regionally polluted continental south (RPCS) air which passed over the more southerly region of China. For each sample, the source region was assigned by manually examining the footprint, since the wind direction at TAP is not necessarily indicative of the longer-range transport. Example footprints are given in theauxiliary material. The prevailing wind direction is from the RPCN region to the northwest, where the air has traveled over Northern China and the Yellow Sea before arriving at TAP. 60% of the samples come from this direction, another 30% of the samples come from the LK local Korean region, and 5% come from each of the RPCS and OBG regions. In the winter months, more than 80% of the samples see air from the RPCN region, whereas in summer the wind direction is more variable, and only about 40% come from this region, with about 40% coming from the local region and about 10% from the RPCS and OBG regions.

[20] At SDZ, we perform a similar analysis using the FLEXPART footprints. The SDZ site is situated northeast of Beijing, on the northwest side of a valley running from southwest to northeast. It has two distinct wind regimes, each accounting for about half of our SDZ samples, which can be readily identified simply by the wind direction at SDZ at the time of sampling [Lin et al., 2008]. For this site, where the sources are nearby and the 1 × 1° scale of the FLEXPART footprints is too large to resolve the nearby transport, the local wind direction is a better identifier of the source region. Winds from the southwest bring air from the Beijing region and North China Plain, which is typically quite polluted, whereas winds from the northeast bring typically cleaner air from more distant sources.

[21] In addition to source region identification, we obtain modeled predictions of CO2ff for each sample from the footprints. Each footprint is convolved with a gridded CO2ff emission flux (in units of kg m−2 s−1) to obtain the contribution of sources in each modeled gridbox. The total modeled CO2ff mole fraction at the receptor is the sum of the contributions of all gridboxes (in ppm of CO2ff). In order to reduce model representation error due to the low 1 × 1° resolution in the gridboxes close to the receptors, we rerun FLEXPART with increased footprint output resolution of 0.1 × 0.1° for the 4 × 4° region around each receptor (the 1 × 1° meteorology resolution is unchanged). The nested regions are 124–129°E and 34–38°N for TAP and 115–119°E, 38–42°N for SDZ.

[22] Our “standard” gridded CO2ff emission flux combines inventory data from several sources to provide the best estimate of emissions for East Asia. It is based on the global CO2ff emission inventory used in CarbonTracker [Peters et al., 2007] (updated to 2010 at http://carbontracker.noaa.gov), which is derived from the CDIAC [Boden et al., 2010] (hereafter referred to as CDIAC), EDGAR [European Commission, 2009] and BP [2010] inventories. Chinese emissions are altered from this, maintaining the same total China emissions, but redistributing the emissions within China according to the provincial pattern of Gregg et al. [2008]. Within each province, the emissions are distributed according to population, resulting in higher emissions than in CarbonTracker for the more developed regions in Eastern China, and lower emissions in the west. For the nested 0.1 × 0.1° regions, we use the same total emissions for that 4 × 4° box, but distributed spatially according to the EDGAR [European Commission, 2009] high resolution product. In our emissions inventory, total Chinese emissions follow the CDIAC values, increasing 6–10% annually from 2004 to 2009, along with an increase of 10.4% in 2010 [BP, 2010] bringing the total increase in Chinese CO2ff emissions to 63% from 2004 to 2010. We also apply the seasonal cycle obtained by Gregg et al. [2008]. For South Korea, emissions increase by 7% over the same period. We also run an alternate simulation “flat emissions” which is identical to the standard emissions scenario except that emissions are held to 2004 levels throughout the period of measurement. Our gridded emission inventories are available for download at ftp.cmdl.noaa.gov/pub/turnbull/FF_Fluxes/.

[23] The model also allows partitioning of the modeled CO2ff values into local and far-field sources. To estimate the local sources, we calculate the modeled CO2ff at the receptor due only to the nested 4 × 4° region. The far-field sources are calculated by assuming zero emissions in the local 4 × 4° region.

3. Results and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Results and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

3.1. Observed Δ14CO2 and Calculated CO2ff

[24] The TAP Δ14CO2 values are typically lower than Δ14CO2 measurements from “clean” Northern Hemisphere midlatitude sites including NWR, UUM and WLG (Figure 2). The one exception is a sample taken on October 22nd, 2004, which had a Δ14CO2 value of 100.7 ± 2.2 ‰, about 35‰ higher than Δ14CO2 values at other sites at the same time. Other species in the same sample are not enhanced relative to background. We hypothesize that this point represents an unidentified analytical problem, discussed in detail in the auxiliary information, and we exclude this sample from further analysis.

