Photochemical ozone budget during the BIBLE A and B campaigns

[ 1 ] Using the measured concentrations of NO, O 3 , H 2 O, CO, CH 4 , and NMHCs along the flight tracks, a photochemical box model is used to calculate the concentrations of the Ox radicals, the HOx radicals, and the nitrogen species at the sampling points. The calculations make use of the measurements from radiometers to scale clear sky photolysis rates to account for cloud cover and ground albedo at the sampling time/point. The concentrations of the nitrogen species in each of the sampled air parcels are computed assuming they are in instantaneous equilibrium with the measured NO and O 3 . The diurnally varying species concentrations are next calculated using the box model and used to estimate the diurnally averaged production and removal rates of ozone for the sampled air parcels. Clear sky photolysis rates are used in the diurnal calculations. The campaign also provided measured concentration of NO y . The observed NO/NO y ratio is usually larger than the model calculated equilibrium value. There are several possible explanations. It could be a result of recent injection of NO into the air parcel, recent removal of HNO 3 from the parcel, recent rapid transport of an air parcel from another location, or a combination of all processes. Our analyses suggest that the local production rate of O 3 can be used as another indicator of recent NO injection. However, more direct studies using air trajectory analyses and other collaborative evidences are needed to ascertain the roles played by individual process.


Introduction
[2] The two phases of the BIBLE (Biomass Burning and Lightning Experiment) campaign carried out during September -October of 1998 (BIBLE A) and August -September of 1999 (BIBLE B) were designed to study tropospheric ozone chemistry in tropical Asia [see Kondo et al., 2002aKondo et al., , 2002b. The instruments provide measured concentrations of NO, NO y , O 3 , H 2 O, CO, CH 4 , NMHCs, several methyl halides and alkyl nitrates, as well as radiometer measurements that provide information on in situ photolysis rates. These data have been compiled by the BIBLE Science team into 1-min merged files for each flight. The merged files provide concentrations of trace gases at the sampling point along the flight track. They provide sufficient constraints (with some additional assumptions) to allow a photochemical model to calculate concentrations of radical species. The calculated radical concentrations can then be used to compute the instantaneous production and removal rates of ozone.
[3] The measured concentrations do not provide sufficient constraints to give unique values for the ozone production and removal rates. Different assumptions would lead to different values [Davis et al., 1996;Jacob et al., 1996;Folkins et al., 1997;Jaegle et al., 1998]. One aim of the campaign is to examine the role lightning and/or biomass burning play in the reactive nitrogen budget [Price et al., 1997;Liu et al., 1999;Levy et al., 1999;Galanter et al., 2000] and how that affects the ozone budget. In this paper, we compare the model calculated [NO]/[NO y ] ratio with the observed [NO]/[NO y ] ratio, and use this along with the model calculated ozone production rate to obtain an indicator for the time elapsed since previous perturbation to the nitrogen species. (We will use [X] to denote the concentration of X.) This is the reason why we chose to use measured [NO] in our calculations and compare the model calculated [NO y ] with the observed values of [NO y ]. Section 2 discusses the overall philosophy of our approach. The model descriptions, the calculation procedures and model results are given in sections 3, 4 and 5.

Process Model and Aircraft Data
[4] The interactions between model and measurements take on many levels of complexity. In theory, if one has the correct chemical mechanisms and transport rates built into a model, one would only need detailed emissions histories of the source gases to simulate all the trace species for comparison with measured values. This point-by-point validation cannot be achieved in practice for many reasons. There is no independent way of verifying that the adopted photochemical scheme is complete. One cannot be sure that the emission histories and transport rates are sufficiently accurate to account for small local variations. Thus, when discrepancies occur, it is seldom possible to attribute them to a specific cause.
[5] A process model assigns values to concentrations of certain species and calculates the concentrations for a subset of the species to test specific processes. The most common of these is the photochemical box model. The approach takes advantage of the fact that certain species have photochemical lifetimes on the order of minutes and shorter. Thus, if the concentrations of all other species are given, the concentrations of the short-lived species are determined by the local solar insolation at the time of the measurement. In practice, the situation is less than ideal because concentrations of some species with intermediate lifetimes (of order days) are not measured.

