Nonmethane hydrocarbons at Pico Mountain, Azores: 1. Oxidation chemistry in the North Atlantic region



[1] Measurements of nonmethane hydrocarbons (NMHC) at the Pico Mountain observatory at 2225 m asl on Pico Island, Azores, Portugal, from August 2004 to August 2005 (in part overlapping with the field campaign of the International Consortium on Atmospheric Research on Transport and Transformation study) were used to investigate NMHC sources and seasonal oxidation chemistry in the central North Atlantic region. Levels of anthropogenic NMHC were characteristic of the marine free troposphere. Their concentrations were low compared to continental sites at higher northern latitudes, but higher than data reported from a similarly located Pacific mountain site at Mauna Loa Observatory, Hawaii. These higher NMHC levels are indicative of a greater influence of the adjacent continents on air composition at Pico. Substantially enhanced NMHC concentrations during the summers of 2004 and 2005 were attributed to long-range transport of biomass burning plumes originating from fires in northern Canada, Alaska, and Siberia. This finding exemplifies the continuing impact of biomass burning plumes on atmospheric composition and chemistry many days downwind of these emission sources. Seasonal cycles with lower NMHC concentrations and lower ratios of more reactive to less reactive NMHC during summer reflect the higher degree of photochemical processing occurring during transport. The NMHC concentrations indicate no significant role of chlorine atom oxidation on NMHC. Ozone above 35 ppbv was measured at Pico Mountain throughout all seasons. Enhanced ozone levels were observed in air that had relatively “fresh” photochemical signatures (e.g., ln [propane]/[ethane] > −2.5). During spring-summer air that was more processed (“older” air with ln [propane]/[ethane] < −2.5) on average had lower ozone levels (down to <20 ppbv). This relationship indicates that conditions in the lower free troposphere over the mid–North Atlantic during the spring and summer lead to net photochemical ozone destruction while air is photochemically aging during transport to Pico. This behavior contrasts to that in the mid–North Pacific where other recent studies have found that the photochemistry is more nearly ozone neutral.

1. Introduction

[2] Atmospheric nonmethane hydrocarbons (NMHC) show considerable variations on spatial and temporal scales. Their concentrations are determined by the strength of emission sources and their atmospheric removal processes, which are mostly due to reaction with the OH radical. Reaction rate constants increase with the molecular size within a given class of NMHC, causing lighter, saturated NMHC to exhibit slower atmospheric decay and longer lifetimes. Their atmospheric concentrations decline at slow enough rates that they can be measured after several days of transport to remote downwind locations. Many individual NMHC have common emission sources and their emission ratios vary comparatively little. This allows changes in absolute concentrations and NMHC ratios to be used as tools to decipher atmospheric transport and oxidation chemistry. Several researchers have investigated this utility and have presented a framework for the interpretation of NMHC concentrations, particularly for observations of the light C2–C5 alkanes [e.g., Jobson et al., 1998; Parrish et al., 2004].

[3] The possibilities for using NMHC data for investigation of atmospheric oxidation processes are particularly promising in situations where observations can be obtained in air that has traveled for extended periods of time without influence from recent emissions or surface processes. Hence, remote islands that are high enough to probe free tropospheric air offer ideal locations for this research. These considerations motivated the measurement of NMHC at the mountaintop observatory site on Pico Island, Azores. Several other previous studies have shown the influence of outflow from the North American continent on atmospheric observations made in the Azores and how measurements there can provide valuable insight into North American emissions and their processing during transport [Honrath and Fialho, 2001; Honrath et al., 2004]. NMHC measurements described in this article commenced in the summer of 2004, as a contribution to the International Consortium for Atmospheric Research on Transport and Transformation (ICARTT) campaign in the North Atlantic Region [Fehsenfeld et al., 2006], and have continued through most times when the station was on power. These data provide one of the few seasonal records of NMHC from a lower free troposphere measurement site.

[4] The first year of NMHC data were analyzed with several objectives: (1) NMHC were used to better characterize the potential influence of local sources on air composition and chemistry at the site and to investigate the degree of local (Pico) and neighboring island influences. (2) Absolute NMHC concentrations, comparison with other sites, and NMHC variability were used to investigate the frequency and impact of long-range transport of anthropogenic emissions from the source regions bordering the North Atlantic and of biomass burning plumes on air composition at the station. (3) NMHC ratio analysis and relationships between NMHC and ozone were utilized to assess the seasonal oxidation chemistry occurring during atmospheric transport across the North Atlantic region by comparison with data from other regions. Furthermore, in the companion manuscript, Honrath et al. [2008] use the NMHC measurements during April 2005 in combination with FLEXPART transport simulations to evaluate the utility of NMHC measurements as indicators of atmospheric processing and photochemical oxidation on an event-specific basis.

