Stable carbon isotopic composition of atmospheric methane: A comparison of surface level and free tropospheric air

. We report CH 4 mixing ratios and (cid:127)513C of CH4 values for remote air at two ground-based atmospheric sampling sites for the period December 1994 to August 1998 and similar data from aircraft sampling of air masses from near sea level to near tropopause in September and October of 1996 during the Global Tropospheric Experiment Pacific Exploratory Mission (PEM)-Tropics A. Surface values of 1513C-CH4 ranged from -47.02 to -47.52%0 at Niwot Ridge, Colorado (40øN, 105øW), and from-46.81 to -47.64%0 at Montafia de Oro, California (35øN, 121øW). Samples for isotopic analysis were taken from 2 ø to 27øS latitude and 81 o to 158øW longitude and from sea level to 11.3 km in altitude during the PEM-Tropics A mission. They represent the first study of 13CH4 in the tropical free troposphere. At --11 km, (cid:127)513C-CH4 was -1%o greater than surface level values. Methane was generally enriched in 13C as altitude increased and as latitude increased (toward the South Pole). Using criteria to filter out stratospheric subsidence and convective events on the basis of other trace gases present in the samples, we find evidence of a vertical gradient in 1513C-CH4 in the tropical troposphere. The magnitude of the isotopic shifts in atmospheric CH 4 with altitude are examined with a two-dimensional tropospheric photochemical model and experimentally determined values for carbon kinetic isotope effects in chemical loss processes of CH 4 C1 on tropospheric mixing ratios is to be detected in comparison to the effect from tropospheric

A second site used for collections of surface air samples is located along the coast of California in Montafia de Oro State Park near San Luis Obispo (35øN, 121øW). This site is protected from the city of San Luis Obispo by low-lying foothills forming a ridge running parallel to the coast. The city of Morro Bay is northeast of the site, while the Diablo Canyon nuclear power plant is -2 km to the southeast. According to the California Energy Commission [1985], which compiled measurements at two wind stations near to the site, Diablo Canyon, California (years 1967-1981), and Santa Maria, California (years 1948-1978), average annual wind speeds were -16.1 and-12.7 km h -1, respectively, at the two stations. Additional data given for Santa Maria indicate that the months from March to June exhibited the highest winds, although only 20% or so above the mean wind speed, while late afternoon to early evening hours exhibited nearly twice the wind speed found for other times of the day. Accordingly, our air samples from Montafia de Oro were collected in late afternoon or early evening during periods when prevailing winds were from the west or northwest.
A third set of samples was from upper tropospheric air collected during the PEM-Tropics A program as part of NASA's Global Tropospheric Experiment (GTE). Air samples used in our study were collected from either a DC-8 or P3-B aircraft over the central and southern Pacific Ocean between 2 ø and 27øS latitude and between 81 o and 158øW longitude. These were collected at average altitudes of between 11.4 and 3.40 km. PEM-Tropics A consisted of a variety of flight patterns, including constant altitude, constant latitude, and vertical-climb flights to study vertical profiles at constant latitude and longitude. We obtained samples from all three types of flight patterns.

