Derivation of tropospheric methane from TCCON CH 4 and HF total column observations

The Total Carbon Column Observing Network (TCCON) is a global ground-based network of Fourier transform spectrometers that produce precise measurements of column-averaged dry-air mole fractions of atmospheric methane (CH4). Temporal variability in the total column of CH4 due to stratospheric dynamics obscures fluctuations and trends driven by tropospheric transport and local surface fluxes that are critical for understanding CH 4 sources and sinks. We reduce the contribution of stratospheric variability from the total column average by subtracting an estimate of the stratospheric CH 4 derived from simultaneous measurements of hydrogen fluoride (HF). HF provides a proxy for stratospheric CH4 because it is strongly correlated to CH 4 in the stratosphere, has an accurately known tropospheric abundance (of zero), and is measured at most TCCON stations. The stratospheric partial column of CH 4 is calculated as a function of the zonal and annual trends in the relationship between CH4 and HF in the stratosphere, which we determine from ACE-FTS satellite data. We also explicitly take into account the CH4 column averaging kernel to estimate the contribution of stratospheric CH 4 to the total column. The resulting tropospheric CH 4 columns are consistent with in situ aircraft measurements and augment existing observations in the troposphere.


Introduction
The most abundant hydrocarbon in the atmosphere, methane (CH 4 ) is a driver of background tropospheric chemistry and a significant radiative forcing gas.However, the long-term trends of atmospheric mixing ratios and fluctuations in the annual growth rate remain unexplained due to an incomplete understanding of its sources and sinks.
Analyses of temporal and geospatial trends of CH 4 require precise, continuous measurements with adequate spatial coverage.Several such monitoring networks, such as World Meteorological Organization (WMO) Global Atmospheric Watch and National Oceanic and Atmospheric Administration (NOAA) Global Monitoring Division, have measured methane for decades.These sites are often in locations primarily intended for background observations, and measurements are confined to the surface, primarily within the boundary layer.The Total Carbon Column Observing Network (TCCON), a ground-based network of near-infrared (NIR) Fourier transform spectrometers (FTS), measures dryair mole fractions (DMFs) of several atmospheric trace gases, including CH 4 , integrated over the entire atmospheric column.The column measurements are sensitive to the free troposphere in addition to the surface, which can allow for better separation of transport from local emissions.Additionally, total column measurements are less sensitive to vertical transport and mixing, and thus meridional or zonal gradients in column measurements can be used to characterize regional-scale fluxes (Yang et al., 2007;Wunch et al., 2011a;Keppel-Aleks et al., 2011).Several TCCON stations are near in situ sites that provide surface, tall tower, and aircraft measurements, which we use to compare the final tropospheric column-averaged CH 4 DMFs.