2. Methods

2.1. Pico Mountain Station

[5] The Pico Mountain observatory is located at 2225 m asl in the summit caldera of the inactive Pico Mountain volcano (38.47 N, 28.40 W), the highest mountain on Pico Island and in the Azores, Portugal. Intensive meteorological measurements [Kleissl et al., 2007] have led to the conclusion that upslope flow affects the Pico Mountain station much less than some other marine mountain observatories, as a result of the latitude, size, and topography of Pico Island. Summertime chemical measurements at the observatory showed very little influence from island emission sources even during uplifting events (or upslope flow), indicating that atmospheric processes at the station have negligible impact from island emissions [Kleissl et al., 2007]. Further site descriptions, data and interpretations from other research, including studies of oxidized nitrogen species, ozone, carbon monoxide and of aerosol properties at Pico Mountain have been presented previously [Honrath and Fialho, 2001; Honrath et al., 2004; Lapina et al., 2006] and in other contributions to the special ICARTT issue [Owen et al., 2006; Val Martin et al., 2006].

2.2. Chemical Measurements

[6] The remoteness of the Pico Mountain site and the limitations for power and for supply of cryogen and consumable gases determined the design of an analytical gas chromatography (GC) system that was tailored toward this unique situation. All consumable gases and blank air were prepared at the site with low-power gas generators. The instrument followed automated startup and shutdown procedures and could be remotely controlled from our Boulder, CO, offices. Outside air was continuously drawn to the instrument from a heated inlet 5 m above ground. Ozone was removed by flowing the sample air through an ozone scrubber prepared from sodium-thiosulfate-impregnated glass wool. After sample drying and NMHC focusing on a Peltier-cooled multistage solid adsorbent trap, NMHC were analyzed by thermal desorption with subsequent GC separation and flame ionization detection (FID). Quantified NMHC included ethane, propane, i-butane, n-butane, i-pentane, n-pentane and isoprene (the ethane record does not begin until in fall 2004 when some modifications in the focusing procedure allowed its quantitative analysis). Sample volumes of 600 ml (10 min collection time) and 3000 ml (50 min collection time) were collected semicontinuously (every few hours). These sampling volumes were alternated for quantification of ethane (in the 600 ml sample) and NMHC > C2 (in the 3000 ml sample), respectively. Typically, a total of 12 ambient air samples, one standard and one blank sample were analyzed per day. Data were electronically transferred to our laboratory for immediate quality control and analysis. More instrument details have been provided by Tanner et al. [2006].

[7] NMHC in ambient air samples were quantified using compound-specific FID response factors. The instrument was calibrated by regular injections of a compressed ambient air sample (breathing grade air, Airgas, Boulder, CO) that was quantified prior to shipment against numerous gravimetrically prepared hydrocarbon standards in the National Oceanic and Atmospheric Administration Earth System Research Laboratory (NOAA-ESRL), Boulder, CO. The NOAA calibration scale has previously been found to be on average within 5% agreement with that of several other laboratories in the U.S., Canada and Europe. This includes results obtained for the 60-component NMHC standard that was used in the round-robin analysis within the Nonmethane Hydrocarbon Intercomparison Experiment [Apel et al., 1994]. The quantifications in the reference gas were also compared against our own laboratory NMHC calibration scale (which was developed from a series of other gravimetrical or cross-referenced NMHC gas standards) and deviations of all quantified NMHC were <10%. A second remote ambient air reference gas (collected at Niwot Ridge, Colorado, and quantified in the same way by NOAA) was injected every 3–4 days for quality control. The primary calibration reference gas was returned to Boulder in spring 2006 and quantified again against the NOAA-ESRL gravimetric hydrocarbon standard scale. That analysis resulted in mixing ratios for the C2–C5 NMHC reported in this study that agreed within −4.2 to 2.6% with the values that were determined 2 years earlier, prior to the shipment to Pico. From these analyses, the stated ±5% accuracy of the NOAA-ESRL calibration scale, and assuming linearity over the whole measurement range, the accuracy error of the Pico measurements was estimated to be within the range of −6.5 to 5.6%. Analytical precision was estimated from 16 measurements of the breathing air reference gas over a 21-day period in April 2005. These measurements resulted in relative standard deviations of 0.7–4.2% at the mixing ratios in this reference gas. From these measurements the overall uncertainty, combining analytical accuracy and precision was estimated to be equal to or less than ±7.7% for all reported compounds, although it should be noted that this value is expected to increase for data approaching the detection limit. Detection limits were determined monthly as 3 times the integrated noise level at the peak retention times or as 2 times the standard deviation of the blank signal (in cases where peaks could be detected in the blank). From these repeated determinations, median detection limits were calculated as 17, 6, 2–4, and 1 pptv for C2, C3, C4, and C5 NMHC, respectively; during summer 2005 the C3 detection limit improved to ∼3 pptv. Ethene, propene, benzene, and toluene, while captured with this system, were excluded from the analysis because of higher and inconsistent blanks, which made their quantification at low pptv levels not feasible.