Sample Collection Procedures
At each site a different equipment system was used for air sampling. Niwot air samples were collected into passivated aluminum gas cylinders (Scott Marin, Inc., Riverside, California) pressurized using an RIX model SA-3 compressor [Lang et al., 1990]. The volume of gas in these cylinders was -793 L (STP) when pressurized to -13,800 kPa by the compressor. Air was drawn through a stainless steel tube that had its inlet secured to a 10 m tower at a height -6.5 m above ground level. Each cylinder was filled to -3450 kPa from the surrounding air and then purged to ambient pressure before the final fill to -13,800 kPa. Two drying tubes in series, each filled with magnesium perchlorate (Mg(C104)2), were located on the high-pressure side of the compressor just before the cylinder inlet. These were used to for •513C of CH 4 on only a small portion of the number of samples taken during PEM-Tropics A flights because of sampling logistics and the time constraints on laboratory analyses of samples. In practice, from 6 to 12 individual canisters collected in sequence were processed as a group (simultaneously) for isotope analysis by our laboratory. This procedure was used to provide a large enough sample for isotope ratio analysis of CH 4. Effectively each set of canisters combined as a group represented an integration over the period of flight time elapsed during sample collection. Overall, the time period of collection for grouped samples was comparable to that of air samples collected for 13CH4 content in other studies that used relatively large air sample volumes in single cylinders or canisters for each individual sample measurement (i.e.,-9-15 min collection time per cylinder as reported by Brenninkrneijer et al. [1995] and -25-30 min collection time per canister for Airborne Arctic Stratospheric Experiment (AASE)-II samples discussed by Gupta et al. [ 1996]). separation column was 1.83 m (3.2 mm OD) of molecular sieve 5A. It was preceded by a 30.4 cm precolumn (3.2 mm o.d.) of the same material. Both columns were at 100øC and used nitrogen carrier gas. Our working CH 4 standard was a cylinder of whole air certified by direct intercalibration with standards maintained by the NOAA/CMDL air-sampling program [Lang et al., 1990]. Relative to the National Institute of Standards and Technology (NIST) scale, atmospheric CH 4 mixing ratios measured using the NOAA/CMDL standards are 0.023 ppm lower. Our measurement precision was -0.005 ppm. The uncertainty in our measurements was 0.25-0.50% for CH 4. All CH 4 mixing ratio data reported are from determinations made at UC, Irvine.
Samples collected at Niwot Ridge were also analyzed for CH 4 and CO by NOAA/CMDL in collaboration with our lab.
Methods and standard gases used for these analyses are detailed by Steele et al. [1987], Lang et al. [1990], and Novelli et al. [1991,1992]. Although CO in Montafia de Oro samples was not measured directly, an approximate value for CO concentration was obtained by a calculation that divided the recovered CO in the vacuum line as determined by manometric measurement to the total amount of air processed assuming 100% recovery of CO.

Analytical Procedures (Sample Preparation)
Preparation of the samples for isotopic analysis was done on a combustion vacuum line that separates CH 4, CO, and CO2 trace gases from whole air (and from each other) while further converting CH 4 to CO 2 and H20 and converting CO to CO2 as the air stream moves toward the pump. Most of the details regarding the vacuum line design and procedure as well as improvements made over time have been reported previously [i.e., Lowe et al., 1991;Tyler et al., 1994a, b]. In brief, a series of cryo traps condensed trace gases such as CO2, H20 vapor, and N20 at liquid N 2 temperature, leaving CO and CH 4 in the air stream. The CO was converted to CO2 using I205 on silica gel, also known as Schtitze reagent (Leco Corp., St. Joseph, Michigan), before its removal using cryo trapping with liquid N 2. The CH4 was then converted to CO2 using platinized alumina pellets heated to 740øC, prior to its recovery using cryo trapping with liquid N2.
Two kinds of sample preparation tests were also made routinely. Taken together, they served to help ascertain the suitability of our analytical methods for keeping measurement data intercalibrated over the long time periods necessary to study isotopic trends. One test consisted of running dry zero air (<0.003 ppm CH 4 as judged by our GC measurements) through the vacuum line and checking for possible accumulation of condensible gases in each of the trapping sections of the vacuum line used to recover CO 2. This test helps determine the stability of the vacuum line by checking for potential contamination from either leaks into the line or outgassing of impurities from the combustion oven or Schtitze reagent. The second test consisted of routinely preparing and then processing on the vacuum line calibration gases of CH 4 in zero air using the same source of CH 4 each time. This test helps assess the long-term stability of our vacuum line, isotope ratio standards, and mass spectrometer and serves as an overall check on our precision of measurement for samples processed using the vacuum line.

