This paper describes the chemical, aerosol, and meteorological measurements taken at Mount Bachelor Observatory (MBO), a new mountaintop site in central Oregon, USA (44.0°N, 121.7°W, 2763 m above mean sea level). During the initial campaign (28 March to 19 May 2004) we evaluated the utility of this location as a site to observe the global atmosphere, especially the free troposphere (FT). We observed some boundary layer (BL)/upslope flow during the daytime, which produced a 37–62% higher average water vapor mixing ratio (WV) compared to radiosonde data taken from similar altitudes in western Oregon. However, recently subsided air masses with low WV contained significantly elevated concentrations of carbon monoxide (CO), total gaseous mercury (TGM), and ozone (O3), (25 ppbv, 0.16 ng/m3, and 13.9 ppbv, respectively), compared to periods with high-WV measurements. These enhancements represent a significant influence from Asian long-range transport (ALRT). One specific time period (“event”) on 25 April produced some of the largest CO, TGM, O3, and particulate scattering (σsp) measurements ever seen along the West Coast due to ALRT. Enhancement ratios between these species are consistent with the major source being Asian industrial emissions. In particular, the ΔTGM/ΔCO enhancement ratio was 0.0045–0.0048 ng/m3/ppbv for all ALRT events, similar to the value previously obtained from pollution plumes directly downwind of east Asia. A smaller pollution event of North American origin produced a ΔTGM/ΔCO value of only 0.0013 ng/m3/ppbv, suggesting that the ΔTGM/ΔCO value is an effective tracer of ALRT. Finally, thousands of kinematic back trajectories were calculated for each event to evaluate sources and transport processes.
 Airborne mercury (Hg), consisting of elemental (Hg0), reactive (RGM), and particulate (PHg) is present in Asian industrial outflow, as determined from recent measurements at Okinawa, Japan [Jaffe et al., 2005], and over the South China Sea [Friedli et al., 2004]. Mercury is an air pollutant of great concern because of its high toxicity and its ability to bioaccumulate in the environment. The atmosphere has been identified as the most important cycling pathway in the global distribution of mercury because of the 6 month – 2 year estimated lifetime of Hg0 [Schroeder and Munthie, 1998; Lin and Pehkonen, 1999]. Additionally, because of the large amounts of coal burned in east Asia presently [Wang et al., 2000] and projected into the future, industrial Hg emissions from east Asia contribute about one half of the global anthropogenic total and are likely still increasing [Pacyna and Pacyna, 2002].
 Previous measurements of mercury species on the U.S. west coast at Cheeka Peak, Washington, in the marine boundary layer (BL) revealed little or no enhancements and weak correlations between Hg0 and CO [Weiss-Penzias et al., 2003]. However, our preliminary results as reported by Jaffe et al.  identified a large mercury enhancement at Mount Bachelor Observatory (MBO) on one occasion during the spring of 2004 due to ALRT.
 Mountain-top free tropospheric observation sites are sparsely located across the northern hemisphere, and can provide unique information about the global atmosphere compared to sites in the BL [e.g., Bodhaine, 1996; Weingartner et al., 1999]. The key questions that will be addressed here are the following: (1) How often and under what conditions is MBO sampling in the FT? (2) How can these air masses be distinguished from North American BL-influenced air masses? (3) What are the chemical signatures (enhancement ratios) that characterize long-range transport from Asia?
2.1. Site Description and Sampling
 Mount Bachelor Observatory (MBO) is located atop a dormant volcano in the Cascade Range of central Oregon, USA (43.98°N 121.69°, 2763 m above sea level (ASL), see Figure 1). A ski area at the mountain operates from mid-November until the end of May, with most of the ski runs, lodges, access roads, and parking lots located on the eastern side of the mountain. Access to the summit is only via an electric-powered ski lift, hiking, or snow vehicle, with the latter occurring ∼once per day, and producing easily observable spikes in nitrogen oxides.
 The sampling site is located on the roof of the summit building and the instrument room is about 15-m below. The roof structure receives ice and snow buildup during stormy periods, which can persist for several days continuously. The roof is manually cleared of ice, but not on a regular basis. For this reason during this initial campaign, our sampling lines were located on the upper part of the leeward (SE) wall. This relatively protected location had to be cleared of ice only a few times during the spring.
 Carbon monoxide, ozone, and nitrogen oxides were sampled through 1/4″ Teflon tubing, with a Teflon 1 μm particle filter at the inlet, which was changed bimonthly. The flow through this sample line was about 3 LPM, with the flow being shared about equally between the three instruments. Submicron particles were sampled through 3/8″ copper tubing with no sharp bends to minimize particle loss in the tube and the inlet tip was fitted with a snow shield. The mercury sampling method is described in its own section below.
