Observations of ozone and related species in the northeast Pacific during the PHOBEA campaigns: 1. Ground‐based observations at Cheeka Peak

,


Introduction
Ozone plays a key role in tropospheric chemistry. It is the primary source for the OH radical which is the most important oxidant in the troposphere. It is also a greenhouse gas and has toxic and phytotoxic effects at levels not far above ambient. Tropospheric ozone has two major sources: stratospheric intrusions and photoche•nical production [Crutzen, 1988]. The relative contributions of these two sources remains an open question. Photochemical ozone production in the troposphere requires nitrogen oxides, hydrocarbons, and light. In most regions of the troposphere the limiting precursor for ozone production is NO.• (NO + NO2). As a result of anthropogenic emissions of NO,, tropospheric ozone has substantially in-In this paper we examine the Cheeka Peak data from the springs of both of 1997 and 1998 to identify the sources, patterns, and relationships among the measured species. In a separate paper we present the aircraft data taken in the same region during the spring of 1999 [Kotchenruther et al., this issue], and in a future paper we will present the results of an analysis of the NOx and 03 photochemistry for the northeastern Pacific.

Experiment
We conducted measurements in two successive spring campaigns, March-April 1997 and 1998, at the Cheeka Peak Observatory (CPO) in Washington State (48.3øN, 124.6øW, 480 m above sea level (asl)). This site is located on the northwestern tip of the Olympic Peninsula immediately adjacent to the Pacific Ocean (see Figure 1). During westerly flow the air masses that arrive at the site are generally unaffected by recent North American emissions and are characteristic of the North Pacific atmosphere [Anderson et al., 1999]. Easterly winds bring air from the more polluted continental boundary layer. The CPO site has been used for atmospheric chemistry and aerosol research for approximately 15 years [Anderson et al., 1999, and references therein]. During these two campaigns we measured 03, NOx, PAN, CO, J(NO2), NMI-tCs, radon, aerosol number density, aerosol absorption, and aerosol light scattering, along with meteorological parameters. Back isentropic trajectories [Harris et al., 1992] were calculated twice each day (0000 and 1200 UTC) for the period of the two campaigns, using data from the European Centre for Medium-Range Weather Forecasts (ECMWF). While our previous work with isentropic trajectories has given us much useful information regarding transport to CPO [e.g., Jaffe et al., 1999] it should be kept in mind that there are a number of possible sources of error associated with isentropic trajectories, including low-resolution meteorological data, nonisentropic transport, and/or subgrid-scale vertical motions. Possible trajectory errors have been discussed in several recent publications [e.g., Kahl, 1996;Stohl, 1998].  [Anderson et al., 1999]. To minimize the admission of seasalt into the instruments for CO, 03, PAN, and NO,• instruments, we added a 2 gm Teflon filter to each inlet, effectively eliminating all particles down to sub-micron sizes. For NOx we used a high-sensitivity chemiluminescence instrument built in our laboratory and previously described by Beine et al. [1997]. NO was detected directly. NO2 was detected as NO following UV photolysis. One instrument cycled between NO and NO2 modes. In addition, during the 1997 campaign we used a second high-sensitivity NO instrument. This allowed us to conduct an informal intercomparison of our NO results. Because our NO,• measurement techniques have been described previously [Beine et al., 1997], only an overview is given here. The flow through the reaction chamber was set to 1 standard liter per minute (sLpm). The NO chemiluminescence detector was automatically calibrated once every 3 hours, except for the first part of the 1997 campaign (March 9 to March 21), when it was calibrated once every 6 hours. For calibration we used a flow of 5 cm 3 min -• STP of 3.00 ppmv NO in N2 (Scott-Marrin, Inc. 3.00 + 0.06 ppmv NO) as a standard addition to the ambient flow. After the campaign, the calibration gas was cross-calibrated against a National Institute of Standards and Technology (NIST) standard (NO in N2, 4.73 + 0.07 ppmv), which led to a correction of the Scott-Marrin standard mixing ratio to 3.21 +_ 0.09 ppmv NO. The median NO sensitivities for the 1997 and 1998 campaigns were 4.23 and 3.84 counts per second (cps)/pptv NO, respectively. The 3 c5 detection limits for NO for the i-minute averages for 1997 and 1998 were 5.37 pptv and 6.09 pptv, respectively, and 1.79 pptv and 2.03 pptv in the hourly averages (9 one-minute measurements of NO and NO2 per hour). The median NO2 sensitivity for the whole campaign was 1.21 cps/pptv NO2 and 1.19 cps/pptv NO2 for 1997 and 1998, respectively. The 3 c5 NO2 detection limits for the one-minute averages were 18.76 pptv NO2 and 19.49 pptv for 1997 and 1998, respectively, and 6.25 pptv and 6.50 pptv for the hourly averages in 1997 and 1998, respectively. The overall uncertainties for NO and NO2 were 10% and 25%, respectively, at levels well above their detection limits.