[25] The calculated CO2ff mole fraction for the remaining 202 TAP samples varies dramatically with synoptic conditions, with a mean of 4.1 ppm and 10th, and 90th percentiles of −0.4, and 15.8 ppm, respectively (Figure 3 and Table 1). Most samples (n = 155) have CO2ff contributions of less than 5 ppm. Negative CO2ff values are seen in 14 samples, this number of non-physical negative results is roughly consistent with the CO2ff (1-sigma) uncertainty of 1 ppm and a sample size of 202.

image

Figure 3. Time series of CO2ff calculated from Δ14CO2at TAP (separated into local and far-field samples) and SDZ, and 1 ppm binned frequency distributions. For TAP, the frequency distribution includes all TAP samples.

Download figure to PowerPoint

Table 1. Measured Values for Each Subset of Data From TAP and SDZa
SubsetnCO2ff (ppm)RCO:CO2ff (ppb/ppm)RCO:CO2 (ppb/ppm)fff (%)
  • a

    For CO2ff, results are given in ppm, and the mean and 10th and 90th percentile values (in parentheses) are reported. For RCO:CO2ff, RCO:CO2, the value derived from a least squares fitting procedure (see text for details) is given, along with the uncertainty in that slope, and the coefficient of determination (r2) is given in parentheses. Fff is the fraction of CO2 variability due to CO2ff, derived from Δ14Cs. The r2 value obtained for the Δ14Cs fitting procedure is shown in parentheses.

TAP All2024.4 (−0.4, 15.8)31 ± 5 (0.3)-60 (0.6)
   Winter1004.4 (0.3, 16.3)26 ± 8 (0.2)11 ± 2 (0.3)90 (0.9)
TAP far-field1442.6 (−0.5, 7.2)44 ± 3 (0.7)-20 (0.1)
   Winter822.6 (0.2, 7.7)44 ± 5 (0.7)30 ± 2 (0.7)70 (0.6)
TAP local low RCO3911.6 (1.5, 30.7)13 ± 3 (0.5)6 ± 1 (0.6)60 (0.6)
   Winter1115.7 (4.1, 43.0)9 ± 3 (0.4)6 ± 1 (0.7)95 (1.0)
TAP local high RCO192.4 (−0.5, 10.2)58 ± 12 (0.8)-10 (0.01)
   Winter44.8 (2.5, 10.2)39 ± 1 (1.0)35 ± 4 (0.9)90 (1.0)
SDZ winter306.7 (−1.8, 31.3)47 ± 2 (1.0)39 ± 3 (0.9)80 (0.9)

[26] We separate the TAP samples into two data sets, local and far-field, using two different methods. First, we flag samples identified by FLEXPART footprints and source regions (Figure 1 and auxiliary material Figure S1) as likely having been influenced by the local Korean surface (53 of the 202 samples). Second, we use the sulphur hexafluoride (SF6) mole fraction as a tracer for local Korean air. SF6is entirely anthropogenic, and extremely inert and long-lived in the atmosphere [Geller et al., 1997]. It is primarily used as a spark quencher in high voltage electrical transmission equipment, and is released to the atmosphere via leakage from these facilities and during its production. The EDGAR [European Commission, 2009] inventory lists large SF6 emissions from Korea (Figure 4), and we see dramatic enhancements over background (background SF6 increased from 5.4 to 7.2 ppt during 2004–2011) of up to 14 ppt in some TAP samples. Samples with SF6 enhancements of more than 1 ppt over (seasonally varying) background are flagged as local (35 samples, of which 30 were also identified by the footprint method). As SF6 emissions from China are also substantial, occasionally air from China might be incorrectly identified as local, although the SF6 plumes will likely have diluted before arrival at TAP. Samples identified by either the footprint or SF6 methods are flagged as local, a total of 58 of the 202 TAP samples.

image

Figure 4. SF6 emissions for East Asia from the EDGAR 2005 inventory [European Commission, 2009].