Ozone Production and Removal Rates
[6] Current understanding of ozone chemistry shows that the production and removal rates of ozone in the troposphere are dominated by HO x and NO x chemistry [see, e.g., Davis et al., 1996;Crawford et al., 1997aCrawford et al., , 1997bKlonecki and Levy, 1997]:

)
In order to calculate the instantaneous production and removal rates of ozone, we must obtain values for  Wennberg et al. [1994] and Fahey et al. [2000] for the stratosphere and Jacob et al. [1996], Folkins et al. [1997], Crawford et al. [1997aCrawford et al. [ , 1997b, Jaegle et al. [1998], andSchultz et al. [1999] for the troposphere. In the BIBLE campaign, the concentration of NO was available from the NO instrument [Kondo et al., 2002a[Kondo et al., , 2002b [Kotamarthi et al., 1997;Crawford et al., 1999;Jaegle et al., 2001]. The ATHOS instrument [Brune et al., 1998] aboard the DC-8 during the SUCCESS campaign provided in situ airborne measured concentrations of OH and HO 2 in the troposphere. The instrument was also deployed during the PEM Tropics-B campaign in 1999 [Brune et al., 1999] [8] Because the radical concentrations in Equation (1) vary with solar zenith angle (SZA), the instantaneous production and loss rates for ozone are also a strong function of SZA. The measurements along the flight tracks were made at various SZAs. For the purpose of identifying how different compositions in an air parcel may affect production and loss rates, it is more useful to use diurnal averaged rates. Strictly speaking, the diurnal rates associated with the sampled air parcel should be calculated using the history (location and solar exposure) of the air parcel following the back trajectory. In practice, there are certain difficulties associated with this. One of these is the lack of information on cloud cover and/or change in albedo along the trajectory. The other is finding the appropriate concentrations to initialize the concentrations in the air parcel. Our approach (to be discussed in section 4) assumes that the air parcel is in photochemical equilibrium with the measured concentrations of O 3 , H 2 O, CO, CH 4 , NMHCs, and the concentration of NO x constrained by observations. Our analyses show that, once the model adopts the measured [NO], the model calculated [OH], [HO 2 ], and the production and removal rates for ozone are not very sensitive to how the other nitrogen species are calculated.

NO y Partitioning and Recent NO Injection
[9] The measured [NO y ] were not used directly in our calculations. Instead, we used the measured [NO] and calculated the concentrations for HNO 3 and other nitrogen species assuming that all the nitrogen species are in photochemical equilibrium. This is similar to the approach adopted by Kiem et al. [1999]. If there have been recent injections of NO (from BIB 8 -2 a biomass burning source or from lightning) in the air parcel, the nitrogen species would not be in equilibrium. Our purpose is to compare the model calculated [NO y ] with the observed values and to use the results of the comparison as a partial indicator for whether there have been recent injections of NO for that air parcel. Several things complicate the interpretation of the comparison. The larger measured [NO]/ [NO y ] ratio may have been due to recent scavenging of HNO 3 . Furthermore, the troposphere is not static. An air parcel sampled at a particular location may not have been at that location for sufficient time to attain equilibrium corresponding to the local conditions. This argues against using comparison of measured and model calculated [NO y ] as the sole criteria. We propose that the production rate of ozone (in percent per day) would provide additional clues on whether recent injection of NO has occurred.

