Comparison of free tropospheric western Pacific air mass classification schemes for the PEM-West A experiment

. During September/October 1991, NASA's Global Tropospheric Experiment (GTE) conducted an airborne field measurement program (PEM-West A) in the troposphere over the western Pacific Ocean. In this paper we describe and use the relative abundance of the combustion products C2H 2 and CO to classify air masses encountered during PEM-West A based on the degree that these tracers were processed by the combined effects of photochemical reactions and dynamical mixing (termed the degree of atmospheric processing). A large number of trace compounds (e.g., C2H6, C3H8, C6H6, NOy, and 03) are found to be well correlated with the degree of atmospheric processing that is reflected by changes in the ratio of C2H2/CO over the range of values from --0.3 to 2.0 (parts per trillion volume) C2H2/(parts per billion volume) CO. This C2H2/CO-based classification scheme is compared to model simulations and to two independent classification schemes based on air mass back-trajectory analyses and lidar profiles of 03 and aerosols. In general, these schemes agree well, and in combination they suggest that the functional dependence that other observed species exhibit with respect to the C2H2/CO atmospheric processing scale can be used to study the origin, sources, and sinks of trace species and to derive several important findings. First, the degree of atmospheric processing is found to be dominated by dilution associated with atmospheric mixing, which is


Figure 1. Mean 500-mbar nomethane carbon (NMC) winds over the Pacific Ocean during PEM-West
A with the following symbols.

Introduction
Long-range transport provides an effective means of exporting long lived carbon, sulfi•r, and nitrogen compounds from the northern henrisphere anthropogenic centers to the more remote Pacific. Chemical signatures in air over the western and eastern Pacific and, to a lesser extent, the central Pacific suggest this region is not immune to the effects of biogenic and anthropogenic emissions of relatively long lived compounds (i.e., lifetinaes of days to months). Understanding the transportation and transformation processes affecting the abundances of nonmethane hydrocarbons (NMHCs), methane (CH4), carbon monoxide (CO), carbon

dioxide (CO2), nitrogen oxides (NOy), and ozone (03) is a key component in understanding the hnpact of biogenic and anthropogenic sources on the tropospheric chemistry within this region. During September and October 1991 the NASA Global Tropospheric Experiment (GTE) conductexl a series of airborne Pacific Exploratory Missions over the western Pacific Ocean (PEM-West A). A complete overview of PEM-West A program goals and individual flight objectives and a description of the complete suite of instruments deployed to meet these objectives is given by Hoell et al.
[this issue] and references therein. One major goal of the PEM-West A program was to accurately define the factors controlling the tropospheric budgets of carbon, nitrogen, sulfur, and ozone within this region. To this end we will use tlfis unique data set to study the relative degrees to which atmospheric chemistry and dynamics are responsible for incorporating anthropogenically and biogenically perturbed air into the more pristine "background air" over the  Figure 1). The subtropical anticyclonic circulation in the central Pacific (located near 30øN and 160øE) was the most dominant feature affecting air originating from several locations. Flow from the Asian continent was mixed into the anticyclone along with air from the northern central Pacific and from the tropical/equatorial central Pacific. In the tropical western Pacific a small cyclonic circulation was apparent to the south-southwest of the subtropical anticyclone. This latter feature is a manifestation of several transient tropical cyclones, which often reached their maximum intensity over this particular region before recurving to the north and northeast. Air masses that had recently resided over China, India, Japan, Taiwan, and Indonesia represent the dominant short-range source of continental air affecting the western Pacific during this study. These air masses would be expected to exhibit different degrees of perturbation depending on the mechanism by wlfich boundary layer air was transported into the free troposphere and the length of time the air residex] within (or the degree it was mixed into) the general circulation patterns shown.
In companion papers, two methods are employed to classify air masses that were encountered in order to characterize their impact on the overall chemical state of the investigate the fraction that various air mass categories contribute to the make up of the total tropospheric cohmm over these regions. Our goal in this paper is to use a chemically based air mass segregation scheme and investi-gate its ability to form a link between these other classification schemes and to provide a means of gaining additional insights into the factors that control the chemical composition over the western Pacific.
