The similarity in the ΔO3/ΔCO in the LT and MT suggests that O3 was not produced during transport from the LT to MT. In order to investigate locations of O3 production in the biomass burning plumes, we calculated photochemical formation (F), destruction (D), and net production (P = F − D) rates using the same box model as that used for OH calculation, as shown in Figure 13. For the region of this study, diurnally averaged F and D terms are expressed as
where RO2 represents peroxy radicals (e.g., CH3O2), [X] is the number density of species X, and kX+Y is the reaction rate coefficient for the X + Y reaction. The F and D values increase linearly with NO and O3, respectively (equations (2) and (3)). F increased with CO in the LT and MT, because NOx and therefore NO increased with CO (Figure 10a). The D values also increased linearly with CO, because O3 increased with CO. As a result, the F - D values were close to zero or slightly negative in the LT and MT, consistent with the similarity of ΔO3/ΔCO in the LT and MT. This tendency is typical for low-latitude air. At 5°–25°N, the median P values at 0–8 km were negative or close to zero for the whole data set sampled during TRACE-P [Davis et al., 2003]. By contrast, at 25°–45°N, the median P values were positive. The lower P values at lower latitudes in March are mainly due to higher H2O, which destroys O3 (equation (3)). From these results, it is very likely that most of the enhanced O3 in the biomass burning plumes was produced before transport to the sampling regions.
5.7.1. O3 Production Efficiency
 The ΔO3/ΔCO for SEA air masses is compared with that obtained in other subtropical and boreal burning regions and is summarized in Table 7. The data over northern Australia were obtained in the boundary layer (<3 km) during BIBLE-B in September [Takegawa et al., 2003b]. The data in Africa were obtained at 0.5–6 km during the Southern African Regional Science Initiative 2000 (SAFARI 2000) [Hobbs et al., 2003; Yokelson et al., 2003]. The ΔO3/ΔCO value of 0.22 for SAFARI 2000 was measured in biomass burning haze 2–4 days old off the coast of Namibia [Yokelson et al., 2003].
Table 7. ΔO3/ΔCO, Emission Ratios (ER), and Ozone Production Efficiency (OPE) for Biomass Burning, Urban, and Petrochemical Plumesa
|Location||Reference||ΔO3/ΔCO||Altitude, km||Age, days||ER (NOx/CO)||ER (C2H4/NOx)||OPE|
|Present study||1, 2||0.20 (±5%)||2–8||2–3||21 (±231%)||0.92 (±234%)||9.5 (±231%)|
|Australia||3, 4||0.12 (±3%)||1–3||0.5–3||18 (±10%)||0.38 (±10%)||6.7 (±30%)|
|South Africa||5||0.22 (±5%)b||1–6||2–4||44 (±36%)||0.38 (±35%)||5.0 (±36%)|
|Canada||6, 7||0.10 (±5%)b||1–3||3–4||7 (±54%)||1.57 (±87%)||15.7 (±54%)|
|United States and Europe||8, 9, 10, 11, 12||0.35 (±15%)||0||0.3||170 (±40%)||0.073 (±30%)||2.1 (±50%)|
|Houston, Texas||13|| || ||0.2|| ||3.6 (±17%)||17.9 (±50%)|
|Houston, Texas||13|| || ||0.2|| ||1.5 (±17%)||10.3 (±50%)|
|Houston, Texas||13|| || ||0.1|| ||2.4 (±17%)||10.9 (±50%)|
 It is remarkable that the ΔO3/ΔCO values in fire plumes in different subtropical regions after 0.5–4 days following emission are similar to within a factor of 2, considering the possible variability in the conditions for photochemical O3 production and loss by deposition. In the measurements over northern Australia and Africa, buildup of O3 occurred near the fires within a half day, before O3 precursors (such as NOx and NMHCs) were consumed and diluted. ΔO3/ΔCO remained nearly constant during transport in the BL over northern Australia for a few days [Takegawa et al., 2003b], supporting comparison of ΔO3/ΔCO observed in air masses with different ages. In order to understand this similarity in terms of photochemistry, we estimated the net number of O3 molecules produced per NOx molecule oxidized (or emitted), known as the O3 production efficiency (OPE). The notion of OPE was first introduced by Liu et al.  to estimate O3 formation rates over the U.S. continent. Here OPE (net number of O3 molecules produced per NOx molecule emitted) has been derived by dividing ΔO3/ΔCO by NOx/CO ER, as was done by Chin et al. . The NOx/CO ER and OPE are given in Table 7. The NOx/CO ER and C2H4/CO ER for SEA were derived from the sum of emissions of NOx, CO, and C2H4 by the burning of savanna/grasslands, forest, and crop residue in the five SEA countries (Table 3). The ERs for other regions are based on in situ measurements in fresh plumes. OPE is controlled by NMHC concentrations at high NOx levels (hydrocarbon-limited) [e.g., Liu et al., 1987; Sillman et al., 1990]. C2H4 is one of the major NMHCs emitted from biomass burning [e.g., Shirai et al., 2003, and references therein; Yokelson et al., 2003] and is an important precursor for efficient O3 production. The average OPE for savanna fires is 7.1 ± 2.4.