Model Description
[10] The AER photochemical box model [Kotamarthi et al., 1997] was used to simulate the concentrations of the radical species along the flight tracks. There are 93 species in the box model including the O x , HO x , NO x , Cl Y and Br Y families. It includes explicit degradation schemes for CH 4 and C 2 H 6 and higher-order NMHCs schemes according to McKeen et al. [1991]. The reaction rate constants were updated according to JPL-97 [DeMore et al., 1997].
[11] In the calculations, concentrations of the long-lived species O 3 , H 2 O, CH 4 , CO, and NMHCs were constrained by the 1-min merged files. In addition, the concentration of NO was also taken from the merged file and fixed at the observed value. In light of the importance of acetone in the HOx budget [see, e.g., Singh et al., 1995], we fixed the concentration of acetone using the correlation between acetone and CO derived from the PEM-West B data [McKeen et al., 1997]. We deemed this to be reasonable given that the BIBLE campaign and PEM-West-B were over similar geographical regions. Neither H 2 O 2 nor CH 3 OOH were measured in the campaign. Their concentrations were calculated assuming local photochemical equilibrium. We ignored halogen chemistry in the calculation by setting the concentrations of Cl Y and Br Y equal to zero. We also ignored heterogeneous chemistry. The heterogeneous reaction of N 2 O 5 + H 2 O producing HNO 3 would have changed the nitrogen partitioning with appreciable effects particularly during winter at high latitudes. Because we fixed [NO] in our calculation, including the reaction would change only [HNO 3 ] and other nitrogen species. Analysis of the HO x budget would show that the concentration of HNO 3 has a small effect on the calculated concentrations of OH and HO 2 . The other reactions (such as BrONO 2 + H 2 O) could also affect the HO x budget. Sensitivity calculations performed using typical values of Cl Y and Br Y show that the effect on the ozone production and removal rates is small.
[12] The point model provided {n j t obs }, which is the set of concentrations for species that were obtained by solving the system of instantaneous equilibrium v where n i t obs (molecules/cm À3 ) is the concentration of the ith species, P i (molecules/cm À3 s À1 ) and L i (s À1 ) are the respective production and removal frequency and are functions of other species concentrations and the solar zenith angle at the time of observation (SZA(t obs )). Species calculated this way included O, O( 3 P), O( 1 D), OH, HO 2 , H 2 O 2 , CH 3 O 2 , NO 2 , PAN, HNO 3 , NO 3 , HNO 4 , N 2 O 5 , HONO, CH 3 OOH, C 2 H 5 OOH, CH 3 CHO, CH 3 CO 3 , CH 3 CO 3 H, C 2 H 2 O 2 NO 2 . The point-by-point calculation was performed only for sampling points at solar zenith angles less than 60°s o that the equilibrium criteria are better met.

Scaling of Photolysis Rates
[13] Two photolysis rates (J(NO 2 ! NO) and J(O 3 ! O( 1 D))) were derived from filter radiometer measurements [see Kita et al., 2002]. The absolute calibration factor for J(NO 2 ! NO) was provided by the manufacturer. Kita et al. used the correlation between the measured radiance and the derived photolysis rate from the PEM-Tropics-A campaign to obtain the photolysis rate. Two scaling factors were defined as the ratio of the derived photolysis rate and the model calculated clear sky photol- . These factors were used to scale model calculated clear sky photolysis rates to simulate cloud effect. R O( 1D) was used for absorbers with cross-section peaking near 300 nm while R NO 2 was used for those peaking near 400 nm.  Kondo et al. [2002a].) Note that values at landing usually show anomaly. In addition, the values for the latter third of flight 13 are for SZA close to 60°(see Figure 2). The rest of the behavior is typical although the magnitude of the scaling factor is on the large side compared to other flights in the same campaign. In all cases, the ratio was close to unity around 4 -6 km. Above this altitude, the ratio was usually larger than 1, indicating the effect of reflection from cloud below the flight track. Below 4 km, the ratio was more likely to be less than 1, indicative of clouds overhead.

