Hydrocarbon ratios during PEM-WEST A: A model perspective

. A useful application of the hydrocarbon measurements collected during the Pacific Exploratory Mission (PEM-West A) is as markers or indices of atmospheric processing. Traditionally, ratios of particular hydrocarbons have been interpreted as photochemical indices, since much of the effect due to atmospheric transport is assumed to cancel by using ratios. However, an ever increasing body of observatonial and theoretical evidence suggests that turbulent mixing associated with atmospheric transport influences certain hydrocarbon ratios significantly. In this study a three-dimensional mesoscale photochemical model is used to study the interaction of photochemistry and atmospheric mixing on select hydrocarbons. In terms of correlations and functional relationships between various alkanes the model results and PEM-West A hydrocarbon observations share many similar characteristics as well as explainable differences. When the three-dimensional model is applied to inert tracers, hydrocarbon ratios and other relationships exactly follow those expected by simple dilution with model-imposed '(cid:127)'background air, and the three-dimensional results for reactive hydrocarbons are quite consistent with a combined influence of photochemistry and simple dilution. Analogous to these model results, relationships between various hydrocarbons collected during the PEM-West A experiment appear to be consistent with this simplified picture of photochemistry and dilution affecting individual air masses. When hydrocarbons are chosen that have negligeble contributions to clean background air, unambiguous determinations of the relative contributions to photochemistry and dilution can be estimated from the hydrocarbon samples. Both the three-dimensional model results and the observations imply an average characteristic lifetime for dilution with background air roughly equivalent to the photochemical lifetime of butane for the western Pacific lower troposphere. Moreover, the dominance of OH as the primary photochemical oxidant downwind of anthropogenic source regions can be inferred from correlations between the highly reactive alkane ratios. By incorporating back-trajectory information within the three-dimensional model analysis, a correspondence between time and a particular hydrocarbon or hydrocarbon ratio can be determined, and the influence of atmospheric mixing or photochemistry can be quantified. Results of the three-dimensional model study are compared and applied to the PEM-West A hydrocarbon dataset, yielding a practical methodology for determining average OH concentrations and atmospheric mixing rates from the hydrocarbon measurements. Aircraft data taken below 2 km during wall flights east of Japan imply a diurnal average OH concentration of -3 x 10 6 cm -3. The characteristic time for dilution with background air is estimated to be -2.5 days for the two study areas examined in this work. for approaches processing


Introduction
An important objective of the PEM-West A experiment is to evaluate the anthropogenic impact to tropospheric chemistry over the Western Pacific Basin [Hoell et [Browell et al., this easily explained by the influence of the large ethane background within the dilution process. By incorporating back-trajectory information within the three-dimensional model analysis, several important results are obtained. Within an idealized setting of the three-dimensional model, a simple technique is derived for determining the relative importance of photochemistry and dilution to the hydrocarbon ratio for a specific region of the model domain. This technique relies heavily on the simplified conceptual dilution model presented by McKeen and Liu [1993], which contains some inherent limitations that must be addressed before the practical extension of this technique can be applied to the observed data set. A limited subset of the PEM-West A data is used as a test application of the technique, from which atmospheric implications and comparison to the three-dimensional model results are discussed.

Model Calculations
The three-dimensional model used in this study is described by Liu  Most importantly, convective transport due to cloud pumping is included as a transport process according to Linet al. [1994], and boundary conditions above 5 km for several key species are set proportional to potential vorticity to better simulate the stratospheric influence on the free troposphere.
