ObserVations of ozone and related species in the northeast Pacific during the PHOBEA campaigns the free troposphere to long-range transport from

. During late March and April of 1999 the University of Wyoming' s King Air research aircraft measured atmospheric concentrations of NO, 03, peroxyacetyl nitrate (PAN), CO, CH4, VOCs, aerosols, and J(NO2) off the west coast of the United States. During 14 flights, measurements were made between 39ø-48øN latitude, 125ø-129øW longitude, and at altitudes from 0-8 km. These flights were part of the Photochemical Ozone Budget of the Eastern North Pacific Atmosphere (PHOBEA) experiment, which included both ground-based and airborne measurements. Flights were scheduled when meteorological conditions minimized the impact of local pollution sources. The resulting measurements were segregated by air mass source region as indicated by back isentropic trajectory analysis. The chemical composition of marine air masses whose 5-day back isentropic trajectories originated north of 40øN latitude or west of 180øW longitude (WNW) differed significantly from marine air masses whose 5-day back isentropic trajectories originated south of 40øN latitude and east of 180øW longitude (SW). Trajectory and chemical analyses indicated that the majority of all encountered air masses, both WNW and SW, likely originated from the northwestern Pacific and have characteristics of emissions from the East Asian continental region. However, air masses with WNW back trajectories contained higher mixing ratios of NO, NOx, 03, PAN, CO, CH4, various VOC pollution tracers, and aerosol number concentration, compared to those air masses with SW back trajectories. Calculations of air mass age using two separate methods, photochemical and back trajectory, are consistent with transport from the northwestern Pacific in 8-10 days for air masses with WNW back trajectories and 16-20 days for air masses with SW back trajectories. Correlations, trajectory analysis, and com-parisons with measurements made in the northwestern Pacific during NASA's Pacific Explo-ritory Mission-West Phase B (PEM-West B) experiment in 1994 are used to investigate the data. These analyses provide evidence that anthropogenically influenced air masses from the northwestern Pacific affect the overall chemical composition of the northeastern Pacific troposphere.

off and approximately 30 min in an outbound ferry leg to the sampling region off the coast of Washington State. Flights were scheduled to maximize the amount of time spent in the experiment area during high Sun conditions. Local noon was at approximately 1230 LT. The outbound ferry leg was used primarily for instrument calibration and to confirm that all instruments were functioning normally. Sampling during each flight was done over the Pacific Ocean in 6 to 10 level flight legs of 20 min duration. There were roughly 5 rain between each leg, which were used to change heading and altitude in order to set up for the next sampling leg. Figure 1 depicts a typical flight track. The only exception to this standard flight track was on April 21 (flight 11), where the King Air flew south down the coast of Oregon and into northern California making measurements at only two altitudes. This flight was used to investigate PAN decomposition in a strongly subsiding air mass off the coast of northern California. After measurements were completed in the sampling region, the inbound ferry leg back to Paine Field was used for further instrument calibrations and to shut down the instruments prior to lan0ing.
Operational forecasting for King Air flights was conducted with the goal of sampling air off the coast of Washington State that had not recently (within 3 days) been exposed to continental influence. Primarily, this meant forecasting for westerly, southwesterly, or northwesterly winds throughout the lower troposphere in the region where sampling would take place. Additionally, flights were not scheduled when frontal systems were present. We avoided fronts in part to minimize the impact of stratospheric air on our measurements, which can occur during tropopause folds, but also to avoid air parcels in which rapid and unpredictable changes in vertical and horizontal motion occurred. Such unpredictability makes the determination of source regions through back trajectories more difficult. Several numerical models, run in forecast mode, were consulted and compared before conducting flights. Additionally, gridded data from atmospheric forecast models were used to generate forecasted back trajectories, and 24 to 72 hour forecasted back trajectories were used to estimate the likelihood that air parcels in the experiment area had avoided recent exposure to the North American continent.

