Assessment of ozone photochemistry in the western North Pacific as inferred from PEM-West A observations during the fall

D 4 Abstract. This study examines the influence of photochemical processes on ozone distributions in the western North Pacific. The analysis is based on data generated during NASA's western Pacific Exploratory Mission (PEM-West A) during the fall of 1991. Ozone trends were best described in terms of two geographical domains: the western North Pacific rim (WNPR) and the western tropical North Pacific (WTNP). For both geographical regions, ozone photochemical destruction, D(O3), decreased more rapidly with altitude than did photochemical formation, F(O3). Thus the ozone tendency, P(O3), was typically found to be negative for z < 6 km and positive for z > 6-8 km. For nearly all altitudes and latitudes, observed nonmethane hydrocarbon (NMHC) levels were shown to be of minor importance as ozone precursor species. Air parcel types producing the largest positive values of P(O3) included fresh continental boundary layer (BL) air and high-altitude (z > 7 km) parcels influenced by deep convection/lightning. Significant negative P(O3) values were found when encountering clean marine BL air or relatively clean lower free-tropospheric air. Photochemical destruction and formation fluxes for the Pacific rim region were found to exceed average values cited for marine dry deposition and stratospheric injection in the northern hemisphere by nearly a factor of 6. This region was also found to be in near balance with respect to column-integrated 03 photochemical production and destruction. By contrast, for the tropical regime column-integrated 03 showed photochemical destruction exceeding production by nearly 80%. Both transport of 03 rich midlatitude air into the tropics as well as very high-altitude (10-17 km) photochemical 03 production were proposed as possible additional sources that might explain this estimated deficit. Results from this study further suggest that during the fall time period, deep convection over Asia and Malaysia/Indonesia provided a significant source of high-altitude NOx to the western Pacific. Given that the high-altitude NOx lifetime is estimated at between 3 and 9 days, one would predict that this source added significantly to high altitude photochemical 03 formation over large areas of the western Pacific. When viewed in terms of strong seasonal westerly flow, its influence would potentially span a large part of the Pacific.


Introduction
Since the early 1970s when the first papers appeared raising the issue of photochemical production of ozone in the remote troposphere [Chameides and Walker, 1973;Crutzen, 1973]; the topic of tropospheric sources and sinks of ozone has been one of intense scientific investigation and debate [e.g., Fabian, 1974;Chameides and Walker, 1973 Figure lb shows the same data set as presented in Figure  l a but plotted as a function of latitude and altitude. We note that only those observations having solar zenith angles of <70 ø are shown since only these data were used in our analysis. With the possible exception of one altitude block in the tropics (i.e., 10-12 km), Figure lb illustrates that a reasonable distribution of data was available covering most altitudes over the latitude range of 0ø-42øN. Table 1 where the mean, standard deviation, median, and max and min values for several photochemical species are presented for the latitudinal breakouts of 0ø-18øN and 18ø-42øN. From these summary data it is apparent that mixing ratios for all five species (i.e., NO, 03, CO, C3H8, C6H6)have significantly greater variability and larger mean and median values for the highlatitude regime. In particular, 03, NO, and the NMHCs have median values which differ by nearly factors of 2 for these two regimes. Reflecting these gradients in 03 and NO, Figure 2c shows that model-calculated values of the photochemical ozone tendency also display a strong latitudinal dependence. In addition to the gradients observed as a function of latitude, significant shifts in trace gas mixing ratios were also seen as a function of altitude. For most species, however, this trend was much weaker than that observed with latitude. It also tended to be quite nonuniform. For some species, low-altitude values exceeded those at highaltitudes; for others the trend was reversed, and in still other cases mid-altitude values were lower than either those at high or low-altitudes [e.g., Blake et Table 1, NO was unique in having a strong positive vertical gradient. As shown in Figure 3, a rather dramatic upward ramping in the NO mixing ratio is seen, with the most abrupt change occurring at altitudes near 6 or 7 km. Although this trend was characteristic of both the high and the low-latitude regimes, it was far more pronounced at high-latitudes. For the entire data set (neglecting 15 data points near the coasts of Japan and China), the median NO mixing ratio is estimated to change by nearly a factor of 6 over the altitude range of 0 to 12 km. Figure 2c, differences in the distribution of ozone precursor species with latitude and altitude had a significant impact on the calculated photochemical formation, destruction, and tendency of 03. For this reason, much of the analysis that follows will focus on comparing and contrasting the photochemical environments of two regions.

