Photochemical production and evolution of selected C 2 –C 5 alkyl nitrates in tropospheric air influenced by Asian outflow

[ 1 ] The photochemical production and evolution of six C 2 –C 5 alkyl nitrates (ethyl-, 1-propyl-, 2-propyl-, 2-butyl-, 2-pentyl-, and 3-pentyl nitrate) was investigated using selected data from 5500 whole air samples collected downwind of Asia during the airborne Transport and Chemical Evolution over the Pacific (TRACE-P) field campaign (February–April 2001). Air mass age was important for selecting appropriate field data to compare with laboratory predictions of C 5 alkyl nitrate production rates. In young, highly polluted air masses, the ratio between the production rates of 3-pentyl nitrate and 2-pentyl nitrate from n -pentane was 0.60–0.65. These measured ratios show excellent agreement with results from a field study in Germany (0.63 ± 0.06), and they agree better with predicted ratios from older laboratory kinetic studies (0.63–0.66) than with newer laboratory results (0.73 ± 0.08). TRACE-P samples that did not show influence from marine alkyl nitrate sources were used to investigate photochemical alkyl nitrate evolution. Relative to 2-butyl nitrate/ n -butane, the measured ratios of ethyl nitrate/ethane and 2-propyl nitrate/propane showed notable deviations from modeled values based on laboratory kinetic data, suggesting additional Asian sources of their alkyl peroxy radical precursors. By contrast, the measured ratios of 1-propyl-, 2-pentyl-, and 3-pentyl nitrate to their respective parent hydrocarbons were fairly close to modeled values. The 1-propyl nitrate findings contrast with field studies in North America, and suggest that air downwind of Asia was not significantly impacted by additional 1-propyl nitrate precursors. The sensitivity of modeled photochemical processing times to hydroxyl radical concentration, altitude, city ventilation times, and dilution is discussed. sinks; 0365 Atmospheric Structure: Troposphere— composition and chemistry; 0368 Atmospheric Composition and Structure:


Introduction
[2] Alkyl nitrates (RONO 2 ) have a relatively low reactivity compared with other components of reactive odd nitrogen (NO y ), which allows them to serve as a reservoir for the long-range transport of nitrogen oxides (NO x = NO + NO 2 ). Within the troposphere the concentration of NO x controls ozone (O 3 ) production [Talbot et al., 2000], and alkyl nitrates can be used as tracers of photochemical O 3 production from anthropogenic precursors [Flocke et al., 1998]. Whereas studies of continental air masses have shown that alkyl nitrates typically comprise less than 10% of NO y [Buhr et al., 1990;Ridley et al., 1990;Flocke et al., 1991;Shepson et al., 1993], alkyl nitrates have been found to comprise a large component (20 -80%) of total NO y in the equatorial marine boundary layer over the Pacific [Talbot et al., 2000;Blake et al., 2003].
[3] Alkyl nitrate sources include marine emissions [Walega et al., 1992;Atlas et al., 1993;Atlas and Ridley, 1996;Hess et al., 1996;Blake et al., 1999Blake et al., , 2003Chuck et al., 2002] and photochemical production from the oxidation of parent hydrocarbons [Darnall et al., 1976;Atkinson et al., 1982;Roberts, 1990;Atkinson et al., 1995;Arey et al., 2001]. In addition, biomass burning has recently been identified as a source of C 1 -C 4 alkyl nitrates [Simpson et al., 2002]. Although methyl nitrate (MeONO 2 ) was the most abundant alkyl nitrate released from Australian savanna burning, biomass burning likely has only a small impact on the global MeONO 2 budget. The importance of the marine alkyl nitrate source decreases with increasing carbon number, and marine emissions are only a minor source of C 4 alkyl nitrates [Blake et al., 2003]. The marine source of C 5 alkyl nitrates is also likely to be small. In photochemical alkyl nitrate production, whereas the branching ratios leading to alkyl nitrate formation increase with increasing carbon number [Atkinson et al., 1982;Arey et al., 2001], the ambient concentrations of parent n-alkanes decrease with increasing carbon number. Together, the net effect of these factors leads to a maximum abundance of C 3 -C 4 alkyl nitrates from the photochemical source, in particular 2-propyl-and 2-butyl nitrate.
[4] Here we use alkyl nitrate measurements collected during the airborne NASA Transport and Chemical Evolution over the Pacific (TRACE-P) field campaign (February -April 2001) to investigate the photochemical production, transport, and chemical evolution of alkyl nitrates in tropospheric air influenced by Asian outflow. We developed the capability to accurately measure C 5 alkyl nitrates shortly before the TRACE-P mission, and the TRACE-P data set provides a timely opportunity to compare field measurements of photochemical C 5 alkyl nitrate production rates with newly published values from laboratory kinetic studies. In photochemical alkyl nitrate production the ratio of a daughter alkyl nitrate to its parent hydrocarbon can be used to determine the extent of photochemical processing within an air mass [Bertman et al., 1995;Flocke et al., 1998;Roberts et al., 1998;Stroud et al., 2001]. Ratios of daughter alkyl nitrates to their parent hydrocarbons were used to investigate alkyl nitrate sources, evolution, and air mass age in tropospheric air samples collected between altitudes of 150 m and 12 km.