Download figure to PowerPoint

[27] The TAP local samples have large and variable CO2ff values (mean 8.5 ± 8.6 ppm, 10th and 90th percentiles of 0.4, 23.2 ppm), consistent with a local source which has been less diluted in the atmosphere (Figure 3). The far-field samples have lower CO2ff mole fractions (mean 2.6 ± 2.4 ppm, 10th and 90th percentiles of −0.5, 7.2 ppm), as might be expected for air which has traveled hundreds of kilometers from the source region, hence diluting the pollution plume.

[28] At SDZ, the sampling spans a much shorter time period (Sept 2009–May 2010) and we have a total of only 32 measurements. Δ14CO2 is often very low at this site (Figure 2), and CO2ff is large, with a mean of 6.7 ± 10.5 ppm and 10th and 90th percentiles of −1.8, 31.3 ppm (Figure 3). Two dominant wind regimes occur at this site. Winds from the southwest bring air from the Beijing region and North China Plain, which is typically quite polluted, and we observe a mean CO2ff of 10 ± 1 ppm. Winds from the more rural northeast usually have less CO2ff, with a mean of 3 ± 7 ppm. As the SDZ data set is small, and we do not see differences in RCO:CO2ff between the southwest and northeast wind directions, we treat the SDZ samples as a single data set.

3.2. Partitioning of CO2 Into CO2ff and CO2bio

[29] For each data set, we approximate the fraction of CO2 enhancement due to CO2ff as fff (section 2.5). The fossil fraction of the source CO2 mixture (fff) for all TAP samples is 60% (Table 1), indicating that slightly over half of the CO2 enhancement at TAP is from CO2ff and the remainder from other CO2 sources. For this location, the influence of ocean CO2 exchange must be quite small, since the air masses have typically passed over the Asian continent, only briefly encountering the Yellow Sea before arriving at TAP. The δ13CO2 values in the TAP also do not indicate significant ocean carbon exchange relative to background (data not shown). Therefore, net terrestrial sources likely make up the remaining ∼40% of the CO2enhancement. Terrestrial sources could include contributions from respiration, open biomass burning and biofuel use (as an alternative or additive to gasoline, along with burning of wood, crop residue, dung, etc. for heating and cooking), and photosynthetic uptake which reduces the net terrestrial source. The TAP far-field samples are even more dominated by terrestrial sources, with fff of 20%. In winter, fffincreases, as might be expected, when photosynthesis is close to zero and respiration is much weaker, but there is still a non-negligible contribution from terrestrial sources, constituting about 10% of the total. The winter TAP far-field samples yield fff of 70%. This is similar to the SDZ samples, which are all from the winter season, and have fff of 80%. The winter terrestrial CO2 source likely includes contributions from biofuel use, and plant, soil, human and animal respiration. For example, Wang et al. [2010] calculated that human respiration in China emits ∼0.2 GtC per year (∼10% of the magnitude of CO2ff), although it may be a smaller percentage in the industrialized region of Northern China impacting our samples. Additionally, China mandates ∼10% bioethanol in transportation fuels (increasing from 7% in 2004 to 15% in 2020) [Slingerland and van Geuns, 2005].

[30] We now more explicitly examine the CO2bio contribution to each individual sample measured. First, we simply subtract CO2ff from the observed CO2 values to reveal what the local seasonal cycle and trend would be in the absence of recently added CO2ff (Figure 5). The result shows an annual increase in background CO2 of about 2 ppm per year, consistent with the global CO2 increase over the same period [Conway et al., 2011]. There is a strong seasonal cycle, with the lowest values in summer, and highest values in winter, consistent with biospheric exchange. We next subtract background CO2 from this value to obtain CO2bio, which is the contribution (positive or negative) of the biosphere to the local signal. Again, we see photosynthetic drawdown of CO2 in most of the summertime samples, and a positive biospheric contribution in the winter.

image

Figure 5. (top) Measured CO2 mole fractions for TAP and SDZ (open diamonds), and after subtracting CO2ff (closed diamonds). The marine boundary layer background CO2 values are shown as a black line. (bottom) CO2bio for TAP and SDZ.