Air Mass Trajectory Analysis
Back isentropic trajectories were calculated to determine the source regions for air masses encountered during aircraft sampling. Forward isentropic trajectories were used to estimate the impacts of air, originating over Asia, on various regions across the Pacific Ocean. Trajectories were obtained from two sources: the National Oceanic and Atmospheric Administration's (NOAA) Climate Monitoring and Diagnostics Laboratory (CMDL) using the European Centre for Medium Range Weather Forecasts (ECMWF) gridded data set and NOAA's Air Resources Laboratory (ARL) using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT4) model, 1997 (http://www.arl.noaa.gov/ready/ hysplit4.html) with the National Center for Environmental Prediction's (NCEP)"final model run" (FNL) meteorological data set. Trajectories of 10 day duration were available from NOAA CMDL starting every 12 hours, and those of 5 day duration were available from NOAA ARL starting every hour. We used NOAA CMDL as our primary source of trajectory data (exceptions are discussed below). There was generally good agreement between the CMDL and ARL trajectories.
Because of the different time duration of NOAA CMDL and NOAA ARL trajectories, 10 and 5 days, respectively, we use only 5 day trajectorie. s from both data sets for consistency. Aircraft data were separated into three categories based on source region as determined from 5-day back isentropic trajectories. The categories were (1) continental North America: those that crossed over the North American continent within 3 days, (2) westerly and northwesterly (WNW): those trajectories originating 5 days upwind north of 40øN latitude or west of 180øW longitude, and (3) southwesterly (SW): those trajectories originating 5 days upwind south of 40øN and east of 180øW. Back trajectories were calculated for each 20 min flight leg using the mean latitude, longitude, and altitude as the starting point. Most flights were conducted between 1700 UT and 2200 UT. However, only 1200 UT and 0000 UT trajectory starting times were available from NOAA CMDL. Therefore both the 1200 UT and 0000 UT back trajectories bracketing the measurement were used to classify the air mass source region. In general, both showed the same source region. Rarely, when differing source regions were indicated, a back trajectory from NOAA ARL was used to "break the tie" and classify the source region. The resulting classifications resulted in a total of 1 continental North American back tra- It is reasonable to assume that the majority of air masses in the WNW regime originated from the northwestern Pacific without experiencing North American influence. However, because trade winds can advect air from North and Central America into the tropical and extratropical North Pacific, this may not be a reasonable assumption for air masses in the SW regime. In order to address this, back isentropic trajectories were calculated starting at 25øN, 140øW, at various altitudes, and over the period of the PHOBEA aircraft experiment. The location 25øN, 140øW represents the average 5 day origin in longitude and the most southerly origin in latitude of air masses with SW back trajectories. Figure 3 shows back trajectories, starting at 25øN, 140øW, 5 km, and for the duration of the PHOBEA aircraft experiment. While we present only trajectories arriving at 5 km, the vast majority of trajectories arriving at altitudes from 1-7 km had a westerly track indicating eastward transport from the northwestern Pacific.
We also calculated forward isentropic trajectories from four Asian cities, to determine their impact on the two regions of the eastern Pacific representing the source regions of the aircraft data. We chose four Asian cities spanning a range of latitudes from 20ø-40øN; they are Hong Kong (22.18øN, 114.23øE), Shanghai (31.25øN, 121.43øE), Tokyo (35.75øN, 139.75øE), and Beijing (39.92øN, 116.33øE). Trajectories were started above these cities at an elevation of 3 km, at 12 hour increments, and for the duration of the PHOBEA aircraft experiment. We calculated the frequency of 5 day forward isentropic trajectories from each of these cities ending in two regions of the eastern Pacific. The two regions were defined by boxes (1) 180.0øW, 124.4øW, 60øN, and 40øN, which represents the source region for WNW back trajectories and (2) 180.0øW, 124.4øW, 40øN, and 25øN, which represents the source region for SW back trajectories. The percentage of 5 day forward trajectories ending in these regions is given in Table 1. These data indicate that the WNW region is generally twice as likely to be influenced by 5 day old air from these cities.

Sources
Because the goals of the PHOBEA experiment are to investigate the chemistry and chemical composition of the remote troposphere, after completing the initial data processing and quality control for each instrument, the entire data set was further quality-controlled to exclude data where there was a reasonable possibility of contamination from local pollution sources. There were two sources of local pollution; exhaust plumes from local shipping and instances when continental air from North America passed over the experiment area.
Ship exhaust plumes were encountered in the boundary layer during flights 3, 5, and 6. These plumes were easily identified in the data record as large spikes in NO (of the order of 50 to 200 parts per trillion by volume (pptv) above the background) and elevated aerosol concentration. In each case of contamination a nearby ship was observed either visually or by aircraft radar. To remove the ship plume contamination from the data record, boundary layer data from flights 3, 5, and 6 were excluded from the analyses presented here unless NO was measured at the background level of 0-20 pptv and contaminated air had not been encountered for at least 2 min.
The influence of continental North American air, as a source of contamination of PHOBEA aircraft data, was determined from an analysis of back trajectories, NO, and aerosol data. This analysis indicated only one flight leg with evidence of continental North American influence.
This occurred on April 16 (flight 9) in the boundary layer where the winds were southeasterly, NO mixing ratios averaged 110 pptv, aerosol concentrations were elevated, and the back tra-  Figure 4. Box plots of ozone made from 1 min averages of the aircraft data. The effects of stratospheric air during flights 3 and 14 are indicated by plotting ozone data both including and excluding these flights. jectory crossed over North America within 24 hours. Hence, in the subsequent analysis, boundary layer data from flight 9 were excluded. In all remaining flight legs, winds ranged from northwesterly to southerly and had back trajectories indicating eastward transport from the remote Pacific (see Figure 2). Air masses originating from the remote Pacific were typically either maritime polar with westerly or northwesterly back trajectories, or maritime tropical with southerly or southwesterly back trajectories.