Tropospheric trends of CH 4 are obscured in total column measurements by variability originating in the stratosphere, especially by vertical shifts of the tropopause.Several methods for accounting for stratospheric variability have been proposed, including incorporating the compact relationship between CH 4 and another chemical tracer in the stratosphere (e.g., Washenfelder et al., 2003;Payne et al., 2009;Angelbratt et al., 2011;Sepúlveda et al., 2012Sepúlveda et al., , 2014;;Wang et al., 2014).Washenfelder et al. (2003) estimate the contribution of variations in stratospheric CH 4 as the product of the hydrogen fluoride (HF) column-averaged DMF and the CH 4 -HF relationship, calculated from the Halogen Occultation Experiment (HALOE) satellite and the JPL MkIV Interferometer data.Wang et al. (2014) similarly use the relationship between stratospheric nitrous oxide (N 2 O) and CH 4 and the fact that tropospheric N 2 O is well known to infer stratospheric variations in N 2 O, and hence CH 4 .Angelbratt et al. (2011) remove the Network for the Detection of Atmospheric Composition Change (NDACC) CH 4 total column variability with a multiple regression model that parameterizes anomalies of several measurements, including HF, carbon monoxide, ethane, and tropopause height.Sepúlveda et al. (2012) use the retrieval algorithm PROFFIT to infer vertical CH 4 profiles directly from the absorption line shapes of the midinfrared (MIR) FTS spectra measured within NDACC, comparing the resulting tropospheric columns with those calculated with a HF proxy method.Extending this study to additional sites, Sepúlveda et al. (2014) estimate a precision of 0.5 % and a systematic error of 2.5 % for daily mean values of tropospheric CH 4 derived from profile retrievals on the MIR NDACC measurements.
Vertical profile retrievals using the TCCON spectra are more difficult than those using NDACC MIR spectra because the NDACC measurements apply spectral filters to narrow the spectral coverage, yielding higher signal-to-noise ratios at higher spectral resolution, at the expense of making simultaneous measurements of some other gases.In addition, the line strengths in the MIR are generally higher and doppler widths are smaller, allowing more degrees of freedom in the vertical retrieval.Nevertheless, profile retrievals are more sensitive to error in the instrument and assumed spectroscopic line shapes than profile-scaling retrievals.Quantifying the variability of stratospheric CH 4 via a chemical tracer is, however, not without challenge, as this method is sensitive to errors in the representation of the relationship between that tracer and CH 4 in the stratosphere and knowledge of their respective averaging kernels.In addition, this method provides no information about vertical structure within the troposphere.
To determine the stratospheric CH 4 component of the FTS-retrieved total column, we propose to use its relationship with HF, which is measured at almost all TCCON sites.Stratospheric CH 4 has a nearly linear inverse relationship with HF, which exists almost exclusively in the stratosphere (Luo et al., 1995;Washenfelder et al., 2003).The photodissociation of chlorofluorocarbons (CFCs) and the resulting carbonyl products produce free fluorine, which can then in turn react with CH 4 and water vapor (H 2 O) to produce HF, the most stable reservoir species of fluorine in the stratosphere (Luo et al., 1994).The reactions producing HF occur in the middle-high stratosphere, leading to a uniformly increasing vertical profile (Luo et al., 1995).CH 4 , by contrast, is transported from the troposphere and is destroyed by O(1D), hydroxyl, and chlorine free radical-initiated oxidation.The resulting nearly linear relationship between HF and stratospheric CH 4 , which is seasonally and zonally consistent, makes HF a useful proxy for the contribution of stratospheric variability to the CH 4 total column.