2.2. Meteorological Observations
 Temperature and relative humidity (T/RH) sensors (Campbell Scientific HMP 45C) were located at a lower height on the leeward wall of the building because they are less durable and could not withstand the heavy accumulation of ice that occurred on the very top of the building. Even at this more protected location, the T/RH housing collected ice on it during stormy periods (Tamb < 0°C). Thus RH measurements > 90% for more than 24 hours continuously and with Tamb < 0°C were discarded from the data set because of probable RH sensor icing conditions. The total number of hours removed is 300 out of 1272.
 Water vapor mixing ratio (WV) was calculated from ambient relative humidity, temperature, and pressure (Vaisala PTB101B pressure transmitter) using an empirical equation from Bolton . When RH measurements were impacted by icing conditions, no WV data are reported. In late 2004 we added a method to directly measure WV from the high-flow gas inlet (Licor 6262). This method avoids the ice buildup problem of the external RH sensor; however these newer data are not included in the present analysis.
 Wind speed was measured with a Taylor Scientific WS-3 Heated Rotor Anemometer, and wind direction with a Taylor Scientific WD-3 Heated Direction Vane. These sensors also experienced icing periods in extreme weather conditions despite substantial heating of 1200 W. We removed 229 out of 1272 total hours (18%) because of icing.
2.3. Chemical Measurements
2.3.1. Carbon Monoxide
 Carbon monoxide was measured with two separate instruments: a Trace Analytical reduction gas analyzer (RGA) (28 March to 22 April) and a Thermo Electron Corporation nondispersive infrared (NDIR) CO analyzer (24 April to 19 May). The reduction gas analyzer is a gas chromatographic method that separates CO from other reductive gases (mainly H2) and then measures the fluorescence of elemental mercury, which is produced from the reaction of CO + HgO(s). The RGA method has a lower method detection limit (MDL) (10 ppbv) than the NDIR method (20 ppbv); however, the RGA was prone to temporary noise increases due to water contamination on the column. For this reason, the RGA is usually only used in our laboratory at the University of Washington, Bothell (UWB), to calibrate and cross-reference CO standards. The RGA made a single measurement every 4 min, whereas the NDIR makes continuous measurements with a response time of 1–2 min, resulting in about twice as much data taken per hour for the NDIR. Thus the total uncertainty in each hourly measurement from the RGA is about 12% and that from the NDIR is about 8%.
 Each instrument was calibrated at least daily with medical-grade breathing air containing a known amount of CO (429 ppbv), as determined both before and after the campaign in the laboratory using the RGA, referenced to a high-level NIST CO standard (9.75 ppmv, SRM# = 2612A) and an ambient-level CO cylinder (177 ppbv) from NOAA-CMDL [Novelli et al., 1991]. We have employed this method of using breathing air for our CO standards and analysis with the RGA against the NIST standard for several years and have found that the breathing air cylinders have minimal drift (<5%) over the period of field use (up to 6 months).
 Ozone was measured using a standard UV absorbance instrument (Dasibi 1008 RS). This instrument was zeroed monthly using an activated charcoal scrubber cartridge and calibrated twice during the campaign with an ozone generator (Columbia Scientific Inc.), referenced to an ozone transfer standard from the Washington State Department of Ecology. The MDL is 1 ppbv and the estimated uncertainty in an hourly average is 2%.
2.3.3. Aerosol Scattering
 Submicron dry aerosol total scattering measurements (535 nm) (σsp) were made with a Radiance Research nephelometer. A 1 μm impactor was used at a flow rate of 10 LPM for the size cut. These values are expressed in m−1 × 106, or Mm−1 corrected to STP (273K and 1 atm). The nephelometer was zeroed monthly using a high-efficiency particle filter, and was calibrated with pure CO2. The sample was heated so that RH < 35% before entering the nephelometer. The MDL for σsp is 0.25 Mm−1 with an estimated hourly uncertainty of 5%.
2.3.4. Nitrogen Oxides
 Total reactive nitrogen oxides (NOy), were measured with a standard chemiluminescence NO detector (Teledyne-API, model 200AU, MDL = 0.1 ppbv). These data were used only to detect the influence of snow grooming activities at the ski area, which were easily recognizable as large spikes, up to 10 ppbv. The rest of the periods, NOy values were very low (mean = 0.16 ppbv) and displayed little correlation with other measurements.
 The mercury measurements were made with a Tekran 2537A analyzer connected to a custom pyrolyzer and heated probe. We elected to add a pyrolyzer and heated probe in order to measure total gaseous (or airborne) mercury (TGM = Hg0 + RGM + PHg) rather than Hg0 because this method significantly reduces the potential for artifacts in the detection. Furthermore, we expect that the TGM signal will be dominated by Hg0 as in our previous measurements at Cheeka Peak, Washington [Weiss-Penzias et al., 2003]. In our design, air was drawn through a quartz inlet which was held at about 130°C by a temperature controller. The probe was connected directly to a pyrolytic column composed of a quartz tube packed with quartz chips, which was maintained at 500°C in a tube furnace. The inlet system did not include any aerosol size segregation step so the upper particle size limit was limited by the efficiency of the inlet tip.