In this paper we will first examine in section 2 the chemical constitution of various air masses in relation to a scale that is defined by the relative abundance of the combustion products ethyne (C2H2) and CO. The relationship between this pair of compounds is then used in section 3 as a surrogate ordinate for exanfining the filnctional dependence of other trace gases relative to the degree that the atmosphere has processed emissions from combustion sources. These functional dependencies are examined in sections 4 and 5 for the relationships relative to air masses classified by isentropic back-trajectory and O3/aerosol lidar analyses. The internal consistency between the chemically and trajectory-based schemes is tested in section 6 followed by a discussion of scientific issues that can be addressed by combining these three different air mass classification techniques.  , especially when they combine to process the compounds emitted into the atmosphere. In the lower troposphere during PEM-West A they find that this general conclusion holds for compounds that are less reactive with OH than normal butane (n-C4H•0, with an OH lifetime of 3.2 days). Their study indicates that in the lower troposphere (<2 km), only the most reactive NMHC compounds (i.e., lifetime due to OH oxidation < 1 day) have relative behaviors that were predictable based on photochemistry alone, whereas those with intermediate lifetimes (1-3 days) were controlled both by the photochemistry and dynamical processes (i.e., mixing) associated with advection, and those compounds with long lifetimes (> 3 days) were primarily controlled by mixing. Their study also indicates that relationships between various compounds can be explored as markers of the relative degree that different air parcels have been impacted by the combinexl effects of different emission sources and differing combinations of photochemical and dynamical processing.

Description of a Chemically Based
To explore the possibility of extending their work to the larger spatial scale of the western Pacific basin (defined here as the western tropical and northwestern midlatitude Pacific and adjacent continental regions), we will need to examine NMHC compounds that have atmospheric lifetimes longer than n-C4H•0. Propane (C3Hs) and ethane (C2H,) are both compounds with longer lifetimes that also have similar biogenic and anthropogenic sources and both tindergo photochenfical loss via H-atom abstraction when reacting with OH and as such are possible candidates for examining larger spatial scale atmospheric processes. Similarly, the combustion products C2H 2 and CO represent a pair of compounds that have common primary anthropogenic sources and a common sink via loss due to O-atom addition upon reaction with OH. The reaction rates for these compounds vary somewhat as a filnction of temperattire and pressure over the altitudes sampled during PEM-West A [Demore et al., 1992]. The dependence on temperature is less for O-atom addition reactions than for H-atom abstraction reactions, but they (the O-atom addition reactions) are ternary reactions and therefore pressure dependent. However, these dependencies are modest for most tropospheric conditions (< _+ 1.5-fold from typical 500-mbar conditions). Based simply on the ratio of chemical lifetimes (i.e., k2H2/kco ,• 3 and k3Hs/k•H 6 '• 5), sinfilar behaviors •night be predicted as these pairs of compounds are photochenfically processed following their input into the atmosphere. In this case, where atmospheric transport is ignored, a shnple conunon photochenfical processing timescale can be derived based on these compound reactions with OH, which can be expressed as zXt= Co/ C OHl ( ) where C o and Ct represent the concentration of the compound of interest at some initial time (0) and at some time later (t), and kc is the compound's reaction rate coefficient with OH. This photochenfical processing time concept can be extended in an attempt to remove the effects of shnple dilution with pure air by using the relative behavior of two compounds with sinfilar sources and sinks, where for the compounds C2H 2 and CO,

at=tt((CH d co)o/( d co)),l/ oH].