Derivation of tropospheric CH 4 columns
TCCON FTS retrievals are conducted with the GFIT nonlinear least-squares fitting algorithm, which determines a vertical scale factor (γ ) of an a priori vertical profile (x a ) based on the best spectral fit of the solar absorption signal.The scaled profile is then vertically integrated, and the resulting column abundance is divided by the vertical column of dry air, calculated using the retrieved column of oxygen (O 2 ) (Wunch et al., 2010(Wunch et al., , 2011a)).
The retrieved integrated column of CH 4 can be expressed as a first order Taylor expansion about the solution γ CH 4 c a CH 4 (Rodgers and Connor, 2003) such that, where ĉ is the retrieved column, γ CH 4 is the retrieved profile scale factor of CH 4 , x is the true profile, and x a CH 4 and c a CH 4 are the a priori vertical profile and column-integrated CH 4 , respectively.We define § as an operator that represents the pressure-weighted integration of the profile: where a is the FTS column averaging kernel, dependent on solar zenith angle and pressure, h is the pressure-weighting function, such that ĉ = h T x (Connor et al., 2008;Wunch et al., 2011b), and i is the index of pressure levels from the surface to the highest level, N. When the vertical column includes water vapor, such as in the case of the priors, the pressure-weighting function incorporates the H 2 O profile to convert x to dry-air mole fractions.
To isolate the tropospheric column of CH 4 , we assume a linear relationship between CH 4 and HF in the stratosphere such that where c trop CH 4 represents the pressure-weighted DMF averaged over the tropospheric column, u is a unity vector the length of the number of vertical levels in the total column retrieval integration, and β is the time-dependent CH 4 -HF slope in the stratosphere.Integrating the vertical profiles, the columnaveraged form of this relationship becomes where c is the total column DMF of the respective trace gases.The βc HF term is negative and represents the amount of CH 4 that has been destroyed in the stratosphere, rather than the stratospheric partial column of CH 4 .We also assume that the a priori CH 4 profiles can be approximated using the linear relationship between CH 4 and HF.By combining Eqs. ( 3) and ( 4), as well as their analogs for the a priori profiles, into Eq.( 1), we derive a tropospheric column-average DMF: Ideally, x HF would be derived from the equivalent of Eq. (1) for HF, but doing so would require inverting the pressureweighted averaging kernel, which does not have a unique solution.Thus, in order to solve Eq. ( 5), we must assume that x HF = xHF = γ HF x a HF and, accordingly, that the shape of the HF profile is known.In general, this is a reasonable assumption because the vertical profile is governed mainly by wellcharacterized chemical production, and, as previously stated, increases uniformly.However, this solution has limitations when the scaled profile deviates from the true profile, such as in the polar vortex.
Substituting γ HF x a HF for x HF , the tropospheric column of CH 4 is derived as follows: All of the terms on the right hand of the equation can be generated from the TCCON data set except for β, which we derive from satellite data.Equation ( 6) can be applied to determine tropospheric DMFs of trace gases other than CH 4 that are correlated with HF in the stratosphere because it does not require assumptions about the relationship between the averaging kernels of the respective gases and is thus a more general approach than that of Washenfelder et al. (2003).