 The Tekran 2537A was configured to sample on a 5 minute collection cycle with a flow of 0.75 slpm, and had an MDL of 0.06 ng/m3. The instrument automatically calibrated to an internal permeation source every 18 hours. The accuracy of the factory determined permeation rate was compared to manual injections of a primary vapor source (Tekran 2505), and agreed to better than 6% with the factory specified rate. The hourly precision was 2% and the mean precision of successive internal calibrations was better than 1%. The overall uncertainty is estimated to be better than 16%. All mercury data are reported in ng per standard m3.
2.4. Supplementary Data
 Vertical profile data were incorporated into our analysis from two radiosonde sites: Medford, Oregon (42.4°N, 122.9°W, MFR), Salem, Oregon (44.9°N, 123.0°W, SLE), and one ozonesonde site: Trinidad Head, California (41.1°N, 124.2°W, THD) (see Figure 1 for locations). MFR and SLE meteorological data were obtained from the University of Wyoming (http://weather.uwyo.edu/upperair/sounding.html). THD ozonesonde data were obtained from NOAA-CMDL (http://www.cmdl.noaa.gov/infodata/ftpdata.html and B. Johnson, NOAA-CMDL, personal communications, 2004).
 Aerosol optical thickness (AOT) data were obtained from the NASA Aerosol Robotic Network (AERONET) for HJ Andrews Forest in Oregon and Rimrock in Idaho (http://aeronet.gsfc.nasa.gov).
2.5. Back Trajectories
 We calculated over 100,000 kinematic 10-day (241-hour) back trajectories from the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT-4) model [Draxler and Rolph, 2003] (hereafter referred to as Hysplit trajectories). The Hysplit model was operated with meteorological input from the NCEP reanalysis data set which has a time resolution of 6 hours, a horizontal resolution of 2.5° latitude by 2.5° longitude, and 18 vertical layers. See section 3.7 for a discussion of this analysis.
3. Results and Discussion
3.1. Overview of Measurements
Table 1 presents an overview (average ±1σ) of all measured parameters. This data set spans the time period 28 March to 19 May 2004 (“spring”), and is broken up into “April” and “May” sections (28 March to 30 April and 1–19 May). The measurements are also categorized into “dry” and “wet,” representing water vapor mixing ratio values less than 2.3 g/kg, and greater than 4.2 g/kg, respectively. The difference between “dry” and “wet” is also shown and discussed below.
Table 1. Mean Seasonal and Monthly Values ±1 Standard Deviation for All Dataa
Ambient T, °C
Wind Speed, m/s
Water Vapor (Calc), g/kg
Water vapor segregated data for the whole season termed “dry” and “wet” is also shown along with dry – wet. Statistically significant (P > 95%) differences are shown in bold.
28 March to 19 May
167 ± 23
1.77 ± 0.12
44.5 ± 7.6
6.9 ± 7.1
−1.5 ± 5.1
60 ± 23
729.2 ± 5.3
10.1 ± 6.0
3.2 ± 1.0
28 March to 30 April
178 ± 18
1.80 ± 0.12
46.3 ± 7.1
7.6 ± 7.9
−2.0 ± 5.4
56 ± 24
729.2 ± 5.8
9.8 ± 6.3
2.9 ± 0.9
147 ± 15
1.70 ± 0.08
41.8 ± 7.5
5.7 ± 5.4
−0.6 ± 4.1
68 ± 19
729.2 ± 4.2
10.6 ± 5.3
3.8 ± 0.8
28 March to 19 May
179 ± 28
1.84 ± 0.16
54.0 ± 8.0
7.7 ± 12
−2.3 ± 5.7
43 ± 26
731.6 ± 3.7
10.6 ± 6.9
1.7 ± 0.5
28 March to 19 May
154 ± 22
1.69 ± 0.08
40.0 ± 5.4
7.3 ± 4.6
2.5 ± 2.7
74 ± 13
730.6 ± 3.7
9.5 ± 6.2
4.6 ± 0.3
Dry – Wet
28 March to 19 May
 There is a distinct seasonal pattern in nearly all parameters, with concentrations of CO, TGM, O3, and σsp being 21%, 6%, 10%, and 33% higher during April compared to May. WV on the other hand, is 31% lower in April. Standard deviations are higher in April across all parameters except for O3. This seasonality is a function of (1) regional warming from April to May, which influences the relative amounts of FT and BL air seen at the site, (2) greater amounts of trans-Pacific transport in April [Liang et al., 2004; Forster et al., 2004], and (3) higher photolysis rates which leads to shorter lifetimes of CO, O3, and TGM (species that react with hydroxyl radical (OH)) during May.