Stratospheric Influence
There are two main sources of ozone in the troposphere; in situ production and transport from the stratosphere. A method of identifying intrusions of stratospheric air into the troposphere has been presented by Cooper et al. [1998]. They used enhanced satellite water vapor images to identify intrusions of dry air during postfrontal conditions and associated these dry air intrusions with high concentrations of ozone of stratospheric origin. They found that the ozone slope with height during postfrontal conditions was 6 to 12 ppbv km -1 in the lowest 8 km of the troposphere, whereas the prefrontal ozone slope averaged 2 ppbv km -• In order to identify instances of direct stratospheric influence in the PHOBEA aircraft data, we performed a similar analysis. We found that during two flights, April 3 and 28 (flights 3 and 14), the ozone slope with height exceeded 6 ppbv km -•' these two flights were also in the vicinity of significant dry air intrusions identified in satellite water vapor images. Ozone mixing ratios on these flights reached 311 ppbv and 132 ppbv, respectively, with very low mixing ratios of water vapor and other measured species.
All back trajectories from flights 3 and 14 fall in the WNW category. Figure 4 shows box plots of ozone mixing ratio with height, grouped in 2 km bins from 0 to 8 km. Box plots of WNW data, both including and excluding flights 3 and 14, are presented. The box plots indicate significant ozone enhancement from 4-8 km when data from flights 3 and 14 are included. However, there is no significant enhancement in ozone from 0-4 km when flights 3 and 14 are included. From this analysis we concluded that data from flights 3 and 14 between 4-8 km are significantly influenced by stratospheric air and excluded those data in the subsequent analyses where we compare WNW with SW air masses.

Data Averages
Each instrument on board the King Air aircraft, in general, made measurements at a different frequency. The frequency of measurements ranged from many times a second (e.g.,     aWNW and SW data have been screened to remove local boundary layer pollution cases (99 out of a total 2236 min) and stratospheric intrusions (144 out of 2236 min) as described in the text. Nmd are the number of statistically independent measurements' N=n are the number of measured 1rain averages. Mixing ratios are in pptv unless otherwise noted. All aerosol measurements were adjusted to their values at STP. bNOx is calculated from the Leighton relationship as described in the text. temperature), to once every 5 min (e.g., PAN), to a few times over the course of a flight (e.g., VOCs). These were consolidated into a master file of 1-min averages. We present in Table 2 a summary of all marine data (i.e., excluding the one continental North American air mass) and two subsets of the marine data: WNW and SW where both subsets were screened to remove local pollution and stratospheric sources as discussed above. Unless otherwise noted, mixing ratios in Table 2 are