As illustrated in
The western North Pacific rim (coordinates: 135 ø to 150øE at 42øN and 112 ø to 148øE at 18øN, labeled hereinafter WNPR) may be characterized as a region influenced by both natural and anthropogenic continental sources. In particular, available evidence points toward this region as having been significantly impacted by high-altitude (e.g., 6-12 km) outflow from the Asian continent and quite likely from other will also prove to be useful. Still more speculative in nature have been suggestions by some investigators that halogen species might play a significant role in controlling tropospheric levels of 03 [Chameides and Davis, 1980;Singh and Kasting, 1988;Barrie et al., 1988;Bottenheim et al., 1990; Cha•field and Crutzen, 1990; Keene et al., 1990;Jenkin, 1993;Jobson et al., 1994 If the latter reaction takes place, it also represents a photochemical loss of 03. An additional minor sink for 03 involves its reaction with OH, (R10) OH + 0 3 •> HO 2 + 0 2 Based on processes (R1)-(Ri0), the rate-limiting reactions controlling ozone can be expressed in terms of the difference between 03 formation and destruction, e.g., 03 photochemical tendency, P(O3). Accordingly, P(O3) can be represented by equation (1) halogens bromine and chlorine the most likely impact would be found in the marine BL. However, even in this environment, evaluation of the source strengths for the reactive forms of these halogens was viewed as being highly speculative in that heterogeneous sources had to be invoked. By comparison, the evaluations for iodine were far more quantitative since the major source for reactive iodine involved the gas phase photolysis of iodocarbons. A summary of these results has been presented below in the text under the "Discussion" section.

Model Description
Two types of box models were used in this study: instantaneous photostationary state (PSS) and time-dependent (TD). Both have been previously described by Chameides et al. [1987Chameides et al. [ , 1989 and Davis et al. [1993]. Changes to these earlier models, as reflected in this study, have been addressed in a companion paper by Crawford  In the instantaneous photostationary state model, all photochemical species are assumed to be in photochemical equilibrium. The use of this model in this study was for purposes of evaluating photochemical products for individual data points. Even so, it is important to recognize that the actual spatial scale encompassed by each data point was typically quite large. It is defined by the product of the aircraft speed and the time resolution employed for the input dam to the model. As noted earlier, the time resolution selected for this analysis was 3 min. Hence, each independent modeling result, reflecting one input dam point, is the average result over a sampling range of nearly 37 km.
Input to the PSS model consisted of fixed values for the mixing ratios of the chemical observables 03, NO, H20, CO, NMHC and the physical parameters temperature, pressure, and UV solar flux. For H2 and ca4, global average mixing ratios were taken, e.g., 0.55 and 1.7 parts per million by volume(ppmv), respectively. As noted above, the concentrations of the short-lived species HO2, OH, O(1D), CH30 , CH302, RO2, NO2, and NO 3 as well as the chemically related species CH20, HO2NO2, and N205 were evaluated by setting the rate of production for each species equal to the rate of destruction. Intermediate-lived species such as HNO3, H202, and CH3OOH were also assumed to reach steady state (at high-altitudes this list would also include N205 and HO2NO2). For each of the latter species, in addition to the known gas phase reactions, a heterogeneous loss was assumed that was similar to that described by Logan et al. [1981]. This computational format has been labeled here as our "unconstrained" or "standard" model run. A second PSS format, involving the use of measured concentrations for the species NO2, H202, CH3OOH , HNO 3 and peroxy-acetyl nitrate (PAN) has been labeled "constrained"; and still a third PSS model version, involving NMHC levels being set equal to zero, is labeled here "No NMHC." Time-dependent box model runs were used to generate diurnal profiles for both radical species and the modeling products F(O3) , D(O3) , and P(O3). These runs were also used to evaluate quasi steady state levels of HNO3 and NOx based on the photochemical recycling of NOy. For this study, Time Dependent Model (TDM)values have been reported for 14 different chemical environments (dam bins) as defined by seven altitude ranges and two latitudinal regimes. Input dam for these runs consisted of median values for CO, H20, 03, NMHCs, T, and P. Typically, between 10 and 140 dam points were available to define these median values.
A critical piece of input to the TDM was the NO mixing ratio. For this study this quantity was established by invoking a constant daytime only NO source flux. This flux was then adjusted so as to give back an NO value corresponding to the median level estimated for a specific altitude/latitude bin. Daytime NO median levels were estimated using dam that were filtered for solar zenith angles of 0 ø to 60 ø. For internal consistency, all other input dam used in our TDM runs were also filtered with a 0 ø to 60 ø zenith angle restriction. Depending on ambient conditions, TDM runs typically required 20 to 100 days to reach a quasi steady state solution.