 Air masses impacted by the Canadian forest fires were transported southward and sampled at 1–3 km over the southeastern United States in July during the 1995 Southern Oxidants Study (SOS-95) [Wotawa and Trainer, 2000; McKeen et al., 2002]. The ΔO3/ΔCO for the Canadian forest fires, together with the other parameters, is summarized in Table 7. The OPE for the Canadian fires was 2 times larger than the average value for the subtropical fires. For these fires, the C2H4/NOx ER was 2–3 times larger and the NOx/CO ER was 2–6 times smaller than those for the Savanna fires.
 OPE was also derived by using ΔO3/Δ(NOy − NOx) for urban plumes [e.g., Chin et al., 1994; Hirsch et al., 1996; Rickard et al., 2002, and references therein] to estimate the net number of O3 molecules produced per NOx molecule oxidized. It is known that OPE estimated by using ΔO3/ΔCO is lower than that estimated by using ΔO3/Δ(NOy − NOx) [e.g., Chin et al., 1994; Hirsch et al., 1996; Ryerson et al., 2001; Rickard et al., 2002, and references therein]. Deposition of HNO3, which constitutes a major fraction of NOy − NOx, leads to overestimates of OPE by the use of ΔO3/Δ(NOy − NOx). On the other hand, the OPE derived using ΔO3/ΔCO should be considered a lower limit because O3 is negatively correlated with CO due to deposition of O3 in the absence of photochemistry [Chin et al., 1994]. The values of ΔO3/ΔCO, ERs, and OPE for urban plumes in the United States are given in Table 7 for comparison. ΔO3/ΔCO was a uniform 0.3 at nonurban sites in eastern North America in summer [Chin et al., 1994]. It was also uniform at 0.3–0.4 in the summer in Atlantic Canada and the spring in the Azores [Parrish et al., 1998]. NOx/CO ER was estimated to have increased linearly from 140 ± 60 pptv/ppbv in 1990 to 200 ± 80 pptv/ppbv in 2000 [Parrish et al., 2002; D. D. Parrish, unpublished results, 2003]. C2H4/NOx ER was derived from the National Acid Precipitation Assessment Program (NAPAP) inventory for the United States in 1995 [Saeger et al., 1989] and the U.S. Environment Protection Agency report for 2003 (http://www.epa.gov/ttn/chief) (D. D. Parrish, unpublished results, 2003). The OPE for urban plumes is lower by a factor of 2–4 than that for biomass burning plumes despite the higher ΔO3/ΔCO ratio. This is due to a much higher NOx/CO ER and a much lower C2H4/NOx ER, which is typical for fossil fuel combustion at high temperatures. High NOx with an insufficient supply of NMHCs reduces OPE, as observed in power plant plumes [Ryerson et al., 2001].
 Measurements of OPE were made using ΔO3/Δ(NOy − NOx) in plumes strongly impacted by petrochemical emissions of alkenes in the Houston, Texas, metropolitan area in August [Ryerson et al., 2003]. The plume ages were 1–4 hours. Because NOy was nearly conserved in these plumes, the effect of HNO3 deposition on OPE was small. It was shown that C2H4 and C3H6 promoted the rapid buildup of O3 within 4 hours. The resulting OPEs of 10–18 in these plumes were much higher than those in urban plumes (about 2) because of much higher concentrations of these alkenes, which rapidly react with OH, efficiently producing HOx. OPE was shown to increase with the C2H4/NOx and C3H6 ER in these plumes. These data are also listed in Table 7 for comparison.