Details of the emissions inventory used within the threedimensional photochemical model are outlined by Liu et al. [this issue]. Anthropogenic emissions of NO x, which are necessary for the calculation of key photooxidants, are based on the spatial distribution of NO x emissions for the western Pacific [Dignon, 1992] Atkinson, 1990] as well as the numbers of observations above detection limits. Unlike the detailed photochemical calculations, the additional complication of inhomogeneous HC emission ratios (relative to CO) is removed for the specific HC simulations by assuming that emission rates are everywhere proportional to that of CO. Table 1  From all indications the model system appears to adequately address transport and its interaction with oxidant photochemistry in the lower troposphere over timescales from -1 hour to a few days and distance scales between -150 and 3000 km. Thus the three-dimensional model results are well suited for examining the interaction between photochemistry and transport for the conditions of the PEM-West A experiment but not appropriate for direct comparison to observed HC, since absolute emission strengths over much of Asia are unknown. We restrict the analysis of three-dimensional model and observed HC to the lowest 2 km of the troposphere for two As discussed below, it is advantageous within the analysis to reduce uncertainties associated with assumed background concentrations by sampling air masses with clear anthropogenic signals. Secondly, although enhanced HC concentrations were observed at higher altitudes, particularly above 7 km, the likelihood that these air masses are a result of convective pumping introduces uncertainty as to their origin, which could be upwind of the model domain [Liu et al., this issue]. Model boundary and initial values (Table 1) were adopted from measurements made below 3 km over remote oceanic areas during the PEM-West A mission. Table 1 also shows the expected lifetime of the chosen HC for a particular average OH concentration in the lower atmosphere ranging from -1 day for toluene to -4 years for CH 4. To help demonstrate the impact of photochemistry on the individual HC, seven additional species are also calculated with emissions, boundary and initial conditions identical to those listed in Table 1, but with photochemical destruction turned off. As shown below, relationships between these inert analogs provide useful information for diagnosing the threedimensional model results and gaging the influence of photochemistry.
As discussed by McKeen and Loeu [1993], an important element needed to unambiguously discern the effects of photochemistry versus atmospheric dilution is the time since emission of the HCs into the sampled air mass. To derive this quantity, for both model results and observations, backtrajectories are calculated isokinetically from the mesoscale model-derived half-hour three-dimensional wind fields, with three-dimensional mapping of grid centers to backward integrated origination coordinates saved at 2-hour intervals. Eight-point interpolation between grid centers is then used to map the trajectory for an arbitrary location, with the shortest time and location of land contact being determined to 2-hour and 60-km resolution. Because of uncertainty in the actual coastline with respect to the model grids, one grid square is added to the land use database, so that land contact here is defined as 60 km or less seaward from the actual coastline.

Model Results
This section is divided into two general parts. Section A discusses various HC relationships common to both the threedimensional model and the observations. From this discussion the effects of atmospheric mixing within the PEM-West A observations will be demonstrated and quantified relative to the rate of OH photochemistry. Section B incorporates the back-trajectory analysis to establish a relationship between model-derived HC concentrations and time from emission.
From the results of this section, a procedure for quantitatively determining the individual effects of photochemical decay and atmospheric mixing is derived and applied to a selected set of PEM-West A observations.

Section A, Basic Hydrocarbon Relationships
Although the three-dimensional model calculation extends over a large region of the eastern Asian continent, only a limited subdomain of the model is useful in simulating the conditions of the PEM-West A experiment. The region enclosed by the dashed lines in Figure 1 is chosen as the study area, and was chosen to coincide, as much as possible, with the study area of Gregory  for comparison. This day and the previous day were periods where moderately high pollution outflow from the Asian continent was simulated by the photochemical-transport model. The choice of a particular day is not important to the analysis, but October 4, 1991, provides a larger range of conditions that vary from moderately polluted to pristine.
As discussed by Brost [1988], regional scale threedimensional photochemical models are limited to large extent by influence of boundary conditions, particularly as one goes higher in the free troposphere. Indeed, most of the three- To facilitate discussion of the interaction between atmospheric mixing and photochemistry, it is useful to consider a highly idealized conceptual model relating the two processes. In a very simplistic approach, the mixing ratio of emitted species X can be changed by chemical loss or mixing with an infinite reservoir of background air: dX -LxX-K(X-X •) ( 1 ) dt where X is the mixing ratio, t is time, X B is the prescribed mixing ratio of X in the background air, L X =kx[OH ] is the loss frequency of species X with OH (bimolecular reaction rate constant kx) and K is a mixing coefficient used to parameterize all the processes that mix species X with background air. There are several major simplifications in equation (1) that limit its applicability. Most important is the fact that mixing with air masses containing mixing ratios of X other than the background value X B is ignored. For inert species this is not a problem, since the mixing process is entirely linear, but the addition of photochemical loss disrupts this linearity, which cannot be addressed in equation (1). Additionally, applying this simple formalism to observed data leads to practical problems, in particular a very strong dependence of the assumed background mixing ratio for the less reactive hydrocarbons. We also must assume that K and L X are constants in order to obtain the analytical solution where X ø is the initial mixing ratio of X upon emission.