A We note that for the species H202 and CH3OOH, only free-tropospheric altitudes were considered due to the enhanced uncertainty associated with cloud washout, precipitation removal, and surface deposition at BL altitudes. When comparing H202 levels at free-tropospheric altitudes, the agreement between model predictions and observations is seen as quite good. For example, the median value for the ratio (H202)mc/(H202)expt is 0.92, and the Pearson "r" correlation coefficient is estimated at 0.74. For CH3OOH the agreement is not quite so good, the model-calculated values being nearly a factor of 1.6 lower than those ob-served; however, the "r" value is still found to be quite significant, i.e., 0.81. Equally important, the level of disagreement lies well within the combined uncertainties (e.g., factor of 2.5) of two of the critical rate coefficients that control the CH3OOH species, i.e. HO 2 4-CH302 and CH3OOH + OH. Other possible sources of systematic error could include (1) the accuracy of the CH3OOH measurements themselves or ( At BL altitudes, where levels of NO were < 10 pptv, this is observed to result in significant net 03 destruction, e.g., -6 to -9x105 molecules/cm3/s. At higher altitudes, where the NO mixing ratio is seen approaching 90 pptv, P(O3) values are observed as high as +6x105 molecules/cm3/s and remain generally positive over the entire leg during daylight hours. Flight 12. Flight 12 originated out of Hong Kong and was designed to look at high-altitude outflow from the Asian continent. The sampling profile was a standard "wall" profile (for derailed description see Hoell et al. [this issue]), and was configured geographically to be approximately 30 km to the east of the island of Taiwan. The setting for flight 12 was quite different from flight 10 in that the highaltitude (z > 6-km) air trajectories indicate that the air mass sampled was less than one day removed from the Asian mainland. On the other hand, the <6-km trajectories generally show a picture not that different from flight 10, namely, that the air mass sampled was marine in origin. As shown in Figure 5b, this assessment is in reasonable agreement with the chemical observations in that low to moderate levels are seen for all three tracer species. At higher altitudes the chemical composition is seen as being far more variable. Like flight 10, the mixing ratios for all three tracers are found to be significantly greater at high-altitudes. For example, during the 10-km sampling run, the mixing ratios for NO, CO, and C3H8 are seen reaching 225 pptv, 175 ppbv, and 175 pptv, respectively. For NO and C3H8 these values are 5 times greater than those observed at 10w-altitude. Although the correlation between the individual tracers is qualitatively not so high as in flight 10, it is still quite significant.
Satellite imagery showed that tropical storm Nat continued to move slowly northeastward and thus had a significant influence on the outflow sampled during flight 12. Embedded thunderstorms were evident along significant stretches of the east China coast. In fact, quite similar to flight 10 approximately 100 lightning signatures were recorded during flight 12. Collectively, the available evidence suggests that during flight 12 the outflow from mainland China was strongly influenced by deep convection. It also appears that this deep convection had a major impact on the concentration profiles of the tracer species NO, CO, and C3H8.
From Figure 5b it can be seen that for both high and lowaltitudes, F(O3) and D(O3) display a wide range of values. Reflecting the elevated levels of NO observed at highaltitude, P(O3) values are seen as being predominantly positive, with some values ranging up to +lx106 molecules/cm3/s. For altitudes < 6 km, involving background marine air, P(Os) ranges between zero and -lx106 molecules/cm3/s. The very low-altitude positive values of P(O3), observed for GMT times of 7.25 and higher, reflect lowaltitude pollution from the southern tip of Taiwan. These observations were made on the return trip to Hong Kong.
Flight 15. Flight 15 is representative of the WTNP sampling regime in that 03 and its photochemical precursor species were typically quite low. This flight was staged out of Guam and involved an "extended wall profile." The flight was designed to sample as far south as the equator following a southeast vector out from Guam. The 10-day back trajectories suggest that at all altitudes the air mass sampled originated over the ocean and had no significant interaction with any landmass. At altitudes < 6 km the chemical data appears to support this picture of a clean marine environment. Typical levels of NO, CO, and C3H 8 are seen from Figure 5c to be 10 pptv, 70 ppbv, and 15 pptv, respectively. However, at altitudes >8 km the chemical tracer data suggest otherwise. For the 8.2 and 11 km data runs, Figure 5c shows that for three different time periods enhanced levels of NO appear. These are seen to be qualitatively correlated with elevated levels of propane and in two cases with elevated levels of CO. Like flights 10 and 12, high-altitude levels of NO and C3H 8 are seen exceeding those at low-altitude by factors in excess of 3. For CO the high-altitude enhancement was closer to a factor of 1.5. As was the case in flight 10, during the high-altitude sampling mn at 11 km, the 95 pptv NO peak was recorded under dusk conditions. Satellite imagery showed that cirrus outflow from Typhoon Pat, located to the north, had reached the general vicinity of Guam. In addition, scattered areas of deep convection could be identified to the southeast of Guam.