3.2. Diurnal Cycles
 In order to determine the degree to which daily changes in upslope/downslope flow on the mountain influenced our measurements, we calculated the average value for each parameter as a function of time of day during the spring campaign. Figure 2 shows the diurnal cycles of CO and σsp (Figure 2a), TGM and wind speed (WS) (Figure 2b), O3 and WV (Figure 2c), and T, and RH (Figure 2d), which are plotted against local time (LT). The variations as a fraction of the mean value for TGM, O3, WV, WS, T, and RH are 7%, 12%, 34%, 22%, 2% (in Kelvin), and 11%, respectively. CO and σsp show no apparent diurnal pattern, which suggests that emission sources in the immediate region are not a significant influence on these measurements. O3, WV, WS, T, and RH diurnal variations reflect some upslope/downslope mountain flow. Temperature reaches a maximum and WS and RH reach a minimum between 1100 and 1400 LT, during the time of highest zenith angle. The minimum in O3 and the maximum in WV occur slightly later at 1500 and 1600 LT, respectively, and are a function of land heating bringing BL air upward to the summit. O3 has a well known vertical gradient in the lowermost troposphere [Logan, 1999; Oltmans et al., 2004] due to loss of O3 to land/sea surfaces in the nocturnal boundary layer [e.g., Brown et al., 2003], and less mixing from the upper troposphere. WV also displays a strong vertical gradient and thus these two parameters can be used to determine when predominantly FT or BL air is seen at MBO.
 Other mountain sites have observed much larger diurnal variability of gas and particle species than what we report here. For example, aerosol scattering measurements varied by a factor of 5 or more at Mauna Loa Observatory during the spring of 1992 [Bodhaine, 1996], which is probably due to more surface heating and a stronger upslope effect compared to MBO. Other measurements in the Swiss Alps have shown that deep alpine valleys are effective at vertically exporting pollutants to heights of at least 4 km during summer clear-sky conditions [Henne et al., 2003]. Mt. Bachelor avoids this phenomenon somewhat, because it is almost a perfect cone-shaped mountain with very little glacially carved terrain, and thus air is not channeled up to the summit in alpine valleys. Thus, although MBO is not as high as other free tropospheric sites (e.g., Mauna Loa, 3.4 km; Jungfraujoch, 3.5 km), it will be shown that MBO can occasionally experience air that is representative of these higher altitudes.
 TGM shows a small diurnal variation of about 0.07 ng/m3 about the mean, with a morning maximum (0800–1200 LT) and an evening/night minimum (1700–0400 LT). The maximum is shifted about 4–6 hours earlier than the WV maximum/O3 minimum cycle, suggesting that TGM is not simply controlled by BL-FT exchange due to daytime heating. Previous measurements have shown a similar phase in diurnal Hg0 patterns (night/morning maximum, afternoon/evening minimum) [Lee et al., 1998; Kellerhals et al., 2003; Weiss-Penzias et al., 2003; Poissant et al., 2004]. Speciated Hg measurements (Hg0, RGM, and PHg), necessary to fully describe Hg dynamics, were begun during the spring of 2005 at MBO, and will be the subject of a forthcoming paper (P. Swartzendruber et al., Observations of reactive gaseous mercury in the free troposphere at the Mt. Bachelor Observatory, submitted to Journal of Geophysical Research, 2006).
3.3. Altitude Comparison With Radiosonde Data
 In this section we present a comparison of the spring MBO data (including “dry” and “wet”) with the closest radiosonde sites at Medford (MFR) and Salem (SLE), Oregon, and an ozonesonde site at Trinidad Head, California (THD) (locations shown in Figure 1). It should be noted that each of these sites is in a different geographical region of the Pacific Northwest: MFR (400 m ASL) is in a valley between the Cascade and Coast mountain ranges, SLE (60 m ASL) is in the Willamette River valley, and THD (100 m ASL) is on the Pacific coast. Because of these geographical differences, the profiles differ somewhat. Figures 3a–3d show the vertical profiles of temperature (T), wind speed (WS), water vapor (WV), and ozone (O3), compared to the MBO spring 2004, “day,” “night,” “dry,” and “wet” (see Figure 3 caption) at 729 hPa (2.8 km). MBO T and WS in Figures 3a and 3b are similar in magnitude to the vertical profiles at the same altitude. The diurnal variation of T is slightly greater than the range of values among the three vertical profiles at the same altitude. WS at MBO is slightly higher than the radiosonde data perhaps because of the summit itself, which could cause higher wind speeds because of ridge effects during occasional high-wind events. MBO WV is 37% and 62% higher than WV at 729 hPa from MFR and SLE, respectively, suggesting that MBO experiences more influence from the boundary layer compared to radiosonde data. The offset between MBO and MFR in Figure 3c indicates that on average, MBO represents a somewhat lower altitude of 787 hPa (2.3 km) than the actual altitude of 729 hPa (2.8 km). This is consistent with our diurnal cycle findings (Figure 2) which show that there is regular mixing from lower altitudes, resulting in higher average WV mixing ratios.