Analytical Procedures (Mixing Ratios)
At UC, Irvine, CH 4 mixing ratios were measured using a Hewlett Packard 5880A gas chromatograph (GC) with packed columns and a flame ionization detector (FID).
The main

Analytical Procedures (Sample Measurement)
The converted CO 2 from recovered CH 4 in air samples was measured by a Finnigan-Mat model 252 isotope ratio mass spectrometer to determine 13C/12C ratios. The units of measurement are %o per mil for/513C with values for both gas standards and samples reported versus Peedee belemnite (PDB) carbonate [Craig, 1957]. The precision of measurement on clean dry CO2 gas standards was _+0.01%o. Overall measurements of an individual CH4 sample from field collections had a precision of +_0.05%o. The reproducibility of isotope measurements from like samples (i.e., multiple samples collected simultaneously) when all possible errors associated with sampling, processing, and measurement were taken into account was <+_0.10%o for 813C of CH 4.
Our working carbon isotope standard was CO2 gas purchased from Oztech Gas Co. (Dallas, Texas). They certified it to be -39.78%0 versus PDB. We compared it to two other CO 2 isotope reference materials. One was a CO2 working reference gas obtained from the National Instititute of Water and Atmospheric Research (NIWA) in New Zealand that had been assigned a value of-47.56-+0.01%o versus PDB by NIWA. The other was a barium carbonate powder recently made available by the International Atomic Energy Agency (IAEA) for use as a CO2 standard following prescribed methods for its preparation. Designated as IAEA-CO-9, it was reported to have a value of-47.12-+0.12%o (2c•, n=10, results of a blind test interlaboratory comparison), although isotope measurements reported by the producer were reported as/513C = -47.23-+0.03%o [Stichler, 1995]. Versus our Oztech gas standard, the NIWA CO2 gas standard had a value of -47.61_+0.03%o while CO 2 gas made in our laboratory from IAEA-CO-9 carbonate obtained directly from the original producer had a value of-47.18_+.04%o (from five prepared standards totaling 11 measurements). Our measured values for atmospheric CH 4 are reported without a correction to account for any offset in our values versus either the NIWA working reference gas or IAEA carbon isotope standard.

Vacuum Line Tests
Contamination or "blank" tests of the vacuum line consisted of running zero air (determined to have <0.003 ppm CH 4 by GC measurements) through the vacuum line and checking for possible accumulation of condensible gases in each of the trapping sections of the vacuum line used to recover CO2. This test was run every few weeks during 1995 and early 1996 and once a week routinely beginning July 1996. The average total recovered gas from blanks was 1-2 gL accumulated in the CH 4 traps and 2-5 gL in the CO traps for -300 L of air processed each time. Blanks for smaller volumes of air processed were smaller.
Dur. ing dates when PEM-Tropics A air samples were processed, blanks were run on 25-45 L of air and found to have -0.7 gL of accumulated product in the CH 4 traps. As judged from the mass spectrometer analysis, these contamination blanks were quite variable in content; that is, they were a mixture of CO 2 and H20, and had /513C values of anywhere from -25 to -50%0. The accuracy of /513C measurements on these blanks was not sufficient to assign a reasonable value to the contamination. A calculated worst possible case scenario for the relatively smaller air samples collected during PEM-Tropics A can be made assuming a fixed blank volume of 0.7 gL at -25%o mixed with recovered CO 2 from either 25 or 45 L of sample air. The calculation indicates that the smaller CH4 samples measured (i.e., CH4 as CO2 recovered from 25 L of air) would be artificially heavier than the larger samples (-45 L of air) by 0.15%o. Furthermore, air samples of 45 L volume would be -0.07%o heavier in 813C of CH4 than samples where 300 L of air were processed with a corresponding 2 gL blank. However, on the basis of the sample blank size relative to amounts of CH4 recovered as CO2, inaccuracy in measuring its /513C value and lack of conclusive proof that the blank observed in zero air processing also appears in air sample processing we made no correction to our/513C-CH4 data. Overall, the worst possible case scenario for vacuum line contamination revealed that neither random nor systematic errors in processing affected our ability to discern differences in/5•3C-CH4 values of the order of the several tenths of per mil potentially caused by the spatial and temporal variability of air samples. Tests of the efficiency of the Schtitze reagent in the vacuum line were made to determine if incomplete reaction between it and CO could cause some CO to be oxidized to CO2 by the platinized alumina and subsequently mixed with CO2 from CH4. oxidation. A series of calibration gases were made by pressurizing 32 L canisters with zero air to an appropriate pressure after first adding an aliquot of CO from a control lecture bottle of research grade CO (99.97% purity, Linde Division, Union Carbide Industrial Gases, Inc., Somerset, New Jersey). These tests revealed that <1.6% of all CO present in a 100 L volume of calibration gas with a CO mixing ratio of-1 ppm went unreacted in passing through the Schtitze reagent in our vacuum line. Less than 1% of all CO in a 220 L volume of calibration gas with a CO mixing of -200 ppb went unreacted. From these tests we concluded that the amount of CO breaking through the Schtitze reagent and combining with recovered CH4 using our vacuum line was negligible for our purposes.