Although no storm scope data were available for flight 15, the chemical and satellite data again seem to suggest that deep convection was a major factor in defining the air composition at high-altitudes. Since marine BL levels of NO, CO, and C3H 8 were typically observed to be quite low, it also suggests that this convection occurred either over an island or some other major landmass. Similarly, some highaltitude segments of tropical flights 14 and 16 also showed trace gas distributions which were consistent with deep convection over a landmass of some type.
The 03 photochemical profile for flight 15 is one that strongly reflects variations in the NOx environment. For example, under daylight conditions, Figure 5c shows that at low-altitudes (z < 5 km) the NO mixing ratio was typically < 10 pptv; as a result the net effect of photochemistry on 03 is one that shows net destruction. Like flights 10 and 12 therefore the low-altitude P purposes of this data analysis we have further filtered our original data runs so as to include only zenith angles between 30 ø and 60 ø . The net effect of this filtering has been to reduce the total number of usable data points from 1085 to 760. However, under this more restricted Z-angle range, the corresponding range for the calculated parameters D(O3) and F(O3) (for similar chemical conditions) has been reduced to factors of 1.5 to 2.0. This filtered data set is shown here in Figures 6a-6c. In this case the data have been "binned" according to altitude and latitude dimensions of 1 km and 9 ø, respectively. In these plots, only bins having three or more data points are shown; and of those bins assigned values, 70% had nine or more data points. Even so, it is important to keep in mind that the data collection period that formed the basis for Figures 6a-6c, as well as the subsequent three dimensional plots, 7a-7h, represent but a chemical snap shot of the western Pacific. In some cases these plots also average away very modest longitudinal gradients. Thus the representativeness of these plots must be viewed with some element of caution.
From an analysis of Figure 6c, two general trends in P(O3) emerge: (1) at altitudes below 6 km, values are consistently negative with the larger of these occurring within the first 3 to 4 km of the surface. (2) Above 7 km P(O3) values tend to be positive, but the magnitude is found to be strongly dependent on latitude. For example, values in the WNPR region tend to be 2 to 4 times higher than those in the tropics (see also Tables 2a and 2b).    process (e.g., 55 to 85 5[), with smaller contributions coming from (R6). At all altitudes and all latitudes the contribution from NMHCs, in the form of (R7), is seen as being no more than 11%. Thus this evaluation further confirms our earlier conclusion that overall NMHC emissions were of minor importance as ozone precursors during the sampling time period of PEM-West A. Given that both HO2 and CH302 radical levels decrease systematically with altitude (e.g., Figures 7d and 7e) and that F(O3) displays only a weak negative trend with altitude (e.g., Figure 6a), NO and peroxy radicals emerge as the major chemical factors controlling the trend in F(O3). Of course, peroxy radicals are themselves chemically coupled to H20. This point is illustrated quite nicely in the form of Figures 7a, 7d,  Although the latter trend is present at all latitudes, the WNPR region clearly shows the largest percent increase.
As in the case of D(O3), we have also found that diel values of F(O3) can be related to measured chemical parameters and the estimated quantity, NOx, by means of equation  Chameides et al., 1987Chameides et al., , 1989]. Unique to this study are the specific results for the western North Pacific, a region heretofore uninvestigated.

P(O3) and Air Mass Type
As discussed in the previous text, NO and H20 are two of the key chemical parameters that appear to most influence the trends in D(O3) and F(O3). For example, at altitudes < 6 km, high levels of H20 and correspondingly low levels of NO lead to values of D(O3) that typically exceed F(O3) , Here we examine the trend in P(O3) from a more synoptic perspective. P(O3) was selected for this exercise since it is this quantity that is most sensitive to changes in environmental conditions.
The air classification scheme used in this analysis (see Table 3) is similar to that described by Gregory  Thus the scheme shown in Table 3 reflects our conclusion that deep convection was a frequent source of elevated levels of NOx for altitudes above 6 km. As noted earlier, these elevated levels of NOx were strongly correlated with our calculated values of P(O3).