Experiment
[5] The alkyl nitrate measurements presented in this paper result from a collaboration between the University of California, Irvine (UCI) and the National Center for Atmospheric Research (NCAR). During TRACE-P we measured six C 2 -C 5 alkyl nitrates: ethyl nitrate (EtONO 2 ), 1propyl nitrate (1-PrONO 2 ), 2-propyl nitrate (2-PrONO 2 ), 2-butyl nitrate (2-BuONO 2 ), 2-pentyl nitrate (2-PeONO 2 ), and 3-pentyl nitrate (3-PeONO 2 ). The measurements were made using a combination of airborne whole air sampling, followed by laboratory gas chromatography (GC) analysis with electron capture detection (ECD). Although we also measured MeONO 2 during the TRACE-P mission, its photochemical production from methane is extremely slow and we do not consider the photochemical production and evolution of MeONO 2 in this paper.
[6] The whole air samples were collected aboard NASA DC-8 and P-3B research aircraft during 38 science flights of the TRACE-P field campaign (24 February -10 April 2001). The air samples were collected at altitudes between 150 m and 12 km over the western and central Pacific Ocean, with main deployment bases in Hong Kong (22°N, 114°E) and Yokota Air Force Base, Japan (36°N, 140°E). A detailed overview of the sampling area, including flight tracks for both aircraft, is given by Jacob et al. [2003]. Individual air samples were collected in 2-L stainless steel canisters each equipped with a stainless steel bellows valve. Prior to the mission, the canisters were conditioned and evacuated, and 10 Torr of degassed, distilled water was added into each canister to quench active surface sites. The air samples were collected roughly every 3 -7 min during horizontal flight legs and every 1 -3 min during ascents and descents. Horizontal sampling times were approximately 1 min and correspond to a sampling distance of about 12 km. Approximately 160 air samples were collected per flight aboard the DC-8, and 140 were collected aboard the P-3B.
[7] After each flight, the filled canisters were transported back to the UCI laboratory. Within 10 days of being collected, each air sample was analyzed for alkyl nitrates and more than 50 other trace gases comprising nonmethane hydrocarbons (NMHCs), halocarbons, and sulfur compounds. The same canisters were reused during the course of the mission, and two identical analytical systems (sharing the same standards) were operated simultaneously in order to improve the canister turn-around time. Because it takes a few days for the analytical systems to equilibrate, they were operated 24 hours a day throughout the mission in order to generate an internally consistent data set.
[8] The reader is referred to Colman et al. [2001] for a detailed description of our analytical procedure. Briefly, to analyze each TRACE-P whole air sample, the less volatile components contained within 1520 ± 1 cm 3 (STP) of canister air (i.e., alkyl nitrates, NMHCs, halocarbons, and DMS) were preconcentrated by passing the air through tubing immersed in a liquid nitrogen bath. The components were revolatilized using a hot water bath and then reproducibly split into five streams, with each stream directed to a different column-detector combination. We have found that the split ratios are highly reproducible as long as the specific humidity of the air is greater than 0.3 kPa at 298 K, which we ensure by the addition of 10 Torr (1.3 kPa) of water into each canister (see above).
[9] The C 2 -C 5 alkyl nitrates were analyzed using two of the five column-detector combinations. The first columndetector combination (abbreviated as ''Restek1701/ECD'') was a RESTEK 1701 column (60 m, 0.25 mm I.D., 0.50 mm film thickness) output to an ECD. The second combination (''DB5-Restek1701/ECD'') was a DB-5 column (J&W, 30 m, 0.25 mm I.D., 1 mm film thickness) connected in series to a RESTEK 1701 column (5 m, 0.25 mm I.D., 0.5 mm film thickness) and output to an ECD. The Restek1701/ECD and DB5-Restek1701/ECD combinations received 7.2 and 6.8% of the sample flow, respectively, and were housed in separate GCs . For the C 2 -C 4 alkyl nitrates the mixing ratios determined from both column-detector combinations agreed very well and were averaged to give a single mixing ratio for each sample. The slope calculated for both combinations differed from the 1:1 line by 1% for EtONO 2 , 2% for 1-PrONO 2 , 2% for 2-PrONO 2 , and 4% for 2-BuONO 2 . In each case the r 2 value exceeded 0.99 and the intercept was between 0.01 and 0.04 pptv. For the C 5 alkyl nitrates, the two combinations showed very good agreement for 2-PeONO 2 (slope = 1.04; r 2 = 0.98) and fair agreement for 3-PeONO 2 (slope = 1.22; r 2 = 0.98). The Restek1701/ ECD results were not used for 3-PeONO 2 because they showed a poorer correlation with the 2-BuONO 2 mixing ratios than did the DB5-Restek1701/ECD results. For similar reasons, the DB5-Restek1701/ECD results were not used for 2-PeONO 2 .
[10] The alkyl nitrate measurements were calibrated using two whole air working standards that were analyzed every four samples and have been previously calibrated against a synthetic standard prepared at NCAR. The alkyl nitrate measurement precision is conservatively 2% for mixing ratios above 5 parts per trillion by volume (pptv) and 10% for mixing ratios below 5 pptv. The accuracy of the alkyl nitrate measurements is 10-20%, and the alkyl nitrate detection limit is 0.01 pptv.

Photochemical Production of Alkyl Nitrates
[11] The photochemical production of alkyl nitrates in the troposphere is initiated by the oxidation of parent hydrocarbons (RH) by the hydroxyl radical (OH) leading to the formation of parent alkyl peroxy radicals (RO 2 ): where k 1 and k 2 are reaction rate constants and the branching ratio a 1 quantifies the fractional yield of a specific parent RO 2 . For example, for C 3 -C 4 parent n-alkanes (propane, i-butane, and n-butane), equation (1) gives rise to two possible RO 2 isomers, depending on which hydrogen is abstracted by OH. (The group rate constant for hydrogen atom extraction from alkanes is larger for a tertiary carbon than a secondary carbon and larger for a secondary carbon than a primary carbon (Table 1).) The branching ratio a 1 is calculated using the group rate constants and group substituent factors derived by Atkinson [1987] and revised by Kwok and Atkinson [1995] (Table 1).
[12] At NO mixing ratios greater than about 30 -40 pptv the subsequent reaction of RO 2 with NO dominates over RO 2 reaction with nitrogen dioxide (NO 2 ), hydroperoxy radicals (HO 2 ), or other RO 2 radicals [Logan, 1985;Finlayson-Pitts and Pitts, 2000]. The reaction of RO 2 with NO has two possible outcomes (equations (3a) -(3b)), one of which leads to the formation of a daughter alkyl nitrate (equation (3b)): where k 3a and k 3b are reaction rate constants and the branching ratio a 2 quantifies the fractional yield of the daughter alkyl nitrate from the reaction of RO 2 with NO: Alkyl nitrates are a minor product of RO 2 + NO reaction.
In laboratory studies, a 2 has been found to range from 0.006 for EtONO 2 [Ranschaert et al., 2000] to 0.126 for 3-PeONO 2 [Arey et al., 2001]. In addition to increasing with increasing carbon number, the magnitude of a 2 is also a function of temperature and pressure [Atkinson et al., 1982;Carter and Atkinson, 1989].

Photochemical Destruction of Alkyl Nitrates
[13] Alkyl nitrates are destroyed by photolysis or by reaction with OH: where J RONO2 and k 4 are the reaction rate constants for the loss of alkyl nitrates by photolysis and OH, respectively. Previous studies have found that the photolysis pathway is the dominant loss process for C 1 -C 3 alkyl nitrates [Roberts and Fajer, 1989;Clemitshaw et al., 1997;Talukdar et al., 1997]. For example, Talukdar et al. [1997] calculated that reaction with OH can contribute approximately 20% of the total loss rate for EtONO 2 and 2-PrONO 2 and less than 10% for MeONO 2 (see Talukdar et al. [1997] for assumed conditions).
[14] The rate of RONO 2 destruction varies with altitude, season, and latitude. For example, J RONO2 is a function of both solar flux intensity (which increases with altitude) and absorption cross section (which decreases with altitude) [Clemitshaw et al., 1997 and references therein;Talukdar et al., 1997]. Overall, the higher solar flux intensities at higher altitudes cause J RONO2 to increase with height in the troposphere [Talukdar et al., 1997]. For 1 April and 40°N, Clemitshaw et al. [1997] calculated that J-values for selected C 2 -C 5 alkyl nitrates are more than a factor of two higher at 10 km than at the surface. By contrast, the reaction of RONO 2 with OH decreases with increasing altitude because of slower rate constants and lower OH concentrations. For conditions comparable to those encountered during TRACE-P (1 April, 40°N), the surface level lifetimes of the C 2 -C 5 alkyl nitrates considered here range from 6 to 17 days (Table 2).