The conceptual picture of air parcel aging embodied by equation (1) is specifically designed for an isolated source embedded within a very large and relatively clean region, such as pollution plumes originating from islands or coastal locations. Since only photochemical decay and dilution are considered, the simple formulation would not apply to conditions where emissions or multiple sources are also influencing concentrations (e.g., a polluted continental boundary layer). Equation (1) offers a simple formulation that incorporates several concepts related to hydrocarbon aging.
The HC aging picture based solely on OH photochemistry [e.g., Roberts et al., 1984;Rudolph and Johnen, 1990] is augmented by the dilution term. Smyth et al.
[this issue] use meteorological scaling arguments and an analogy to the concentration decay observed with exponential dilution flasks [Lovelock, 1961] to infer rough estimates of atmospheric dilution. The exponential decay within a dilution flask system is equivalent to equation (2) for an inert species with no background and the dilution constant K replaced by the flow velocity divided by the flask volume. Liu, 1993;McKenna et al., 1995] and the qualitative argument of HC clock resetting by discrete mixing with air parcels of a different age [Parrish et al., 1992] are implicitly incorporated within the continuous dilution term.

Species Versus Species Relationships
When a similar equation is written for species Y, ratios can be derived from equation (2) that are independent of time and/or the mixing parameter K, depending on assumptions specific to each HC. For example, if species X and Y are inert, their ratio to the amount above background is a constant and equal to the emission ratio (X0-xB)/(Y0-yB). Similarly, if the photochemical decay rates (L X y) are much larger than the mixing parameter K, time can b• eliminated between the two solutions, yielding a linear photochemical decay on a log-log plot. Figures 2a and 2b show the limiting cases for the dilution of inert species (solid line) and the photochemical OH-kinetic relationship (dashed-dotted line) for the C3H 8n-C4H 10 pair. The placement of the intersecting point in both of these figures is somewhat arbitrary, but the observed and three-dimensional modelled data clearly fall between these two limiting cases. A general solution of Y versus X, including both photochemical and background terms, has a complicated dependence on the assumed initial mixing ratio as well as background conditions. If the two species have photochemical loss rates strong enough to assume that background concentrations are zero, then the relationship simplifies to  _+0.27. This number implies that the concentration changes of hydrocarbons with reactivities less than or equal to n-butane over the western Pacific are influenced equally or slightly more by dilution than by photochemistry. The results are quite similar when comparing other highly reactive alkanes to C3H 8, Again, using the OH reactivity with n-C4H10 as a reference, the dilution coefficient to loss ratio (K/Ln_butane) determined from regression slopes of i-C4H10, n-C5H12, i-CsH12, and n-C6H14 are 1.14, 1.30, 0.93 and 1.27, respectively. As the reactivity increases, the uncertainty becomes larger because there are fewer points measured above detection limit and even fewer points with mixing ratios above the n-C4H10 discrimination limit of 30 pptv (e.g., 67, 35, 41, and 5 points for i-C4H10, n-C5H12, i-C5H12, and n-C6H14, respectively). Using n-C4H10 or i-C4H10 as the least reactive HC in the regression determination also yields quite similar results (equivalent K/Ln_butane = 1.2 +0.4) but with additional uncertainty due to fewer points and the smaller difference between the mixing slope of 1 and the photochemical slope between the more reactive alkanes. The highly reactive alkanes therefore yield a reasonably consistent value of the relative importance of dilution and photochemistry, regardless of the choice of HC used in the regressions. Because dilution is apparently so effective for the PEM-West region, our results generally imply that the use of hydrocarbon ratios as photochemical markers, or as indices of reaction time, should only be applicable when the hydrocarbons are more reactive than n-C4H10. Additionally, the high reactivity of these HC requires either sampling of very polluted conditions or extremely sensitive and accurate measurements under less polluted conditions in order to accurately use them as photochemical clocks. Ratios of less reactive species still provide a measure of atmospheric processing, but changes in a HC ratio are influenced more by physical mixing than by photochemistry. Lastly, rough estimates of K can be derived for reasonable assumptions of [OH].