Isotope Gas Standard Tests
Throughout the time period of sample line processing of atmospheric samples we analyzed a series of calibration gases of CH4 in air. These calibration air samples were made in one of two methods. One method was to remove an aliquot of CH 4 from a control cylinder of research grade CH4 (99.99% purity, Linde Division, Union Carbide Industrial Gases, Inc., Somerset, New Jersey) and to inject it into an air canister. It was then diluted using zero air to obtain -30-60 ppmv CH4. Using this method, canisters were measured 18 times between June 1995 and August 1997 and found to have a/513-CH 4 value of-35.28_+.012%o (lo) for all measurements. A second method used a cylinder of zero air with an added component of 60.2 ppmv CH 4 (Scott-Marin, Inc., Riverside, California). This cylinder was analyzed 10 times between February 1996 and July 1997 and found to have a/513-CH 4 value of-19.53_+0.07%o (lo) for all measurements. These results are a good indication of the stability of our processing line over time. They confirm that there were no contamination processes occurring that were varying with time and shifting measured/513C of CH 4 values from samples significantly.

Air Sample Tests
On two Niwot Ridge air samples, duplicate aliquots from the same air cylinder were analyzed. The sample from March 15, 1996, was processed and measured twice and had a/5•3-CH 4 of -47.38%o each time. The sample from April 12, 1996, was analyzed in a similar manner and had/5•3-CH4 values of-47.37%0 and -47.38%o. These tests showed that processing multiple aliquots of air sample and/or multiple measurements of an air sample on the mass spectrometer were unnecessary to obtain sufficient measurement precision for our study.

Intercalibration of Air Sample Measurements With Another Laboratory
Our CH 4 data sets from Niwot Ridge and Montafia de Oro are intercalibrated with data obtained by NIWA at Scott Base, Antarctica, and Baring Head, New Zealand [e.g., Lowe et al., 1991Lowe et al., , 1994Lowe et al., , 1997Lassey et al., 1993] through an on-going intercalibration program. The intercalibration program includes not only the isotopic standards mentioned earlier but also the effects of sample containers, collection procedures, and trace gas processing prior to isotopic measurements of CH 4. A subset of the air samples collected at Montafia de Oro and Baring Head are measured by both laboratories while still in their original canisters through a sample exchange program.
For 16 air samples processed and measured by both laboratories (eight samples from Montafia de Oro and eight from Baring Head) the level of agreement for •513-CH4 was +0.01e0.12%o. This determination excludes one sample from Montafia de Oro for which the measured difference in •513 of CH 4 found by the two laboratories was 0.29%0. If we further exclude the largest outlier in the average of 16, from a sample collected at Baring Head, the level of agreement between the two labs becomes-0.01+0.09%o (n=15) with the negative value indicating that our reporting of •513C is lighter by 0.01%o for the same sample measured by the New Zealand researchers.
In principle, we could adjust either Montafia de Oro or Baring Head data by 0.05%0 (making our determinations 0.04%0 heavier for the same samples measured by NIWA) to account for the offset between the two labs in measuring the NIWA reference gas. However, the uncertainty in the direct comparison of air samples (_+0.09%0) is larger than the offset. Furthermore, our offset in measuring barium carbonate provided by the original producer is 0.05%0 in the opposite direction from the NIWA gas when compared to the original value assigned by the producer (i.e., not the IAEA value). We prefer to ignore corrections of the order of a few hundredths of a per mil until we participate in an appropriate blind test interlaboratory comparison.