[8] Ozone was determined using a commercial ultraviolet absorption instrument (Thermo Environmental Instruments, Inc., Franklin Massachusetts, Model 49C), and CO was determined using a commercial instrument modified by the addition of a zeroing system (Thermo Environmental, Inc., Model 48C-TL). The absence of O3 loss in the inlet line was verified once per day; CO instrument calibration checks were performed at the same time. More details on these chemical measurements are provided by Honrath et al. [2004] and Owen et al. [2006].

2.3. Site Characterization

[9] Observations of isoprene were used for investigating the influence of emissions from Pico Island on the NMHC distribution and chemistry at the site. Isoprene was not detected (<1 pptv) in either winter or nighttime samples. During spring, isoprene was occasionally observed during the day. Occurrences and mixing ratios of isoprene increased during late summer; e.g., during August 2005, isoprene was detected on 60% of all afternoons, with a maximum mixing ratio of 26 pptv observed on 1 August 2005 [see Kleissl et al., 2007, Figure 8]. There is little vegetation growing on the upper ∼700 m of the slopes of Pico Mountain and the most plausible explanation for isoprene observations at the observatory is the upslope transport of air from lower island elevations. A correlation analysis between NMHC and isoprene in identified upslope events was used to investigate possible anthropogenic signatures in upslope air. N-butane was chosen as an anthropogenic tracer as butane is abundantly used on the island for domestic cooking and heating and there are no known biogenic butane sources. Human activities on Pico and most broadleaf vegetation biomass are colocated in the elevation zone (sea level to ∼1500 m) that typically is within the marine boundary layer height. This analysis was done by comparing isoprene and n-butane data in two subsets of samples. On days when isoprene was detected at the station, the mean isoprene mixing ratio (with standard deviation) during the 12–14 h (local time) window (which was the time when maximum daily values were observed) was 4.0 ± 5.7 pptv. On these same days during 22–6 h isoprene was not detected in a single sample (<1 pptv). In the same subsets of samples, n-butane was 17.1 ± 21.1 pptv during 12–14 h, and 17.4 ± 21.6 pptv during 22–6 h. Since no increase in n-butane was evident in the elevated isoprene samples, it was concluded that the identified upslope air did not have any anthropogenic signature from local sources. Most likely, the upslope air originated from elevations several hundred meters below the observatory, which are more sparsely vegetated but will likely still see some isoprene emissions. These analyses suggest that upslope air did not originate from the populated areas of the island, which are at much lower elevation along the coastline. Kleissl et al. [2007] provide a more detailed discussion of flow regimes at Pico Mountain. As no systematic enhancements of NMHC other than isoprene were seen in air that was identified as upslope flow, NMHC data were not further selected according to flow conditions.

[10] Since the alkanes and alkenes dropped to their lowest seasonal levels during the summer, during midday to early afternoon isoprene at times became the second most abundant (after ethane) NMHC in air sampled at the observatory. Given the much faster OH reaction with isoprene than with other identified NMHC, isoprene, even at these relatively low pptv levels, makes a major contribution to the overall OH reactivity from NMHC. For instance, for the 2 days with the highest isoprene mixing ratios (23, 26 pptv) during 2005, considering all C2–C5 NMHC quantified in our measurements, and using upper estimates of 3 pptv for ethene and 2 pptv for propene, we calculated that the OH reactivity from isoprene contributed, 94% (day of year (DOY) 213) and 84% (DOY 222) to the overall OH reactivity from NMHC at their mixing ratios measured on those days. This is the strongest local influence on atmospheric chemistry of air sampled at the station that we have identified so far. However, given the short atmospheric lifetime of isoprene it is clear that the episodic occurrences of isoprene are solely due to small-scale local effects and do not have an impact on the interpretations of the long-lived NMHC, whose oxidation is predominantly determined by their chemistry during long-range transport.