 Average O3 at MBO (44.5 ppbv) is significantly lower than the March–May 1997–2004 O3 profile from Trinidad Head (57 ppbv) at 729 hPa, shown in Figure 3d. According to this comparison, the “equivalent altitude” of MBO is approximately 900 hPa (1330 m), much lower than that determined from the MFR WV comparison. However, this may not be an accurate comparison because THD is in a region of greater mean subsidence associated with the Pacific high. Modeled long-term vertical velocity values for April–May at 700 hPa are +0.55 Pa/s and −0.015 Pa/s for THD and MBO, respectively (available online at http://www.cdc.noaa.gov). In other words, THD at 729 hPa generally experiences air from higher altitudes because of greater regional subsidence compared to MBO, resulting in lower WV and higher O3.
 Thus, in order to generate a WV data set that is comparable to that from the same altitude over the MFR radiosonde site, we would retain 44% of the MBO data with WV < 3.1 g/kg. Both data sets would then have a mean WV value of 2.3 g/kg. For the following analysis however, we will segregate the data by a more conservative WV cut-point of 2.3 g/kg, giving an average MBO WV value of 1.7 g/kg. This corresponds to the driest, subsiding air masses observed at MBO. In Figure 3c, comparing MBO “dry” with the MFR radiosonde data, gives an elevation for MBO of approximately 685 hPa (3.4 km). Average O3 in these “dry” data (Figure 3d) also indicates that higher-altitude air masses are primarily being sampled. These data include 183 hours, which is 20% of the hours with valid WV data. These data are considered free tropospheric and are assumed to have had no contact with the North American continent in recent days. In contrast, we also examine the wettest 20% of the springtime data (“wet”), which correspond to all measurements with WV > 4.2 g/kg. These data yield an “equivalent altitude” of 900 hPa (1330 m) according to Figure 3c, and are thought to be representative of the planetary boundary layer. Note that O3 is significantly lower in these data as well (Figure 3d).
3.4. Comparison of Dry and Wet Segregated Measurements
 The “dry” and “wet” data sets differ considerably, which confirms our conceptual view of airflow at MBO; “dry” data represent air masses that have recently descended, have little influence from regional emissions, and may contain ALRT pollutants, and “wet” data have recently come from lower altitudes, may contain pollutants of North American origin, and may have been influenced by the clean marine boundary layer. Table 1 quantifies these differences. First, CO and TGM are significantly enhanced in the “dry” data because of the inclusion of several ALRT pollution plumes, higher background concentrations, and the natural seasonality of these two species (more hours in April). Second, ozone is also enhanced in the “dry” data, but in addition to ALRT, this may reflect the greater stratospheric influence in recently descended air. Additionally, O3 in the “wet” data may have been removed by surface deposition to the sea surface and vegetation. Third, σsp is not significantly enhanced in the “dry” data, which may reflect the shorter lifetime of σsp compared to CO and TGM. Also, local transport is probably more important for σsp than it is for CO and TGM. Fourth, “dry” air masses are considerably colder and have higher pressure, suggesting that these air masses are associated with anticyclonic flow. A comparison of wind direction distributions for the two data sets, shown in Figure 4, reveals that the prevailing wind direction during these “dry” periods is from the north, which confirms anticyclonic flow. “Wet” air masses have predominantly W and SW wind directions. Because of the sparse population in these directions, and close proximity to the Pacific Ocean (<300 km), the “wet” data do not show significant enhancements of CO or σsp. The strongest pollution events during the campaign (25 April and 13 May) were observed during dry subsiding air masses, and are attributed to distant pollution sources.
3.5. Comparison With Other Free Tropospheric Measurements in Pacific Region
Table 2 compares CO, TGM, O3, σsp, and WV measurements from MBO with past measurements in the FT along the West Coast and across the Pacific. Also shown at the bottom of Table 2, are measurements from Cheeka Peak Observatory (CPO) in Washington State from 2002, a site at 500 m in the marine boundary layer, which is compared to our “wet” air measurements from MBO. Previous CO measurements from the PHOBEA campaigns show a relatively consistent mean value of about 140 ± 15 ppbv between 2 and 4 km, which is almost 40 ppbv lower than our “dry” CO value of 179 ± 28 ppbv. This significant jump in FT CO concentrations is consistent with recent observations from the MOPITT instruments [Shindell et al., 2005] and NOAA-CMDL flask samples (P. Novelli, personal communication, 2005). Notice how TGM and σsp were also enhanced at MBO relative to past measurements, suggesting that higher levels of pollution transport occurred in 2004 compared to previous years. That this enhancement at MBO could be due to local sources is unlikely, since if this were the case, MBO “wet” data would show higher levels than MBO “dry” data. One explanation for the higher levels at MBO in 2004, then, is the large Siberian forest fires that occurred during the summer of 2003. Direct transport from these fires to the Pacific Northwest was observed [Jaffe et al., 2004; Bertschi and Jaffe, 2005] and may have resulted in an increase in background CO concentrations persisting into the spring of 2004.