Recovery of CH4 From Air Samples
The yield on the vacuum line for recovery of CH 4 converted to CO 2 by combustion was 100.0_+ 1.2% (1 o, n=42) for all Niwot Ridge samples processed. A similar type of yield calculated for all samples processed from Montafia de Oro was 99.6-+1.3% (lo, n=108). Typical volumes of CH 4 recovered were -540 gL from processing -300 L of air from Niwot Ridge. The amount of CH 4 recovered from Montafia de Oro air samples were correspondingly smaller because only -180 L of air were processed using 32 L canisters.

Methane mixing ratios and •513C of CH 4 values from Niwot
Ridge Using the measurement data, we looked for relationships between the isotope or mixing ratios and source and sink changes such as simple seasonal changes from winter/summer differences in OH abundance. In both Figures 1 and 2 the data were fit with an unweighted least squares fit (1-harmonic) by setting the periodicity at 1 year ((o=2•/365) but allowing for four parameters to be calculated from the best fit to a general equation of the form y = a + bx + csin(rox) + dcos(rox) Trends in the data can be seen both from a visual inspection of the data points and from the calculated curve fits shown.
At  wind was usually from the west and was most often <8 km h -1 (see Figure 3a). More importantly, even when wind was from a southern or southeastern direction, it did not appear to affect the •513C-CH4 values (see Figure 3b). In contrast, Bollinger et al. [ 1984] studied NO x mixing ratios at Niwot Ridge and determined that there was a strong correlation between meteorological conditions and NO x at the site. Low NO x was associated with westerly winds (more likely in winter), while high NO x was clearly associated with Denver metropolitan pollution (more likely in summer). Events driving high-NO x periods included daytime warming causing a valley-to-mountain circulation with accompanying upslope winds (winter) and increased convection with easterly winds (summer). Our current wind data from Niwot Ridge also contrast with our previous experience at the site from samples collected between 1989 and 1992. At that time, morning winds were routinely stronger, and upslope winds from the southeast, biased by urban air, were likely if we sampled in afternoon or evening. However, in our 1995-1997 data set, removing data from samples taken during southerly and easterly winds does not appreciably change the results. Overall, it appears that the combination of taking early to late morning samples only during 1995-1997 and having mostly relatively calm days, thus far, has resulted in an