From the threedimensional model, a diurnally average OH concentration of 2 x 106 cm -3 is determined over the study area outlined in Figure   1 Table 1 for the three-dimensional model results in Figure 3b. Although general analytical solutions relating two such HC can be derived, they depend on the assumed background, which is undefined. There may exist a unique set of background concentrations that would yield a meaningful regression analysis from the observed data. However, such an exercise may overstate the validity of the simple dilution model, especially for measurements over a region as large as the western Pacific where a background may have strong latitudinal and temporal dependencies. Comparing Figures 2a and 3a, there also appears to be a much tighter correlation between C3H 8 and n-C4H10 than between  The reactive species relationships show more scatter, due to different photochemical and mixing histories of individual air masses, as mentioned previously. In both figures the relationship between reactive species remains similar to the relationship for the inert species but are shifted to lower concentrations. The effect of chemistry thus appears to smear the basic relationship curve and also to displace the points as well as the backgrounds nearly along the strictly photochemical decay slope represented by the dashed-dotted line in Figure 3b. This implies that the mixing process dominates the C2H 6 -C3H 8 relationship, while the first-order effect of chemistry is simply to displace the background.

Ratio Versus Species Relationships
With a little algebraic manipulation of equation (2), similar relationships for the two limiting cases can be derived between an HC species and the ratio of two HCs. For the inert species X= ZB(BX/Z+E x/z(Y/z'BY/z) ) where E I/Z is the emission ratio of species I to species Z and B I/Z is the ratio of the background value of species I to that of species Z. For the other limiting case where L X y Z >> K, a linear slope on a log-log plot again defines the p}iot'$chemical decay relationship. As in the case for one HC versus another, the general solution for the case of both dilution and photochemistry being significant is complicated by a nontrivial dependence on initial conditions and the assumed background concentration, but the restrictive use of highly reactive HC (with zero assumed background) leads to a workable relationship between a species and a ratio: [z The slope of X versus Y/Z on a log-log plot again defines the exponent term which can be related to the loss rate of n-C 4H10 .
It is easy to show that if Y and X are the same species, then the regression slope determined from the log(X/Z) versus log(X) yields the exact same result as simply determining the regression slope from log(X) versus log(Z), so there is no additional information obtained by examining n-C4H10/C3H 8 versus n-C4H10 for example. Figure 4a shows the limiting cases, observations, and best fit line for n-C4H10/C3H8 versus C3H 8. As before, only samples in which n-C4H10 is > 30 pptv are used in the regression fit. The regression slope yields a K/Ln_butane ratio of 2.1 +.5, nearly twice as large as previously determined.  Figure 5b, which can simplistically be ascribed to variations in relative emissions between the three species that are not included in the three-dimensional model. The sharp tailing-off of the model results, an artifact of the imposed boundary conditions, is not apparent in the observations. Clearly, the regional scale model is an inappropriate tool for studying relationships between longlived HC, particularly for clean conditions. As in the species-species relationships, the three-dimensional model results for the inert species follows the simplified mixing form quite well in Figures 4b and 5b. The reactive points are smeared compared to the inert relationship but preserves the essential shape. Similar to the discussion related to Figure 3b Okinawa source. Thus an explanation of the scatter within Figures 6a and 6b would arise from the difference in HC decay between a point source emitting into relatively clean conditions and diluting rapidly, as opposed to a broad regional source with only peripheral contact with background air. However, the three-dimensional model OH concentration is higher within regional pollution plumes which also may contribute to the scatter.
As in the species-species and species-ratio relationships, the general solution to the ratio-ratio relationship when both background and photochemical loss are significant is dependent on background assumptions. However, unlike the previous cases the ratio-ratio relationship is not dependent on the absolute amount of HC upon emission, only the emission ratios. For the species used in Figures 6a, and 6b, the denominator has a significant background value and the numerators have negligible background values (with the appropriate windowing). In this case, manipulation of equation ( plots but also the rate that hydrocarbon relationships move along the parametric curves. The 2500 pptv background case represents an extreme limit to the relationship, since C2H 6 does not change with time. For this particular case, the slope of the log-log relationship is identical to that determined by regression in Figure 2a, and the time to reach an arbitrary, low ratio is a minimum. With lower backgrounds the parametric curves start out with steeper slopes, but after about two n-C4H10 lifetimes, the slopes become nearly equivalent to that of the invariant C2H 6 case. This is of course expected, since C2H 6 will be close to background and more invariant as time increases, having less of an influence on the ratio relationship. Thus the dilution model implies that the slope of this log-log ratio should approach the value (K+Ln_butane)/(K+Lpropane) for those points well removed from the emission source.   