During the 1997 campaign a second NO instrument (called "NO-II") was used to provide NOy calibrations for the PAN instrument and to provide an additional quality control and intercomparison with the primary NOx instrument. This NO instrument was very similar to the primary NO chemiluminescence detector and shared calibration gas with the primary instrument. The second instrument spent most of its time measuring ambient NO; it was switched manually to NOv mode only for calibrations. Thus, for most of the 1997 campaign, we have duplicate NO measurements. We compared the hourly averaged NO mixing ratios at all times where each instrument reported data and found a slope of 0.983 (for NO-II response divided by the primary NO response) with an Rsquared value of 0.978, N:1032 for the full data set. Because each instrument ran on separate measure-zero cycles, which were not synchronized, there are some differences at high levels when there was more ambient variability. Comparing only the data below 75 pptv, we get a mean, median, and root mean square difference (NO primary minus NO-H) of-4, -4, and 6 pptv, with a R 2 for the regression of 0.91. The difference at low levels is mostly due to a positive offset in the NO-II instrument, which is apparent from the nighttime NO data. The data used in this paper are from the primary NO instrument which exhibited no nighttime offset. CO was measured with a commercial nondispersive infrared absorbance instrument (API-300, Advanced Pollution Instruments, San Diego, California) modified to reduce water vapor interference and improve detection limits [Jaffe et al., 1998]. Water vapor was removed from the sample stream with a nafion gas dryer in reflux mode. In this configuration the 3 o CO detection limit was 42 ppbv for a 20 minute average. The overall uncertainty in the hourly averages is 6% at a mixing ratio of 150 ppbv. Previously CO measurements made with this instrument were intercompared against observations by the NOAA-CMI)L group at a remote site [Jaffe et al., 1998]. The two measurements had a mean absolute difference of 3.1%. Ozone was recorded with a commercial ultraviolet absorption instrument (Dasibi 1008 RS). The instrument was calibrated before and after the campaign using a standard ozone calibrator (Columbia Scientific Inc.), and the total uncertainty is approximately 2%. PAN measurements were made using a gas chromatograph with electron capture detector (GC-ECD, Hewlett-Packard 5890). The 30 m (0.53 mm D) megabore capillary column (J&W DB-210) was kept at 35øC, and the ECD was kept at 50øC. The carrier and makeup gas were 5% CH4 in Ar flowing at 15 mL/min through the column and 35 mL/min through the ECD. Samples (1 mL) were injected every 15 min. Calibrations were performed approximately every other week using a PAN standard which was generated by dynamic dilution using a diffusion tube containing PAN in tridecane [Gaffney et al., 1984]. In the 1997 campaign the dynamic PAN standard was quantified as NO following reduction by CO in a 350øC catalytic gold converter [Bollinger et al., 1983]. Thus the PAN measurements were referenced to the same NIST traceable standard as was used for the NO measurements.