Alkyl Nitrate Evolution
[15] Assuming that OH reaction with the parent hydrocarbon is the rate-limiting step in photochemical alkyl nitrate formation, equations (1) -(3) and (5) -(6) can be simplified to sequential alkyl nitrate production and destruction [Bertman et al., 1995]: Here, k prim , k sec , and k tert are group rate constants for H-atom abstraction from -CH 3 , -CH 2 -, and >CH-, respectively (all k values are for 298 K and have units of 10 À12 cm 3 molec À1 s À1 ); F(X) is the group substituent factor for substituent group X (F(-CH 3 ) is 1.00 by definition).
where k A = k 1 [OH] and k B = J RONO2 + k 4 [OH] are pseudofirst order rate constants, and the factor b = a 1 a 2 takes into account the branching ratios that lead to the alkyl nitrate formation, assuming no peroxy radical self-reaction. The time evolution of a daughter alkyl nitrate to its parent hydrocarbon in an air mass can then be modeled by [Bertman et al., 1995]: where the initial concentration of the alkyl nitrate is assumed to be zero.
[16] If we assume that photochemical losses are negligible compared to production (such as in very fresh air masses) and consider only chemical terms (i.e., dilution acts on each species equally and there is negligible contribution from marine alkyl nitrate sources), then the measured mixing ratios of two daughter alkyl nitrates i and j that originate from the same parent hydrocarbon are related by [Flocke et al., 1998]: The left hand side of equation (10) is termed the measured ratio, m meas , because it can be quantified by atmospheric measurements. The right hand side is the predicted ratio, m pred = b i /b j , which is based on the branching ratios a 1 and a 2 . The C 5 alkyl nitrates 2-PeONO 2 and 3-PeONO 2 both originate from n-pentane, and the slope of 3-PeONO 2 vs. 2-PeONO 2 is expected to equal b 3-PeONO2 /b 2-PeONO2 for freshly polluted air masses. By contrast, equation (10) does not apply for alkyl nitrates <C 4 because they are formed from the decomposition of other, larger alkoxy radicals in addition to their parent hydrocarbons [Flocke et al., 1998;Roberts et al., 1998].

Results and Discussion
[17] This paper focuses on three aspects of photochemical alkyl nitrate production and evolution. In all three cases, TRACE-P alkyl nitrate measurements are compared with predictions based on laboratory kinetic studies. First, the photochemical production of 3-PeONO 2 and 2-PeONO 2 from n-pentane in relatively fresh air masses (where photochemical losses are likely negligible compared to production) are used to assess laboratory predictions of C 5 alkyl nitrate production. Second, RONO 2 /RH ratios are used to quantitatively investigate photochemical processing within air masses sampled during TRACE-P. Third, RONO 2 /RH ratios are used to detect additional sources of alkyl nitrate precursors in Asian outflow. The two young, polluted plumes used in the first and second investigations are described below, following an overview of the alkyl nitrates sampled during TRACE-P.

Overview of TRACE-P Alkyl Nitrate Measurements
[18] Of the 5493 whole air samples collected aboard the NASA DC-8 and P-3B research aircraft during TRACE-P, the percentage of measurements in which alkyl nitrates were detectable exceeded 99% for all reported alkyl nitrates C 4 , 97% for 2-PeONO 2 , and 84% for 3-PeONO 2 . The mixing ratios of the C 2 -C 5 alkyl nitrates decreased with increasing altitude (Table 3), consistent with their surface sources and their short atmospheric lifetimes relative to the vertical tropospheric mixing time of $1 month. The maximum abundance of 2-PrONO 2 and 2-BuONO 2 indicates the importance of photochemical alkyl nitrate production during TRACE-P.
[19] The TRACE-P alkyl nitrate mixing ratios can be compared to studies in urban continental, rural continental, and marine environments (Table 4). For example, in Karachi, Pakistan during winter, Barletta et al. [2002] measured mean ground-level C 2 -C 4 mixing ratios that are 2 -3 times larger than the mean values reported between 0 and 2 km during TRACE-P. In Athens, Greece during summer, Glavas and Moschonas [2002] measured mean 2-PeONO 2 and 3-PeONO 2 mixing ratios that are 5-9 times higher than the mean mixing ratios sampled between 0 -2 km during TRACE-P and comparable to the maximum TRACE-P mixing ratios measured between 0 and 2 km. In rural Alabama during summer, Bertman et al.
[1995] measured mean alkyl nitrate mixing ratios that are Units: 10 À12 cm 3 molec À1 s À1 for k 1 and k 4 ; 10 À6 s À1 for k A , k B , k 4 [OH], and J; days for lifetime. We emphasize that the 10 km J-values were calculated without applying temperature-dependent absorption cross-sections and are likely overestimated. Diurnally averaged [OH] of 8.8 Â 10 5 and 3.2 Â 10 5 molec cm À3 were used for calculations at 0 km and 10 km, respectively. ''n/a'' = not available. b Atkinson [1997]. Although an average temperature of 230 ± 5 K was measured between 9.5 and 10.5 km during TRACE-P, we used 245 K to simulate upper tropospheric temperatures because temperature-dependent rate constants are only available over the restricted range 245 -330 K. c Atkinson et al. [1997]. Entries marked with an asterisk are from Atkinson [1990]. Temperature-dependent rate constants are only available over the restricted temperature range 290 -380 K, and k 4 could not be calculated for upper tropospheric temperatures. d Diurnal J-values were calculated using absorption cross-sections reported by Roberts and Fajer [1989] and conditions appropriate for TRACE-P (1 April; 40°N; ozone column = 350 Dobson Units; see text). similar to those measured between 0 and 2 km during TRACE-P, with the exception of EtONO 2 , which was three times higher in the Alabama study. At the remote Mauna Loa Observatory in Hawaii during late spring, Elliot et al. [1992] measured upslope mixing ratios that are generally much smaller than those observed during TRACE-P. Stroud et al. [2001] measured airborne alkyl nitrate mixing ratios over the northeastern plains of Colorado during summer, between altitudes of 0 and 7 km. Their mean alkyl nitrate mixing ratios lie in between the TRACE-P means for 0 -2 km and 2 -7 km, although the results are not statistically different. Overall, individual alkyl nitrate measurements during the TRACE-P mission covered the range from clean remote air to polluted continental air. The mean TRACE-P measurements were of similar magnitude to alkyl nitrate mixing ratios in rural continental air from the United States.