This pair of compounds has yielded usefid information about the relative degrees to which different air masses have been processed within the atmosphere overlying remote regions [e.g., Sandholm et al., 1992]. Over the western Pacific the mixing ratios of C2H • and C3H8 were correlated with the observed ratio C2H2/CO (Figures 2a and 2b) as might be expected based on the relative photochemical processing time relationship described above. Even though these data display a high degree of correlation, there is a tendency toward nonlinearity at high and low values. The apparent plateau in C2H 6 and C3H 8 for mixing ratios of C2H2/CO larger than 1.5 (pptv/ppbv) is suggestive of the influence of variable source emission ratios of C2H, and C3H8 relative to those of C2H 2 and CO. The apparent plateau in mixing ratios below C2H2/CO values of about 0.5 (pptv/ppbv) suggests that some limiting value is reached that might represent the "background" mixing ratios for the region. For Call6 and C3H8 this lower linfit would imply that the western Pacific background mixing ratios during the PEM-West A period -,,/ere approximately 400 pptv and 18 pptv, respectively. The high degree of correlation that is displayed by the data in Figures 2a and 2b does suggest that common atmospheric processing mechanisms were operating on these compounds over the range of CaH2/CO ratios from about 0.5 to 1.5 pptv/ppbv. However, the above discussion also indicates that we must take into account that instead of indicate q-la about the mean for (a) C2H 6 (r 2 = 0.96); (b) C3H8 (r 2 = 0.94); and where the least squares best fit line (thin line) is depicted for those data having associated C2H2/CO ratios betwecn 0.5 and 1.5 pptv/ppbv. In the above discussion we have also ignored the in situ photochemical production of CO from CI-I• and NMHCs which has the same effect as reducing the rate constant for CO, 0(co), in equation (2). In practice we can neglect the production of CO from reactive NMHCs, such as isoprene, which are oxidized in the continental boundary layer and cannot be distinguished from direct emissions of CO [e.g., Altschuller, 1991]. These effects are included in the observed initial value of CaH2/CO. Production of CO from less reactive NMHCs and CH4 can be estimated from Figure  2  C2H2/CO ratios (-•0.3 pptv/ppbv) are well below expected values based on either mixing or chemistry alone acting on this processing timescale. Third, the C2H2/CO ratios observed during PEM-West A span a broad enough range of rabies (i.e., nearly an order of magnitude) to be useful as an ordinate for examining the functional dependence of other trace gases with respect to the chemical space defined by the ratio of C2H2/CO. Based on the discussion above, it should also be apparent that the C2H2/CO-based air mass classification scheme is limited in its ability to distinguish between different combinations of conditions which could lead to the same apparent level of atmospheric processing. This ambiguity stems from the possibility that differing mixtures of continental air with different trace gas levels could be entrained into different "background air" under different tneteorological conditions to yield sinfilar levels of apparent atmospheric processing as expressed on the C2H2/CO scale. Even though this limitation will prevent the C2H2/CO-based scheme by itself from unatnbiguously identifying the sources impacting a particular air mass, it may provide a means by which perhaps otherwise chemically similar air masses may be coidentified compared with respect to other tracer signattires.

Examination of Trace Gas Relationships With Respect to the C2H•/CO Ordinate
On the basis of the discussion above, the ratio of C2H2/CO appears to be a viable candidate for segregating air masses in relation to the relative degree to wlfich they have been atmospherically processed and impacted by combustion sources. To assess this possibility, we selected several key compounds for analysis with respect to ratios of C2H2/CO; these are depicted in Figtires 4a-4h. Measurements made in the boundary layer (or mixed layer) are omittexl from this analysis, and data taken between 2-to 7-kin are separated from those taken between 7-to 12-kan.

This altitude separation is based on the results of Talbot et al. [this issue] and Gregory et al. [this issue]
, which indicate that air of continental origin often had enhanced mixing ratios in the 4to 8-hn altitude regime. The compounds CH4, C2H6, and C3H8, wlfich have both anthropogenic and biogenic sources, show an excellent correlation with respect to C2H2/CO. This tendency is also exhibited for the anthropogenically produced compound C6H6. Over the range of C2H2/CO values from 1.5 to 0.5 (pptv/ppbv), mixing ratios of CH4 decrease from approximately 1745 to 1710 ppbv. This small relative change (i.e., 2%) is not unexpected for a compound that has a long tropospheric lifetime (--•10 years). Also as expected, mixing ratios of the more reactive compound C2H6 (lifetime • 2 months) exlfibit a greater decrease over this same range of C2H2/CO values, going from about 780 pptv to 460 pptv, and the even more reactive compounds (lifetimes of 1 to 3 weeks) C3H8 and C6H6 decrease by more than a factor of 3. For these compounds with primarily near-surface emission sources, there is also little difference in the observed nfixing ratios between the 2-to 7-km and the 7-to 12-hn altitude regimes when expressed against the C2H2/CO ordinate.