During the 1998 campaign the PAN calibrations were made with a commercial PAN calibrator. The commercial calibrator (Meteorologie Consult GmbH, Glash•itten, Germany) generates known mixing ratios of PAN from the photolysis of acetone in the presence of NO and 02. Since the conversion is nearly quantitative, the PAN mixing ratio can be directly related to the NO mixing ratio. For this test the NO calibration gas was the same as that used for the NO instrument, and a PAN conversion efficiency of 100% was assumed for the commercial calibrator. The PAN detection limit was 6 pptv, and the uncertainty of the measurements was 18%.
The photolysis rate of NO2 was measured using a spherically integrating radiometer (Meteorologie Consult GmbH, Glash[itten, Germany) based on the design of Junkermann et al. [1989]. The sensor measures the actinic flux (4•) in the 300-400 nm band using two filtered radiometers, one for downwelling and one for upwelling radiation. The instrument is calibrated by the manufacturer annually with a chemical actinometer at the Forschungszentrum-J[ilich, Germany. The sensor was installed at the top of a 12 m tower to minimize shadow effects. This is the same instrument that was used in our previous studies [e.g., Beine et al., 1997]. The uncertainty is given by the manufacturer as 8%, considering both cloud and zenith angle effects [Volz-Thomas et al., 1996;Beine et al., 1999] The aerosol instrumentation is described in detail by Anderson et al. [1999], so only a brief summary is given here. The aerosol instruments sampled the air through a vertical duct 40 cm in diameter from the top of the 10 m tower. The nature of the aerosol inlet results in a size cut that varies with wind speed. At wind speeds up to 10 m/s, encompassing 90% of our observations, the aerosol inlet collects particles with diameters of at least 5 um. A condensation particle counter (CPC) was used to measure particle number density for particles with a diameter greater than 10 nm. We used an integrating nephelometer (model 3563, TSI Inc., St. Paul, Minnesota) to measure total light scatter and backscatter at three wavelengths. Light absorption was measured using absorption photometers (model PSAP, Radiance Research, Seattle, Washington) that are calibrated to a wavelength of 550 nm [Bond et al., 1999]. For both the nephelometer and the PSAP instruments a valve switched sequentially between coarse (d<10 um) and fine (d<lum) mode aerosol. In this paper we will use only the fine mode data for total scattering (C•sp) and absorption (C•ap), both at 550 nm. During both the 1997 and 1998 campaigns, three different PSAP instruments were compared. The agreement between the three instruments was within 4%, averaged over the entire campaign. The total uncertainty for c5 and c• depends on the air mass. In marine sp ap air the uncertmnty averaged 14% and 56%, respectively, but the relative precision is much better at 7% and 33%
A plot of these data versus wind direction (not shown) indicates that the highest values for nearly all parameters arrive at the site with easterly winds. This comes as no surprise given the high density of emissions from the busy Vancouver-Seattle-Tacoma corridor. Because the focus of this work is on the atmospheric chemistry of the North Pacific environment, we segregated the data conservatively to consider only periods when we were fairly certain that recent emissions from North America had not impacted our observations. In this paper we will use the term "marine" to refer to air masses which have not crossed over or been influenced by North American emissions within at least the past 3 days, and "continental" to refer to the portion of the data for which a North American influence is likely. Note that marine air masses can still contain continental and/or anthropogenically emitted compounds resulting from long-range transport [e.g., Jeffe et el., 1999].
To separate marine and continental air masses, we used the wind direction, the wind speed, and the 10-day back trajectories. We classified as marine, periods with (1) wind directions between 150 ø and 300 ø, (2) wind speeds greater than 2 m/s, and (3) trajectories which had no land contact for at least 3 days prior to reaching the station. For 1997, 814 out of 1451 hours were classified as marine (56%), and for 1998, 566 out of 1432 hours were classified as marine (40%). Table  2 shows the data for periods classified as marine. NMHC mixing ratios for marine flow (as defined above) are given in Tables 3a and 3b. Figure   An example of a 10-day period with several spikes in NOx and CPC during marine periods is shown in Figure 3. It is important to reiterate that these data have already been selected for "marine" classification based on wind direction, speed, and back trajectories. During the periods with high NOx mixing ratios the average wind direction remains nearly due west, in the middle of the marine sector (150ø-300ø). For the elevated periods in 1997 at DOY 81.9, 82.9, and 85.4 (shown in Figure 3), winds were steadily from the west or southwest in each case.