Measurement uncertainties
The c trop CH 4 error is calculated by propagating the uncertainties of the retrievals, which in Eq. ( 6) are associated with the vertical scale factors, and β, which is described in Sect.2.2.These errors are propagated as the sum of the squares of the standard errors for each term.Because the vertical scale factor errors are derived assuming that the residuals from the spectral fits have a Gaussian distribution, systematic artifacts that are stable from spectrum to spectrum inflate the errors calculated from the spectral fit.To derive the measurement precision, the uncertainties are scaled to account for the variations in DMFs calculated from successive spectra within 5 min.The mean and median tropospheric CH 4 precisions vary from 0.04-1 and 0.01-0.03%, respectively, at individual sites and are 0.1 and 0.004 % across all sites and years.By comparison, the precision in Washenfelder et al. (2003) is 0.5 %, although this improvement is partially attributable to advancements in the retrieval methodology and instrumentation.
The data have been corrected for laser sampling errors (Dohe et al., 2013;Messerschmidt et al., 2010), and the associated uncertainties of those corrections are summed in quadrature with the measurement precision.

Determination of CH 4 -HF slope
Vertical profiles of CH 4 and HF mole fractions were developed from level 2, version 3.0 and 3.5 retrievals from the Atmospheric Chemistry Experiment Fourier Transform Spectrometer (ACE-FTS) instrument on the Canadian SCISAT-1 satellite.SCISAT-1 orbits in low Earth orbit with an inclination of 74 • , offering coverage of tropical, mid-latitude and polar regions from 85 • N to 85 • S (Bernath, 2005;De Mazière et al., 2008;Mahieu et al., 2008;Waymark et al., 2014).Data were taken from February 2004 through September 2010 for version 3.0 and October 2010 through December 2012 for version 3.5.These data sets differ only in that the a priori pressure and temperature profiles in the latter are taken from the global Canadian Meteorological Center model rather than the regional one, which began to provide unphysical profiles starting October 2010 (Waymark et al., 2014).The ACE-FTS profiles are additionally filtered to exclude occultations with physically unlikely profiles and all individual CH 4 and HF retrievals with statistical fitting errors above 5 %.Because the HF abundance in the troposphere is essentially zero, any coincident retrievals of CH 4 and HF were assumed to reside in the stratosphere; therefore we did not designate a pressure level threshold to isolate the stratosphere.Data above 70 km were excluded for consistency with TCCON retrievals, although CH 4 concentrations are generally depleted at that altitude.The CH 4 -HF relationship exhibits a strong altitude dependence, with steeper CH 4 -HF slopes in the upper stratosphere (Fig. 1).Annual slopes follow the long-term trend from Washenfelder et al. (2003), given the expected trajectory of HF concentrations in the stratosphere (Fig. 2).
Tracer-tracer relationships in the stratosphere tend to be dependent on latitude, with the tropics exhibiting different slopes than the mid-latitude "surf zone" and the polar regions (Luo et al., 1995).While ACE-FTS coverage at high latitudes is extensive, tropical coverage is sparse; thus, to ensure a large enough number of data points in the tropics for robust statistical analysis, we binned CH 4 and HF mole fractions in 30 • zonal bands.The tracer relationship demonstrates a clear zonal trend: the slopes are less steep in lower latitudes, and the Northern Hemisphere slopes are more steep than their zonal counterparts in the Southern Hemisphere (Figs. 1, 2).To determine statistically robust values for β, the CH 4 -HF slope was computed for bootstrap subsamples of 1000 individual retrievals from each year and zonal band.In order to minimize the effect of outliers in the determination of the slope, we applied an iteratively re-weighted least squares regression with a Tukey's bi-weight function, weighting data points by pressure (Hoaglin et al., 1983).The means and 2σ standard deviations of the resulting probability distributions were taken respectively as the values and errors of β (Table 1).For 2013, we calculated the annual growth rate of the CH 4 -HF ratio in the northern mid-latitude region (30-60 • N), chosen because the surf zone is well-mixed and thus has the most robust tracer relationships, and added it to the respective zonal values for 2012.The error for β in 2013 was computed as the sum in quadrature of the error for β in 2012, the standard error of the annual growth rate, and the 2σ standard deviations of the interannual variability of each zonal band.While temporal trends in β do indicate seasonal variability, the impact on the slopes is not sufficiently statistically robust from year to year to incorporate a seasonally varying β.The sensitivity of the tropospheric methane calculation to β differs by site, but generally varies by 0.1-1 ppb for β of 10.