Download figure to PowerPoint

[31] A few of the samples show large positive CO2bio contributions in summertime, which is initially counterintuitive since photosynthetic drawdown is expected to be strong in summer causing large negative CO2bio contributions. We examine the four samples with CO2bio contributions of greater than 12 ppm in detail to understand possible causes of these anomalous values. All four samples were previously identified as having a local Korean signature by both the back-trajectory and SF6 methods, and show large enhancements over background in most of the measured trace gas species (Table 2). This is most notable for CH4 and N2O, suggesting agricultural sources. RCO:CO2ff (discussed in section 3.3) calculated individually for each of these four samples ranges from 11 to 19 ppb/ppm, also consistent with a local source, and straddling the middle of the range of instantaneous (single-sample) RCO:CO2ff values observed for the local samples (16 ± 7 ppb/ppm, section 3.4). We considered the possibility that there was a nuclear 14CO2 source (nuclear reactor or burning of 14C enriched waste) “contaminating” these samples, resulting in an underestimate of CO2ff and therefore an overestimate of CO2bio. While we cannot completely rule this out, it seems unlikely, since we would expect that a variable 14CO2 source would occasionally result in 14CO2 values higher than background, which we do not observe. Furthermore, the nuclear reactors in South Korea are all pressurized water reactors (PWRs), which produce most of their 14C as 14CH4, with only a small expected contribution to 14CO2 [Graven and Gruber, 2011]. We also rule out a strong biomass burning source, since open biomass burning produces much higher RCO:CO2ff values (∼50–100 ppb/ppm) than observed. It is most likely that these CO2bio values are the result of strong agricultural sources, on dates when harvests have dramatically decreased photosynthetic uptake, while the respiration source remains strong. At least two of these samples are consistent with broad-scale Korean harvests, which typically begin in mid-September [Chung et al., 2004].

Table 2. Enhancement Over Background (Δ) of Various Species for Four TAP Samples With High CO2bio Contributionsa
Date27 Sep 200621 Oct 200629 May 20088 Aug 2009
  • a

    The measured mole fraction in each sample is given in parentheses where available. All four samples are identified by back-trajectories and SF6 mole fraction as having a local Korean source.

Hour (local time)14:0015:0015:0014:00
CO2ff (ppm)15.7 ± 1.012.0 ± 1.018.6 ± 1.14.6 ± 1.0
CO2bio (ppm)16.7 ± 1.015.8 ± 1.024.5 ± 1.020.6 ± 1.1
ΔCO2 (ppm)32.4 (409.4)27.7 (407.8)43.1 (432.1)25.1 (406.1)
ΔCO (ppb)209.9 (315.7)229.6 (345.2)286.8 (409.5)48.7 (134.3)
ΔCH4 (ppb)869.1 (2703.1)106 (1946.5)394.9 (2229.6)159.8 (1982.4)
ΔSF6 (ppt)6.0 (12.1)8.6 (14.7)12.8 (19.4)5.2 (12.1)
ΔN2O (ppb)2.7 (323.3)2.5 (323.2)6.1 (328.3)2.2 (325.0)
RCO:CO2ff (ppb/ppm)13.3 ± 0.919.2 ± 1.615.4 ± 0.910.7 ± 2.4

[32] Overall, our results indicate that CO2bio contributes substantially to the observed CO2 enhancement at both TAP and SDZ, even in winter, and that even for sites with large pollution signals, CO2 enhancements cannot be reliably interpreted as entirely due to CO2ff.

3.3. CO Enhancements and Emission Ratios in China

[33] There is a large and variable enhancement in CO (ΔCO) relative to background in both the TAP and SDZ samples. We see strong correlations between ΔCO and CO2ff, with a distinct difference in the observed regression slopes for samples previously identified as local and far-field (Figure 6 and Table 1). Correlations and emission ratios are calculated using a linear least squares method, taking into account the uncertainties in both coordinates [Press and Teukolsky, 1992]. RCO:CO2ff is calculated as the slope of the regression of ΔCO and CO2ff. For the TAP far-field source, we estimate RCO:CO2ff as 44 ± 3 ppb/ppm (Table 1). The SDZ RCO:CO2ff is slightly higher at 47 ± 2 ppb/ppm, but the two are not significantly different (p = 0.6).

image

Figure 6. Emission ratios (RCO:CO2ff) from TAP and SDZ. TAP samples are split into three subsets described in detail in the text. Error bars are the one-sigma uncertainty in CO2ff; CO measurement uncertainty is smaller than the symbol size.