3. Results and Discussion

3.1. NMHC Mixing Ratios and Comparison to Other Data Sets

[11] Absolute levels, NMHC ratios, and variability of NMHC were compared with previously reported data from selected other locations for characterization of the influence of upwind emission sources and long-range transport on air composition and chemistry at the Pico Mountain station. Plots with the individual sample data (representing a total of 1958 analyzed air samples) for ethane, propane, and n-butane from August 2004 to September 2005 were presented by Tanner et al. [2006]. For a better illustration of the seasonal changes of NMHC, here we combined these data to whisker plots for monthly subsets of available measurements (Figure 1). As a first approximation the seasonal cycle of NMHC background mixing ratios can be described with sinusoidal fit curves [Rudolph, 1995], however higher-resolution data have also shown that with decreasing NMHC lifetime observed seasonal cycles deviate increasingly from this behavior, where the winter maxima become increasingly narrow and the summer minima increasingly broad [Goldstein et al., 1995; Swanson et al., 2003]. The Pico data do not quite have the temporal resolution and high number of data points to clearly demonstrate this behavior. A further constraint is that with increasing molecule size an increasing fraction of the data (in particular of summer values) fall below the detection limit. Therefore we only applied a sinusoidal regression function (least squares fit regression to the diurnal mean data), defined by y = A + B sin (day + C), to the ethane and propane measurements. These regression functions are the best description of the seasonal behavior of NMHC at the observatory and calculated A values of 985 pptv for ethane, and 185 pptv for propane are our best estimates for the annual mean mixing ratios of these two NMHC at the station; B values of 474 (ethane) and 143 (propane) define the amplitude of the seasonal cycle.

Figure 1.

Whisker plots of monthly data for ethane, propane, i-butane, n-butane, i-pentane, and n-pentane. The 5, 25, 50, 75, and 95 percentiles are indicated by the horizontal lines of each box; the vertical lines extend to the minimum and maximum observed values. The width of the box indicates the time period over which data were acquired for a given month. The thin vertical lines show the windows that were applied in the seasonal (winter, spring, summer, and fall) analysis. Data from a remote boreal forest site in Canada [Jobson et al., 1994a] and 25, 50, and 75 percentile data from Mauna Loa Observatory [Greenberg et al., 1996] are shown for comparison. Sinusoidal fit curves are included for ethane and propane, but not for the higher alkanes as their annual cycle increasingly deviates from this behavior.

[12] The measured NMHC show a distinct seasonal cycle with highest mixing ratios in the late winter and lowest values in the summer. This behavior is driven by the seasonal changes in NMHC removal rate by the OH radical, whose concentration is linked to the latitudinal solar radiation cycle. High variability in NMHC mixing ratios was observed at any given time of year. It is noteworthy that all of these features show relations and dependencies upon the individual NMHC reactivity with OH and the resulting NMHC lifetime. The longest-lived NMHC, ethane, (having an atmospheric mean lifetime of ∼2 months) shows the relatively smallest amplitude between the mean winter and summer mixing ratios and the smallest relative variability on short (e.g., weeks) time scales. These features increase with increasing molecule size (shorter OH lifetime, e.g., ∼3 days for n-pentane). The seasonal maximum and minimum of ethane occur the latest of all compounds (early March and early September, respectively, determined from the timing of the minimum and maximum of the best fit curve). Because of its slower OH reaction, ambient ethane levels respond with a longer delay to the seasonal OH cycle. Heavier NMHC were found to maximize as early as mid-January and minimize as early as mid-July. The similarity of the seasonal behavior of NMHC between the remote island data from Pico and a series of continental Northern Hemisphere sites [e.g., Jobson et al., 1994a; Goldstein et al., 1995; Gautrois et al., 2003] further underscores that the levels of these NMHC in background air are determined by their hemispheric sources and seasonal oxidation chemistry.