Validation of methodology
Equation ( 6) incorporates two major assumptions: that the CH 4 -HF relationship is linear, and that the retrieved HF column is a close approximation to the true HF column.To test these assumptions, we compared tropospheric CH 4 DMFs derived directly from ACE-FTS CH 4 profiles to those calculated by substituting the most recent TCCON priors (GGG 2014) and ACE-FTS CH 4 and HF profiles into Eq.( 5).For this analysis, the ACE-FTS trace gas profiles up to 70 km interpolated onto a 1 km vertical grid were considered the true profiles x HF and x CH 4 , and assuming γ CH 4 ≈ 1, we solved Eqs. ( 1) and ( 5) for c trop CH 4 .Because the minimum retrieval altitudes were at least 5.5 km and on average 9.5 km, mole fractions of CH 4 and H 2 O near the surface were extrapolated using TCCON priors.Occultations with individual CH 4 , HF and H 2 O errors greater than 10 % were excluded for latitudes poleward of ±30 • .In the tropics, the error threshold was relaxed to 40 % in order to ensure a large enough data set for results to be meaningful.We then compared the calculated tropospheric methane column-averaged DMF to the ACE-FTS CH 4 profiles integrated to the tropopause.For the intercomparison, the integrated ACE-FTS profiles were smoothed with the TCCON CH 4 averaging kernel and priors (Connor et al., 2008;Wunch et al., 2011b).The tropopause altitude was calculated using National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) Reanalysis local noon temperature profiles (Kalnay et al., 1996), from which the TCCON priors are generated, for consistency.
As Fig. 3 illustrates, the temporal and zonal dependencies of the tropospheric methane calculation are well characterized, with a few notable exceptions.The consistency of the bias across years (slope = 0.99-1) indicates that the annual variability of β is accurate, although the asymmetric scatter of the residuals in the northern tropics could be a result of the smaller number of data points included in the determination of β.Additionally, the seasonal variability associated with descent within the polar vortices, not currently captured by the HF priors, accounts for the outliers apparent in higher latitudes.Because the southern polar vortex is stronger and more persistent than in the north, the calculated tropospheric column exhibits a much larger spread in the southern polar zonal band.
The underestimation of tropospheric CH 4 in the Northern Hemisphere and slight overestimation in the Southern Hemisphere is partially a result of several assumptions that are necessary because coincident TCCON measurements do not exist for all zonal regions over the time series.TCCON averaging kernels are highly dependent on the solar zenith angle and surface pressure at the time of measurement; however, the solar zenith angles at the surface during ACE-FTS occultations are close to 90 • , and the surface pressure is unknown.To address the latter, we assume that the pressure at the lowest point in the ACE-FTS profile is the surface pressure.We also use the solar zenith angle calculated at the latitude, longitude and time of the occultation, which, while accurate, does not test one of the main advantages of this methodology, which is to adjust the tropospheric CH 4 calculation for seasonality and latitude, which impacts each zonal band differently and thus creates an offset.
The method validation also requires the assumption that γ CH 4 = 1 in Eqs.(1) and (5).A bias in the a priori profiles between the Northern and Southern Hemispheres would lead to differences in the retrieved values of γ CH 4 , indicating that the TCCON a priori profiles are slightly too low in the Southern Hemisphere relative to the Northern Hemisphere.While γ CH 4 is generally within 1 % of 1, TCCON measurements tend to have larger values for γ CH 4 in the Southern Hemisphere compared to the Northern Hemisphere, and a difference of 0.01 between hemispheres can shift the residuals from the one-to-one line by up to 4 ppb, or about one quarter of the offset in the mid-latitudes.Because of the low sensitivity of the tropospheric CH 4 calculation to changes in β, inter-hemispheric biases in β determined from ACE-FTS would have to be considerable to explain the offset.

Results
Tropospheric column-averaged DMFs were calculated for TCCON sites in Sodankylä (Fig. 4a), Bremen (Fig. 4b), Park Falls (Fig. 4c), Lamont (Fig. 4d), Izaña (Fig. 4e), Darwin (Fig. 4f), Wollongong (Fig. 4g) and Lauder (Fig. 4h) using the 2012 version of the GGG software.Location information for each of these TCCON sites can be found in Table 2.As we would expect, the tropospheric column-averaged DMFs of CH 4 are higher than the total column DMFs.Many of the low outliers in the total column that are a result of the stratospheric variability no longer appear in the tropospheric DMFs.The intraday variability of the tropospheric DMFs are generally equivalent to those of the corresponding total column DMFs, although the tropospheric standard deviations are, in some cases, significantly larger than those of the total column.Sites in the tropics are especially susceptible to both larger errors for a single measurement and larger daily variances due to the higher HF errors caused by H 2 O interference (e.g., Darwin, Fig. 4f).Additionally, the tropospheric calculation removes most of the effects of the seasonal cycle of stratospheric variability.While the magnitude of the impact on the seasonal cycle of CH 4 varies from site to site, the tropospheric column calculation generally shifts the peak of CH 4 from late fall to winter and the minimum from spring to late summer (Fig. 5).At Lamont and Park Falls, this impact is especially apparent, with 2-month lags in the maxima and minima in the tropospheric versus total columns.The detrended seasonal cycles of the tropospheric CH 4 also exhibit fewer short-term fluctuations, except in the case of Izaña, which is located on a mountain at about 2.4 km and thus is more sensitive to the free troposphere.