Download figure to PowerPoint

[34] Our observed emission ratio is consistent with inventory-based estimates of emissions from China. We obtain RCO:CO2ff from inventories by convolving each reported national CO inventory with the CDIAC national total CO2ff inventory (Figure 7) [European Commission, 2009; Kopacz et al., 2009; Zhang et al., 2009; Tanimoto et al., 2008; Ohara et al., 2007; Yumimoto and Uno, 2006; Allen et al., 2004; Wang et al., 2004; Tan et al., 2004; Streets et al., 2003; Carmichael et al., 2003; Heald et al., 2003]. Our observed overall 2004–2010 RCO:CO2ff of 44 ± 3 (Figure 6 and Table 1) is consistent with two recent inventory estimates of 48 ppb/ppm for 2005 [Tanimoto et al., 2008] and 43 ppb/ppm for 2006 [Zhang et al., 2009]. Although our observed year-to-year differences in RCO:CO2ff are not statistically significant, they track the reported inventory decrease from 2005 to 2006 quite well (Figure 7). In contrast, the EDGAR CO inventory for 2005 appears to be substantially too low, with a value of 29 ppb/ppm [European Commission, 2009], and the EDGAR inventory for 2000 is also substantially lower than other inventories for the same year, at 40 ppb/ppm (Figure 7).

image

Figure 7. RCO:CO2ff for China. Left y axis shows the RCO:CO2ff values in units of ppbCO/ppmCO2, and the right y axis indicates the corresponding inverse values RCO:CO2ff−1 in units of molCO2/molCO. The TAP value for 2010 is excluded as it is dominated by a single outlier and is off-scale at 220 ± 110 ppb/ppm. Open green symbols are measured RCO:CO2 from other observational studies [Suntharalingam et al., 2004; Wang et al., 2010], and closed green symbols are the same data corrected to RCO:CO2ff assuming that 20% of the CO2 enhancement is from sources other than CO2ff. EDGAR emission inventories [European Commission, 2009] are black circles, other inventories are black diamonds [Kopacz et al., 2009; Zhang et al., 2009; Tanimoto et al., 2008; Ohara et al., 2007; Yumimoto and Uno, 2006; Allen et al., 2004; Wang et al., 2004; Tan et al., 2004; Streets et al., 2003; Carmichael et al., 2003; Heald et al., 2003].

Download figure to PowerPoint

[35] We also compare our RCO:CO2ff results with other observational studies (Figure 7). Two studies have made observations of the ratio of CO to total CO2 (RCO:CO2), making the assumption that CO2 enhancements in polluted plumes are due only to CO2ff. Suntharalingam et al. [2004]made observations from an aircraft off the coast of China as part of the TRACE-P experiment, and obtained an RCO:CO2 estimate of 55 ppb/ppm. Wang et al. [2010] used wintertime measurements of CO and CO2 from 2004 to 2008 at Miyun station, about 35 km southwest of SDZ in the same valley, and obtained RCO:CO2 estimates of 35–58 ppb/ppm (screened for air masses from the North China Plain region, their “NCN samples”). They calculated that human respiration could bias RCO:CO2 low by ∼9% at their site, but did not correct for this in their reported RCO:CO2 values, and assumed that no other sources contributed to CO2 variability during the winter. However, SDZ sees pollution events of similar magnitude to those at Miyun (CO2 and CO enhancements of up to 40 ppm and 1500 ppb, respectively), and yet other CO2 sources account for 20% of the CO2 enhancement (section 3.2). In the TAP far-field samples, where the enhancements are smaller, other CO2 sources contribute 25% in winter.

[36] The raw RCO:CO2 from both previous observational data sets are consistently slightly lower than inventory estimates, and lower than our central estimate based on observed RCO:CO2ff, particularly in 2007 and 2008. However, we can also estimate RCO:CO2ff from the RCO:CO2 observations, increasing RCO:CO2 by 20% to account for CO2 from other sources observed in our data. With this correction, the Suntharalingam et al. [2004] observation is closer to the inventory estimates for 2001, and the Miyun observations show slightly better overall agreement with our observations, particularly in 2007 and 2008. These results suggest that estimates of RCO:CO2 should be interpreted with caution, since even in highly polluted regions, the effect of CO2from non-fossil sources can substantially bias the results.