To identify deep convection events, we took an approach very similar to that described earlier in our analysis of flights 10, 12, and 15. However, in addition to the meteorological and chemical data presented in that analysis, we also found it useful to include the chemical tracers: 03, H20, DMS, SO2, C2H2, C2H 6, CH4, PAN, CH3I, (NOx)mc , and NOy. Again, as seen from Table 3, this has led to our identifying a total of 11 air mass classes. Each of these classifications is based on our having identified significant data segments from two or more flights, each showing similar characteristics.    Figure 9. For purposes of comparing with P(O3), we have grouped Smyth et al's. results into six bins. These bins have been defined by first separating their results into lower free-tropospheric and upper free-tropospheric groupings. Each of these groups, in turn, was subdivided into three subgroups based on the magnitude of the C2H2/CO ratio. These subgroups are identified in Figure  9 by their median values of the ratio C2H2/CO. They define air mass types that range from well-processed air (i.e., near background) to relatively fresh anthropogenic emissions. The corresponding values of P(O3) are shown as ranging from -2x1½ to +3.5x10 s molecules/cm3/s. As seen from Figure 9, the observed trend appears to parallel that predicted from the other air mass classification schemes. For example, upper tropospheric values of the ratio are seen as having positive values of P(O3); whereas lower free-tropospheric ratios correlate with negative P(O3) values. Within an altitude grouping, one might also expect that the higher the value of the ratio the higher the value of P(O3). This, in fact, is what is observed for the high-altitude block. On the other hand, for the lower free-tropospheric grouping the results appear to be somewhat anomalous. In this case the highest ratio does not correlate with the highest P(O3) value; however, the separation between P(O3) values for the three differera ratios is also seen to be quite small and probably lies well within the uncertainties of the measurements and/or modeling calculations.

Critical-NO and NOx. That level of NO at which photochemical production of ozone from processes (R5), (R6), and (R7) is exactly balanced by photochemical ozone destruction due to processes (R4), (R9), and (R10) defines the "critical" NO level. Thus it represents the NO level
where P(O3) changes sign.

In the current analysis, values of NO(crtt) were estimated based on median values for all model input parameters.
These evaluations were carried out for 13 altitude/latitude data bins as shown in Table 4. The results show NO(crit) values ranging from 6 to 17 pptv, with the median value being close to 11 pptv. No simple trend appears to be present in these results except perhaps in the case of the WNPR region. Here it is seen that for altitudes above ! km there is a trend of decreasing NO(crit) values with increasing altitude.
On average, ambient NO levels for altitudes < 4 km are found to be 1.5 to 3 times lower than NO(crit); whereas for altitudes above 8 km they tend to be !.5 to 4 times higher.  Table 4 shows that for both latitude ranges, NOx(crit ) generally decreases with increasing altitude. The stronger trend in NOx(crit ) most likely reflects this parameter being a more conserved quantity than NO. As noted earlier in the text, the partitioning of NOx between NO and NO: is a strong function of altitude.

In contrast to NO(crit) values, with the exception of one or two low-altitude points,
From Table 4 the values of NOx(crit ) can be seen ranging from 7 to 50 pptv, giving a median value of 20 pptv. In the only other study in the Pacific for which NOx(crit ) values were evaluated (e.g., [Liu et al., 1992], levels of 55 pptv were reported for an altitude of 2-3 km. At a similar altitude, the PEM-West A results suggest NOx(crit ) levels of between 25 and 39 pptv.
NOx lifetime. Lifetime estimates for NOx as a function of altitude and latitude are presented in Table 4. For these evaluations the daytime gas phase reaction of NO2 with OH was taken as the only major loss process for N Ox. The results indicate that lifetimes for NOx can range from ! to 1.5 days for low-altitudes and 3 to 9 days at high-altitudes (i.e., 8-i2 km). By contrast, high-altitude lifetime estimates for NO2 fall into the range of 1 to 2 days. The difference between these two estimates reflects the large shift in the partitioning of NOx toward NO at high-altitudes. This shift is due both to the much slower rate of conversion of NO to NO:, via its reaction with 03, and to a somewhat enhanced NO: photolysis rate at high-altitudes.
The extended lifetime for NOx at high-altitudes suggests that deep convection events 3000 to 4000 km inland from the coast could still represent potentially important sources of NOx for the western Pacific. Similarly, taking the nominal high-altitude fall-season winds observed during PEM-West A • representative (i.e., 55 km/h), within 9 days significant amounts of NOx, generated from deep convection near the Asian coast, could virtually cross the entire Pacific. Although large mixing factors would be involved, considering the absence of other major primary NOx sources over the open ocean, deep convection along the Asian coast could prove to be one of the more important sources of NOx to the North Pacific. In fact, evidence supporting the importance of long-range transport of highaltitude NOx can be found in the NO data recorded in the tropics (see, for example, earlier discussion under "Ozone Photochemical Trends: Flight Track Analysis").