Comparing WNW With SW Data
The WNW and SW data are compared by evaluating the level of significance (LOS) between the average mixing ratios at the same altitude. The LOS was computed from equations The LOS values listed in Table 3 indicate the confidence level at which the difference in average mixing ratio becomes significant (e.g., a LOS equal to 0.95 would indicate a difference in average mixing ratio between species in the SW and WNW regimes significant at the 95% confidence level).
Missing LOS values indicate less than two independent measurements in either WNW or SW regimes. The LOS averaged over all altitudes is also presented. It is important to note that an LOS analysis assumes that the data are normally distributed. For those data that were significantly skewed, LOSs were also computed from the lognormal distributions. However, the LOSs from the lognormal distributions did not differ significantly from those listed in Table 3, therefore we report only LOSs from the measured distributions. Numerous species show significantly enhanced mixing ratios in WNW back trajectories when compared to SW back trajectories. We can explain these enhanced mixing ratios by examining the factors which influence the concentrations of species measured. These factors are as follows' the back-ET AL.: AIRBORNE OBSERVATIONS DURING PHOBEA  i-butane, n-butane, PAN, and aerosol CPC) have a high average LOS between mixing ratios measured in the WNW and SW regimes, WNW mixing ratios being significantly higher. However, the average LOS for MeI was low, reflecting that its primary source is oceanic. The average LOSs for i-pentane, n-pentane, and ethene were also low, reflecting that their lifetimes are short (-1-3 days assuming an average OH con-centration of lx10 6 molecules cm -3) compared to the others and are significantly shorter than the average transport time across the Pacific (--10 days). Hence the lifetime is short enough that Asian continental emissions likely do not survive transport in either WNW or SW regimes to any great degree. The only remaining species, benzene, had a low average LOS, which was inconsistent with the high LOSs of the other anthropogenically produced trace gases of similar lifetime. PAN mixing ratios are higher in the WNW regime due to lower temperatures that result in slower losses, and high PAN may also explain the elevated NOx mixing ratios in the WNW regime. Aerosol number concentrations were also somewhat higher in WNW trajectories. SW back trajectories are likely more remote from anthropogenic sources, hence there is a longer time period for cloud and depositional processes to reduce the aerosol number concentration relative to air masses with WNW back trajectories.

Enhanced mixing ratios in WNW trajectories of those species with a high LOS cannot be a priori attributed to East Asian emissions. These higher mixing ratios likely reflect the combined influences of latitudinal gradient and faster transport of Asian continental emissions to the WNW region.
Apart from the patterns in LOS discussed above, moderate to strong differences were also seen in NOx (average LOS equal to 0.75), ozone (average LOS equal to 0.96), and J(NO2) (average LOS equal to 0.73).

NOx mixing ratios were higher in air masses with WNW back trajectories. NOx has its primary source in urban areas and has a lifetime of less than 1 day in the lower troposphere and 4-7 days in the upper troposphere [Brasseur et al., 1999].
Therefore it is unlikely that direct anthropogenic emissions from the East Asian region influenced our measurements in the lower troposphere. As mentioned above, higher PAN mixing ratios in the WNW regime likely play some role in elevating NOx mixing ratios. However, while we have attempted to screen our data from the direct effects of ship emissions, the effects of older diffuse emissions in our boundary layer data cannot be discounted [Corbett et al., 1999].
Ozone mixing ratios were significantly lower in air masses with SW back trajectories. Ozone production rates should be lower in the SW regime due to lower NOx mixing ratios, and ozone destruction rates should be higher due to increased actinic fluxes at lower latitudes and higher water vapor concentrations. Additionally, the SW region is less subject to intrusions of ozone from the stratosphere [Elbern et al., 1998]. While we have attempted to remove the effects of recent stratospheric intrusions, we cannot account for older intrusions that have become well mixed into the background. These likely play some role in enhancing ozone in the WNW regime.

J(NO2) was higher in air masses with SW back trajectories because synoptic conditions leading to SW flow were generally associated with high pressure in Oregon and northern
California and clearer skies, particularly at lower elevations. J(NO2) also tended to increase with height during all daytime flights mainly due to the effects of low-level clouds.

al. derived equation (4) as the relative photochemical age tr of an air mass tr = {ln([A]/[B]) -ln([A]o/[B]o) } / {(k•-kA)[OH]average} (4) where [A]/[B] is the measured ratio of hydrocarbons A and B, [A]0/[B]0 is the ratio of A and B in the source region, ka and k• are the second order rate constants for the reaction of A and B with OH, respectively, and [OH]•ver,ge is the average OH concentration encountered by the air mass.
We calculated tr for each VOC sample using the measured ratios of propane/ethane and ethyne/CO during PHOBEA and by using the following assumptions; source region concentrations were the average Ineasured ratios from 0-2 km of propane/   Nonetheless, if the average age is calculated for each trajectory regime, then there is good agreement between the two methods. The average age estimated by each method and delineated by trajectory type is listed in Table 4. Each calculation estimates the average age to be roughly 8-10 days for air masses with WNW back trajectories and 16-20 days for air masses with SW back trajectories.  Figure 6) a latitudinal gradient of higher mixing ratios poleward is evident for CO, ethane, and propane in the Asian continental outflow. However, it is interesting to note that the median PAN mixing ratios were higher between 15øN-40øN than at more northern latitudes. This likely indi-cates a large source region for PAN between 15øN-40øN relative to higher latitudes.