The average NOx mixing ratios during the 1997 measurement period were significantly larger than those measured during the 1998 campaign. This is mostly because there were 46 hours with NOx mixing ratios greater than 500 pptv during the 1997 measurements, but only 17 during the 1998 campaign. However, these 17 hours brought much higher particle counts and aerosol absorption, as compared to the high-NOx There are several possible explanations for these relatively high NOx and CPC values. First, it is possible that complex meteorological patterns could bring air masses back to the station from North America. However, based on our screening criteria, these would have had to have been transported over the Pacific for at least 3 days, and this seems an unlikely explanation given the short lifetime for NOx and particle number density.
That these elevated periods are not due to recirculated North American emissions is shown by the absence of a CO enhancement in these air masses. For example, we can calculate the CO enhancement expected if these periods were due to recirculated regional sources. This can be done using   While the NOx' enhancement from this plume would be huge compared to a background value in the 50-100 pptv range, the CO enhancement would be insignificant compared to its background value of 150 ppbv. It is possible that NMHC observations would also show a signature of ship exhaust, but in searching through our data of discrete canister samples, we do not find any that coincide with periods of high NOx and CPC. Thus we believe that ship exhaust is the most likely cause for most of these periods based on the following: (1) the steady westerly winds during these high NO,, periods; (2) the absence of other significant nearby sources in the marine  . This one events explains why these species were elevated in the 1998 "high NOx" cases shown in Table 4.

This period had back trajectories which originated from the
Sea of Okhotsk and northeastern Asia, very near to regions where significant biomass burning was occurring at the time (see discussion below). However, it is difficult to reconcile the trajectories, which take at least 5-7 days to cross the Pacific, with elevated NOx and CPC, unless the emissions were extremely high. Unfortunately, we do not have MHC data for this event, so it is difficult to be more specific on the nature of the source or sources for this event.

In a companion paper, Kotchenruther et al. [this issue] describe the vertical profiles made in the spring of 1999 off the coast of Washington State using the Wyoming King Air. Because flights were only conducted during strong westerly flow conditions, very little evidence of North American pollution was observed.
Thus it is reasonable to compare the aircraft data from the 0-2 km layer with the Cheeka Peak marine data, keeping in mind that these data were collected during different years. These data are shown in Table 5. To make the comparison more meaningful, the NO data from CPO are shown for the hours of 1800-2400 UTC (1000-1600 PST), which is when the bulk of the aircraft data were collected.

Segregation by Trajectory Source Region
To help identify the role that latitudinal gradients and transport play at CPO, the marine data set was segregated based on the 10 day back isentropic trajectories. Our goal was to quantify the mixing ratio o.f each species segregated by trajectory type. Four trajectory classifications were used, those arriving from (1) midlatitudes; (2) high latitudes; (3) lowlatitudes, or (4) the Asian continent. In our previous work , only two classifications were used; Asian trajectories and non-Asian trajectories.
To categorize each trajectory, we first considered whether the 10 day transport crossed over the Asian continent (defined by the box from 0ø-50øN and 100ø-150øE). Those trajectories which crossed over the Asian industrial region were classified as "Asian." Following this step, we classified each trajectory based on whether the point 5 days backward in time was located in the low-, mid, or high-latitude regi9n, defined by the latitude lines at 35øN and 55øN. It should be apparent that both the midlatitude and Asian trajectories are arriving from the same direction, the only difference is the overall speed.
Trajectories were available for 0000 and 1200 UTC each day. We generated a 12 hour average from the hourly marine data centered on the trajectory time (+/-6 hours), provided there were at least three valid points in the average. Each 12 hour average was then paired with a trajectory type. Using the 12 hour averages has the added advantage of eliminating autocorrelation in the data and thus simplifies the statistical analysis, compared to using the hourly data. Results of this analysis are presented in Table 6 for the 1997 and 1998 data.