[13] A number of other NMHC records have been presented in the literature. Here we selected two particular data sets for comparison and to highlight the most prominent features in the NMHC data from Pico Mountain. The data included in Figure 1 are measurements made from September 1991 to August 1992 during the Mauna Loa Observatory Photochemical Experiment–2 (MLOPEX-2, at 19°N, 155°W) [Greenberg et al., 1996] and from April 1990 to October 1992 at the continental, remote boreal site in Fraserdale, Ontario (50°N, 82°W) [Jobson et al., 1994a]. The MLOPEX-2 data are of particular interest as they allow a comparison of the conditions in the mid-Pacific with the Atlantic Pico site. Similar to the Pico Mountain observatory, Mauna Loa (MLO) is a remote mountaintop island location, where free tropospheric air is sampled that has traveled over the ocean for several days. MLO has a more prominent diurnal upslope-downslope cycle than Pico and data in identified upslope air shows a clear influence from island emissions. Data presented by Greenberg et al. were divided into the occurrences of these two flow regimes. Included in Figure 1 are the 25/50/75 percentiles for the time periods spanned by the width of the boxes during downslope (i.e., free tropospheric air) flow. Upslope data for MLO typically were higher, with relative enhancements increasing with decreasing molecule lifetime. Pico data are consistently higher for all NMHC and during all seasons. The relative differences between Pico and MLO NMHC mixing ratios increase with decreasing NMHC lifetime, e.g., while ethane mixing ratios are ∼20% higher, n-butane values at Pico are more than 5 times higher than at MLO, which reflects on average a shorter source-receptor transport time to Pico compared to MLO. In contrast to MLO, the comparison with Fraserdale shows that Pico experiences overall lower NMHC mixing ratios than this low-elevation continental site. Again, differences in these two data sets become more pronounced with increasing molecular weight, but in this case with the Pico data becoming increasingly lower.

[14] The general trend with [NMHCMLO] < [NMHCPico] < [NMHCFraserdale] is likely due to several reasons. Probably of greatest importance is the distance of these sites from NMHC sources, in particular to continental areas, which is about two times greater for MLO than Pico. The longer transport results in longer photochemical processing times and more depleted NMHC concentrations at MLO compared to Pico. Second, aircraft profiles have shown that NMHC mixing ratios generally decline with height within the free troposphere [e.g., Blake et al., 1997]. Both MLO and Pico, because of the small island size and high elevation, behave to some extent like tower platforms. MLO is about 1200 m higher in elevation than the Pico Mountain station. Consequently NMHC mixing ratios in downslope air are expected to be lower at MLO than at Pico. Third, NMHC mixing ratios in the lower troposphere decrease toward lower latitude [Rudolph, 1995]. Again, this dependency implies higher NMHC levels at Fraserdale (50°N), followed by Pico (38°N) and MLO (19°N). This spatial distribution also reflects chemical oxidation since OH has a latitudinal gradient. Further comparisons of the Pico data with several other NMHC data sets from higher northern latitudes in Canada, the Atlantic Region and Europe (as summarized by Gautrois et al. [2003]) show that Pico NMHC levels are without exception lower, both during the winter and the summer, compared to the further northern locations. One other point to consider is that possible temporal trends in NMHC may bias this site comparison as both the MLO and Fraserdale data are 12–15 years older than our Pico measurements. Unfortunately, reports of NMHC trends at remote background sites are scarce and do not allow a conclusive evaluation of long-term trends of NMHC concentrations. Measurements made in Finland have shown decreasing levels of shorter-lived compounds and increasing trends of longer-lived NMHC [Hakola et al., 2006]. In source regions in Europe and the U.S. NMHC emissions and resulting ambient air mixing ratios have generally been decreasing over the past decade [U.S. Environmental Protection Agency, 2003; Stemmler et al., 2005; Plass-Duelmer and Berresheim, 2006], which suggests that Pico, being downwind of these areas (mostly U.S.) would be experiencing somewhat lower NMHC levels now than in the recent past.

[15] Figure 2 compares the cumulative distributions of the NMHC mixing ratios during fall 2004 (22 September to 20 December), winter 2004–2005 (21 December to 19 March), spring 2005 (20 March to 20 June) and summer 2005 (21 June to 21 September). In these analyses the median value is located at the center of the y axis with a logarithmic scale extending both to higher and lower values such that lognormally distributed data define a linear distribution on the graph. The y axis scale is designed to be linear in units of the standard deviation of the distribution. Lacking data for particular NMHC in the lower percentage ranges result from respective fractions of these data falling below the instrument detection limit (for instance, i- and n-pentane were below the detection limit in ∼4% of the measurements during fall 2004, whereas during summer 2005 ∼85% and 60% of chromatograms could not be quantified for i- and n-pentane, respectively). The regression line slopes through these distributions indicate the variability of the atmospheric mixing ratios of the respective compounds. Deviations from linearity indicate higher mode contributions to the distribution, which may imply different behavior of NMHC data in air sampled from different sources or at different times. Calculated linear regression coefficients ranged from 0.95 to 0.99, indicating that most of the variability is lognormally distributed. There are no obvious differences in the quality of the fit between the seasons, which may indicate similarity in source strengths and removal mechanisms between seasons. An interesting feature is that at the high end of the concentration range of each NMHC, measured mixing ratios are often lower than expected from a purely lognormal distribution. At this point we are not certain of the interpretation of this feature, but it does provide further evidence for the absence of strong local sources, which would be expected to yield data deviating toward higher concentrations. Steeper slopes are observed for longer-lived NMHC (e.g., ethane) as their longer lifetimes reduce the relative variability caused by emission and aging influences. Regression line slopes are consistently lower for the summer, which indicates higher variability resulting from the shorter seasonal lifetime, lower absolute concentrations, and the relatively stronger influence of perturbations from different histories of transport and photochemical aging.