Comparing PHOBEA Data With That From NASA's PEM-West B Experiment
The bowed shape of PAN, CO, ethane, and propane profiles in Figure 6 is consistent with long-range transport of anthropogenic trace gases from the northwestern Pacific. We surmise that the vertical profile of these trace gases in the springtime 1999 Asian continental outflow is at least qualitatively similar to that in 1994. Once boundary layer air mixes into the free troposphere over the northwestern Pacific, stronger westerly winds above the boundary layer can transport it more rapidly to the northeastern Pacific. The highest mixing ratios of these gases during PHOBEA were measured between 2-6 km. PHOBEA WNW mixing ratios from 0-2 km were lower than from 2-6 km because removal rates of these compounds are faster in the boundary layer. In addition, air masses measured between 0-2 km were slower in transit across the Pacific than those between 2-8 km. PHOBEA WNW mixing ratios from 6-8 km are also lower than those from 2-6 km likely because fewer air masses have mixed up to that height in their transport across the Pacific. Table 5 compares the integrated column concentrations from 0-7 km of PAN, CO, ethane, and propane measured in the Asian continental outflow during PEM-West B and in the northeastern Pacific during PHOBEA. When comparing these data sets, it must be noted that they represent different years, 1994 versus 1999, and different latitude ranges. Each of these compounds is known to have had a strong latitudinal gradient, and the emissions rate of these compounds is likely to have changed somewhat over the 5 year time frame. Having said that, the column concentrations of these four species were in general larger during PEM-West B than in PHOBEA, which is consistent with photochemical removal and dilution with cleaner air masses during long-range transport. Table 6 presents the linear correlation matrix (r values) for selected species and includes those with a LOS greater than 0.9. In general, the correlations between VOCs are good, but they are poorly correlated with water vapor and ozone. Water vapor is primarily affected by air and sea surface temperatures, which increase with decreasing latitude, and ozone is a secondary pollutant, which is not expected to correlate with primary pollutants far from their source regions. Hence the poor correlation of VOCs with water vapor and ozone is not surprising.

Dominant Correlations
The other major correlations were between aerosol scattering and absorption (but not with number concentration, which is not shown), with correlation coefficients ranging between 0.88 and 1.0. Aerosol scattering and absorption were also loosely correlated with PAN (r values ranging between 0.6 and 0.7). However, aerosol scattering and absorption were uncorrelated with the VOCs listed in Table 6. PAN, on the other hand, showed a moderate correlation with the VOCs (r values ranging between 0.6 and 0.7).

Implications of Transpacific Pollution Transport
In a previous publication we have shown that episodic transport events can bring enhanced concentrations of industrial pollutants from Asia to the northeastern Pacific within 5 days . In this report we use the PHOBEA aircraft data to show that industrial emissions from Asia influence the entire lower troposphere in the northeastern Pacific, not only in episodic transport events, but also in the background concentrations of ozone and other species ob-

served. Guderian et al. [1985, and references therein] have
shown that ozone mixing ratios as low as 35 ppbv can damage florae in controlled studies and that the damage to plant growth increases linearly with increasing ozone. The average marine mixing ratios of ozone in the PHOBEA aircraft measurements were 47 ppbv in the 0-2 km layer and 60 ppbv in the 2-4 km layer. Air masses at these altitudes likely encounter the florae of the Pacific northwest, where the highest mountain peaks can extend above 4 km (e.g., Mount Rainier at 4.4 km) and high alpine forests can range between 2 and 3 km. Hence it is possible that marine ozone from the northeastern Pacific has a detrimental impact on the biosphere of the northwestern United States and Canada. However, the issue is complicated because alpine florae may be adapted to higher ozone concentrations and the largest ozone impact on plants occurs during the summer growing season, whereas our measurements were in the spring. A modeling study by Berntsen et al. [1999] has shown that about 4 ppbv (roughly 10%) of the ozone in air arriving to North America during spring is due to current emissions of NOx from Asia. As a result of continued rapid growth in NOx emissions from China [Streets and Waldhoff, 2000] ozone in the northeastern Pacific is predicted to continue increasing [Berntsen et al., 1999;Jacob et al., 1999]. Other modeling studies have suggested that PAN transported from continental regions provides a significant contribution to NOx in the northeastern Pacific through PAN's thermal decomposition [Moxim et al., 1996;Horowitz and Jacob, 1999]. NOx from PAN would help sustain higher ozone concentrations in air masses transported across the Pacific. In future publications we will employ a photochemical model constrained with the PHOBEA data to estimate the ozone photochemical tendency in the northeastern Pacific and explore the impacts of PAN thermal decomposition and increasing Asian NOx emissions on ozone in the northeastern Pacific.