Shanghai Plume
[20] A fresh, well-defined plume was encountered at an altitude of 330 m on 21 March 2001 from 1335-1355 local time (0435 -0455 UTC) during DC-8 Flight 13 ( Figure 1). This was a local flight deployed from Yokota Air Force Base and flown over the Yellow and East China Seas. The plume was characterized by high levels of short-lived compounds such as ethene (lifetime $ 1 -2 days) and propene (lifetime $ 8 -12 hours) (680 -3050 pptv and 5 -110 pptv, respectively). However, the very short-lived hydrocarbon trans-2-butene (lifetime $ 4 hours) did not exceed its detection limit during the plume encounter. The plume contained the highest pentyl nitrate levels of the mission, with mixing ratios up to 30 pptv (2-PeONO 2 ) and 20 pptv (3-PeONO 2 ) ( Figure 2a). These mixing ratios are comparable to mean mixing ratios measured in the city of Athens during summer [Glavas and Moschonas, 2002]. Based on the observed mixing ratios of compounds with major anthropogenic sources, such as carbon monoxide (CO) and the industrial tracer tetrachloroethene (C 2 Cl 4 ), the Shanghai plume was the most polluted air mass sampled during the TRACE-P mission. During the plume encounter, C 2 Cl 4 mixing ratios exceeded 120 pptv, compared with a median C 2 Cl 4 mixing ratio of 5 pptv during the remainder of the flight.
[21] Five-day backward trajectories provided by Florida State University (FSU) suggest that the plume had encountered the city of Shanghai on the Asian coast roughly 18 hours prior to sampling. The high levels of short-lived alkenes suggest that the Shanghai plume was relatively fresh and is a good candidate for investigating the photo-chemical production of C 5 alkyl nitrates. The requirement to consider only chemical terms in applying equation (10) is satisfied because the marine source of C 5 alkyl nitrates is expected to be small (section 1). In addition, dilution is presumed to act equally on 2-PeONO 2 and 3-PeONO 2 because of their similar lifetimes (Table 2).

Seoul Plume
[22] The second-highest pentyl nitrate mixing ratios during TRACE-P were encountered during a spiral into the marine boundary layer aboard P-3B Flight 19, a local flight deployed from Yokota Air Force Base and flown over the Sea of Japan (Figure 1). The spiral occurred on 3 April 2001 between 1310 and 1320 local time (0410 -0420 UTC). Mixing ratios as high as 16 pptv (2-PeONO 2 ) and 12 pptv (3-PeONO 2 ) were measured at altitudes below 900 m ( Figure 2b). Five-day backward trajectories show that the air mass had traveled over Beijing roughly 1.5 days prior to sampling, then had been advected over Seoul 0.5 -1 day prior to sampling. As with the Shanghai plume, mixing ratios of the very short-lived hydrocarbon trans-2-butene did not exceed its detection limit during the plume encounter, but elevated mixing ratios of propene (5-25 pptv) and ethene (140-790 pptv) were again detected.
[23] To investigate possible influence of Beijing air (which would contribute relatively aged components) on A total of 5493 samples were collected at altitudes between 150 m and 12 km, mainly over the western and central Pacific. At low altitudes including the marine boundary layer (0 -2 km), the number of samples (n) was 1732; in the lower free troposphere (2 -7 km) n = 2835; in the upper troposphere and lower stratosphere (7 -12 km) n = 926. LOD = limit of detection.  ) and Japan/Korea (methyl bromide or CH 3 Br) were examined. The plume showed a 150% increase in CH 3 Br mixing ratios relative to the rest of the flight, with only minor increases in H-1211 (7%) and CH 3 Cl (8%). By contrast, the Shanghai plume showed large enhancements of both H-1211 (225%) and CH 3 Cl (115%). These results suggest that the plume encountered during P-3B Flight 19 was strongly influenced by pollution from Seoul, with little influence from Beijing pollution sources further upwind. The backward trajectories transit over Beijing around midnight local time, at altitudes above 700 hPa (2.5 km). We believe that the Seoul plume shows little influence from Beijing because Beijing's pollution was contained within a compressed nighttime boundary layer, and the trajectories passed over Beijing at an altitude above the main pollution layer. The freshness of the Seoul emissions and the large pentyl nitrate mixing ratios suggest that the Seoul plume is also a good candidate for investigating the photochemical production of C 5 alkyl nitrates. Detailed evidence that supports using the Shanghai and Seoul plumes in this investigation is given in section 4.5.

Photochemical Production of C 5 Alkyl Nitrates
[24] The measured slopes of 3-PeONO 2 vs. 2-PeONO 2 during the Shanghai and Seoul plume encounters were 0.60 ± 0.01 pptv/pptv (r 2 = 0.998) and 0.65 ± 0.02 pptv/ pptv (r 2 = 0.996), respectively (Figure 2a -2b). That is, for every 100 molecules of 2-PeONO 2 that were formed from the oxidation of n-pentane, 60-65 molecules of 3-PeONO 2 were produced. The slopes from the Shanghai and Seoul plumes show excellent agreement with a slope m meas = 0.63 ± 0.06 pptv/pptv from a ground-based field study in Germany [Flocke et al., 1998]. They also lie within the uncertainty range of m meas = 0.68 ± 0.09 pptv/pptv based on upslope measurements at the Mauna Loa Observatory [Atlas et al., 1992]. However, the Mauna Loa air samples are likely more aged than the Shanghai and Seoul plumes and as a result are not expected to be good candidates for comparison with m pred (see section 4.5.1).  [25] The measured slopes can be compared with the predicted slope m pred = b i /b j , which is based on the branching ratios a 1 and a 2 (Table 5). Here, m pred was calculated using both old and newly published values of a 1 and a 2 in order to investigate the magnitude by which m pred is affected by the new branching ratios ( Table 6). The branching ratio a 1 was calculated from the structure-activity relationships described by Atkinson [1987], using the parameters derived by Atkinson [1987] and updated by Kwok and Atkinson [1995] ( Table 1). The reader is referred to Finlayson-Pitts and Pitts [2000] for a detailed sample calculation. For the alkyl nitrates listed in Table 5, the new structure-activity parameters caused the largest change for the primary alkyl nitrates: a 1 decreased by 14% for 1-PrONO 2 , whereas a 1 increased by 2 -6% for secondary RONO 2 , and decreased by 2% for 3-PeONO 2 .
[26] Arey et al. [2001] present new values of alkyl nitrate formation yields from parent RO 2 molecules (a 2 ) and from parent RH molecules (a 1 a 2 ) for selected C 5 -C 8 alkyl nitrates. The C 5 values cited by Arey et al. [2001] are based on work by Atkinson et al. [1995]. For hexyl, heptyl, and octyl nitrates, the sum of isomeric alkyl nitrate yields from a given parent n-alkane are about 32, 39, and 29% lower, respectively, than data previously reported from the same laboratory by Atkinson et al. [1982]. Although the new total pentyl nitrates yields are about 10% lower than the earlier data, the results are not statistically different to within the stated uncertainties. Arey et al. [2001] suspect that wall losses (especially during calibration) may have contributed to the higher yields in the earlier study, in contrast to the new studies in which wall losses during calibration were verified to be minor or negligible.
[28] The slope of 3-PeONO 2 versus 2-PeONO 2 during Shanghai plume encounter is significantly smaller than the newest predicted slope from laboratory kinetic studies, whereas the slope measured during the Seoul plume encounter does not significantly differ from m pred = 0.73 ± 0.08 (Table 6). Overall, the ratio between the production rates of 3-PeONO 2 and 2-PeONO 2 from n-pentane measured during this study (m meas = 0.60 -0.65) shows better agreement with predicted slopes using older values of a 2 (m pred = 0.63 -0.66) than with predicted pentyl nitrate production rates using the most recent values of a 1 and a 2 (m pred = 0.73 ± 0.08). These comparisons suggest that the ratio between the production rates of 3-PeONO 2 and 2-PeONO 2 from n-pentane likely lies near the lower limit of the reported uncertainty range for the most recent value of m pred = 0.73 ± 0.08. Here a 1 is the formation yield of a parent alkyl radical from the reaction of a parent hydrocarbon with OH; a 2 is the formation yield of a daughter alkyl nitrate from the reaction of a parent alkyl peroxy radical with NO; b is a 1 a 2 . b Here a 1 was calculated using rate constants for the abstraction of hydrogen atoms from C 3 -C 5 n-alkanes. These rate constants were determined using the group rate constants and group substituent factors derived by Atkinson [1987] and revised by Kwok and Atkinson [1995].  The measured slopes from TRACE-P were calculated for the young, highly polluted Shanghai and Seoul plumes.