The above characteristics of hydrocarbons relative to C2H2/CO are expected because they and CO have a common primary source in the continental boundary layer and share a common atmospheric sink due to OH reactions. Indeed, for all compounds depicted in Figures 4a-4d the overall lack of a statistically significant difference (outside -t-1 a) between the data from different altitudes indicates that for these compounds the degree of atmospheric processing represented by the ratio of C2H2/CO is relatively independent of altitude. This temlency in the hydrocarbon compounds relative to C2H2/CO is partly due to atmospheric nfixing or dilution by background air, and although atmospheric mixing itself is expected to be a purely physical process, the rate of dilution should also be proportional to the photochemical sink because mixing ratios of hydrocarbons in the background air are inversely proportional to their sinks. Thus Figures 4a-4d indicate that C2H2/CO is indeed a good measure for the degree of atmospheric processing. Any trace species with a functional dependence or "trend" which deviates significantly from the trends of hydrocarbons discussed above would indicate different characteristics in its sources and/or sinks. As an example, even though the relationship is not so robust as others, aCO2 nfixing ratios do appear to reach a minimum in less processed air, a trend that could suggest biogelfiC CO 2 uptake dominates over CO 2 emissions from Asia (see also discussion by Collins et al. [this issue] and where the a refers to subtracting off the value of the smallest observext nfixing ratio in order to accentuate trends in the data). This effect would produce a negative correlation between aCO 2 and C2H2/CO. On average, the largest values of CO 2 occur at the lowest values of C2H2/CO , which should represent air that is far removed from continental sources of C2H 2 and CO or biogenic sinks of CO 2.
Unlike the hydrocarbon compounds, NO exlfibits a significant positive altitudinal dependence with respect to its relationship to the C2H2/CO ratio (see Figure 4g). Some portion of tiffs altitudinal dependence is due to the relatively rapid (i.e., timescale of •ninutes) photostationary state partitioning between NO and NO 2. The photochenfical cycles controlling the partitioning between NO and NO 2 favors a larger fraction of NO at higher altitudes due to a slowing of the reaction rates foreting NO 2 at higher altitudes (i.e., colder temperatures slowing the reaction NO + 03 • NO 2 q-02; and smaller HO 2 concentrations slowing NO + Figure 4g. However, measured values of NO• exhibit an altitudinal tendency that is similar to that seen in NO.

Several possible reasons exist for the difference between model-calculated and measurext values of NO 2 (figures not shown here; see discussion by Crawford et al. [this issue].
Since all three parmeters (i.e., NOxmeasuted, NOx calculated, and NOm,a•a ) exhibit overall similar tendencies with respect to C2H2/CO values (i.e., NO• and NO values increasing with increases in C2H2/CO values) and the NO values have no similar current controversy about their validity, we will limit our discussion to solar zenith angle fdtered NO values until these issues can be more fully resolved. Even so, our discussions will need to reflect anticipated changes with respect to the NO• partitioning between NO and NO2 as a function of temperature and ozone levels.
The trend in NOy with respect to C2H2/CO also appears contrary to initial expectations. This is especially true ff mixing ratios of NOy, which represent the sum of the odd-nitrogen-containing compounds, are conserved as various forms of reactive odd nitrogen become chemically processed in the troposphere (e.g., NOt--PAN, NOx•HO2NO2, and NO•---HNO•). Conservation of reactive nitrogen during transformations within the NOy pool would suggest that NOy mixing ratios should tenrain relatively constant as a function of changes in C2H2/CO ratios. In addition, the NOy primary atmospheric sink, namely, the "irreversible" processes associated with rainout/washout (e.g., loss of HNO• and particulate NO•) and dry deposition, has no direct relationship with the degree of atmospheric processing indicated by changes in the ratio of C2H2/CO.