Table 2 have drastically different values than the kinetic slopes for OH reactions. This is particularly true when i-C4H 10 is the numerator of the ordinate. In this case, the C1 reaction with i-C4H 10 is nearly identical to that of C3H 8, making the expected slope of the C1 kinetic ratio close to zero. The slopes for the numerator-abscissa combinations of the second row in

Section B, Inclusion of Back Trajectories
In To adequately test the photochemical-mixing relationship proposed in equation (1), a sufficient sampling of points with clear signals of anthropogenic influence from a relatively isolated region is desirable. Additionally, it is also desirable to have a large range in the time from emission in order to test the simple dilution model over an appropriate dynamic range. For data collected above 2 km, most back trajectories pass over land at altitudes that are not impacted by anthropogenic activity along the coast, so the data used in this analysis are restricted to -200 samples that were collected below 2 km altitude during flights 6-13. From these data, only 70 of the back trajectories are determined to have land contact times of 4 hours or greater during the September 20 to October 6 period within the model domain. Many of these back trajectories either originated from relatively clean coastal regions or show anthropogenic influence from an isolated source region but exhibit very little variation in the contact time from land. The clearest signals of anthropogenic influence with a wide range of land contact times to a small origination region is from the data collected east of Japan on flights 7 (9/24/91) and 8 (9/25/91). Figure 8 shows the flight tracks of these two flights and the back trajectories for several of the data-collection locations. Eleven of the 70 possible back trajectories originate within a 180-km region close to and north of Tokyo with the height of the origination point determined to be between 1.4 and 2.3 km above sea level. Hydrocarbon values with back trajectories originating 60 km north or south of this region were significantly lower for a given land contact time. The back trajectories also indicate that the winds were fairly steady and persistent during the travel time, so that strong meteorological variations should not be an added complication to consider.
Equations (1)  due mostly to the two low HC sets at 32 and 54 hours from contact, the exponential decay appears to describe the dependence rather well. Moreover, the increasing slope of the decay as hydrocarbon OH reactivity increases is quite obvious. The chemical time constants for ethane and acetylene are long enough that the decrease of these species should be due mostly to mixing, while the more reactive species will have an additional contribution from OH loss. It therefore appears plausible that the observed data can be used to estimate L X , K, and their relative importance to the exponential decay. However, equation (5) contains the initial concentration upon emission and the background concentration in addition to L X and K as unknowns. Figures 9a and 9b, a slope and intercept at t=0 would be determined that best fit the data. However, the initial condition is only with respect to a particular back trajectory's contact with the coastline. Between the time of an air parcel's injection of HC and the time it reaches the coastline, small-scale transport and dilution are expected to significantly influence the HC. Since the intercept or initial condition at t=0 determined from the regression fits are expected to have little connection with the original emissions, they are ignored in this analysis. The dilution and photochemical loss information of immediate interest is contained in the regression slopes, which depend on the undefined background values specified within the formalism. As discussed in the previous section, one can safely assume that for hydrocarbons with sufficiently high OH reactivity the background is negligible. For HC with OH reactivities slower than butane, the PEM-West A data imply there is always a measurable amount originating from ubiquitous hemispheric or continental sources, which may have a measurable impact on the absolute amount observed near more recent emissions. Since equation (5)  Model hydrocarbon results are obtained and back trajectories are calculated for the area east of Japan outlined in Figure 8 for 0600 UT, September 4, 1991, roughly the midtime of most of the observations in Figure 9. Analogous to the filters applied to the observed data, only data below 2 km and with back trajectories originating from the same 180 km of coastline are used in the three-dimensional model analysis. Figures 10a and 10b show the decay plots for the same hydrocarbons shown in Figure 9. The stronger decay slopes for more reactive hydrocarbons are evident in the figures.