Constrained Parameters
[14] The different panels from Figure  3.4. Model Calculated n OH t obs and n HO 2 t obs [15] Without direct measurements of OH and HO 2 in the campaigns, there is no easy way to verify whether the assumptions made in the calculations (concentration of acetone; equilibrium assumptions for H 2 O 2 , and CH 3 OOH; photolysis scale factors) are appropriate. The same model was used to analyze data from PEM Tropics-B. With H 2 O 2 and CH 3 OOH fixed at the observed values, the model calculated median concentrations of OH and HO 2 were within 15% of the measurements. The calculated values for OH and HO 2 also accounted for about 70% and 90% of the observed variance respectively. Sensitivity analyses using the PEM Tropics-B data show that the calculated OH and HO 2 would be 20% larger when H 2 O 2 and CH 3 OOH are calculated assuming photochemical equilibrium.
[16] Examination of the budget indicated that the production of HOx is dominated by the reaction of O( 1 D) with H 2 O below 8 km. Above 8 km, photolysis of H 2 O 2 , photolysis of CH 2 O, and the reaction of CH 3 O 2 with NO play comparable roles. Below 5 km, the removal of HO x results from equal contributions from the reaction of OH with CH 4 , the self reaction of HO 2 forming H 2 O 2 , and the reaction of CH 3 O 2 with HO 2 . The first two reactions continue to play an important role above 5 km, while the third becomes less important with the reaction of OH with HO 2 playing an increasing role. The partitioning between OH and HO 2 is controlled by NO, O 3 , and CO. Below 6 km, the local concentration of OH is determined by the balance between production from O( 1 D) + H 2 O, and the removal by reaction with CO. Above 8 km, production of OH is dominated by the reaction of HO 2 with NO while the removal by reaction with CO continues to be key. The production of HO 2 is dominated by the reaction of OH + CO in the troposphere. The removal is dominated by reaction with NO at high altitudes and by the reaction with CH 3 O 2 and the formation of H 2 O 2 at low altitudes. Thus the budget analyses suggest that once the observed NO con-   centration is used in the computation, the results are not very sensitive to how the other NO y species are determined. Clearly, the results will be better constrained if measured values are available for H 2 O 2 and CH 3 O 2 . The calculated values for n OH t obs and n HO 2 t obs along the flight tracks for selected flights are shown in the first row of Figure 3.

Diurnal Averaged Production and Removal
Rates for Ozone

Method
[17] The results from the point model ({n j t obs }) provide the instantaneous equilibrium concentrations of the species at local time at the sampling point. In the real atmosphere, some of the species are not in equilibrium because their lifetimes are of order days. There are several ways one can account for the diurnal variations of those species and obtain diurnal averaged ozone production and removal rates. One way is to use the results from the point model as initial values for the diurnal box model and simulate the diurnal behavior of the radical species assuming the air parcel is stationery at the same spatial position. The results will depend on the treatment of cloud cover. The decision has to be made whether to assume the same cloud cover that occurs at the sampling point persists over several days, or to assume a climatological cloud cover, or to use clear sky photolysis rates. We chose to perform the calculation using clear sky photolysis rates since it is unlikely that the same cloud cover at the sampling time would have persisted over the several previous days.
[18] We used the model to propagate the initial values (n j t obs ) by solving  (3). As a result [NO y ] is preserved, i.e. n NO y = n NO y t obs . Since n i t obs n i t obs are in instantaneous equilibrium at the sampling points with adjusted photolysis rates, n i (t obs ) calculated using clear-sky photolysis rates may no longer equal n i t obs after a 1-day propagation. Thus some adjustment has to be made if we wish to continue to use the measured NO concentration to constrain the model results. We run the model for 20 days so that NO x = NO + NO 2 is in approximate diurnal equilibrium with NO y . It is difficult to obtain exact diurnal equilibrium between NOx and NO y because of a slow conversion between NO x and HNO 3 due to weak feedback on OH. We then scaled the NO y species by the NOx concentrations (t) > n NO y tobs . Finally, we propagate the solution for 1 more diurnal cycle to obtain all the species {n j II (t)} that will be used to calculate the production and removal rates of ozone. Note that the measured J(O 3 ! O( 1 D)) values were not used in the diurnal calculation.
[19] As discussed in section 1, other methods for computing the diurnal behavior are equally valid. An alternate method would be to adjust n NO (t obs ) + n NO 2 (t obs ) after each 24-hour propagation to equal n NO t obs + n NO 2 t obs . An upward (downward) adjustment can be interpreted as an external source (sink) for NO x in the model. One would have to choose whether to make the adjustment at one instance in time or spread the production/removal over a 24-hour period. However, if one runs the model for several days to achieve approximate diurnal equilibrium, the results should be similar to our method as long as the equilibrium NO x /NO y ratio is not very sensitive to NO y concentration.