Table 2. Comparison of All-Spring and WV-Segregated Data From MBO With Other Measurements in the 2–4 km Altitude Range Along the West Coast and Across the Pacific During the Springa
Also shown are the measurements from Cheeka Peak, Washington, from 2002, a site in the marine boundary layer.
 Three episodes of pollution transport are detailed in this section: two ALRT events on 25 April and 13 May, and one local pollution event on 27 April. The highest concentrations of CO, TGM, O3, and σsp from the entire campaign occurred on 25 April. A close-up of this time period is shown in Figure 5a. Maximum hourly averages were 296 ppbv for CO, 2.52 ng/m3 for TGM, 78 ppbv for O3, and 65 Mm−1 for σsp, while WV was at a minimum of 1.8 g/kg. These pollutant concentrations are comparable to the largest plumes seen during ITCT-2K2 and PHOBEA aircraft campaigns [Nowak et al., 2004; Price et al., 2004]. In addition, this is the first documented TGM enhancement in a trans-Pacific ALRT plume. The peak of the event occurred at night during strong subsidence, enhanced pressure (max. = 737 hPa), low wind speeds (min. = 0.5 m/s), and a negative correlation between CO and WV (r = −0.72). Temperature climbed to 10°C during this event, which may have resulted in the thermal decomposition of PAN and enhanced O3 formation [Hudman et al., 2004].
 The 13 May event is shown in Figure 5b. This event is characterized by lower peak concentrations than the 25 April event: CO = 200 ppbv, TGM = 1.87 ng/m3, σsp = 45 Mm−1, and O3 = 72 ppbv. These lower concentrations are partially a result of less subsidence (maximum pressure = 731 hPa), yielding a weaker CO/WV correlation (r = −0.53).
 The 27 April event was the smallest of the three pollution episodes, shown in Figure 5c. CO, TGM, O3, and σsp have maxima at 180 ppbv, 1.8 ng/m3, 49 ppbv, and 20 Mm−1, respectively, and in contrast to the 25 April and 13 May events, WV reaches a maximum of 5.5 g/kg. While these concentration enhancements are not great, this event displays a strong positive correlation between CO and WV (r = 0.85) and between CO and σsp (r = 0.78), suggesting that polluted air from the BL was observed. Meteorologically, this event was characterized by rapidly cooling temperatures (from 10 to 3°C), falling pressure (from 740 to 734 hPa), and increasing wind speeds (from 7 to 20 m/s). These conditions suggest a cold front was approaching the site, quite different from high pressure subsidence observed for the other events. The main peak of the event occurred on 27 April at 1000 UT, and displays a small increase in TGM, and a loss in O3. The latter observation may be a result of loss of O3 to deposition, if in fact the air masses had contact with the surface.
 Linear correlation coefficients and slopes for TGM/CO, O3/CO, and σsp/CO for each event are shown in Table 3. Differences between the ALRT and local events are evident in these data. The ALRT events are believed to be composed mainly of Asian industrial pollution with very little biomass burning or dust emissions. This conclusion is drawn by comparing the enhancement ratios (ERs) between species (i.e., the slopes of the linear correlations) with ERs obtained from previous measurements of Asian pollution. First, the TGM/CO ERs have been postulated to be sensitive indicators of ALRT because of the large mercury emissions in China; a value of 0.0056 ± 0.0016 ng/m3/ppbv is thought to be indicative of Asian industrial emissions [Jaffe et al., 2005]. Indeed, this is the same ER that was observed by Friedli et al.  in a plume downwind of Shanghai, China. Because of the long lifetimes of CO (1–6 months) and Hg0 (6–24 months), we would expect to see this ER conserved during trans-Pacific ALRT. ΔTGM/ΔCO from both the 25 April and 13 May events is in fact remarkably consistent at 0.0045–0.0048 ng/m3/ppbv. Second, the Δσsp/ΔCO values from both ALRT events (0.47 Mm−1/ppbv) agree with the ERs from ALRT events (0.5 Mm−1/ppbv) between 0 and 5 km observed during the PHOBEA campaigns [Price et al., 2004]. Previous ALRT events containing large amounts of biomass burning emissions have resulted in Δσsp/ΔCO values that are about a factor of 2 larger than what we observe here [Bertschi et al., 2004]. Third, the ΔO3/ΔCO value for both ALRT events is 0.22 ppbv/ppbv, which would be classified as “Type-II” according to Price et al. , and falls within the 0.1–0.5 range that are typical ERs from industrial pollution.
Table 3. Correlation Coefficients and Slopes (Using Hourly Averages) for Two Long-Range Transport Events, 25 April and 13 May, and One Local Pollution Event, 27 Aprila
Slopes are only shown for r-values > ∣0.3∣.