NOx-O3 chain length and unit production rate. The NOx-O3 chain length may be defined as the number of 03 molecules produced photochemically per NOx molecule oxidized. Liu et al. [1987] were the first investigators to discuss this quantity and, subsequently, labeled it "ozone production efficiency" [Fehsenfeld and Liu, 1993]. We will here refer to this quantity as the NOx-O3 chain length. Under quasi steady state conditions it can be estimated from equation (6):

kll[OH][NO 2]
The numerator in tlfis equation consists of what earlier was defined as F(O3), and the denominator represents the loss of NOx due to the daytime reaction of NO 2 with OH, i.e., (R15):

(R!5) NO 2 + OH + M > HNO3
Although other possible loss charonels could be considered (i.e., NO3/N2Os/aerosot nighttime chemistry), because of a lack of quantitative details concerning these loss processes [Wayne et al., 1991], we have restricted NOx losses to reaction (R15) only. (Estimates of the possible error associated with this approximation indicate that it is probably less than a factor of 2 for PEM-West A conditions.) The "primary" chain length as defined by equation (6) is important in that given adequate time, it defines the amount of 03 that could be photochemically produced from each NOx molecule released into the troposphere before being converted into a more stable NOy form. In the lower troposphere, because of the short lifetime of both NOx and the product HNO 3, the primary chain length can be used in combination with known NOx emission rates to directly estimate 03 production [Liu et al., 1987;McKeen et al., 1991]. However, in the upper free-troposphere, because of a much extended lifetime for HNO 3, this species may be recycled back to NOx to initiate new 03 production. Thus as reported by Liu et al. [1995], the "ensemble chain length" for 03 production is defined in terms of the product of the primary chain length and the total number of times a given NOx molecule cycles through HNO 3 or some other NOy species (for additional detail on this point see discussion under "NOx Sources"). This means that to quantitatively evaluate the contribution of a given NOx source to photochemical 03 formation, both the primary chain length and the recycling efficiency must be known. As discussed below, the magnitude of both of these terms can be a function of altitude as well as other factors such as the NOx concentration itself. Although the primary chain length may give an indication of 03 formation from one NOx molecule before that molecule is oxidized, it is also of some value to examine the near term, diet, 03 production efficiency from NOx. We will here refer to this quantity as the 03 production per unit NOx per day and give it the symbol AP, as per Liu et al. [1987]. Thus AP is simply CL/rNox. Liu et at. have reported modeling results for summer surface conditions which suggest that AP should be nearly independent of NOx at levels less than 1 ppbv when all other parameters are held constant. As shown in Table 4, we report AP results for the PEM-West A data for surface conditions as well as for the middle and upper free-troposphere. The 0 to 12 km results show that for the WTNP and the WNPR regions, AP decreases with altitude by factors of 2 to 3. Furthermore, this decrease with altitude appears to be mostly driven by an increase in the NOx lifetime which, in turn, reflects changes in the partitioning of NOx as well as changes in HOx levels. For example, sensitivity tests at 8-10 km involving variations in NOx of plus and minus a factor of 3 show that the average change in AP was only a factor of 1.4.

The effect of the NOx concentration on chain length becomes apparent once it is recognized that elevated NOx not only promotes reactions (R5), (Rt), and (R7) but also leads to a reduction in
Thus based on actual high-altitude observations, the PEM-West A results suggest that for the remote troposphere, given adequate time, one gets only st.ightty more mileage from a given NOx molecule at high-altitude than for altitudes near the surface. Furthermore, because of the long reaction times involved at high-altitude, transport processes become quite important in defining the influence of any newly formed 03 on actual observed 03 levels. By contrast, in the remote troposphere diet 03 production per unit NOx is significantly higher at low-altitudes. However, insofar as F(O3) is concerned, the lower value of AP at high-altitudes is typically compensated by much higher levels of NOx and by the strong partitioning of NOx in favor of NO.

NO,, Sources
The results from this study as well as those from several earlier studies [e.g., Liu et al. 1980;Chameides et al. 1987Chameides et al. , 1989and Ridley et al. [1987] have shown that NO is the rate-limiting precursor in the formation of 03. The issue of a) 0-18 N Lat. ]. Thus the primary focus of this text will be to compare and summarize some of the more important findings of these efforts. Figures 10a and 10b show altitude versus (NOx)mc/C3H8 plots where the data have been binned to help visualize the traderlying trends. In this cas. e, the results show quite clearly that for both the WNPR and the WTNP regimes, median values of the ratio (NOx)mc/C3H8 increase with increasing altitude by nearly a factor of 2. Since the major source of propane is combustion and natural gas emissions [Blake et al., , 1994, this issue] and the lifetime of propane is 3-5 times longer than NOx, such profiles would seem to argue in favor of the major source of NOx being located at high-altitudes. Similarly, when the ratio (NOx)mc/C3H8 is plotted against a second ratio, C2H2/CO, as shown in Figures 1 l a and l

Diurnal Photochemical Trends and Ozone Budgets
The evaluation of the O3 budget was carried out using diurnal profiles for D(O3) and F(O3), as shown in Figures  12a and 12b. Table 5 gives the diurnal average rates for D(O3) and F(O3) as estimated from these profiles. The diurnal profiles for P(O3) are shown in Figures 13a and 13b; the average values of P(O3) based on these profiles are again given in Table 5. Recall that it is the diurnal average value of P(O3) that most directly relates to the impact of photochemistry on ambient levels of 03. From Table 5 we see that in all cases but one the daily change in 03, either positive or negative, is smaller than 2 ppbv. Thus changes in local ambient ozone levels, as driven by photochemical process, are predicted to have been very gradual.