Results
[20] The calculated values for n OH II (t obs ) and n HO 2 II (t obs ) along the flight tracks are shown in Figure 3 along with n OH tobs and n HO 2 tobs . At most altitudes, the differences between the point model values and the diurnal model values are largely due to the clear-sky (in the diurnal model) versus adjusted photolysis rates (in the point model) in the calculations. The differences above 12 km are particularly large, approaching a factor of 2 (see flights 3 and 13). At those altitudes, the model calculated concentrations of CH 3 O 2 NO 2 are large and the model calculated NO y is much larger than the observed NO y . We will explore this in future studies.
[21] The concentrations {n j II (t)} were used in the expression in equation (1) to calculate the diurnally average production rate (hPi) and removal rate (hLi) for ozone. A sample of the results is shown in the second row in Figure 3. The calculated mean and median values are plotted in Figures 4 and 5 for BIBLE A and BIBLE B, respectively. In each case, we sorted the data that satisfy the solar zenith angle criteria by altitude and computed the mean and median values for each altitude. The standard deviations and the percentiles are given in the corresponding figures as indicators of the spread of the values. The difference between the median and the mean values provide an indication of the distribution. The mean production value is typically 1.5 times the median production value, suggesting that the distribution is skewed toward larger values. In contrast, the mean removal rate is only 10% larger than the median values. This is consistent with the fact that the production term depends more directly on NO concentrations which show large variability. For BIBLE A, the median net value is negative below 7 km. The altitude behavior of the net tendency in BIBLE A is similar to results calculated by Crawford et al. [1997a] for the tropics corresponding to the ''high NOx'' regime that are influenced by continental outflow. The tendency is positive above 6 -8 km and negative below. Similar behavior was calculated using data from the tropical South Atlantic during Trace-A [Jacob et al., 1996], from the tropical North Pacific during PEM-West A [Davis et al., 1996a], from the tropical South Pacific during PEM-Tropics A [Schultz et al., 1999]. The median net ozone tendency is positive below 3 km in BIBLE B because most of the data were obtained over Australia where the effects from bio-mass burning are large [Takegawa et al., 2002].
[22] The integrated column mean and median values (from 1 km to 14 km) are given in Table 1. The data below 1 km is excluded because most of them were taken during take-offs and landings. We divided the data into three sets, representing the Ferry flight north of the equator, flights over Indonesia between the equator and 10S, and flights over Australia south of 10S. Model results based on GCM simulations [Levy et al., 1999] reported that the tendency term due to chemistry in the tropical free troposphere is +163 Tg O3/yr. This is equivalent to +4 Â 10 10 molecules/cm 2 /s. Crawford et al. [1997a] reported values of +0.1 Â 10 10 molecules/cm 2 /s and À20 Â 10 10 molecules/cm 2 /s for the ''high NO x '' and the ''low NO x '' regimes respectively. Schultz et al. [1999] calculated À18.4 Â 10 10 molecules/cm 2 /s using data from PEM-Tropics A for the tropics. Clearly, the calculated integrated tendency is sensitive to the NO x concentrations in the air sampled in each study. The value given by Levy et al. applies to the whole troposphere and includes regions with high NO x and large ozone production. In contrast, the Crawford et al. and Schultz et al. studies used data from remote regions where ozone removal dominates.