Note that we have omitted the NMHCs in Table 6 because there are too few samples. Interpretation of the 1997 results is fairly straightforward. The trajectory segregation shows the impact of long-range transport from Asia and the latitudinal gradients for several species, for example, PAN and CO. For PAN the lower mixing ratios in lower-latitude air masses is particularly apparent, a result of the higher temperatures all along the transport path.
Low-latitude trajectories arrive at CPO about 2øC warmer, on average, than the midlatitude trajectories. Comparing the periods of midlatitude trajectories with periods of Asian trajectories, we find that there are statistically significant differences (P>95%) for CO, PAN, Rn, aerosol scattering, and aerosol absorption. These results are essentially the same as we have previously reported . Comparing the periods with low-latitude versus midlatitude trajectories, we find statistically significant differences (P>95%) for Rn, 03, PAN, aerosol scattering, and absorption. For CO the lowlatitude cases are significantly different from the mid-latitude cases at the 86% confidence level. For the high-latitude cases, aerosol scattering and absorption are significantly lower than the midlatitude cases, but at somewhat lower confidence levels P = 94 and 89%, respectively. Our interpretation of these results is that, in general, lower latitudes bring lower mixing ratios of most continentally emitted species, compared to the midlatitude trajectories. The high-latitude trajectories arrive with generally lower amounts of aerosol scattering and absorption and similar amounts of CO, Rn, and PAN. Although, the Asian trajectories have the highest concentrations of all species (except O3), the impact of Asian sources is certainly seen in the "midlatitude" trajectories as well.
For 1998 a somewhat different pattern emerges. Only PAN shows a statistically different mixing ratio during the period with Asian trajectories (higher PAN) and low-latitudes trajectories (lower PAN), compared to the midlatitude cases. However, the statistical significance is lower, 89% in both cases. This observation seems to indicate that during the spring of 1998, the latitudinal gradients were substantially reduced or eliminated. Also during 1998, O3 was elevated in the Asian and low-latitude cases, compared to the midlatitude cases (P = 82 and 86%, respectively). These observations, along with the fact that several species were significantly elevated compared to 1997, suggests that the spring of 1998 was significantly different from the spring of 1997.

Interspecies Correlations
Correlations between measured species can give information on the similarity of sources and/or sinks. Table 7a shows the correlation coefficient R, using the hourly averages for those relationships with an R value greater than 0.35. For the relationships shown, there are between 300-700 hourly data pairs. Considering the autocorrelation in the hourly averages, significant up to a time lag of approximately 10 hours, this results in 30-70 degrees of freedom for each correlation. Thus an R value equal or greater then 0.35 corresponds to a significant correlation at a probability of 95% or higher. Table 7b shows the correlation coefficients for NMHC species and CO. For the NMHC comparison we used a 3 hour average CO JAFFE ET AL.: GROUND-BASED OBSERVATIONS DURING PHOBEA There are a number of patterns which are brought out by this analysis. In general, a number of these species including PAN, CO, Rn, C•sp, and C•,p were significantly correlated in both the 1997 and 1998 data. 03 was weakly correlated with CO in the 1997 data set (R=0.28), although this is significant with a confidence of 98%; however, CO and 03 were uncorrelated in the 1997 data set. Most likely this reflects the multitude of sources responsible for tropospheric 03. In contrast, at sites in the northeastern corner of North America, downwind, and much closer to large continental sources, Parrish et al. [1993Parrish et al. [ , 1998 have observed a strong correlation between CO and 03 during summer outflow. Under these conditions the bulk of the observed 03 is probably due to photochemical production associated with North American NOx emissions. At more remote sites a CO-O3 relationship is not generally found [e.g., Jaffe et al., 1998]. That the CO-PAN correlation was significant in both years and that PAN is generally correlated with aerosols implies that industrial emissions are the dominant source for both species.