[37] Our observations, the inventories, and the RCO:CO2 observations are all broadly consistent in showing a downward trend in RCO:CO2ffuntil 2006 that is indicative of improved fossil fuel combustion efficiency, and then a flattening from 2007 onwards, although it should be noted that the year-to-year differences in the observational data sets are not statistically significant. Our results suggest that the recentTanimoto et al. [2008] and Zhang et al. [2009] inventories for China are accurate, and that efforts to improve combustion efficiency and reduce CO emissions have been effective.

3.4. CO Enhancements and Emission Ratios in Korea

[38] The TAP local samples show two distinct populations. One has a low RCO:CO2ff of 13 ± 3 ppb/ppm, and the other has a substantially higher RCO:CO2ff of 58 ± 12 ppb/ppm. The population with the low RCO:CO2ff values is characterized by large CO2ff values and variable RCO:CO2ff (Figure 6), consistent with local emissions from varying nearby sources in South Korea that are not yet strongly diluted or well mixed. In particular, occasional samples may be dominated by emissions from a large coal-fired power plant about 20 km to the northeast of TAP (Taean Power Station). Power plants typically produce very little CO (due to efficient combustion and pollution controls), and therefore samples dominated by emissions from this power plant can be expected to have very low RCO:CO2ff values, as observed in a few local samples in Figure 6. The second population with high RCO:CO2ff values is likely dominated by air from North Korea, which is believed to produce much higher RCO:CO2ff than South Korea [Zhang et al., 2009].

[39] There has been less study of South Korean than Chinese emissions, particularly for CO. Figure 8 shows the available inventory estimates and our observations for South Korea. Suntharalingam et al. [2004] estimated RCO:CO2ffof 15.4 ppb/ppm for Korea in 2001 from inventories, but did not make any observations of Korea-specific plumes. The EDGAR 2005 inventory estimates RCO:CO2ff for Korea as 11.6 ppb/ppm, and Zhang et al. [2009]obtained an inventory-based estimate of 2.6 ppb/ppm for 2006 (using CDIAC CO2ff). Our observed RCO:CO2ff for South Korea of 13 ± 3 ppb/ppm is consistent with Suntharalingam et al. [2004] and EDGAR inventories, but markedly higher than the Zhang et al. [2009]inventory. The Q. Zhang estimate uses a different data source (National Institute of Environmental Research of Korea) and could be missing some source categories, since it is much lower than inventory-based RCO:CO2ff for any other region, including Japan, Europe and North America. Our data do not indicate a temporal trend in RCO:CO2ff for South Korea, although the small data set and consequentially large uncertainties do not preclude a small trend. Our observed RCO:CO2ff value is similar to those observed for other developed and urbanized nations. For example, observations from Sacramento, USA in March/April 2009 showed RCO:CO2ff of 14 ± 2 ppb/ppm [Turnbull et al., 2011], observations from the northeast USA from 2004–2010 show RCO:CO2ff of 12.5 ppb/ppm (Miller et al., submitted manuscript, 2011), observations from Heidelberg, Germany from 2002 to 2009 had a mean RCO:CO2ff of 14.6 ppb/ppm [Vogel et al., 2010], and observations from the Netherlands for 2006–2009 had RCO:CO2ff of 5–15 ppb/ppm [Van Der Laan et al., 2010].

image

Figure 8. RCO:CO2ff for South Korea from our observations and from inventories. Left y axis shows the RCO:CO2ff values in ppbCO/ppmCO2, and the right y axis indicates the corresponding inverse values RCO:CO2ff−1 in units of molCO2/molCO (inverse scale). EDGAR [European Commission, 2009] inventory is shown as a black circle, other inventories are black diamonds [Zhang et al., 2009; Suntharalingam et al., 2004].

Download figure to PowerPoint

[40] The data set for North Korea is very small (19 samples from 2004 to 2010), and the samples likely include some influence of South Korean emissions. Nevertheless, our RCO of 58 ± 12 ppb/ppm (Table 1, TAP local high RCO samples) is consistent with the Zhang et al. [2009] inventory for 2006 of 66 ppb/ppm, but is much higher than the EDGAR 2005 estimate of 36 ppb/ppm.