Uncertainties Associated With the Method
[23] Crawford et al. [1997b] provided expressions for diurnal averaged ozone production and removal rates.  Crawford et al. [1997aCrawford et al. [ , 1997b chose averaged cloud-correction factors for their photochemical model. The chosen values ranged between 0.8 and 1.0 below 5 km and between 1.0 and 1.18 above 5 km for the extratropics [Crawford et al., 1997b]. In their calculations, all photolysis rates were multiplied by the same cloud correction factors that were specified as functions of altitude. We compared our values for hPi and hLi for BIBLE A with the values calculated using equation (5) in Figure 6. Points for hPi are typically within 50%. The points that differ for more than a factor of 2 are from around 10 km where [NO] (5).
[24] Figures 7 and 8 show the percent contributions of the various terms that make up the production (hPi) and removal (hLi) rates of ozone. With the assumptions we made, the model results show that the reaction HO 2 + NO constitutes 60%-80% of hPi, with CH 3 O 2 + NO contributing the remaining 20% -40%. The next term is from the photolysis of O 2 , which is minimal. Contribution from RO 2 reaction is at most a few percent from a few sampling points with unusually large NMHC concentrations. The hLi term is dominated by the reaction of O( 1 D) + H 2 O in the lower troposphere, responsible for 80%. The reactions of O 3 + HO 2 and O 3 + OH make up the rest. In the upper troposphere, the reactions of the HOx radicals with O 3 add to 80%, with the O( 1 D) reaction contributing 20%.
[25] We now examine how the assumptions adopted in our calculations may affect the calculated production and removal rates for ozone. The first is the assumption that the nitrogen species are in equilibrium with the observed [NO]. The question can be raised how this assumption affects the model calculated [OH] and [HO 2 ], and how they in turn affect the ozone production and removal rates. As pointed out in section 3.5, the model calculated [OH] and [HO 2 ] are not very sensitive to this assumption as long as observed [NO] is used in the calculations. As discussed in section 3.4, we estimated that the [OH] and [HO 2 ] should be within 50% of the actual values.
[26] The second assumption has to do with using clear sky photolysis rates for the diurnal calculations. The information on the derived J(NO 2 ! NO) was used in the NOx scaling in equation (4) at the sampling point. The information on the derived J(O 3 ! O( 1 D)) was not used. Had we used an average cloud correction factor with values between 0.8 and 1.2 similar to Crawford et al. [1997b], the hPi and hLi values would be smaller by about 10%. Finally, we do not have reliable information to estimate the uncertainty in CH 3 O 2 . However, the budget analysis showed that the CH 3 O 2 term should contribute up to 50% (Figure 7). Thus we estimate that hPi should be within a   Figure 3 represented by the thick blue line. Note that [NO y ] meas and [NO y ] equil are in reasonably good agreement (within a factor of 2) except above 12 km and below 2 km. At high altitudes, [NO y ] equil is dominated by n CH 3 O 2 NO 2 II (t obs ). Near the ground n HNO 3 II (t obs ), dominates. The high concentrations above 12 km calculated for CH 3 O 2 NO 2 result from the adopted reaction rate constants. . Mean and median values for ozone production, removal, and net production rates as functions of altitude for BIBLE A. The bar on the mean profile represents the standard deviation, calculated separately for points larger and smaller than the averaged values. The box on the median profile represents points whose values are 25% larger and smaller than the median. The whisker gives the 5% to 95% of the data. Values plotted are average (mean) values for each altitude bin. The two curves in each of the left panels are production (solid) and removal rate (dotted). Note that the points are plotted at slightly different altitudes for clarity.
With recommended temperature dependence of the equilibrium constant for CH 3 O 2 NO 2 , the decomposition rate at 12 km is a factor of 10 smaller than the rate at 11 km. We varied the rate within the uncertainty limit cited in JPL-97 to determine how [CH 3 O 2 NO 2 ] may change. The calculated concentration of CH 3 O 2 NO 2 becomes negligible when the faster rate is adopted. The difference at low altitudes is most likely due to the fact that our method does not account for scavenging of HNO 3 .