The term containing the initial concentration represents an offset to the time dependence in equation (5). In a simple regression fit to the points of
As discussed in the previous section, the simple dilution model is able to reproduce relationships between nonreactive species quite accurately due to the linear nature of the dilution process in the absence of photochemical loss. Thus the background concentrations are well defined by the boundary conditions, there is an unambiguous determination of the terms on the left hand side of equation (5) from the threedimensional model results, and the slope of a linear least squares fit to the logarithm of the amount of inert material above background versus contact time determines the dilution coefficient. Applying this procedure to the inert analogs of the species listed in Table 1 results in a derived K of 0.016 + .004 hour -1 for the three-dimensional model, or a characteristic lifetime for dilution of 2.6 +0.6 days, independent of the species used in the least squares fit. Although this quantity is a characteristic of the numerical model's net transport for only a particular location and 2-day period, it is interesting to compare this characteristic time with the photochemical lifetimes in Table 1. The timescale for the dilution process is faster than that of photochemistry for all species with OH reactivity less than n-butane, which explains why photochemistry appears to have only a secondary effect in determining the three-dimensional model HC relationships in Figures 3b and 5b. Once K is known, the three-dimensional model average [OH] can be determined unambiguously from linear least square fits of n-butane and toluene decay, since the background mixing ratios of these two species are set to be near zero. Subtracting K from the best fit slopes and dividing by OH reactivity yields an [OH] of 2.5 (+0.9) x 106 and 2.1 (_+0.5) x 106 cm -3 from n-butane and toluene, respectively. The

Back Trajectories and Hydrocarbons From Observations
With these three-dimensional model results as a guide, the simple dilution model can be applied to the observations. First, the background of a given hydrocarbon is defined, based on back-trajectories below 2 km that have not had land contact for at least 96 hours from all the observations between 9/22/91 and 10/6/91 (-26 samples). The average over the lowest quartile of these clean, marine-air values is chosen to represent the background concentration in keeping with the threedimensional model results. These statistically determined backgrounds and the associated relative background uncertainty for the Tokyo plume are given in the last two columns of    . This iterative procedure converges so long as there is reasonable separation in the reactivities of the two hydrocarbons and there is not excessive scatter in the hydrocarbon decay slopes. The iterative procedure is also independent of the initial guesses to [OH] and K, provided they fall within realistic limits of zero and the maximum possible value deduced from the decay slope of either species. As mentioned previously, the intercept value is allowed to be determined in the regression fit but is ignored in this analysis. The lines in Figures 9 and 10  and trajectory origin that are not included in these uncertainty estimates. As discussed with regard to the three-dimensional model results, uncertainty due to the background assumption can be significant but is easily comparable between the various HC by the relative background uncertainty defined previously. From the three-dimensional model results, a species with a relative background uncertainty less than 0.15 will have a smaller uncertainty associated with the background than with the statistical scatter of the data. Measurement uncertainties of species having concentrations close to the detection limit (e.g., toluene, n-C6H14, and pentanes) should be reflected somewhat in the statistical scatter of the data.
Compared to the relatively large uncertainty generated by the statistical scatter attributable to the small sampling size (11 points), the other sources of uncertainty are expected to be minor. Nonetheless, the uncertainties quoted in Tables 3a and  3b  In fact, the decay slope of toluene is less than that of n-hexane or the pentanes, which does not seem possible since its OH reactivity is the fastest of these species. Assuming the kinetic data for toluene is not off by more than 50% and there is no selective input of toluene along the trajectory, the only possible explanation is a systematic bias in the toluene measurements, which is close to the detection limit for many of the samples used in this analysis. Similarly, [OH] values based on n-pentane are systematically higher than for other HC; however, the large uncertainty and the fact that samples were close to the detection limit precludes a definitive difference to be ascribed to n-pentane in relation to the other HC. All of the characteristic dilution time determinations based on CH3CC13 as the least reactive HC are systematically higher than with other HC. In this case, the relative uncertainty in the background is higher than all other HC. An 8% increase in the adopted background would reduce the characteristic dilution times by 25% and feed back into the [OH] determinations based on this species. In contrast to the three-dimensional model results the observations of CO, CH 4, and C2H 6 have small relative background uncertainties and should be the most reliable species to use as the least reactive HC.