25 April 2004 (n = 29)
13 May 2004 (n = 18)
27 April 2004 (n = 18)
 ERs from the 27 April local event are quite different from the ALRT events and are another way to distinguish between “dry” and “wet” air masses. In particular, the ΔTGM/ΔCO value is considerably less (0.0013 ng/m3/ppbv) than the value that is attributed to Asian industrial pollution. A smaller value in local pollution is consistent with a lack of mercury sources in the region around MBO [Jaffe et al., 2005]. O3 also behaves quite differently in this local event, being uncorrelated with CO. This suggests a lack of photochemical production and/or loss to surface deposition. σsp is well correlated with CO in this event, but the Δσsp/ΔCO ratio is less than that observed during the ALRT events. This may have been caused by precipitation associated with the approaching cold front, which would remove particles but not CO and result in a lower ER.
3.7. Pollution Transport Events: Back Trajectory Analysis
 Kinematic back trajectories from Hysplit were used to verify source regions for the 25 April, 13 May, and 27 April pollution events. Our analysis is patterned after the work of Stohl et al. [2002, 2003], who introduced the term “retroplumes.” The idea is that thousands of trajectories must be calculated, each one representing the path of a single particle, in order to have a large enough sample size that can begin to represent air mass transport. These trajectories were used to calculate how many hours a particle of air had spent in a predefined “east Asian” box (20–45°N by 100–130°E by 0–10 km, see Figure 9), and we refer to those hours as “residence time” as a function of altitude in the box. The size and placement of the box roughly corresponds to the region of large CO concentrations as revealed from the MOPITT instrument (data available online at http://www.eos.ucar.edu/mopitt/). We calculate hourly trajectories whose start locations are every 100 m of altitude from 0 – 2900 m above ground level (AGL), and every 0.125° in a 0.5° latitude × 0.5° longitude box centered on MBO. This results in a total of 25 trajectories generated per hour for every 100 m of altitude. Hysplit calculates ground level in the grid square containing MBO to be approximately 830 m above sea level, putting the actual summit of Mt. Bachelor at about 1940 m AGL.
 The results of these calculations are shown in Figure 6. Residence times in the east Asian box are given by different colors and line styles as a function of altitude in the east Asian box. From these plots we observe that there are peaks in residence times that roughly correspond to the 25 April and 13 May ALRT events. The peak on day 116 (25 April) is the largest of the whole time period confirming that the source of this large pollution event was within the east Asian box. The residence time calculations also show that the 13 May event was much smaller (note different y axis) and was composed of two peaks – a feature that shows up in our data (see Figure 5b). There is also agreement with the 27 April event in that there is a period where residence time in the east Asian box goes to zero just before 1200 UT.
 There is an interesting result in Figure 6 that points out the potential limitation of using trajectories for this type of analysis. Note that the residence times for trajectories <1 km in the box are very low and almost insignificant. Residence times for <2 km and <3 km show much more structure. This indicates that trajectories do not resolve BL-FT exchange processes in the east Asian box very well, which is probably a function of coarse wind field resolution, and trajectory errors associated with turbulent flow and convection [Stohl, 1998]. Thus we will focus on trajectory residence times <3 km in the east Asian box as representing air masses having been in contact with the source region.
 The peak in trajectory residence times for both ALRT events shown in Figure 6 is slightly offset (∼12 hours early) compared to the maximum concentrations observed at MBO (compare to Figures 5a and 5b). Figure 7 shows that the agreement between trajectories and observations is much better when a starting altitude of about 1500 m AGL is used for MBO, instead of the 1900–2900 m AGL used in Figure 6. 1500 m AGL is approximately 440 m lower than the actual height of MBO and shows the importance of running trajectories at many altitudes in order to capture plumes that might be descending over time.
 Further evidence of the subsiding air mass on 25 April is given by total column aerosol optical thickness (AOT) measurements at 500 nm, taken at two sites in the Pacific Northwest, shown in Figure 8. Here we see that the peak in AOT occurred at approximately day 116.0, which corresponds to the peak in trajectory residence times between 2500 and 2900 m AGL in Figure 7. Thus, even though Rimrock, Idaho is to the east of MBO, a large part of the plume evidently passed over this location at higher altitudes before descending to MBO.
 Back trajectory diagrams showing long-range and local transport are shown in Figures 9–11 for the 25 April, 13 May, and 27 April events, respectively. Figure 9 shows that for the 25 April event, air masses picked up pollutants from northeastern China, Korea and/or southern Japan about 9–10 days back. Then the air masses were lifted 5–7 days back into the midlatitude westerlies, indicated by trajectories at altitudes > 10 km. These high-altitude trajectories actually cross Europe 10 days back, indicating very fast transport. Finally, the trajectories show descending air masses to MBO starting about 1.5 days back.