Evolution of 3-PeONO 2 /2-PeONO 2
[29] In this section we quantitatively show why young air masses such as the Shanghai and Seoul plumes are the best candidates of the TRACE-P data set for comparing the photochemical production of C 5 alkyl nitrates with laboratory predictions. The evolution of 3-PeONO 2 /2-PeONO 2 within the Shanghai plume is used as an example. The sensitivity of the modeled 3-PeONO 2 /2-PeONO 2 ratio to the applied initial conditions is also discussed.

Shanghai Plume Evolution
[30] To predict the time evolution of 3-PeONO 2 / 2-PeONO 2 within the Shanghai plume, equation (9) was rearranged to solve for the daughter pentyl nitrate: We note that dilution is neglected in this equation (see section 4.5.2). Equation (11) was then solved for both 3-PeONO 2 and 2-PeONO 2 under conditions appropriate for the Shanghai plume, from t = 0 to t = 30 days. The newest values of a 1 and a 2 from the literature were used to calculate b (Table 5). A diurnally averaged surface-level OH concentration of [8.8 ± 4.7] Â 10 5 molec cm À3 was used in calculating k A and k B . This OH value was determined for TRACE-P conditions (35 -41°N, 0 -1 km) using a model previously described by Crawford et al. [1999]. Diurnal J-values are also required to determine k B , and they were calculated assuming a clear sky using absorption cross sections reported by Roberts and Fajer [1989] and conditions appropriate for TRACE-P (1 April, 40°N, ozone column = 350 Dobson Units; Table 2), also using the model described by Crawford et al. [1999]. The reader is referred to Clemitshaw et al. [1997] and Talukdar et al. [1997] for a comparison of alkyl nitrate ultraviolet absorption cross-sections obtained from various laboratory studies since the late 1980s.
[31] The time evolution of n-pentane is also required to solve equation (11) and was determined as follows. First, the initial concentration of n-pentane, [n-pentane] 0 , was calculated using: During the Shanghai plume encounter, the average mixing ratio of n-pentane was 365 ± 202 pptv (1 pptv = 2.46 Â 10 7 molec cm À3 at the Earth's surface). The 18-hour transit time of the Shanghai plume was used to approximate the elapsed time t (the age of the Shanghai plume is examined in detail in section 4.5.2). The diurnally averaged OH concentration of 8.8 Â 10 5 molec cm À3 gives k A = 3.5 Â 10 À6 s À1 . Solving equation (12) under these conditions gives [n-pentane] 0 $ 460 pptv (1.1 Â 10 10 molec cm À3 ). Equation (12) was next rearranged to its more familiar form: allowing the depletion of n-pentane to be modeled from t = 0 to t = 30 days (Figure 3a). The calculated n-pentane values were then used to solve equation (11) for 2-PeONO 2 and 3-PeONO 2 . We note that once a value of OH has been selected, the only two values that change with time in equation (11) are time itself, t, and [n-pentane], which decreases with time.
The way in which [n-pentane] and t interact with each other determines the shape of the 2-PeONO 2 and 3-PeONO 2 time evolution curves.
[33] This modeled increase in the ratio of 3-PeONO 2 / 2-PeONO 2 with photochemical processing is consistent with measured 3-PeONO 2 /2-PeONO 2 ratios during TRACE-P. The slopes of 3-PeONO 2 /2-PeONO 2 during the Shanghai and Seoul plume encounters (0.60 ± 0.01 and 0.65 ± 0.02, respectively) are lower than a measured slope of 0.80 when data points from all the TRACE-P flights (which comprise more aged air masses) are included (Figure 2a -2c). In Figure 2c the elevated mixing ratios in the top right quadrant of the graph represent the young Shanghai and Seoul plumes, and they lie below the fitted line.
[34] The key point to take from Figure 4 is that when selecting field data from which to calculate m meas for comparison to m pred , younger air samples are more appropriate because they are expected to better approximate m pred . Under the conditions used to generate Figure 4, the ratio of 3-PeONO 2 /2-PeONO 2 is predicted to increase 5% after one day of processing and 11% after two days. For the <1 day old Shanghai plume (section 4.5.2), the measured slope of 0.60 is expected to be within 5% of its initial value at t = 0. As a result young, polluted air masses such as the Shanghai and Seoul plumes are considered to be the best candidates of the TRACE-P field data set for comparison with predicted slopes from laboratory kinetic studies. Because the slope of 3-PeONO 2 /2-PeONO 2 is predicted to increase exponentially with time, we suggest that air mass age of less than one or two days is a more useful criterion for applying equation (10) than our earlier, qualitative assumption that photochemical loss is negligible compared to production.