Figure 2.

Cumulative distribution of NMHC mixing ratios at the Pico Mountain station during the four measurement seasons. The linear scale of the y axis in these graphs reflects the number of standard deviations from the median value in each data set. For better graphical presentation this has been plotted as the percent contribution, where the median (50% of distribution) corresponds to 0 on the number of standard deviation scale. The regression equations in these graphs use the value of the standard deviation scale as the y value (e.g., y = −1 at 16.9% on the percent of distribution scale, y = 0 for 50% on the percent of distribution scale, and y = 1 for 84.1% on the percent of distribution scale) and the ln (mixing ratio) as the x value. The ln (median value) in pptv of each distribution can therefore be calculated by setting y = 0.

3.2. Biomass Burning Influences on NMHC Concentrations at Pico

[16] As discussed in detail in other contributions to the ICARTT issue, the summers of 2004 and 2005 were characterized by an unusually high occurrence of boreal wild fires at high northern latitudes. Three previous publications [Val Martin et al., 2006; Lapina et al., 2006, 2008] investigated nitrogen oxides, carbon monoxide emission ratios, and ozone chemistry in identified boreal biomass fire plumes originating in North America and Siberia, and transported over 6–15 days to the Azores. The majority of the fire plumes observed at Pico had a well defined, detailed structure and could be identified from the short-term variability in CO and NOy. Comparison of NMHC data from within and outside of the 2004 and 2005 fire events (see Lapina et al. [2008] for the dates and times of these events and for additional information on event identification) consistently show enhancements of NMHC levels during these identified episodes of biomass burning. A summary and comparison of the C2–C4 NMHC data for 2005 is given in Table 1. Mixing ratios for the NMHC in fire plumes increased significantly, with medians up to a factor of 3 higher for propane and the butanes. Comparison of propane fire event data for summer 2004 (not shown) with the data from fire events in 2005 shows overall even higher mixing ratios during 2004, suggesting that the identified fire events in 2004 brought air with higher NMHC enhancements to the station than in the following year. The enhanced NMHC mixing ratios in the fire plumes during both years underscore the conclusions derived from observations of CO, NOy and black carbon, that boreal biomass burning emissions continued to affect atmospheric composition and oxidation chemistry after 1–2 weeks of transport to the Azores region.

Table 1. Comparison of Ethane, Propane, and Butane Isomer Mixing Ratios Seen Outside and Within Fire Events During Summer of 2005
No Fire EventFire EventsNo Fire EventFire EventsNo Fire EventFire EventsNo Fire EventFire Events
Number of samples31533292312902529025
Mean mixing ratio ± standard deviation (pptv)527 ± 177834 ± 16836 ± 3680 ± 241.0 ± 3.44.5 ± 2.68.9 ± 10.911.9 ± 5.8

3.3. Evaluation of the Influence of NMHC Sources and Their Distance From Pico Using NMHC Variability-Lifetime Relationship

[17] The relationship between the variability of NMHC and their lifetimes can be used to characterize the degree of influence of local emissions on the air composition at a given site [Jobson et al., 1998, 1999]. Here we use this analysis to further investigate potential local emissions and transport from other Azores islands versus those from distant regions on air composition at Pico. The variability of NMHC (expressed as σlnx, the standard deviation of the natural logarithm of all measurements) has been found to show linear behavior when plotted against the estimated atmospheric lifetime (τ) in a double-logarithmic plot. The regression line through these data gives the relationship σlnx = A τ–b. The derived A and b coefficients from the regression line analysis have been used to characterize the exposure to emission sources or remoteness of measurement locations [Jobson et al., 1998, 1999]. Using one other atmospheric component with known atmospheric lifetime, best fit analysis through the combined data has also yielded estimates for mean OH radical fields during transport of air to the measurement site [Ehhalt et al., 1998; Williams et al., 2000, 2001].