 The OPE is plotted versus C2H4/NOx ER for biomass burning (boreal and subtropical fires), urban, and petrochemical plumes in Figure 14. The OPE increases with C2H4/NOx ER in biomass burning plumes as well as in urban-petrochemical plumes, although the slopes are somewhat different. The positive OPE-C2H4/NOx ER correlations suggest that O3 in the biomass burning plumes, as well as in urban plumes, was produced in hydrocarbon-limited regimes. Although more alkenes are available in biomass burning plumes than in urban plumes for the same amount of NOx, major O3 production occurs at locations where NOx concentration is still high, as observed over northern Australia [Takegawa et al., 2003b]. At these locations, the rate of O3 production depends also on the concentrations of alkenes. The range of C2H4/NOx ER for the petrochemical plumes partly overlaps with those for biomass burning. It is remarkable that the OPE for the petrochemical plumes in this C2H4/NOx ER range agrees with those for biomass burning to within 50%. The difference of 50% is small considering the difference in the conditions for O3 production, including differences in contributions of other NMHCs, H2O concentration, solar actinic flux, air mass age, and boundary layer height. This similarity in the OPE demonstrates that C2H4/NOx ER is a critical parameter that controls OPE in biomass burning plumes as well as fossil fuel combustion plumes. This result is understandable because C2H4 is the major constituent of highly reactive hydrocarbons both in biomass burning and urban emissions. In turn, improvement in the estimates of the emissions of C2H4 and C3H6 is critical in improving assessments of the effects of biomass burning on O3.
5.7.2. O3 Flux
 A net increase in the O3 mixing ratio was assessed as shown in Table 6: in the BL, LT, and MT, δO3 = 20, 26, and 18 ppbv, respectively. O3B decreased from 31 ppbv in the LT to 27 ppbv in the MT. Air transported from tropical latitudes was sampled more frequently in the MT, as discussed in section 4. The O3B of 31 ppbv is similar to the median O3 value for maritime air shown in Figure 6, and δO3/O3B was 0.8 and 0.6 in the LT and MT, respectively. From these δO3 values we have estimated the O3 flux from SEA to the western Pacific caused by biomass burning. Air impacted by biomass burning was transported from SEA to the Pacific along westerlies, mostly at the latitude range of 17°–30°N, as seen from the wind vectors shown in Figure 3. O3 flux across the meridional plane at 0–8 km along 120°E at 17°–30°N was calculated as
where na is air number density, U is the average westerly component of the wind velocity given from the ECMWF data during TRACE-P, η is the probability of sampling the SEA air, excluding low-H2O data, f is the fraction of the SEA data with CO > 110 ppbv, and S is the cross section of the meridional plane. Because U increases with altitude (Figure 3), increases in O3 at higher altitudes are more efficient in transporting O3. These values at different altitudes are given in Table 8, together with the estimated uncertainties. η maximized in the LT and decreased with altitude. The O3 flux in the BL was very small due to weak and unstable westerlies. The estimated F(δO3) is 50 (±133%) Gg O3 day−1.
Table 8. Parameters Used for Calculating the Net O3 Fluxa
| ||0–2 km||2–4 km||4–6 km||6–8 km|
|U, m/s||0.9 ± 5.0||8.7 ± 5.7||20.0 ± 10.3||27.4 ± 13.3|
|η||0.19 ± 0.11||0.40 ± 0.20||0.23 ± 0.22||0.01 ± 0.04|
|f||1.00 ± 0.00||0.91 ± 0.12||0.64 ± 0.33||1.00 ± 0.00|
 On the other hand, the total O3 production rate over peninsular SEA can be derived independently, assuming that ΔO3/ΔCO is uniform (0.20 ppbv/ppbv) for all biomass burning-impacted air. The total CO emitted from SEA was estimated to be 17.9 (±155%) Tg CO yr−1 for the year of 2000 [Streets et al., 2003b]. All of the SEA air masses used for the present analysis were sampled in March 2001. Considering that CO emissions in March constituted 46% (±22% relative error) of the total annual emission for 1997–2001 (Figure 1b), the emission rate in March is directly given as 267 (±157%) Gg CO day−1. This is transformed to a total O3 production rate of 73 (±157%) Gg O3 day−1. The F(δO3) derived above constitutes 68% (±206% relative error) of the estimated total O3 production rate, suggesting that a majority of O3 produced in SEA was transported to the western Pacific.
 In order to assess the overall accuracy of the derived O3 flux, comparison was made with a value derived independently. O3 fluxes in different types of air masses during TRACE-P were estimated using O3 data obtained by lidar on board the DC-8 [Browell et al., 2003]. Air masses characterized by high O3 and low aerosol (HO3 category) in this study basically correspond to SEA air, and the O3 flux of HO3 peaked at 26°N. The O3 flux for HO3 air was integrated over the latitude range of 17°–30°N and an altitude range of 0–8 km resulting in 104 Gg O3 day−1. For comparison with this flux, δO3 in equation (4) needs to be replaced with the median values of O3 impacted by biomass burning (O3).
The F(δO3) was estimated to be 114 (±135%) Gg O3 day−1, which is in reasonable agreement with that estimated by Browell et al. , considering some differences in the sampling regions for the P-3B and DC-8 and in the criteria defining biomass burning-impacted air masses.