For  Figures 14a-14c show the values of D(O3), F(O3), and P(O3) displayed as diurnal-averaged column-integrated quantities. These results clearly identify the free-troposphere as being by far the largest contributor to the North Pacific ozone budget. For example, it can be estimated that the column-integrated 03 formation and destruction fluxes for the free-troposphere (i.e., 1-12 km) are 5 to 8 times larger than for the marine BL ( 0-1 km). Thus the PEM-West A data support the earlier conclusions of Fehsenfeld and Liu [1993] that photochemical activity within the BL has a minimal impact on the global budget of 03. Only in the event that significant BL sources of NOx were present and this NOx was convectively transported to high-altitudes would the importance of BL NOx levels to the 03 budget be enhanced [e.g., Cha•eld and Delany, 1990; Pickering et al., 1990;Fehsenfeld and Liu, 1993].
From Figure 14b we also see that for the WNPR region the 03 column production flux (i.e., 31x10 •ø  The near balance between column photochemical production and destruction (i.e., Figure 14b)  Since the column average lifetime of 03 for the tropical regime is estimated to be even shorter (i.e., 16 days) than that for the WNPR region and the seasonal changes are also smaller, one would expect that steady state conditions would also prevail in the former regime. But this requires that the column-integrated budget for 03 be approximately balanced. Figure 14a, however, shows that the 0 to 10 km column destruction rate exceeds production by about 80%, leaving a significant 03 deficit of nearly 12x10 •ø molecules/cm2/s. At these low-latitudes it is highly unlikely that any direct influx of 03 from the stratosphere could balance this deficit. It is possible, though, that an influx of O3-rich midlatitude tropospheric air could compensate for this deficit.
An alternative hypothesis is that the deficit cotfid be balanced by additional 03 photochemical production at altitudes between 10 km and the tropopau,se. To explore this possibility, model simulations were carried out in which increasing levels of NO were added to altitudes ranging from 10 km to the tropopause (i.e., 17 km). The results show that increases in 03 formation start leveling off as NO mixing ratios approach 150 pptv. Based on the "flight track analysis" for flight 15 presemed earlier in the text such elevated levels of NO do not appear to be unrealistic. In fact, it is quite pausible that relatively high NO mixing ratios  preparation, 1995], the tropical 03 column deficit is reduced to -4x10 •ø molecules/cm2/s, or 13% of the total column destruction rate. This is well within the uncertainties of the calculations and may therefore represent a viable explanation. The fact that a deficit still remains, however, would seem to argue against there being yet another major 03 destruction mechanism operating in the troposphere. Recall that earlier in the text the possibility that tropospheric iodine might catalytically destroy 03 was briefly discussed. Using the PEM-West A CH3I data as one indicator of the marine iodine source strength, Davis et al.
[this issue] have further explored the iodine-O3 question in terms of three possible iodocarbon source scenarios. Based on a simple one-dimensional model that incorporated a firstorder diffusion process to describe vertical transport, this analysis resulted in three different Ix (Ix = I + IO + HOI + HI +21202 q-INOx) altitudinal distributions. For the upper troposphere the Ix levels were 0.5, 1.5, and 7 pptv. Box modeling runs based on these Ix levels indicated that only at the 1.5 and 7 pptv level was there significant 03 destruction. Changes in total column-integrated 03 destruction were 1.8x10 •ø and 9x10 •ø molecules/cm2/s, respectively. The Ix value corresponding to the higher destruction was estimated based on a marine iodine source region having high biological productivity. However, during PEM-West A all sampling in the tropics occurred over low productivity marine water. Thus Davis et al. concluded that the impact of iodine on the column-integrated destruction of 0 3 in the tropics was probably no greater than 6 %.