Sensitivity to Initial Conditions
[35] Based on the estimated initial n-pentane mixing ratio of 460 pptv for the Shanghai plume, 2-PeONO 2 and 3-PeONO 2 are predicted to peak at values of 13 pptv and 11 pptv, respectively (Figure 3a). By comparison, the average mixing ratios of 2-PeONO 2 and 3-PeONO 2 measured in the Shanghai plume were 17.5 ± 6.9 pptv and 12.4 ± 4.1 pptv, respectively. That is, the modeled peak mixing ratios are similar to values measured within the Shanghai plume, to within the natural variability of the measurements. However, when making these comparisons it is important to recognize that the RONO 2 concentrations measured in the Shanghai plume may not be the peak values. Instead, the measured mixing ratios could lie before or after the plume's peak values.
[36] Direct comparison between the measured and modeled results is also impeded because an assumption in the application of equation (11) is that the initial concentration of the alkyl nitrate is zero, whereas in reality alkyl nitrate production would have begun (1) in Shanghai before the ventilation of the plume from the city and (2) possibly further upwind in air that was advected through Shanghai. In the latter case, influence from upwind air appears to be unlikely. Because ethyne is relatively short-lived (lifetime $2 -3 weeks) compared to CO (lifetime $ 2 months), the ratio of ethyne/CO can be used to assess the amount of atmospheric processing (photochemical reaction + dynamic mixing) within an air mass . A ratio of 4-5 pptv/ppbv indicates unprocessed air; a ratio of 1 -2 represents processed air; and a ratio of less than 1 indicates very processed air (E. Browell, NASA-Langley, personal communication, 2002). The median ethyne/CO ratio during the Shanghai plume encounter was 6.3 pptv/ppbv, with individual values ranging from 5.0 to 9.4 pptv/ppbv (the CO measurements were made by Glen Sachse, NASA-Langley). These results show that the Shanghai plume samples were unprocessed and were not significantly impacted by more aged components from further upwind.
[37] The extent of photochemical processing within Shanghai depends in part on the city's ''ventilation time,'' the time required for an emitted pollutant to be transported downwind of the city. Streets et al. [1999] report on air quality in Shanghai, and based on their work the most frequent wind speed class in Shanghai during the spring season is 3 -8 m s À1 , with an average of 5.5 m s À1 or roughly 20 km hr À1 (J.-H. Woo, University of Iowa, personal communication, 2002). Surface winds measured in Shanghai the day before sampling (20 March 2001) were higher than average, at roughly 10 m s À1 . Pollutants that are emitted from central Shanghai need to travel roughly 50 km downwind to be ventilated from the city. Assuming a relatively constant wind direction, pollutants emitted in central Shanghai were likely transported downwind of the city in less than 2 hours on 20 March, and the age of the Shanghai plume is expected to be only a few hours older than its 18-hour transit time to the sampling location over the Pacific Ocean. The photochemical production of alkyl nitrates within Shanghai is expected to remain low as a result of Shanghai's multiple daily ventilations. However, a background alkyl nitrate concentration of even a few pptv in Shanghai violates the assumption of zero initial alkyl nitrate concentration in the application of equation (11), given that a few pptv represents a reasonable fraction of the mean 2-PeONO 2 and 3-PeONO 2 mixing ratios measured within the Shanghai plume (17.5 ± 6.9 pptv and 12.4 ± 4.1 pptv, respectively).
[38] Dilution competes with chemistry on all timescales, and its effect on the Shanghai plume during its 18-hour transit over the ocean must also be considered. In previous studies the ratios of various hydrocarbons have been shown to change as a function of both photochemistry and atmospheric mixing [McKeen and Liu, 1993;McKeen et al., 1996]. In the comparison between 2-PeONO 2 and 3-PeONO 2 , dilution is presumed have a similar effect on  Table 1. both species because of their comparable lifetimes (Table 2), and the effect of dilution is expected to largely cancel in the ratio of 3-PeONO 2 /2-PeONO 2 .
[39] Finally, we consider the OH concentration that was applied in equation (11). Based on [OH] = 8.8 Â 10 5 molec cm À3 (section 4.5.1), equation (11) predicts maximum 2-PeONO 2 and 3-PeONO 2 mixing ratios after 4 and 5 days, respectively. When the concentration of OH used to calculate k A and k B is doubled, the pentyl nitrate lifetimes with respect to OH are halved, and the pentyl nitrate concentrations peak two days earlier (Figure 3b). That is, the time at which the pentyl nitrate peaks are predicted to occur is highly sensitive to the amount of OH that is used to solve equation (11).
[40] The measured OH mixing ratio in the Shanghai plume was 0.12 -0.19 pptv (3.0 -4.7 Â 10 6 molec cm À3 ) (H. Harder, Pennsylvania State University, personal communication, 2002). In the Shanghai plume's estimated 18 hour backward trajectory we estimate that roughly 7 hours (from 0630 to 1330 local time) were spent in daylight. Although OH exhibits a strong diurnal variation, its average mixing ratio was fairly constant between 0930 and 1330 during low-altitude flight legs of Flight 13 (0.16 ± 0.02 pptv; http://www-gte.larc.nasa.gov/trace/gte_mrg1.htm# TRACE-P), suggesting a widespread area of high OH concentration over the East China Sea on the day of sampling. If we apply 0.16 pptv from 0930 to 1330, and an estimated average OH concentration of 0.08 pptv from 0630 to 0930 (from 0 pptv at sunrise to 0.16 pptv at 0930), then the average OH concentration during the 18 hour plume travel time was 1.2 Â 10 6 molec cm À3 (1 pptv = 2.47 Â 10 7 molec cm 3 at the Earth's surface). This is 35% higher than the applied value of 8.8 Â 10 5 molec cm À3 . The plume likely spent a couple hours in Shanghai before being ventilated from the city at roughly 1930 local time on 20 March, and additional reaction with OH is expected to have occurred within Shanghai that late afternoon/early evening. Despite the uncertainties in these calculations (such as variations in daytime OH levels and the exact amount of time spent in Shanghai before ventilation), the average OH level in the Shanghai plume's history was likely larger than the OH value that was applied in the model.
[41] In summary, the modeled pentyl nitrate mixing ratios are highly sensitive to the amount of OH that is assumed in the history of the air mass. In the case of the Shanghai plume, a direct comparison between the measured and modeled pentyl nitrate mixing ratios was not possible because background pentyl nitrate mixing ratios within Shanghai were not quantified, such that the condition of zero initial alkyl nitrate concentration could not be verified. In addition, it was not known whether the Shanghai plume was sampled at/near the peak pentyl nitrate concentrations. These unknowns precluded using the Shanghai plume measurements to validate the time and magnitude of the pentyl nitrate peaks that were predicted using equation (11).

Evolution of C 2 -C 5 Alkyl Nitrates
[42] In photochemical alkyl nitrate evolution, the ratio of a daughter alkyl nitrate to its parent hydrocarbon increases as an air mass ages because the daughter alkyl nitrate is produced at the expense of its parent hydrocarbon. Here RONO 2 /RH ratios are calculated for the measured C 2 -C 5 alkyl nitrates and compared to modeled curves based on laboratory kinetic data.