PEM-Tropics Air Samples '
We measured air samples from six different flights during the PEM-Tropics A program. The data appear in Table 1. An explanation of Table l's format follows. As mentioned earlier, we combined air samples from several 2 L canisters collected sequentially to obtain enough air for •513C-CH4 analysis. Because of this, the location and altitude of each group of samples providing a •513C-CH4 value in Table 1 are listed as a range of values. Similarly, the CH 4 mixing ratio for each •513C-CH4 value is the average value of all canisters combined as a group as determined by a calculation that divided recovered CH 4 from vacuum line sample processing by the total amount of sample processed assuming 100% recovery of CH 4. In-flight determinations of CH 4 mixing ratio of individual canisters were unavailable at the time we made our analyses, and sampling logistics prevented us from making these measurements ourselves.
Ideally, we would have liked to analyze air samples from groups of canisters collected at nearly constant altitude because It is helpful to look at the data in Table 1 graphically. In Figure 5a, a two-dimensional (2-D) plot of median altitude versus median latitude for all longitudes held arbitrarily constant shows the approximate location of air parcels used to make each measurement. Numbers in Figure [Newell et al., 1996;Wu et al., 1997]. Even so, the troposphere was found to have a layered structure [e.g., Newell et al., 1996] under certain conditions with some of the layering arising from the very events (e.g., convection) known to provide fast mixing. For that reason we implemented criteria to filter out stratospheric subsidence and convective events on the basis of other trace gases present in the samples.
We looked for corroborating evidence for convective plumes and stratospheric subsidence in air samples from Table 1 using companion measurements of 03, CO, H20 vapor, C2H 6, C3H 8, and C2C14 made at the time of sampling (NASA PEM-Tropics A, archived data, 1997). We assumed that instances of high 0 3 with accompanying low CO, H20 vapor, and CH 4 mixing ratios at a given altitude were an indication of stratospheric subsidence, while instances of high 0 3 with elevated mixing ratios of CH4, CO, C2H6,, C3Hs,and C2cl 4 (present in urban plumes but not in biomass burning plumes) at a given altitude were an indication of convective events. Low 0 3 and surface level mixing ratios of CH 4, CO, C2H 6, C3H 8, and C2c14 were assumed to arise from convective events without accompanying biomass burning or urban air plumes. The degree of change in these values from that expected at altitude provides a measure of the strength of the influence of that event on the air sample. On each flight for which we had samples we plotted the observed concentrations for these compounds versus time for the entire flight track as well as for the portion of the track from which our samples were collected (data not shown). Using the chemical data described above, we determined that several air parcels providing data points in Table 1 were indicative of some degree of transport from another region. As identified by reference number in Table 1, stratospheric subsidence had affected points 1 (strongly) and 11 (weakly), while convection from the surface had affected points 6, 7, 9, and 14 (all weakly, from urban plume), points 8 and 10 (moderately, from biomass burning), and point 13 (weakly, from a clean convective event). Although these determinations are qualitative, they aid in interpreting Table 1  We must stress that in comparing the levels of mixing ratio for various compounds to that expected for air unaffected by fast mixing events the use of terms such as strongly, moderately, and weakly is subjective. In no case were CH 4 mixing ratios in our PEM-Tropics A samples elevated appreciably above levels for background southern tropical tropospheric air at the surface. This

Thus our PEM-Tropics A air samples measured for 13CH4 content provide evidence that a pattern of 1513C-CH4 values exists
in the southern tropical troposphere that follows a trend toward isotopically heavier CH 4 with increasing latitude and altitude.

Modeling Results and Discussion
New measurement data reported here for the years 1995-1998 corroborate the basic features of the observational data discussed by Gupta et al. [1996], i.e., that there is a large õ13C-CH4 difference (-1%o) between surface and near-tropopause values and an even larger difference between surface and aged air in the stratosphere. They also extend the observational evidence because of the direct comparison of intercalibrated surface and upper air samples measured by our laboratory and NIWA and because they indicate that a gradient in 1513C of CH 4 at Southern hemispheric low-latitude tropospheric altitudes exists as well.
Previous model-calculated 1513C-CH4 values have shown poor agreement with observations in determining the difference between surface and higher-altitude air [Bergamaschi et al., 1996;Gupta et al., 1996Gupta et al., , 1997. In this section we use the latest determinations of carbon KIEs and available observational data, including the most recently reported, to see if a better fit between observations and calculations is now possible and to determine what the causes for any remaining discrepancy might be.

Kinetic Isotope Effects on Atmospheric Methane
Qualitatively, the difference in •513C of CH 4 found in comparing surface to higher-altitude air is largely because CH 4, which has only surface sources, is increasingly exposed to chemical sink processes as it ages. Some of these sinks isotopically fractionate CH 4