Concluding Remarks
In this paper we have presented the most comprehensive suite of aircraft measurements of trace gases and aerosols in the northeastern Pacific troposphere to date. The data are consistent with transport from the northwestern Pacific in 8-20 days. The impacts of this transpacific transport extend throughout the 0-8 km column.
After screening the data to remove local pollution and two cases of direct stratospheric input, the marine data were segregated into two regions based on 5 day back isentropic trajectories. The majority of these back trajectories, 71 of 93, originated north of 40øN latitude or west of 180øW longitude (termed westerly and northwesterly, WNW), and the remainder originated south of 40øN latitude and east of 180øW longitude (termed southwesterly, SW). The relative number of air masses with 5 day origin in the WNW and SW regions were consistent with climatology. Further trajectory analyses indicated that the majority of air from both regions was likely to have originated from the northwestern Pacific, without recent exposure to the North American continent.  aThe r values in parentheses are those whose absolute value is greater than or equal to 0.80. The linear correlations were determined from leg averages, but excluding instances where the data was contaminated by local pollution or stratospheric sources (see text). STP) through the inlet. The NO instrument smnpled air prior to the metal bellows pump at ambient pressure. The metal bellows pump supplied air to the Dasibi, and a cabin vent in front of the Dasibi insured that the instrument was at cabin pressure.

The Dasibi was calibrated in the laboratory against a CSI
Photocal 3000 ozone generator prior to installation on the plane; and calibrated again on board the aircraft through the sampling inlet. The two calibration slopes differed by 4%, indicating a 4% loss of ozone when sampled through the aircraft inlet and bellows pump. The detection limit of the Dasibi during these flights was about 2 ppbv; the overall uncertainty of the data was of the order of 1 ppbv.

A2. NO
The NO instrument was constructed and tested in our research laboratories at the University of Alaska-Fairbanks and the University of Washington-Seattle. The measurement of NO used the principle of chemiluminescence; ambient NO was reacted with excess ozone in a low-pressure dynamic flow reaction chamber, and photons from decay of the product, NO2*(2B•), were detected by a photomultiplier tube [Kley and McFarland, 1980;Drummond et al., 1985;Ridley and Grahek, 1990].
NO was quantitatively measured by switching between four modes of operation: measure, zero, calibration, and calibration zero. A 20 min measurement cycle was used during the experiment to coincide with the flight legs. It consisted of 9 min for measurements, 6 min for zeroing, and 5 min for calibration.
Air at ambient pressure was sampled from the 3/8 inch D Teflon sample line (as described above) and entered the NO instrument through a 1/4 inch D Teflon tube. The sample then passed through a 47 mm Teflon filter holder containing a 1.0 gm Teflon filter. The air sample was mass flow controlled at 1 standard leter per minute (sLmp). Upstream of the mass flow controller the airstream was at ambient pressure, and downstream it was 82 torr. The sample then passed over a water vapor addition system. This system was made from a 47 mm Teflon filter holder that was used as a 30 mL water reservoir and heated to 70øC. The filter holder contained a 1.0 gm Teflon filter that allowed water vapor to pass but isolated liquid water from the sample stream. Water vapor addition was approximately 1 g h -•. The humidified sample stream then entered a zeroing volume consisting of a 4 foot segment of 3/8 inch D Teflon tubing with a total volume of 86.9 mL. Subsequent to the zeroing volume, the sample stream entered the reaction chamber and mixed with -4% 03 in 02 from the ozone generator. The reaction chamber was based on the design of Ridley and Grahek [1990], only differing slightly from their design in its mounting to the PMT coeler. The Ridley and Grahek design was simplified by eliminating one flange, resulting in a closer fit and one less window. The reaction chamber was temperature-controlled at 30øC. During operation the measured reaction chamber pressure was between 5 and 6 torr, which compared reasonably well with expectations (4 torr) given the 200 Lpm pumping speed of the vacuum pump and some loss of efficiency due to tubing connections. The sample stream exited the reaction chamber through a 1 inch D stainless steel bellows tube connected to a two-stage vacuum pump equipped with a reactive gas trap filled with Hopcalite catalyst for ozone removal. The pump used Fomblin perfluorinated pump oil.
A red RG 610 Schott filter was placed between the PMT and the reaction chamber, blocking wavelengths shorter than 610 nm. The PMT was a dry ice cooled, red-sensitive, Hamamatsu R-1333. The signal from the PMT was amplified by a photomultiplier amplifier discriminator, converted to a transistor-transistor logic (TTL) signal, and read by a PCM-DAS16D/16 data acquisition card installed on a laptop computer. The computer read the count rate every 0.1 s, accumulated 10 readings, and output the signal as counts per second.
The ozone generator used a high-voltage discharge corona to generate roughly 4% 03 in 02. The pressure inside the ozone generator was kept at approximately 1 atm, and it received a flow of 0.200 sLpm 02 maintained by a mass flow controller. The bulk of the 03 in 02 mixture entered the NO instrument and was directed into the reaction chamber.
The signal was zeroed in "zero" and "calibration zero" modes by diverting roughly 5% of the flow from the ozone generator into the sample stream just prior to the zeroing volume, with the remaining 95% of the ozone flow continuing to the reaction chamber. Zeroing efficiencies between 95 and 99% were achieved.
Calibrations were performed with a NO calibration standard from Scott-Marrin Inc (1.001 ppmv + 2%, NO in N2).