3.5. Evaluation of Bottom-Up CO2ff Inventories Using Model-Data Comparison

[41] The modeled TAP CO2ff mean of 3.9 ppm is similar to our observed mean of 4.4 ppm, and for the TAP far-field samples, the means of model and observations are both 2.6 ppm. However, the model does a poor job of predicting CO2ff on individual days and the modeled distribution is more strongly skewed than the observations (Figure 9). At SDZ, the model performs markedly better (Figure 9), with a modeled mean of 7.0 ppm CO2ff, versus the observed mean of 6.7 ppm, and r2 for observed versus modeled CO2ff of 0.65.

image

Figure 9. (left) Modeled versus observed CO2ff values for TAP far-field and SDZ. Grey line is the 1:1 line. (right) Histograms of observed and modeled CO2ff for TAP far-field and SDZ. Negative observed CO2ff values are included in the zero bin.

Download figure to PowerPoint

[42] Model representation error due to the proximity to Seoul does not appear to explain the difference between the model and observations for TAP. When we separate the model results into far-field and local contributions, we find that the samples we had previously defined as far-field are indeed dominated by far-field influences in the model, and no significant influence of local emissions is seen. Despite the failure of the model to adequately represent the daily observations, the model predicts the range of values at TAP quite well (Figure 9) as well as the interannual variability in mean CO2ff values (Figure 10). We believe that the dominant source of error impacting the sample-by-sample model versus observed CO2ff at TAP is the atmospheric transport. This is likely because the CO2ff emissions are not uniformly distributed, but are focused into pollution plumes from urban and industrial regions and thus relatively small errors in the model transport can bias their detection (or not) at distant receptor sites such as TAP. It also appears that the model may not be dispersive enough in some seasons, driving the high modeled values in some samples (Figure 9). On an annual basis, these errors appear to be averaged out. In contrast, SDZ is much closer to the sources and small biases in model transport will exert less influence on detection at the receptor site.

image

Figure 10. Annual mean CO2ff in TAP far-field samples from observations (red); the model (black); the model with emissions held flat at 2004 levels (gray). Error bars are the standard deviation of the individual observations.

Download figure to PowerPoint

[43] At SDZ, where the model appears to perform reasonably well, we can evaluate how well the model predicts the observed CO2ff values on a sample-by-sample basis. The slope of modeled versus observed CO2ff at SDZ is 0.9 ± 0.2, not significantly different from 1.0 (the result if the model perfectly reflects the observed values). A paired Student's t-test of the individual SDZ modeled and observed values gives p = 0.15, also indicating that they are not significantly different from one another.

[44] The poor representation of individual TAP observations by the model precludes examining the TAP data in the same manner as for SDZ. We can nevertheless use the model to examine trends in the annual mean of the TAP far-field data.Figure 10 gives annual mean observed and modeled CO2ff values for the TAP far-field samples for the years 2004 to 2010. For most years, we have 24–30 samples during the year, with the exception of the first and last years, 2004 and 2010, when we have only 5 and 10 samples, respectively. Two model runs were performed: the “standard” model with CO2ff emissions with the annual increases from CDIAC of 6–10% yr−1 for China, and the “flat” model holds CO2ff emissions constant at 2004 levels for the entire time period.

[45] Emissions of CO2ff from China reportedly increased 63% from 2004 to 2010 [Gregg et al., 2008; Boden et al., 2010; BP, 2010] (and extrapolated 2010 BP value). Initial expectations are that this increase would be observed in the TAP far-field samples (identified as having a Chinese source), yet no such trend is apparent in the individual CO2ff observations (Figure 3), and neither the observed annual mean CO2ff values, nor the modeled annual mean values show an upward trend in emissions despite the imposition of substantially increasing emissions in the “standard” simulation. However, the interannual variability in the observed annual mean CO2ff is reflected in the modeled results, increasing from 2004 to 2007, and then dropping in 2008–2010, and the modeled SDZ mean is consistent with our observed SDZ mean (Figure 10). Since the standard model CO2ff emission flux increases through time, the lower modeled values in 2008–2010 must be due to interannual variability in atmospheric transport of the emitted CO2ff, and cannot be ascribed to reduced CO2ff emissions. The flat model run follows the same pattern, but values are lower than either the standard model or observations (Figure 10). We use two different tests to compare the observed and standard model mean CO2ff values: a paired Student's t test and the Wilcoxon signed rank test [Press, 2007]. The Wilcoxon signed rank test is non-parametric and does not require the underlying data to be Gaussian. Both tests indicate that there is no significant difference between the observed and standard model annual mean CO2ff values (p = 0.22 and p = 0.25 for paired Student's t test and Wilcoxon, respectively). However, the observed annual mean values are significantly higher than the flat model (p = 0.02 for both tests). That is, our observations are consistent with an increase in emissions of 8% per year from 2004 to 2010, and are not consistent with flat emissions at the 2004 level.