Reaching a consensus for each KIE is very important because any proposed significant change in the KIEs of the sink processes alters the isotopic balance of the summed CH 4 sources and estimates of the various strengths and distributions of CH 4 sources to the atmosphere. Changes in the two most important
KIEs also alter the calculated vertical profile of •513C of CH 4 significantly. Gupta et al. [1996] show that although C1 radicals are more concentrated in the stratosphere, some effect on CH 4 from reaction with C1 will be observed at lower altitudes because of vertical mixing between higher and lower altitudes. Although the OH reaction dominates CH 4 loss at all altitudes, beginning at -16 km, C1 abundance is great enough that direct reactions with CH 4 cause some of the 13C enrichment in CH 4 The effect of C1 on •513C-CH4 may be quite disproportionate relative to its abundance compared to OH at all altitudes depending on the KIEs of the two reactions in question.  Figure 6. The adopted source magnitude and •513C-CH4 data appear in Table 2. It is important to realize that any reasonable set of source magnitudes and distributions will be sufficient to study mixing  60øN and 90øN, 1513C values were decreased to -65%o. and isotope ratios of CH 4 both vertically and latitudinally as shown by Gupta et al. [1996]. It is not necessary to use a source function that provides an exact match to observed values for these ratios at the few locations reported on globally in order to do a study of the vertical gradient of CH 4. In fact, the source function chosen to approximate observed values at the surface will depend heavily on the choices of magnitudes of the various loss processes and their associated KIEs, which are also subject to uncertainty. mixing ratios were within _+0.01%o of their values for the previous year. Once steady state was obtained we examined the CH 4 mixing and carbon isotope ratios calculated at appropriate grid points, months, and altitudes and compared them to observational data.

Model Results and a Comparison to Observations
The experimentally determined KIEs of Cantrell et al. [ 1990] and Saueressig et al. [1995] along with the source strengths and carbon isotope ratios in Table 2  Differences found in comparing model-calculated CH 4 mixing ratios throughout the troposphere to the surface were very small.
For example, at an altitu& of 3.25 km in September, no changes toward lower mixing ratios were calculated to have occurred over the southern high latitudes (i.e., 0.00 ppm below 40øS); only very small changes toward lower mixing ratio were calculated for the tropics (-0.01 ppm between -30 ø and 20øS,-0.02 ppm between -10øS and 10øN, and -0.02 ppm from 10 ø to 20øN) and from -0.02 to -0.04 ppm at higher northern latitudes. At 10.25 km in September, changes of-0.02 ppm at southern high latitudes and changes in the tropics very similar to those found in going from the surface to 3.25 km (0.00 to -0.02 ppm between -30øS and 10øN) were found. The northern high latitudes, however, showed a larger decrease in mixing ratio with increasing altitude (--0.05 to --0.06 ppmv between 20 ø and 80øN) such that the difference between 10.25 km and the surface was larger than that between 3.25 km and the surface.
The relative change in mixing ratio calculated for Southern Hemispheric latitudes is reasonably close to data reported here and by Lowe et al. [ 1991 ]; neither set of measurement data shows appreciable change in mixing ratio with altitude up to -10 km. Measurements reported by Brenninkmeijer et al. [1995], Gupta et al. [1996], and our data for point 1 from air samples taken in the tropopause or lower stratosphere (i.e., -11-13 km), however, showed that mixing ratios can be decreased from surface values by appreciable amounts (-0.05 to -0.10 ppm). Our model calculates appreciable September changes in mixing ratio at high isotopic CH 4 at various altitudes. The lessening of the KIE effect of C1 by -32%o (which matters mostly above 16 km but affects the total column CH4 to some extent) is approximately balanced by the increase of the KIE effect of OH by -5%o (which is the most important loss process at all altitudes).  [1997] measured a/5•3C-CH4 value of--42.7%o. This means that comparisons between model-calculated enrichments with altitude and observed enrichments with altitude, already discrepant in the tropopause, worsen with increased altitude. At 24 km the discrepancy is several per mil as compared to -1%o at 10 km (see Figure 7d).