However, NOy appears to follow the same magnitude of change in mixing ratios with respect to C2H2/CO values as that exhibited by the moderately long lived compound C2H 6.  Table 1 gives a listing of the classification schemes adopted in these studies, which can be broken into the general air mass origin regions depictexl in Figure 6   Japan, designated C anti c) anti southern continental (originating from over India and southeast Asia, designated I and i) classifications. These cases were taken, based on an inspection of the trajectories, to characterize the general outflow patterns encountered by the aircraft and where the tipper and lower case designations separate less than 2-day from 2-to 4-day air mass travel thnes from land. The air masses originating from the northern continental regions (C and c) have on average larger trace gas nfixing ratios than those originating frown the southern continental regions (I and i). In both cases the air masses that were encountered within approximately 2 clays or less from land (C and I) had trace gas mixing ratios that were only slightly larger (--• 1.5fold) than vahses found in what could be considered the "chenfically youngest" (i.e., larger C2H2/CO values) northem marine air masses (N) that had travel thnes of at least 10 clays from land sources. This shifilarity seemingly indicates that witlfin about 2 clays, continental air masses appear to blend in with the chenfical signattires found in the circnlation patterns over the western Pacific and suggests atmospheric processing can efficiently and perhaps quickly reduce trace gas nfixing ratios toward the average values found in the work of Gregory et al.'s northern marine air mass classifications. Perhaps contrary to initial expectations, the more moderately aged continental classifications (c and i) have quite different chenfical signattires, which appear to reflect some hnpact from anthropogenic inputs in one case (c) and more "agexl" marine air in the other (i). In all of these cases, use of the C2H2/CO ordinate as a "chenfically based" air mass classification scheme appears to coxnplhnent the trajectory-based schexne by providing a means to flirther delineate air masses within an otherwise shnilar coarse scale dynamical framework (e.g., "aged" marine air or continental air).   The air that they classified as having been convectively pumpexl into the western Pacific troposphere is seen to generally fall into the two distinct categories identifiexl by the lidar-based groupings. The air characterizexl as continental convective outflow (C) has nfixing ratios that are consistent with those found in the trajectory-based classification of continental air masses. In particular, nfixing ratios of most compounds are shnilar to those found in the northem continental 2-to 4-day and the southern continental < 2-<lay classifications. Shnilarly, the convective outflow classification (O) also follows our expectations for air masses influenced by convection over the marine environment. On average, NMHC mixing ratios for tiffs classifica-

Evaluation of the Consistency of Trajectory and Chemically Based Schemes
In contrast to the lidar-based scheme the air masses classified by the trajectory-based scheme are well segregated in the chemical space or coordinate system (i.e., C2H2/CO ) provided by the chemically based atmospheric processing scheme. The consistency displayed between these latter two schemes is not surprising if one recalls that the chenfically based scheme is designed to provide a ineasure of the degree of total atmospheric processing experienced by an air mass starting from the moment C2H2, CO, and other trace species are emittexl. Tlfis design should be well matched by the trajectory-based scheme, wlfich takes the Asian continent and the Pacific rhn countries as the reference point for an air mass origin. A valuable test of the consistency between the chemically basexl scheme for classifying air masses and the trajectory-based schelne is to colnpare the air mass travel time from land derived from the latter with the degree of atmospheric processing froln the former. To do tlfis, we need to express the degree of atmospheric processing, including photochenfical reactions anti nfixing in an equivalent pseudo-OH reaction thne as def'med in equation (2) for C2H2/CO (denoted process time herein after). Tlfis process time places all factors that affect the degree of atmospheric processing on the thne coordinate of the OH reaction with C2H 2 and CO (i.e., the corresponding time for reaction with OH to remove an equivalent amount of the compound). In equation (2), reaction rate constants at 500 mbar are usexl, i.e., 6.0 x 10 -13 cm3/s for kc2m marl 1.95 x 10 '•3 cm3/s for 1%o [Demore et al., 1992]. Furtherlnore, kco is rexlucexl by 20 % to account for the production of CO from NMHCs and CH 4 as discussed earlier.
The first step of the test is to show that the travel thne froln land and the process thne of various air masses are of the same order.
Froln our earlier discussion it is obvious that both schemes give shorter thnes for continental air masses than for lnarine air masses. This is also true on a fmer scale; for example, the continental air masses with shorter travel tinges froin land (I and C in Table 1 and Figtires 7a-7h) also have shorter process thnes than those of the continental air masses with longer travel thnes (i and c in Table 1 and Figtires 7a-7h).
It is interesting to note that the sequence of the process time depends also on the latitude of the origin of the air mass, as indicated by the fact that air masses at higher latitudes tend to have shorter process tinges (or less atmospheric processing). Assunting a constant C2H2/CO vahle ha the enfissions, there are two likely factors that would contribute to tiffs phenomenon. One is that the atmospheric sink is smaller at lfigher latitudes due to smaller OH concentrations. It is also possible that air masses at lower latitudes tend to go through greater convective nfixing or dilution resulting in longer process times (or more atInospheric processing).
The second step of the test is to compare the vahles of travel tinge from land to the indicatexl process time. Since the process thne inchIdes the equivalent amount of OH reaction "time" due to dihltion (denoted dilution time herein after), it is expected to be greater than the travel thne and the difference between them is a measure of the degree of dilution.