A comparison of 03 budget estimates from this study with model estimates by Fehsenfeld and Liu [1993], shows reasonable agreement between the two studies. For example, using Fehsenfeld and Liu's altitude criteria for the freetroposphere, 1-12 lcm, we estimate an average columnintegrated 03 formation of 24x10 •ø molecules/cm2/s. (In this evaluation an assumed value for F(O3) of lx10 •ø molecules/cm2/s was used for the altitude interval of 10-12 km in the tropics). When converted to similar units, Fehsenfeld and Liu's results would suggest an average value of 30x10 •ø molecules/cm2/s, i.e., a factor of 1.25 times higher than this study. We note that these authors, with the exception of an observational altitude profile for the NO mixing ratio, based their results on generic global data.

Summary and Conclusions
When examined in terms of photochemical ozone processes, the PEM-West A data set was best described in terms of two geographical domains: the western North Pacific rim (WNPR) and the western tropical North Pacific (WTNP). The first region is one that was influenced by both natural and anthropogenic continental sources. Highaltitude outflow from the Asian continent as well as from other northern hemispheric continents appear to have been involved. By contrast, the tropical regime, for altitudes less than 10 km, can be viewed as a region whose chemical fingerprint reflected either aged/well-processed continental air or air masses that had their origin in the tropical/equatorial Pacific.
In all cases the photochemical destruction of ozone, D(O3), was found to decrease more rapidly with altitude than photochemical formation, F(O3). Thus the ozone tendency, P(O3), typically was negative at low-altitudes (e.g., < 6 km) but positive for altitudes > 6 to 8 km. The most important chemical factor controlling the altitude trend in D(O3) was the H20 mixing ratio. In most cases, (R4) was the dominant 03 loss reaction, although at the highest altitudes the contribution from (R9) and (R10) increased significantly, with (R9) in some cases becoming dominant. The trend in F(O3) with altitude showed very modest decreases, reflecting the fact that decreases in HOx radical levels with altitude were substantially offset by increases in the mixing ratio of NO. For altitudes < 4 lcm the two most important ozone formation processes were identified as (R5) and (R6); whereas for altitudes > 4 lcm (R5) was the dominant process (i.e., 55-85 %). At all altitudes and all latitudes the contribution from (R7) was 11% or less. This observation indicates that NMHC emissions were typically of minor importance as ozone precursor species during the time period of PEM-West A.
A synoptic analysis of the PEM-West A database by several different investigating groups resulted in five different air mass classification schemes.
These were examined here in terms of their respective values of P(O3). The general trend that emerged showed that the largest positive values occurred for BL air, within 2 days of mainland Asia or Japan and for high-altitude air parcels (e.g., > 7 lcm) influenced by deep convection/lighming. Significant negative values of P(O3) were found when encountering clean marine BL air or relatively clean lower free-tropospheric air parcels.
When median values of the ratio NOx/C3H 8 were plotted against altitude or the ratio, C2H2/CO, the resulting profiles were found to be most consistent with the major sources of NO• being located in the upper troposphere. Results generated by other PEM-West A investigators as well as further work by these authors suggest that one of the major contributors to this high-altitude NO• pool was deep convection, especially that associated with lighming. Other contributing high-altitude primary NO• sources appear to include aircraft emissions and stratospheric intrusions. Much more difficult to assess was the degree to which recycled NOy contributed to the observed levels of NOx. Potential uncertainties in measured NOy and HNO 3 as well as possible incompleteness in the model chemistry represent the primary reasons for a lack of a more definitive statement on this important topic. In spite of these shortcomings, it still may be argued that the overall importance of recycled NOx was higher in the western tropical Pacific than for the Pacific rim region.
Diurnal-averaged column-integrated photochemical formation and destruction fluxes for the WNPR region were shown to exceed those for NH dry deposition and NH stratospheric injection by factors of nearly 6. For this same region a near balance was found between photochemical 03 production and destruction, suggesting that this region was near steady state. Ozone colunto lifetime arguments, together with small seasonal changes in total column 03, suggest that the tropical regime should also have been near steady state. In fact, the column-integrated fluxes show that photochemical destruction exceeded production by nearly 80%. Two hypotheses were put forward in an effort to explain this deficit. The first involved the possibility that O3-rich air could have been transported from midlatitudes into the tropics; the second proposed that the unsampled atmospheric column from 10 to 17 lcm might have provided the additionally needed photochemical 03. The latter hypothesis requires relatively high levels of NO (e.g., 150 pptv); however, these do not appear to be totally out of line with those estimated from tropical lighming. In this context, results from the present study indicate that NOx would have an extended lifetime at altitudes of 8-12 km of 3 to 9 days and even longer for still higher altitudes. This suggests that for some seasons of the year, deep convection over regions of Asia and Malaysia/Indonesia could lead to significant enhancements in high-altitude 03 formation that might extend well out into the North Pacific. When coupled with very strong high-altitude westerly flow, its influence could potentially span a large part of the Pacific.