[28] If we ignore the results for 12 km, HNO 3 is the dominant contributor to [NO y ] equil with PAN playing a substantial role in certain situations. Also plotted with the bottom panel in Figure 3  ] would signal that the air parcel is not in photochemical equilibrium. This could occur for a number of reasons. First, it could be due to recent injection of NO into the parcel. Alternatively, this could be a result of the recent scavenging of NOy rather than injection of NO. This most likely has the largest impact below 5 km. Finally, a sampled parcel could have been recently transported from another location and thus is still adjusting to the new photochemical environment.   [30] We next discuss how the model calculated production rate for ozone hPi can be used as another indicator for recent injection of ozone precursors. The values for hPi along the flight track are plotted in the second row in Figure 3. The results are plotted in units of percent per day, which is more directly related to the photochemical age of the air parcel as ozone adjusts to injected precursors. Equation (1) shows that an air parcel with recent injection of NO and RO 2 would have a high value for ozone production.  Figure 6. Comparison of the diurnal averaged production (hPi), removal (hLi) and net (hPi-hLi) rates of ozone with values obtained using the Crawford et al. [1997b] parameterization. Values are calculated using BIBLE B data.
The production value will remain high until NO x is converted to HNO 3 and the RO 2 are removed by some chain termination step. In the mean time, ozone concentration will be increasing. This argument suggests that, immediately after the injection, the production rate is high while the ozone concentration is still low. Thus hPi in %/day, which corresponds to the net production rate divided by the ozone concentration, should be a better indicator of recent NO injection.
[31] Figure 9 shows an example of the scatterplot of the ratio of NOy ½ equil NOy ½ meas versus hPi for sampling points between 8 and 9 km from BIBLE A. There were 394 points from the 1min file that satisfy the criteria having SZA less than 60°. Twenty-four of the points have hPi values larger than 20%/ day and are not displayed in the Figure 9. Only 5% of the points shown in Figure 9 have NOy ½ equil NOy ½ meas < 1.2. Figures 9b and 9c show the probability distribution functions of the sampled points in hPi and in NOy ½ equil NOy ½ meas , respectively. Based on these results and the fact that the production rate for ozone in an aged parcel is of order 5%/day, we will argue that only points with NOy ½ equil NOy ½ meas > 2 and hPi >5%/day are likely to have experienced recent NO injection. Points from flights over the oceans (flights 3, 14 and 15) and flights over Australia (flights 4 -6) have small values of hPi. Flights out of Bandung over land area have large values of hPi and are mostly indicative of air mass still adjusting to recent injection of NO.

Future Directions
[32] In this paper, we presented a method for calculating the diurnal averaged production and removal rates of ozone along the flight track. We suggested that the production rate may be a useful indicator of the time elapsed since the last injection of NO. We have yet to make use of the available information on back-trajectories of air parcels. It is possible to estimate the source terms for NO x from lightning and biomass burning [see, e.g., Koike et al., 2002] and use the photochemical air trajectory model to compute the expected concentrations of the trace species at the sampling point [see, e.g., Kita et al., 2002]. Comparison of the calculated   (b) Normalized frequency of occurrence in ozone production rate (%/day) for two groups of data segregated according to whether ½NOy equil ½NOy meas is larger or smaller than 1.5. The frequency of occurrence is normalized to 100 for each group. (c) Normalized frequency of occurrence in ½NOy equil ½NOy meas for two groups of data segregated according to whether the ozone production rate is larger or smaller than 5%/day. The frequency of occurrence is normalized to 100 for each group.
values with the measured values should provide information on the origin of the air parcel and help validate the estimated source strengths used in the calculation. Such analyses will better constrain the (unmeasured) concentrations used to initialize the air trajectory calculations. With the constrained initial conditions, the photochemical air trajectory model can then be used to compute the average net production for ozone for the parcel history.