[18] While previous studies have applied this relationship to characterize data sets from mostly shorter campaigns, the wealth of the Pico data offers an opportunity to test for possible seasonal variations in this behavior using a full year of data. In Figure 2 the variability of each NMHC during each of the four seasons is reflected by the slopes of the regression line fit to the cumulative distribution plots, with the inverse of each slope giving the corresponding σlnx. An estimate of the seasonal lifetime, τ, of each NMHC was obtained by averaging the lifetime corresponding to each NMHC measurement made during the season. The lifetime was calculated from the product of the NMHC OH reaction constant and the OH radical concentration estimated according to [Goldstein et al., 1995]:

equation image

using values of A° = 1.6*106 cm−3 and B = 0.80 for 800 hPa, 36.0°N, 27.5°W [Spivakovsky et al., 2000], and t is time of year in days. Reaction rate constants were adjusted to the temperatures measured at the Pico Mountain station for the respective measurement times. Note that this local lifetime represents an estimate for the conditions at the receptor site. The true NMHC lifetime may deviate from this estimate dependent on the conditions encountered during transport to Pico.

[19] The variability-lifetime relationships with linear regressions providing solutions for σlnx = A τ–b are shown in Figure 3. In each seasonal data set NMHC σlnx values are well correlated with the seasonal lifetime estimates; R2 values range from 0.91 to 0.99. The calculated A and b values for the four subsets of seasonal data range from 1.4 to 1.9 for A and 0.39 to 0.44 for b. The absence of detectable differences between the seasonal data sets suggests a similar behavior of the influences that determine NMHC variability throughout the year. The best fit linear regression analysis through the combined seasonal data yielded σlnx = 1.56 τ–0.38. The exponent b in this equation has been taken to describe the relative importance of source and sink terms in determining the regional variability of species concentrations. Interpretation of observed values from different sites has shown that b approaches 0 near urban areas, where the variability is strongly influenced by differences in the strength of local emission sources. Values of b close to 1 were found in stratospheric data, where the variability is dominated by chemical loss alone [Jobson et al., 1998, 1999]. Interestingly, the Pico value is close to results from three other free tropospheric data sets from marine environments, even though those data resulted from shorter observational periods; b values in data from the Arctic Boundary Layer (ABLE3A), the equatorial Atlantic (TRACE-A), and the western Pacific (PEM-West B) experiment all ranged from 0.46 to 0.53 [Jobson et al., 1999]. This comparison illustrates a remarkable consistency of these variability-lifetime analysis results between the comparatively short marine troposphere aircraft campaigns and the much longer, continuous, seasonal Pico data. These findings further indicates the suitability of Pico Mountain for studying the atmospheric chemistry of the remote, lower free troposphere.

Figure 3.

The standard deviation of the natural logarithm of NMHC mixing ratios during the four seasons at their estimated seasonal OH lifetime. In the top right corner results for the two-sided linear regression analysis on seasonal subsets of the data in linear coordinates are given in the same colors as the corresponding data points in the graph.

3.4. Analysis of Atmospheric Processing of NMHC Using NMHC Ratios

[20] The relationships between concentrations of different NMHC observed at Pico are dependent upon the NMHC source emission ratios and upon the atmospheric processing that occurs during transport from the emission region to Pico. In this section we present analyses of several NMHC ratios that are particularly useful for providing atmospheric processing information. NMHC ratios are relatively insensitive to changes from mixing during transport; consequently NMHC ratios are better indicators for studying chemical processing during transport than are NMHC concentrations alone. However, such analyses rely on the assumption that variations of emission ratios of NMHC pairs are relatively small in source regions. A good body of NMHC data from urban areas supports this assumption; however there have been a few reports that point toward seasonal changes of NMHC emission ratios in source regions [Greenberg et al., 1996; Swanson et al., 2003; Lee et al., 2006]. Thus, a crucial step in each analysis will be the evaluation of emission ratios. In the ratio analyses presented here we will follow a common convention: the less reactive NMHC will be placed in the denominator of the ratio, so that the ratio will decrease as NMHC removal processes progress.