Comparison to Washenfelder method
The derivation introduced here improves on the previous calculation of Washenfelder et al. (2003) by explicitly including the CH 4 averaging kernels in the estimate of stratospheric loss and including the recent ACE-FTS satellite data set, which allows for the analysis of temporal and zonal dependencies.To assess the impacts of these additions to the tropospheric CH 4 column, we calculated the tropospheric CH 4 DMFs using the Washenfelder et al. ( 2003) derivation (Eq.4) and the annual northern mid-latitude values of β (Table 1, column 6) for all sites.The daily standard deviations tend to decrease modestly with the updated methodology (Fig. 6), although intraday variability is reduced by up to 40 ppb.The inclusion of the CH 4 averaging kernel adjusts the air-mass dependence of the tropospheric CH 4 calculation, thereby reducing the amplitude of the CH 4 seasonal cycle.This improvement is especially apparent during winter at the high-latitude sites (Fig. 7), because the solar zenith angles are large and, therefore, the CH 4 averaging kernels have a strong dependence on altitude.Additionally, calculating the CH 4 -HF relationship as a function of latitude allows for more meaningful geospatial comparisons.

Comparison to in situ measurements
Following the method for numerical integration of in situ profiles derived in Wunch et al. (2010), smoothed columnaveraged DMFs determined for from several aircraft campaigns (Table 3).Additional information on the TCCON calibration, including instruments, can be found in Wunch et al. (2010), and the WMO calibration scales used for the instruments can be found in Dlugokencky (2005).Aircraft profiles were integrated to the tropopause, determined using the flight temperature profiles.Aircraft errors are calculated as the sum in quadrature of the respective 2σ instrument errors and the estimated uncertainties associated with the profile not reaching the tropopause and the surface.FTS columns were calculated with the aircraft calibration factors for CH 4 determined  et al. (2004).Additionally, because the derivation method is predicted to vary linearly, we calculate the slope assuming a y intercept of zero.
The FTS tropospheric columns show general agreement to each other (Fig. 8), with a slope close to within error of the one-to-one line and a slope and error similar to that of total column CH 4 (Wunch et al., 2010).The tropospheric column calibration curve has a slight hemispheric bias, with Southern Hemisphere sites above the fit line and Northern Hemisphere sites below, with the INTEX-NA campaign, over Park Falls, WI, as the only exception.
Additionally, we compared the tropospheric CH 4 to longterm in situ flask measurements collected at the Atmospheric Radiation Measurement Program (ARM), Southern Great Plains (SGP) site, near the Lamont TCCON station, and analyzed by the NOAA Earth System Research Laboratory (ESRL).Surface measurements are collected from a 60 m tower, typically once per week on one afternoon, and aircraft samples are collected approximately biweekly with a flight path centered over the tower.The integrated aircraft DMFs are generally higher than the TCCON tropospheric columns, which provide a lower bound to the flask measurements (Fig. 9a).The partial aircraft columns, restricted to the free troposphere (approximately 3-7 km), are more consistent with the TCCON tropospheric columns.The calibration curve reinforces the distinction between the aircraft tropospheric and partial tropospheric columns when compared to the FTS DMFs at Lamont (Fig. 9b); while the best fit slopes, calculated as in (Fig. 8), are equal within measurement error, the slope of the free troposphere partial column has a smaller offset from the FTS-aircraft one-to-one line.
In situ measurements at the surface are also useful for regions without large local surface sources and if the troposphere is well-mixed, as in New Zealand.We compared Lauder FTS measurements to in situ data at the Baring Head National Institute of Water and Atmospheric Research of New Zealand (NIWA) facility, about 600 km northeast of the TCCON site (41.4 • S, 174.9 • E, 85 m a.s.l.).The Baring Head flask measurements are collected on a stationary platform at a sampling height of 10 m, analyzed with a flame ionizing detector, and calibrated with the NOAA04 scale (Lowe et al., 1991).The surface measurements are similar to the tropospheric columns, both in terms of the DMF values and the timing of the seasonal cycle (Fig. 10).The Lauder tropospheric columns are somewhat higher in the late summer and early fall, which could be a function of local CH 4 sources near Lauder, changing wind directions impacting the covariance between the two sites, or seasonal HF variability not captured in the tropospheric column derivation.Given the relatively large discrepancy of about 10 ppb between the two data sets during those months and the low sensitivity of the tropospheric column to small changes in β, the last of these explanations is the least likely.