Figure 10 shows very tightly confined back trajectories for the 13 May event. The probable source region for this event is quite similar to that of the 25 April event and explains the relatively uniform interspecies ERs between events (see Table 3). These trajectories also show considerable lifting of air masses between days 5–7 to altitudes > 10 km, which if stratospheric air were mixed in, could explain the weaker correlations with O3 in this event. Once again, subsiding air masses are well represented by these trajectories between 0–3 days back.
Figure 11 and its inset show transport over North America during days 0–2 for the 27 April event. From these trajectories it appears that the source of this weak pollution event was most likely western Oregon (Eugene or Medford), or perhaps northern California (Sacramento). High wind speeds and turbulent BL flow probably diluted this plume considerably. Figure 11 also shows that a few trajectories enter the east Asian box and thus might influence the pollutant concentrations somewhat. However, most of the low-altitude trajectories from days 6–10 originate over the subtropical Pacific Ocean (15–25°N) and thus these air masses are subject to considerably more photochemical aging than higher altitude, northerly air masses.
 Our results from this initial measurement campaign at Mount Bachelor Observatory indicate that this site at 2.8 km ASL is suitable for observing the free troposphere. On average, our measurements reflect a mixture of BL and FT air, as determined from diurnal cycles of WV and O3 and a comparison with radiosonde data, but these periods can be effectively characterized using WV as a segregation tool. By focusing on the driest and wettest 20% of the data, we find that these represent very different types of air masses with different chemical compositions and transport histories. For example, CO, TGM, and O3 are significantly enhanced in the dry data by 25 ppbv, 0.16 ng/m3, and 13.9 ppbv, respectively. This indicates that ALRT has a large influence during the spring at this site, and that transport at higher altitudes is more efficient than at lower altitudes. The dry data also display characteristics of recently subsided air masses from higher altitudes with colder temperatures, higher pressures, and northerly wind directions. We estimate the average altitude of the dry air masses is 3.4 km on the basis of a comparison with the WV vertical profile from Medford, Oregon.
 Concentrations of CO, TGM, and σsp measured at MBO are higher than those measured from a similar altitude during previous aircraft campaigns. This finding is corroborated by other CO observations, including satellite and CMDL flask observations. The difference indicates more influence from Asian sources compared to previous years, perhaps in part because of the large Siberian forest fires of 2003.
 Three pollution events are analyzed in detail during the spring campaign: 25 April and 13 May (ALRT) and 27 April (local). Even though the 25 April event was much larger than the 13 May event, both displayed similar features. For example, the ratio of enhancement for each event between TGM and CO was 0.0045 – 0.0048 ng/m3/ppbv, similar to the value thought to be representative of Asian industrial pollution. Furthermore, the Δσsp/ΔCO enhancement ratio for each event was 0.47 Mm−1/ppbv, which suggests little biomass burning or dust influence. Meteorologically, each event was associated with very dry air and high pressure indicating subsidence from higher altitudes. These characteristics of ALRT are in contrast to the 27 April local pollution event, which showed a strong positive correlation between WV and CO, and was associated with falling pressure and temperature. The ΔTGMΔ/CO enhancement ratio was only 0.0013 ng/m3 for this event reflecting the lower density of mercury sources in the local region.
 Hysplit kinematic back trajectories were used in a retroplume approach to determine the source regions and transport pathways for these events. For each trajectory a residence time in a predefined east Asian box was calculated for different altitudes in the box. We find that large peaks in residence time occurred during the times of the observed ALRT events, and that during the local event there was almost zero trajectory residence times in the Asian source region. For both ALRT events trajectories showed how the plume descended in the FT to the altitude of MBO over time. This finding is also supported by aerosol optical thickness measurements taken in the Pacific Northwest during the same time period. The source region for the ALRT events appears to be northeastern China, Korea, and/or southern Japan about 10 days prior to arrival at MBO. Pollutants were evidently lofted to the midlatitude westerlies and transported eastward before subsiding over our region. The source region for the local event is most likely the cities of Eugene or Medford, Oregon, or Sacramento, California. However, this plume must have been diluted because of turbulence in the BL since the enhancements during this event were much smaller than for the ALRT events.
 Thus MBO has proven itself as a site for observing direct transport of Asian air pollution with very little influence from local emissions. Using a combination of meteorological observations, chemical and aerosol measurements and their enhancement ratios, and kinematic back trajectories, we can effectively resolve time periods of predominantly BL and FT air masses.
 The authors would like to thank Mount Bachelor Ski Area Inc., Isaac Bertschi, and Carol Higgenbotham for their assistance with field operations; Bryan Johnson for ozonesonde data; and Lyatt Jaeglé for helpful discussions. We also appreciate the thoughtful comments of two anonymous reviewers. This work was supported by EPA “Science To Achieve Results (STAR)” program (R-82979701) and NSF Atmospheric Chemistry program (ATM-0425262).