[46] Furthermore, both our standard model and flat model use CO2ff emissions flux with Chinese emissions distributed according to the provincial distribution obtained by Gregg et al. [2008]. Other fossil fuel inventories, such as CarbonTracker [Peters et al., 2007] and EDGAR [European Commission, 2009] distribute the Chinese national emissions according to population (except for power plant emissions which are placed in their reported physical location), and this results in substantially lower emissions in the Northern China region to which our TAP far-field and SDZ samples are most sensitive. We find that our observed TAP far-field annual mean CO2ff values are significantly higher than a modeled result using CarbonTracker CO2ff emission flux (p = 0.01), which differs from our standard model only in the spatial distribution of emissions within China.

4. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Results and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[47] The TAP and SDZ sites show large and variable CO2ff and CO enhancements. The majority of samples from the TAP site have footprints dominated by emissions from Northern China, and a smaller number “see” air which has the imprint of North and/or South Korea.

[48] Samples from SDZ and TAP far-field (China) have RCO:CO2ff values consistent with available inventory estimates and other observational data sets from China, suggesting that recent efforts to improve combustion efficiency in China have been successful. Our observations indicate RCO:CO2ff for South Korea consistent with inventories and RCO:CO2ff for other industrialized nations.

[49] CO2 enhancement at TAP is about half from anthropogenic CO2ff and half from other sources. There are substantial contributions of CO2bio in our samples at TAP and SDZ, even in winter, indicating that it cannot be reliably assumed that CO2 enhancement in polluted air masses is entirely due to CO2ff.

[50] Using the best available CO2ff emission inventory for China, including spatial distribution according to reported Chinese provincial emission estimates, and reported emissions increases of 8% yr−1 (63% from 2004 to 2010), a modeled prediction at our sites is consistent with our observations. Although it is tempting to conclude that the CO2ff emission flux we use in the model and reported emissions are accurate, we are at this time not able to sufficiently assess the magnitude of biases in model transport to confirm this. Potential biases in the model transport include the underlying meteorology, particularly wind speed and boundary layer height, and the parameterization of vertical mixing in the model. Nevertheless, this result indicates the promise of top-down atmospheric observations to constrain urban and regional fluxes, as modeled transport is improved, and the observational network becomes denser.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Results and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[51] This work would not have been possible without the NOAA/ESRL Cooperative Sampling Network run by the Carbon Cycle Greenhouse Gases group, and the joint sampling by CMA at SDZ and WLG, and at TAP by KCAER (CATER 2006-3103). Molly Heller was instrumental in obtaining the air samples. Pat Lang and Kelly Sours made the greenhouse gas measurements. Chad Wolak and Patrick Cappa prepared the14CO2 samples. Much helpful advice from all in the Carbon Cycle Greenhouse Gas group ensured the data quality.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Results and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Results and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Auxiliary material for this article contains two figures.

Auxiliary material files may require downloading to a local drive depending on platform, browser, configuration, and size. To open auxiliary materials in a browser, click on the label. To download, Right-click and select “Save Target As…” (PC) or CTRL-click and select “Download Link to Disk” (Mac).

Additional file information is provided in the readme.txt.

FilenameFormatSizeDescription
jgrd17605-sup-0001-readme.txtplain text document3Kreadme.txt
jgrd17605-sup-0002-fs01.pdfPDF document128KFigure S1. Example FLEXPART footprints for the different regions.
jgrd17605-sup-0003-fs02.pdfPDF document23KFigure S2. FLEXPART footprint for 22 October 2004.
jgrd17605-sup-0004-t01.txtplain text document1KTab-delimited Table 1.
jgrd17605-sup-0005-t02.txtplain text document1KTab-delimited Table 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.