Conclusions
Inadequate constraints on the global CH 4 budget have been a long-standing problem, and understanding recent trends depends on reliable and frequent observations of tropospheric CH 4 concentrations.By explicitly taking into account the averaging kernels of CH 4 and incorporating temporally and spatially varying estimates of the CH 4 -HF relationship, the methodology described here refines earlier tracer proxy methods for estimating stratospheric CH 4 .The tropospheric column measurements of CH 4 derived from TCCON total column-averaged DMFs provide a useful addition to existing data sets used to analyze the global methane cycle and verify chemical transport models.
While the CH 4 -HF relationship is robust, the calculation of β still has limitations.The slight non-linearity and seasonal variability of the CH 4 -HF relationship could impact the estimation of stratospheric CH 4 loss.Further analysis of ACE-FTS and other high-frequency stratospheric measurements could produce a statistically significant seasonal cycle to apply to β.

Figure 1 .
Figure 1.CH 4 (ppb, y axis) vs. HF (ppb, x axis) from ACE-FTS measurements taken between 2004-2012, binned by zonal bands.The slopes of the linear regression are in the upper right-hand corner, and number of data points (N) are listed below each plot.

Figure 2 .
Figure 2. Long-term CH 4 -HF slopes derived by Washenfelder et al. (2003) and annual-mean slopes from ACE-FTS measurements.The inset shows the time series of zonal pressure-weighted ACE-FTS slopes (β), with error bars denoting the 2σ standard error.For each year, zonal slopes are offset from each other for clarity.

Figure 3 .
Figure3.Validation of the tropospheric column-averaged CH 4 derivation using HF as a proxy.The calculated tropospheric CH 4 (y axis) uses the TCCON priors, CH 4 averaging kernel and ACE-FTS vertical profiles to determine the value that the ground-based FTS would retrieve.The integrated tropospheric CH 4 (x axis) applies the pressure-weighting function and TCCON CH 4 averaging kernel and priors to the extrapolated tropospheric ACE-FTS profile of CH 4 .The black lines depict the one-to-one line (solid) and the linear regression of the calculated vs. integrated tropospheric CH 4 with a zeroed y intercept (dashed).The r 2 values correspond to the linear regression.Note the different DMF ranges in the Northern vs.Southern Hemispheres.

Figure 6 .Figure 7 .
Figure 6.The differences between daily tropospheric CH 4 standard deviations using the Washenfelder et al. (2003) and updated methods.(Positive values correspond to larger Washenfelder et al. (2003) standard deviations.)Box ends, midline and whiskers illustrate the quartiles, medians and twice the interquartile ranges, respectively.Outliers beyond the the range are denoted by diamonds.

Figure 8 .
Figure 8. Tropospheric CH 4 column comparison for TCCON vs. aircraft profiles.Error bars denote the 2σ standard deviation from the daily median (FTS) and the estimated instrument errors and tropospheric uncertainty of the measurements (aircraft).

Figure 9 .Figure 10 .
Figure 9. (a) Daily median column and aircraft CH 4 DMFs at Lamont.Flask partial tropospheric columns integrate CH 4 from the minimum to the maximum altitudes of the aircraft measurement.Flask free troposphere columns integrate CH 4 from the minimum to the maximum altitudes of the aircraft within 3-7 km.Total and tropospheric column DMFs are the same as in Fig. 4. (b) Tropospheric CH 4 column comparison for TCCON vs. in situ profiles.

Table 2 .
TCCON sites, coordinates, altitudes, and locations used in this analysis.

Table 3 .
Aircraft Overflights.TCCON site locations and aircraft campaign dates and altitude ranges are listed.