Photochemical Processing Times
[43] Before attempting to quantify the amount of photochemical processing within an air mass using RONO 2 /RH ratios, dilution needs to be considered, in addition to the sensitivity of the modeled evolution curves to the kinetic data required to solve equation (9) (k 1 , k 4 , a 1 , a 2 , [OH] and J ). In the latter case, the modeled ratios of 2-PeONO 2 /n-pentane vs. 2-BuONO 2 /n-butane are used as an example. The modeled slope is found to vary only slightly depending on whether older or newer branching ratios from the literature are used (Figure 5a). Although the position of the line is quite insensitive to changes in the OH concentration, the times associated with a particular position on the line are seen to change with the OH concentration (Figure 5b). [44] Whereas recent alkyl nitrate field studies have been performed at ground-level [Bertman et al., 1995;Flocke et al., 1998;Roberts et al., 1998], the TRACE-P data set includes samples collected throughout the troposphere from 150 m to 12 km. Stroud et al. [2001] also measured alkyl nitrate mixing ratios in the lower and middle troposphere (0 -7 km). At present it is not possible to fully quantify the effect of altitude on alkyl nitrate evolution because k 4 values are not available for RONO 2 + OH reaction at temperatures colder than 290 K . By comparison, an average temperature of 230 ± 5 K was measured in the upper troposphere (9.5 -10.5 km) during TRACE-P http:// www-gte.larc.nasa.gov/trace/gte_mrg1.htm#TRACE-P). However, we can briefly consider the effect of altitude on the remaining kinetic parameters used to solve equation (9). The hydroxyl radical concentration and k 1 both decrease with increasing altitude (decreasing temperature), leading to slower alkyl nitrate production ( Table 2). Although the lower [OH] values at higher altitude also lead to slower alkyl nitrate destruction by OH, alkyl nitrate J-values increase with altitude (section 3.2). In this study J-values were calculated at 10 km under TRACE-P conditions, although the absorption cross sections were applied assuming no temperature dependence (Table 2). Because including temperature-dependent cross sections causes J RONO2 to increase more slowly with increasing altitude [Talukdar et al., 1997], the J-values calculated at 10 km are expected to be too large and the amount of alkyl nitrate destruction by photolysis at 10 km is likely overestimated. As stated above, temperature-dependent k 4 and J-values are required in order to quantify how the pseudo-first order rate constant for alkyl nitrate loss (k B = k 4 [OH] + J RONO2 ) changes with altitude.
[45] Because the rate constants k 1 and k 4 are known for surface temperatures, photochemical ages can be considered for air masses sampled near the Earth's surface in which the initial alkyl nitrate concentration is known to be zero. Although the Shanghai and Seoul plumes were sampled at low altitude, the initial alkyl nitrate concentration within Shanghai and Seoul was likely not zero (section 4.5.2). As a result, the age of these plumes cannot be accurately quantified based on RONO 2 /RH ratios. Further, in air samples collected over the western Pacific during the Pacific Exploratory Mission (PEM) West A, McKeen et al. [1996] found that dilution was so effective that hydrocarbon ratios should be used as photochemical markers only for hydrocarbons more reactive than n-butane (OH lifetime $ 3 days under PEM-West A conditions). Under TRACE-P conditions the lifetimes of 2-PeONO 2 and 3-PeONO 2 are 6 and 8 days, respectively, and the lifetime of their parent n-pentane is 3 -4 days. Therefore for TRACE-P, 2-PeONO 2 / n-pentane and 3-PeONO 2 /n-pentane are not good candidates for use as photochemical markers. Instead, the measured ratios of 2-PeONO 2 /n-pentane vs. 2-BuONO 2 / n-butane for the Shanghai and Seoul plumes have been superimposed on evolution curves that were generated under surface-level conditions only to illustrate the sensitivity of estimated photochemical ages to the amount of OH used in the model (Figures 5a -5b). Under an applied diurnally averaged [OH] of 8.8 Â 10 5 molec cm 3 , the predicted photochemical ages of the Shanghai and Seoul plumes are roughly 2 -3 days and 4 -5 days, respectively ( Figure 5a). When the applied [OH] is doubled to 1.76 Â 10 6 molec cm 3 , the estimated ages of the Shanghai and Seoul plumes are roughly 1-2 and 2-3 days, respectively (Figure 5b).
[46] Overall, the amount of photochemical processing within an air mass can be quantified using equation (9) but only under strict conditions. The modeled ages are highly sensitive to the amount of OH that is applied, and the model is limited to surface-level conditions. Although knowing the ventilation time of a city may allow the elapsed time (since the parent hydrocarbon was emitted from a polluting source) to be quantified, equation (9) only applies if the alkyl nitrate concentration prior to t = 0 was zero. In addition, the influence of dilution must be considered. Whereas sensitivity to initial conditions may not allow equation (9) to be applied quantitatively, it can be used qualitatively to study the relative extent of photochemical processing among different air masses, as discussed below.

Literature Comparison
[47] Alkyl nitrate evolution was modeled using equation (9) and the results were compared with observations from all TRACE-P flights. The newer branching ratios from the literature were used in this analysis. To satisfy the condition of zero initial alkyl nitrate concentrations, air samples displaying large oceanic sources of alkyl nitrates were not used. Because oceans are a major source of MeONO 2 , ratios of MeONO 2 /RONO 2 can be much higher in air masses that have recently been at low altitude over the Pacific Ocean compared to those that have been influenced by urban/industrial emissions [Blake et al., 2003]. Blake et al. used ratios of MeONO 2 /EtONO 2 , MeONO 2 /2-PrONO 2 , and MeONO 2 /2-BuONO 2 to distinguish between air masses influenced by equatorial oceanic sources and those influenced by northern hemispheric urban/industrial sources. In this study, as in the study by Blake et al., plots of MeONO 2 / EtONO 2 , MeONO 2 /2-PrONO 2 , and MeONO 2 /2-BuONO 2 showed clear marine and nonmarine ''wings'' (data not shown), and the marine wings were removed from the data set. As noted in section 4.1, the influence of the marine source on the TRACE-P alkyl nitrates was less prominent compared with their photochemical source.
[48] The ratio of each daughter alkyl nitrate to its parent hydrocarbon was plotted against 2-BuONO 2 /n-butane using the edited data (Figures 6a -6e). The ratio of 2-BuONO 2 / n-butane is chosen for the abscissa because n-butane is one of the most abundant hydrocarbons emitted and because 2-BuONO 2 is among the most abundant alkyl nitrates formed [Roberts et al., 1998]. In addition, 2-BuONO 2 is not formed in significant quantities from other decomposition pathways [Roberts et al., 1998]. The dark grey triangles in Figure 6 represent TRACE-P field data collected near the surface (0 -1 km), and the light grey squares show the remaining data collected between 1 and 12 km. Lower ratios (in the bottom left of each panel) represent fresh emissions, and higher ratios (in the top right of each graph) indicate photochemically processed air. For the surface data, individual points were grouped into bins each containing 10% of the edited data by sorting the data based on 2-BuONO 2 / n-butane ratios. An average ratio is plotted for each bin together with error bars that represent ±1s. The solid line in each panel is the modeled curve based on equation (9). As discussed in section 4.6.1, equation (9) can only be solved for surface-level conditions, and the modeled curves shown in Figure 6 represent surface conditions.
[49] The measured RONO 2 /RH ratios generally lie above the modeled line for C 2 -C 3 alkyl nitrates (Figures 6a -6c) and below the modeled line for C 5 alkyl nitrates (Figures 6d -6e). Bertman et al. [1995], Roberts et al. [1998], and Stroud et al. [2001] observed similar results during field studies in Pennsylvania and Alabama, in Nova Scotia, and over 29 -12 Colorado, respectively, with the exception of pentyl nitrate results from Pennsylvania which lied above the modeled curve, possibly as a result of chromatographic coelution problems. Whereas the C 2 -C 3 alkyl nitrates have shorter lifetimes than their parent hydrocarbons, the C 4 -C 5 alkyl nitrates have longer lifetimes than their parents [Roberts et al., 1998]. As a result, mixing with older air masses lowers the RONO 2 /RH ratio for C 2 -C 3 alkyl nitrates, and raises the ratio for C 4 -C 5 alkyl nitrates. However, similar to findings by Roberts et al. [1998], this is opposite to the overall effects that are observed in Figures 6a -6e, in which C 2 -C 3 results are larger than predicted whereas C 5 results are smaller than predicted.