From Table 1 it can be seen that the travel times froln land are relatively well defmexl for the northern continental 2-to 4-day (c) and southern continental 2-to 4day (i) classifiexl air masses. So these air masses have been selected first for the second step of the test. The corresponding median vahles of C2H2/CO for these air masses in Figtires 7a-7h range from 0.8(i) to 1.6(c) pptv/ppbv. Table 1 when their process times and dilution thnes are calculated starting from the time of enfissions. Although we fifily expect the process thne to be greater than the travel time from land, the prexlonfinant fYactional contribution of the dilution thne to the overall process tinge is surprising. to the free troposphere is due to convective processes [GMel, 1983], which tend to be episodic and scattered, the entrainment may occur far upwind of the landfall point. Obviously, it is possible for this effect to account for a large part of the difference between the process time and the travel time from land calculated above. A key question now is, after taking out the upwind eftbet, how much of the difference is due to dihttion. Unfortunately, we do not have enough hffonnation on the time and location of convective events over land during PEM-West to estimate this effect.

Discussion
The preceding comparisons of the trajectory, lidar, and atmospheric processing sche•nes for classifying air masses indicate that the atmospheric processing scheme gives a meaningtiff relative scale that can be used in conjunction with the other two schemes. More hnportantly, combining the results of the three schemes (i.e., Figures 7 and 8

[this issue]), it is reasonable
to conchide that tlfis difference could be due to enfissions upwind of eastern Asia. During the period studiexl, the tipwind region was central and western Asia and Europe as the prevailing wind in the free troposphere was mostly westerlies near the western boundary of the model domain. Of course, the effects of boundaries can contribute significantly near areas where the wind is inward from the boundary. In fact, significant mnounts of the continental boundary layer tracer species that were observed in the upper troposphere during PEM-WA could owe their origins to regions far upwind of eastern Asia.
When we take the above discussion into account, we fmd evidence that the difference in calculated process times between two different air •nasses becomes more comparable to the differences between their travel times from land. For exmnple, the difference in mexlian process time between southern continental less than 2-day air masses (I) and southern continental 2-to 4-day air •nasses (i) is about 10 days; and the difference in median process tinge between northern continental less than 2-day air masses (C) and northern continental 2-to 4-day air •nasses (c) is only about 4 days. Both have relatively large tincertainties comparexl to the process thnes starting from enfissions because there is large variance in the C2H2/CO values of (C) and (I) air masses. The difference in process thne vah•e of 4 days between (C) and (c) is only about 2 times the travel tinge from land; that is, the latter is about equal to the dilution tinge. In this case, photochemical reactions during the travel time from land contribute approxin•ately as nmch as dilution to the trends in Figtires 7a-7d. Similar arguments can be made for other pairs of air mass types.
To answer the •nore general question of the dihltion tin•e's contribution to the process time, we can first exanfine points near the model's largest C2H2/CO vah•e (---2.6) which most likely crone from land areas that are relatively close to the points of observation. In other words, the travel times for these points from land should be relatively short. From Figtire 7b they can be dexlucexl to be less than 2 days, which is in agreement with the model esthnate. The corresponding process time for points with C2H2/CO of 2.6 is about 15 days. This means that dih•tion can contribute the equivalent of up to 13 days to the process thne. Furthermore, since the model's largest (C2H2/CO --•2.6) values have a short travel tinge, the dihltion must be prin•arily due to the entrainment of ambient air during the vertical transport of near-surface trace species to the free troposphere and for compounds with moderate to long lifetin]es (> 2 days). Photoche•nistry and horizontal transport processes contribute to the process tinge at a later stage.