Comparison Between the 1997 and 1998 Observations
For a remote observation site such as CPO we expect that a variety of natural and anthropogenic sources contribute to the observed mixing ratios. However, these sources, and the transport are not necessarily constant from year to year. To examine these variations, we conducted a statistical comparison of the 1997 and 1998 data sets. In doing this comparison it is necessary to consider that the hourly averaged data has a significant autocorrelation associated with it. While there are several ways to take this into account, the most straightforward method is to use the 12 hour averaged data set, described above, which has no autocorrelation. Species with significant differences are shown in Table 8. Robust differences are seen for CO, ethane, and acetylene. For 03 the difference between the 2 years, approximately 2%, is nearly the same as our total uncertainty in the measurements; thus this difference should be treated with more caution. For the NMHC species the differences are large enough to be greater than the measurement uncertainty; however, some care should be taken in interpreting these since there were not a large number of samples taken during 1997 and the samples were not uniformly spaced through the 2 month period.  aFor this analysis we used the marine data set and a 3 hour running average for CO. All correlations are significant at better than 95% except for CO-ethane (94%), CO-propane (76%), and CO-i-butane (72%). Here ns, not significant. aThe P value gives the probability that the two species were significantly different in the 1997 and 1998 data sets. and other trace species and are believed to have had a widespread influence on trace species throughout the North Pacific (G. Carmichael, personal communications, March 2000).

Impacts of Biomass Burning on the North Pacific
Although an emission inventory for these fires is not yet available, the CO output from the northern Indochina fires may be at least as large as the Indonesian fires.
Additionally, between April and September 1998, substantial biomass burning also occurred in Mongolia and far eastern Russia. The fires started in the region surrounding Lake Baikal (approximately centered on 110øE x 55øN) during April and May and subsequently moved eastward, so that in the period July-October they were concentrated in far eastern Siberia, adjacent to Sakhalin island (approximately 140øE x 55øN). These fires were probably larger than any other event Isentropic trajectories show that air arriving to CPO has frequently come from eastern and northeastern Asia within the past 10 days. Thus, we expect that biomass burning sources in these regions are an important source for many trace species at CPO. For example, as mentioned above we saw one case of elevated CO, 03, PAN, aerosols, and other species which appears to be associated with rapid transport from northern Asia during a period when substantial biomass burning was occurring (DOY 111, 1998). To the extent that this CO was transported to higher latitudes, where lifetimes during winter will be much longer, this would enhance the global CO burden. We believe this mechanism is plausible to explain some of the 1998 CPO data, even when trajectories are not coming directly from Asia. While many studies have identified a biomass burning "signature" based on observations of CO and NMHCs in relatively fresh plumes [e.g., Blake et al., 1994Blake et al., , 1996, these emissions will change dramatically over time as a result of mixing and chemical processing during transport. However, for longer-lived species, such as CO, ethane, ethyne, and propane, we might expect to see the ratios of these species pre- both enhanced biomass burning and enhanced transport of industrial emissions is needed to explain the higher mixing ratios we observed at CPO during 1998.

Summary and Future Work
In this paper we have described the observations of CO, 03, NOx, PAN, NMHCs, Rn, aerosol scattering, aerosol absorption, and aerosol number density at the Cheeka Peak Observatory during the PHOBEA measurement campaigns during the spring of 1997 and 1998. From these data we examine how the mixing ratios of these species in the eastern Pacific are controlled by long-range transport, ships, and latitudinal gradients. There are a number of significant differences between the 1997 and 1998 observations in terms of the relationships to source regions and the mean mixing ratios.
We attribute these differences to the strong E1 •,lino present in late 1997 and 1998 and the much larger biomass burning emissions in late 1997 and early 1998. In addition to the CPO observations, vertical profiles of a nearly identical set of chemical compounds were made in this region during the spring of 1999 using the Wyoming King Air and are described by Kotchenruther et al. [this issue]. Taken together, the PHOBEA ground and aircraft data are currently the most comprehensive database of tropospheric chemistry in the northeastern Pacific during spring. Future work using this database will focus on an analysis of the tropospheric 03 budget in this region and developing a better understanding of the factors that control long-range transport from distant source regions.