GTE
[50] The average EtONO 2 /ethane vs. 2-BuONO 2 / n-butane ratios measured during TRACE-P differ from the modeled curve at short photochemical processing times and show some indication of leveling off at longer residence times, although the scatter in the data are too large to clearly show the approach to a steady state (Figure 6a). By comparison, the EtONO 2 /ethane vs. 2-BuONO 2 /n-butane ratios from Alabama were about a factor of 10 above the predicted line and were used as a strong indication of additional sources of EtONO 2 , such as the decomposition of larger alkoxy radicals to form ethyl radicals [Bertman et al., 1995]. Measured ratios from the Nova Scotia field study lied closer to the predicted line and tended towards it at longer photochemical processing times (higher ratios) [Roberts et al., 1998]. Roberts et al. suggested that the difference between the Alabama and Nova Scotia data sets likely arose because the Nova Scotia site (on the eastern edge of the continent) was more often influenced by aged air than the Alabama site. Older air could contain NO x levels that are below the threshold for RONO 2 production, and could mix in air masses with lower EtONO 2 /ethane ratios. The strong tendency of the TRACE-P data towards the predicted line at longer residence times suggests that these air masses were influenced by mixing with older air masses. The deviation from the line at lower ratios is consistent with an additional source of ethyl radicals, similar to the findings by Bertman et al. [1995] and Roberts et al. [1998]. The presence of additional EtONO 2 precursors is consistent with the unexpectedly high levels of acetaldehyde that were observed during TRACE-P (D. Jacob, Harvard University, personal communication, 2002).
[51] For the surface-level data the slope of 1-PrONO 2 / propane versus 2-BuONO 2 /n-butane agrees very well with predicted values, with a small (20 -30%) offset above the predicted curve (Figure 6b). The largest ratios (oldest air masses) lie closer to the predicted line than smaller ratios, again consistent with mixing with older air masses at longer residence times. However, as with EtONO 2 /ethane, the scatter of the data is too large to discern a statistically significant trend. By contrast to the TRACE-P results, Roberts et al. [1998] observed a factor of 8-10 offset for 1-PrONO 2 /propane vs. 2-BuONO 2 /n-butane, which they attributed to a possible local source of 1-PrONO 2 precursors such as 2-methyl pentane. Similarly, Stroud et al. [2001] observed a factor of 2 -10 offset above the predicted line. The TRACE-P field measurements do not show a similarly large offset from the predicted line, indicating that impact from additional 1-PrONO 2 precursors was not a large factor in air sampled downwind of Asia.
[52] The 2-PrONO 2 /propane vs. 2-BuONO 2 /n-butane surface-level data points show a small (25 -45%) offset above the modeled curve, whereas data collected between 1 and 12 km lie further above the modeled curve for younger air masses and closer to the line for older air masses (Figure 6c). This suggests that the air masses sampled between 1 and 12 km were influenced by additional sources of 2-PrONO 2 precursors, which could also help explain the high acetone levels that were observed during TRACE-P (D. Jacob, Harvard University, personal communication, 2002). We note that although the modeled curve was generated for surface conditions, it can be qualitatively compared to measured ratios at higher altitudes. For example, halving k A and doubling k B causes the modeled RONO 2 /RH ratios to shift to lower values but still along the same curve. Bertman et al. [1995], Roberts et al. [1998], andStroud et al. [2001] found that their field data lied above the predicted line by factors of 2-5, 2 -3, and 2 -4, respectively.
[53] In contrast to the C 2 -C 3 alkyl nitrates, the C 5 alkyl nitrate results generally lie below the predicted line, although the offset is small (5 -35%; Figures 6d -6e). Roberts et al. [1998] and Stroud et al. [2001] also found good agreement between measured and modeled results, with measurements $50% lower than the predicted curve. Based on similar findings from measurements in Alabama, Bertman et al. [1995] suggested that the chemistry of C 4 and C 5 alkyl nitrates is reasonably well understood. Here, the similar results obtained from independent groups sampling over different global source regions suggest that the photochemical sources of 2-butyl-, 2-pentyl-, and 3-pentyl alkyl nitrates are well represented in alkyl nitrate evolution models such as equation (9).

Conclusions
[54] The photochemical production and evolution of six C 2 -C 5 alkyl nitrates was investigated using selected data from 5500 whole air samples collected downwind of Asia. The airborne samples covered an altitudinal range of 150 m to 12 km and were collected during the NASA TRACE-P experiment (February -April 2001). 2-Propyl nitrate and 2-BuONO 2 were the most abundant C 2 -C 5 alkyl nitrates, indicating the importance of photochemical (rather than marine) alkyl nitrate production during TRACE-P. Individ- Figure 6. (opposite) Measured and modeled time evolution of ratios of daughter alkyl nitrates to their parent hydrocarbons versus 2-BuONO 2 /n-butane, for TRACE-P samples that do not show a marine alkyl nitrate source. (a) EtONO 2 /ethane; (b) 1-PrONO 2 /propane; (c) 2-PrONO 2 /propane; (d) 2-PeONO 2 /n-pentane; (e) 3-PeONO 2 /n-pentane. Dark grey triangles are individual TRACE-P samples collected near the surface (0 -1 km); light grey squares show the remaining samples collected between 1 and 12 km; black squares are average ratios each containing 10% of the nonmarine TRACE-P data (±1s for 2-BuONO 2 /n-butane in the x-direction, and ±1s for the plotted RONO 2 /RH in the y-direction). The solid lines show modeled evolution curves using kinetic data from the literature. ual alkyl nitrate mixing ratios ranged from clean marine to polluted continental, and mean alkyl nitrate mixing ratios were similar to those measured in and over rural areas of the United States.
[55] The TRACE-P field measurements were compared with model predictions of alkyl nitrate production and evolution. The photochemical production and destruction of 2-PeONO 2 and 3-PeONO 2 combine in such a way that the ratio of 3-PeONO 2 /2-PeONO 2 is predicted to increase exponentially with time. As a result, young unprocessed air masses were found to be the best candidates for comparing photochemical production rates of C 5 alkyl nitrates with laboratory predictions. In young, highly polluted air masses such as the Shanghai and Seoul plumes, the ratio of 3-PeONO 2 /2-PeONO 2 was 0.60-0.65. These results agree with predicted ratios from older laboratory kinetic studies (0.63 -0.66), and they lie in the lower end of the uncertainty range from the most recent laboratory kinetic studies (0.73 ± 0.08).
[56] Modeled photochemical ages based on RONO 2 /RH ratios are highly sensitive to OH concentration and are expected to be sensitive to altitude, although the dependence on altitude could not be quantified. During TRACE-P, the influence of dilution and non-zero initial alkyl nitrate concentrations precluded the use of RONO 2 /RH ratios as quantitative photochemical markers.
[57] The photochemical evolution of alkyl nitrates and their parent hydrocarbons was investigated using TRACE-P data that did not show influence from marine alkyl nitrate sources. Marine influence was determined using the ratios of MeONO 2 /RONO 2 , which can be much higher in air masses that have recently been at low altitude over the Pacific as compared to air masses influenced by urban/ industrial emissions. Evidence for additional Asian sources of EtONO 2 and 2-PrONO 2 precursors, beyond what is expected from ethane and propane chemistry, was found based on deviations of measured RONO 2 /RH ratios from modeled evolution curves. The additional EtONO 2 and 2-PrONO 2 precursors could help to explain the high levels of acetaldehyde and acetone that were observed during TRACE-P. The measured ratios of 1-PrONO 2 , 2-PeONO 2 , and 3-PeONO 2 to their respective parent hydrocarbons agreed fairly well with model predictions, suggesting little impact by additional precursors from Asian sources.