These restilts clearly indicate that (1) dihltion or atmospheric mixing play a controlling role in determining the value of C2H2/CO; (2) trentIs of trace species relative to C2H2/CO , such as those displayed in Figtires 7a-7h, are prhnarily controlled by atmospheric mixing; and (3) for compounds less reactive than C6H6, photochenfical reactions during the travel tinge from land contribute only a nilnor part to the trends. Nevertheless, it is hnportant to recall the point made earlier, i.e., "although at•nospheric mixing itself is purely a physical process, the rate of dih•tion is proportional to the photochenfical sink because nfixing ratios of hydrocarbons in the background air are inversely proportional to their sinks." Without the sinks, atmospheric mixing would not yield any trend. Thus through atmospheric mixing, photochemical reactions do play a controlling role in determining the trends after all. Although this statement appears to contradict statement 3 above, this is not so because the term "reactions" here refers to those operative in the background air over its collective thne Iristory before the tinxe of observations, whereas the tern• "reactions" cited in the earlier statement refers to those operative only along 3x10: Figure 9. Model-calculatexl •nixing ratio of C2H 6 versus the ratio C2H2/CO between 1.94 kan and 8.14 km. the air xnass trajectory between the time of enfission and the time of observations. The answer to why the atmospheric dilution is so dominant in the degree of atmospheric processing when the process time is calculated starting from enfissions but not when the difference in process thne is calculated between two types of air masses probably lies in the vertical nfixing process. After trace species are enfitted into the surface layer over land, they are transported vertically by turbulent mixing and convection, somethnes into the tYee troposphere. These vertical nfixing processes tend to be accompaniexl by entraimnent of a large quantity of eanbient air wlfich may contain a significant eanount of background air. As a restlit, there is substantial dilution during vertical mixing. After the trace species are transportexl into the free troposphere by vertical mixing, they are then usually transported horizontally by prevailing winds, which tend to be steady with comparatively little turbulence and hence far less dilution. Therefore the answer to the question is that most of the dilution occurs during the vertical nfixing early after emission.
The above comparisons can also be used to obtain vahmble information on the sources and sinks for a number of key species. Relatively straightforward conclusions can be drawn from Figtires 7a-7d with regard to the sources and sinks of CH4 and three nomnethane hydrocarbons (NMHC). Their consistent trends relative to C2H2/CO suggest the following: (1) major sources of CH 4 and the three NMHCs are located in the continental boundary layer (inchiding the Pacific rhn); (2) major sinks of these species are due to OH reactions; and (3) there is no significant oceanic source for these species. We have also made the san•e type of plots (not shown) for CO, C2He, CFCs, HCFCs, other NMHCs, and a number of model parmneters (e.g., 03 production and destruction rates). No major surprise is found except for CH3CC13 which shows little trend relative to C2H2/CO (I '2 = 0.1 for 2-7 kin; r 2 = 0.02 for > 7 lon). Since the litetinge of CH3CCI 3 against OH reaction is shorter than CH4, which does exhibit a clear relationship relative to C2H2/CO , its (CH3CCI3) lack of a trend nmst be due to a relatively small source in Asia comparexl to sources in other parts of the northern henrisphere. For CO2, Figtire 7e reveals little additional information than what is already discussexl in the previous section, except there is an interesting indication of slightly smaller nfixing ratios for air masses at lower latitudes, suggesting a greater continental biogenic sink there. We also find that the modeled gross photochemical 03 production rates vary as a fi•nction of the degree these air masses have been atmospherically processed with average rates ranging from approximately 0. Since about 60% of lightning occurs between 20øS and 20øN [Hameed and Dignon, 1988], as opposed to subsonic aircraft emissions and stratospheric intrusions having a larger occurrence at midlatitudes, the latitudinal distribution of NO can be used to differentiate the relative hnportance of the three sources. In Figtire 7g the similarity in NO mixing ratios between northern (c) and southern (I) continental less than 2-day types of air masses also suggests that lightning is important in maintaining relatively high levels of NO in the latter air masses. This point is further supported by the complete lack of an altitude gradient in PAN mixing ratios when expressed against the C2H2/CO ordinate (see Figure lib). In addition, this view is also supported by comparisons of the lidar-based and chemically based schemes. Both the continental convective outflow (C) and the marine convective outflow (O) show enhanced NO levels relative to NMHCs (Figure 8). Specifically, the average ratio of NO/C2H • is 0.06 and 0.1 for (C) and (O) types of air mass, respectively. This is higher than the other types of air masses. In fact, the average ratio of NO/C2Ha of clean Pacific air (P) is only 0.05 at a sinfilar altitude (near 8 km) to the (O) air masses, suggesting that lightning in convecti9e activities may be a significant source of the NO (or NO0 in these air masses (see also Browell  The process thne corresponding to a mexlian C2H2/CO value of 0.9 pptv/ppbv is about 30 days, which is long enough for an air mass with 13 nffs zonal wind speexl to circle the whole globe at 30 ø latitude. Since the average zonal wind speexl near 10 Ion altitude is about 15 riffs at midlatitudes in the fall, it is obvious that the continental boundary layer trace gases, such as the CO and NMHCs observexl in the upper troposphere, may originate from regions far upwind of the Asian continent. This is supported by restilts from a regional scale model calculation that give lfiglfiy patchy distributions for trace gases transportexl into the upper troposphere by convective processes from the boundary layer of eastern Asia [Liu et al., tiffs issue]. The relatively uniform distribution of continental boundary layer tracer species observed in the upper troposphere requires that they have origir•s far upwind of eastern Asia. In fact, the upper troposphere over the western Pacific observed during PEM-West A may contain a significant contribution from air masses transported by convective processes from the boundary layer of all land areas in the northern henrisphere.