During calibrations a mass flow controller delivered 4 cm 3 min -• STP of NO in N2 to the sample stream.
For safety considerations the NO instrument could not run continuously between flights on board the King Air. Installing and removing the instrument from the plane before and after each flight was also not practical. Hence it was necessary to restart the instrument before each flight. A warm-up time of 3 hours was sufficient for zero mode signal stabilization. However, the sensitivity of the instrument to NO took much longer to stabilize. Generally the sensitivity did not stabilize for the duration of a flight. Hence we measured the sensitivity frequently, once during each flight leg, and the sensitivity at any given time was determined by linear interpolation between each calibration. The sensitivity generally increased throughout a flight and ranged between 3.5 and 4.5 counts per second (cps) per pptv NO for a 5 hour flight. On the basis of laboratory tests the sensitivity stabilized after 12-24 hours of operation.
Following the 3 hour warm-up, the zero level was stable, except it tended to increase as a function of altitude due to the increasing effects of cosmic rays on the PMT. The increase was typically 10% from sea level to 8 km.
The NO instrument signal was quality-controlled to remove anomalous spikes in the data caused by interference from the aircraft's conununications system and corrected for instrument switching response time constants, ambient humidity fluctuations, and NO loss in the sample lines due to reaction with ambient ozone. The mstrument's zero signal and sensitivity showed no dependence to variations in ambient pressure. However, ambient humidity changes, while mitigated by the use of a water vapor addition system, still had a small effect. The effects of ambient humidity fluctuation on zero signal and sensitivity were characterized in our laboratory at the University of Washington. Corrections to the NO data due to humidity fluctuations were small, generally less than 1%. Signal spikes caused by the communications system accounted for less than 1% of the overall data; the other corrections were of the order of 5% or less in total. Extensive leak testing of the instrument and sample lines revealed no detectable cabin air leaks into the instrument. Determining the presence of signal artifacts is critical for the measurement of NO when its concentration is low [Drummond et al., 1985]. Signal artifacts are an instrument zero offset, most easily detected as nonzero measurements in the absence of analyte. Artifacts can be caused by, among other things, interfering compounds which chemiluminesce, significant fluctuations in ozone concentration from the ozone generator, and cabin air leaks into the sampling lines.