In deriving a best estimate of [OH] and the characteristic dilution time, estimates within Table 3  reactive alkanes (C 3 to C 6) are therefore used as the observational basis for deriving the relative dilution rate and testing the applicability of the simple dilution model for the lower tropospheric data set in the vicinity of eastern Asia. To eliminate any possible bias from measurement iraprecision or inaccuracy, only data significantly greater than the detection limits are included in the analysis. In terms of the highly reactive hydrocarbons, several features of the threedimensional model results and observations are similar and consistent with the simple dilution model. According to the dilution model for species with insignificant backgrounds, regression fits to the log of one species versus another or the log of a ratio versus the log of a species yields a quantitative determination of the dilution coefficient, relative to the loss rate of a particular hydrocarbon. Comparing one species versus another, both the threedimensional model results and the observations are consistent with an average rate of dilution roughly equivalent to n-butane oxidation for the western Pacific lower troposphere during solstice. This result is also consistent among the various possible combinations of highly reactive alkane pairs afforded by the hydrocarbon observations. Since dilution apparently has such a large effect, this result has negative implications for the use of hydrocarbon ratios as photochemical clocks, particularly with hydrocarbons having OH reactivity less than butane. Similarly, hydrocarbon ratios used as markers of atmospheric processing such as C2H2/CO or C3Hs/C2H 6 should be considered largely a measure of atmospheric dilution with photochemistry playing only a minor part. The role of dilution in affecting other photochemical relationships, i.e., between NO NO CO, and 0 3 is an obvious direction for future study. y' x, ' For negligible backgrounds the dilution model predicts that the regression slope determined from the logarithm of one ratio versus that of another depends only on the photochemical loss rates and not on the history of dilution and mixing that occurs downwind of the source. The observed data fit the expectation of dominant OH kinetics affecting the hydrocarbon ratio relationship, and this result is consistent between various combinations of reactive alkanes used in the ratios. Equivalent regressions to the three-dimensional model results also yield the OH photochemical rate relationship imposed by the model formulation. This is a clear indication that the observations do conform to the expectations of the dilution model, but comparing one ratio to another yields no information on the relative importance of photochemistry and dilution.
No attempt has been made in this study to fit a large set of the less reactive hydrocarbons to the dilution model as was done for the more reactive alkanes. This is because a background value must be assigned for each hydrocarbon, and the simple dilution model cannot be expected to adequately treat the statistical nature of real-world atmospheric mixing processes. Although this is also true of for the highly reactive hydrocarbons, the atmosphere is always mixing emissions toward the zero background for these HC, except in the transient event of plume overlap. In the case of less reactive species the background concentration is probably a complicated function of latitude, longitude, and season. Given this conceptual difficulty and the fact that any regression analysis depends strongly on the chosen background, only qualitative comparisons between threedimensional results and various HC relationships from the measurements of less reactive HC are examined. However, one valuable interpretation of ratio versus ratio relationships previously studied should be mentioned. For the case of n-C4H10/C2H 6 versus C3H8/C2H 6 the dilution model predicts that the regression slope should lie between the photochemical and the inert-dilution limits simply because of the high background for C2H 6. If an appropriate background were subtracted from C2H 6 in the ratios, then the expected photochemical decay slope could theoretically be determined by the modified regression slope. The inclusion of back-trajectory analysis for both the three-dimensional results and the observations allows for the absolute determination of rates by introducing time into the analysis of hydrocarbon decay measurements. Within the entire PEM-West A hydrocarbon data set, only a very limited number of points (eleven) were determined to have arrived from a contiguous source region with sufficient anthropogenic signal and a sufficient range of land contact times to allow meaningful statistics. These points originated from the Tokyo region and determined to be sampled from 2 to 4 days after last contact with the coastline. Clearly, a measurement platform and experiment specifically designed to study the evolution of pollution plumes into an ocean basin would be desirable to obtain better statistics and further test the analysis procedure presented here. Designed more as a survey experiment, it is only by happenstance that the PEM-West A mission was able to encounter and collect the available samples from the Tokyo plume used in this analysis. The decrease of inert species from the three-dimensional model for a sample of points originating ,from the same location results in the determination of a dilution coefficient that has a characteristic time of-2.5 days, independent of the inert species chosen in the regression. With this dilution rate the three-dimensional model decreases of nbutane and toluene are both consistent, to within 20%, of the 2.1 x10 6 cm -3 average OH determined from averaging the model OH field directly. Applying the dilution rate and OH concentration to the other three-dimensional model species that have a significant background shows that the best choice of a background value to use in regression fits corresponds to the lowest values of that hydrocarbon in the model domain.
With these three-dimensional model results as a guide, a methodology is developed to derive the dilution coefficient and OH concentration from the 11 samples based on fitting the decay rates of observed hydrocarbon pairs. The results are reasonably consistent between various hydrocarbons, and yield an average OH of 3.2 x10 6 cm -3 for the 3-day period September 23-25, 1991, in the lowest 2 km, for the study area east of Japan shown in Figure 8. The characteristic time for dilution is determined to be -2.8 days, which is quite consistent with the three-dimensional model results from the same area. Assuming the average OH for the western Pacific is 1.5 to 2 x 10 6 cm -3, the dilution coefficient determined from the regression of all the hydrocarbon observations below 4 km is also quite consistent with this estimate.