Further Discussion of PEM-Tropics A Samples and Vertical Gradients
In general, we expect the lower troposphere to be wellmixed vertically because of convective events and air mass subsidence as well as diffusion. Ehhalt [1974] first showed that vertical profiles of CH 4 mixing ratio were uniform from monthly samples taken both 200 km offshore from Santa Barbara in the Pacific Ocean and over Scottsbluff, Nebraska, in 1965-1967. Nevertheless, the influence of subsidence and convection events on air samples measured to determine CH 4 mixing and 1513C value depends on how much time has elapsed and how far away the event is from the air parcel sampled. The underlying chemical composition in the absence of such events may still provide a structure for the •j13C-CH4 gradient that is measureable in the troposphere.
Our PEM-Tropics A air samples measured for 13CH4 content provided evidence that a pattern of •j13C-CH4 values exists in the tropical troposphere that follows a discernable trend both vertically and latitudinally. Although the vertical trend was not seen by Lowe et al. [ 1991 ] in samples taken at and near the South Pole, both the PEM-Tropics A data and South Pole data show a moderate correlation between •513C-CH4 and mixing ratio at all altitudes.
Therefore both data sets are consistent with the wellknown phenomenon of isotopically enriched CH 4 from chemical loss processes. Only the degree to which both CH 4 ratio and •513C-CH4 are affected by loss at different altitudes is different in the two regions studied. One plausible explanation why there might exist a tropospheric gradient in •513C-CH4 when no appreciable one exists for the CH 4 mixing ratio comes from the great difference in the KIEs of CH 4 reaction with OH and C1. It may be that 1513C-CH 4 values in the troposphere are more sensitive to changes in chemical loss processes than CH 4 mixing ratio values are because of the indirect effect of stratospheric C1 on tropospheric air.
Stratospheric-tropospheric air exchange can bring aged CH 4 into r the troposphere f om above with a C content relatively more affected than its mixing ratio by the C1 chemical loss process in the stratosphere. In effect, stratospheric C1, with its very large carbon KIE compared to OH, has an increasingly important and more easily detectable effect on tropospheric •513C-CH4 values as altitude increases or as all forms of chemical loss take place in the troposphere since it affects the troposphere only indirectly through exchange of stratospheric and tropospheric air. However, a corresponding situation for the vertical gradient of tropospheric mixing ratios of CH 4 does not exist because the effect of stratospheric C1 on tropospheric mixing ratios is too small to be detected in comparison to the effect from tropospheric OH.

Conclusions
Air Measurements of 1513C-CH4 in air collected from PEM-Tropics A revealed a trend in •j13C-CH4 both vertically and latitudinally in the free tropical troposphere that is discernible in spite of any stratospheric subsidence and fast transport effects such as convective events. Such a vertical trend is theoretically possible on the basis of the larger KIE associated with CH 4 loss through reaction with C1 than through reaction with OH and on the fact that unlike CH 4 loss from reaction with OH, which occurs at all altitudes and accounts for almost all CH4 loss, direct loss of CH4 by reaction with C1 occurs only in the stratosphere. Hence the long-term effect of the slow exchange between stratospheric air and tropospheric air may be to bring highly 13C enriched CH4 in stratospheric air to lower altitudes in a layered pattern in spite of stratospheric subsidence and convection.
Currently, no multidimensional photochemical model calculates values for bl3C-CH4 in reasonable agreement with observations at any altitude appreciably above the surface, even with CH4 source functions that provide good isotopic agreement at the surface. This is true even though the models can be made to provide excellent to very good agreement between calculated and observed mixing ratios at all altitudes. The model and parameters we have used here, which match calculated and observed data well at the surface, badly undercalculate the enrichment of isotopic CH 4 with altitude well into the stratosphere. Our model does not provide a verification for the b•3C-CH4 gradient with altitude observed in the tropical free troposphere from PEM-Tropics A samples.
It will be very important to obtain more isotopic data from both air samples collected at fixed surface sites and aircraft samples. Continued measurements at surface sites will provide information on long-term trends and seasonal changes in atmospheric CH 4 that can be related to its sources and sink processes. More upper air samples (such as those obtained by aircraft and balloon-borne cryogenic samplers) will help resolve current problems in establishing the proper magnitudes of KIEs for various CH 4 sink processes and reconcile measurements and model calculations involving b13C-CH4 and mixing ratios.
Isotopic measurements of CH 4 made by each research group must be intercalibrated with others to fully make use of all available data.