KOTCI-t]ENRUTtt
There are two standard methods to monitor instrumental artifacts. Because NO is quickly converted to NO2 by ambient ozone in the absence of light, the most reliable method is to measure NO at night in a region remote from sources of NO. Any significant deviation from 0 pptv likely reflects instrumental artifacts. A less reliable method is to measure the NO concentration in cylinders of high-quality synthetic zero air. Zero air cylinders, however, may contain mixing ratios of up to a few pptv NO, thus making these tests somewhat less conclusive.
Artifact tests were made during one night flight and with zero air measurements before each flight. The mean NO mixing ratio measured on the night flight was -0.1 + 1.6 pptv (mean plus or minus one standard deviation, N = 63 one-min averages). Figure 7 depicts NO measurements during the night flight. Zero air measurements had more variability than the night flight measurements, but also centered around 0 pptv NO. From 16 measurements made throughout the aircraft campaign, the mean NO mixing ratio in zero air was -0.3 + 2.9. On the basis of these tests the NO instrument was considered free of significant artifacts during this experiment.
The 2c• detection limit of this instrument was on average 2 pptv NO for a 9 min signal integration time and 4 pptv for a 1 min signal integration time. The instrument precision was on average +4 pptv at the 95% confidence interval for a 1 min signal integration time. An ambient sample was injected every 5 min onto the precolumn by switching a 10 port valve (Valco, model W). By using a second 4 port valve (Valco, model W), the first 45 s of the precolumn eluent were sent to a bypass channel to elute the (oxygen) solvent peak which tends to overload the detector. At the same time, pressure fluctuations due to sample injection were abated. Forty-five seconds into a run the 4 port valve was switched, and the precolumn and analytical column were then in series. After 2.2 min into the run, the 10 port valve was switched back to load a subsequent ambient sample in the sample loop, while at the same time the precolumn was backflushed to remove higher boiling compounds.
A Campbell Scientific Inc. (CSI) 21X data logger controlled the instrument and acquired the data. All instrument signals were stored in a CSI storage module for download to laptop computer at the completion of each flight. In-flight span/calibration of the gas chromatographic (GC) system was made every 20 min, during each sampling leg, using a commercial PAN calibrator (MetCon). Multipoint calibrations were also performed on the ground using the same instrument. The calibrator generated PAN from the photolysis of acetone mixed with NO in air. The detection limit was linearly dependent on ambient pressure and was determined from laboratory measurements to be 15 pptv at 760 torr and 45 pptv at 400 torr. PAN measurements below the detection limit of the instrument were set to one half the detection limit for statistical calculations.
The PAN instrument precision and accuracy were also pressure-dependent. At 760 torr the precision was better than 50% at 20 pptv and better than 10% at 40 pptv. However, because of a pressure correction factor, precision was worse at higher altitudes. For example, this correction factor was 3 at 6.5 km, resulting in a precision which was better than 50% at 60 pptv at that altitude. Taking all uncertainties into consideration (including flow controller calibrations, standards, efficiency of PAN production in the calibrator unit), the accuracy was estimated at 30% below 100 pptv at 760 torr, with a similar scaling as the precision at lower pressures.

A4. CO, CH4, and VOCs
A total of 46 pressurized whole air samples were collected in stainless steel canisters for GC-flame ionization detection and electron capture detection analysis of CO, CH4, and VOCs [Blake et al., 1994]. Air samples were drawn from a 1 cm D rearward facing stainless steel inlet, then through 2 m of stainless steel tubing and a stainless steel metal bellows pump at a pumping speed of roughly 15 Lpm (at sea level). The sample lines on the high-pressure side of the bellows pump were pressurized and purged several times before filling the canisters to between 1.38x105-2.76x10 s Pa. An average of three samples were taken per flight.
Two of the forty-six whole air samples contained anomalously high VOC levels, several orders of magnitude higher than all other samples. These two samples were inconsistent with other measurements made at the same time (e.g., NO and aerosols), which indicated a relatively clean air mass. We assumed these two canisters became contaminated during handling and were excluded during our analysis. Light absorption at 550 nm (Oap(550)) was measured with a differential transmission absorption photometer (Radiance Research, model PSAP). This device was calibrated by the manufacturer, and calibration corrections, including a correction for response to light scattering, were made as recommended by Bond et al. [1999].

AS. Aerosol
The number density of particles larger than 10 nm diameter was measured with a condensation particle counter (CPC; TSI, Inc., model 3010). The PSAP and CPC measurements were made immediately downstream of the nephelometer. Upstream of the nephelometer was a ball valve and high efficiency particulate air (HEPA) filter arrangement that allowed manual switching to filtered, particle-free air. All aerosol concentrations have been corrected to STP.

A6. J(NO2)
The NO2 photodissociation frequency was measured utilizing two J(NO2) radiometers (MetCon). Each radiometer provided a nearly isotropic response over a solid angle of 2rr steradians and used a Schott 2xUG3, UG5 filter combination.
The theory behind these measurements is described by Junkermann et al. [1989]; Volz-Thomas et al. [1996] estimated the uncertainty of a similar instrument to be 8% under clear-sky conditions. J(NO2) data measured with radiometers are subject to three corrections; for spectral response, angular response, and temperature. The spectral correction is small, but depends on altitude and solar zenith angle. Discrete correction values from the manufacturer were fitted to a polynomial expression allowing for correction at all altitudes and zenith angles <80 ø . This correction was of the order of 0-5%. An angular correction is necessary at zenith angles larger than 80 ø [Volz-Tho-