ppbv/d at smaller C•H•/CO values (--*0.3 to 0.4 ppt/ppbv) and increase with increasing varies of C•H2/CO to an average of approximately 8 ppbv/d at the largest C•H•/CO values
The above f'mding on the transport of continental trace species to the tipper troposphere is extre•nely valuable tBr testing the vertical transport parameterization usexl in models. The relative values of continental boundary layer tracers such as NMHCs in the stratospherically influencexl air masses can be usexl to evaluate quantitatively the vertical transport schemes of models which usually are highly paratneterized and difficult to evaluate for their accuracy. Many important tropospheric issues cannot be adequately addressexl until vertical transport processes are accurately simulated in models. For exeanple, tropospheric ozone budget depends critically on the vertical transport of NO x from the continental boundary layer to the free troposphere [Liu et al., 1987;Pickering et al., 1990;Jacob et al., 1993]. Another exatnple is the transport of sulfilr species and aerosols to the free troposphere which are important to the radiation budget.

Summary
We have employexl a chenrically based air •nass classification scheme using the combustion products C2H 2 and CO as tracers of continental source emissions.
For the data taken during the 1991 PEM-West A program, the degree of atmospheric processing that reflects the combinexl effects of chenristry and nrixing appears to be well representexl by the measured ratio C2H2/CO. A large number of compounds having free tropospheric photochenrical lifethnes t¾om clays (e.g., NOx and C6H6) to years (CH4) were fBund to display similar tendencies with respect to the degree of atmospheric processing indicated by the ratio C2H2/CO. Comparison of the C2H2/CO-basexl atmospheric processing scheme with two other independent schemes, i.e., isentropic back-traj•tory-based scheme and lidar O• and aerosol-based scheme, for classifying air masses indicates that the C2H2/CO-based scheme is successfill in providing a consistent and meaning fid measure of the degree of atmospheric processing for the majority of the air masses classified by the latter two schemes. In addition, the C2H2/CO-basexl scheme appears to segregate between the chemical differences in air masses with greater than 10-day travel times from sources and for those modifiexl by convective events. The C2H2/CO-basexl scheme and the trajectorybasexl scheme are more compatible with each other than with the lidar-based scheme because the former two schemes use the land as their reference point.
The C2H2/CO-based scheme also compares very well with restilts calculated by a three-dhnensional regional model except for a slfift in the distribution of model points toward smaller values of C2H2/CO, which is attributed to the effect of emissions upwind of the model domain. The comparison indicates that most of the initial atmospheric processing following enfissions of trace Sl•,•cies near the stirface is due to entrainment of mnbient air during the vertical transport of the trace species from the surface to the free troposphere. Photochemistry and horizontal transport processes contribute to the atmospheric processing at a later stage.
The complimentary nature of these three sche•nes enable them to be combined in order to obtain vahmble infornmtion on the sources and sinks for a nmnber of key species. First, a significant noncontinental source(s) of NO (and NO 0 in the free troposphere is indicated and is attributed to either the enrissions frmn lightning or the r•ycling of NOy transported through deep convection. Second, for the western Pacific, continental boundary layer tracer species, such as hydrocarbons, are vertically transportexl into the upper troposphere as efficiently as they are transportexl into the •nidtroposphere. Third, the CO and NMHCs observexl in the upper troposphere may reflect a significant contribution from convectively transported stirface enrissions from northern henrispheric land areas other than Asia. Finally, we believe that restilts of the C2H2/CO-basexl scheme can be extremely valuable for the quantitative evaluation of the vertical transport processes that are usually parmneterizexl in models.