3.4.1. Ratios of Isomeric Pairs of NMHC

[21] The OH reaction rate constants of the isomeric pairs i-butane and n-butane are very similar; consequently, the ratio of these two compounds is expected to change comparatively little during oxidation by OH. The i-butane/n-butane correlation in the data from Pico, differentiated by the time of year is shown in Figure 4 (top left). Here we use a double-logarithmic plot to better display the behavior at both low and high concentration values. A tight relationship between the concentrations of these two isomers is obvious. Other than for a few outliers and an increase in scatter at lower mixing ratios (which likely is due to a loss of precision at lower mixing ratios), the two butane isomers show no obvious change in their correlation throughout the course of the year. Subsets of the data for the four seasons showed no difference in their correlation; a two-sided linear regression analysis of the four, linearly plotted seasonal data subsets gave slope (with 95% confidence interval), intercept (in pptv), and R2 values of 0.52 ± 0.04, −1.6, 0.91; 0.49 ± 0.01, 3.9, 0.98; 0.49 ± 0.02, −1.0, 0.89; and 0.44 ± 0.08, −0.3, 0.55, for fall-summer. The regression through all data gave 0.52 ± 0.01, −0.9, 0.96, and the geometric mean ratio (and standard deviation) was 0.46 ± 0.12. Similar values (range 0.37–0.55) have been reported in data from a multitude of other sites in both continental and marine environments [e.g., Bottenheim and Shepherd, 1995; Bottenheim et al., 1997; Greenberg et al., 1996; Parrish et al., 1998].

Figure 4.

(top left) Mixing ratio of i-butane versus n-butane and (top right) ratio of i-butane/n-butane versus n-butane and (bottom left) mixing ratio of n-pentane versus i-pentane (left) and ratio of n-pentane/i-pentane versus i-pentane. The circles with the error bars in Figure 4 (right) show the geometric mean and standard deviation of the data within 10-percentile bins of the data.

[22] Figure 4 (top right) investigates the behavior of the [i-butane]/[n-butane] ratio as a function of total concentration (of n-butane). The 10-percentile bins of the data do not show any significant difference in the [i-butane]/[n-butane] ratio, indicating constant behavior over the range of mixing ratios observed at the station. It is noteworthy that the constant butane isomer ratio contrasts with results from the polar marine boundary layer, where episodic, concentration-dependent increases in [i-butane]/[n-butane] have been observed during springtime solar sunrise. These increases have been explained by the influence of chlorine atom oxidation, which is ∼50% faster for the n-isomer [Jobson et al., 1994b; Hopkins et al., 2002]. The absence of increasing [i-butane]/[n-butane] ratio with increased processing in the Pico data indicates that chlorine atom oxidation plays no significant role during transport of air from continental source regions to Pico. The Pico data further demonstrate that low variability in the [i-butane]/[n-butane] ratio is seen in all seasons throughout the troposphere of the Northern Hemisphere, with the exception of the springtime polar marine boundary layer.

[23] The results of this analysis for the two pentane isomers, which also have similar OH reaction rate constants, indicates behavior different from the butane isomers (Figure 4, bottom left). The two-sided regression analysis through the linearly plotted data yielded slope (with 95% confidence interval), intercept and R2 values of 0.69 ± 0.08, 0.6 pptv, 0.93. The geometric mean ratio (and standard deviation) of [n-pentane]/[i-pentane] for all included data was 0.78 ± 0.31. The geometric mean ratio is higher than the regression slope because increasingly higher [n-pentane]/[i-pentane] ratios were observed moving from the higher mixing ratio percentiles (mostly winter data) toward the lower percentiles of the data (mostly spring and summer data). The geometric mean ratio increased from 0.64 ± 0.11 in the upper 10 percentile of the data to a value of 1.15 ± 0.51 in the lowest 10 percentile. A change in the seasonal behavior of the pentane isomers is also apparent in the cumulative distribution plots (Figure 2). While the i-pentane distribution shows the higher values in the fall and winter samples, the summer distribution is reversed, with i-pentane being found at lower mixing ratios. The pentane isomers are the only pair of compounds that exhibit such a switch of positions during the course of the year. Also, during the summer there were a considerable number of samples that had relatively high n-pentane (8–45 pptv), but with i-pentane below the detection limit of 1 pptv. A loss of correlation of pentanes with other NMHC, and between the two pentane isomers, is also evident in the lower R2 values for the summer data. For summer samples with both i- and n-pentane data (n = 34) above the detection limit, the i-pentane mixing ratio and standard deviation was 3.7 ± 2.0 pptv, and the corresponding value for n-pentane was 3.0 ± 1.5 pptv. Unfortunately, the interpretation of the pentane data is somewhat limited by the measurement sensitivity, which was not sufficient to detect either one or both of the pentane isomers in many of the summer samples. For n-pentane, only 41%, and for i-pentane an even smaller fraction (14%) of the summer measurements were above the detection limit. Only 11% of the summer samples had detectable levels of both pentane isomers. Using this fraction of the data, and assuming lognormal distribution (which most of the data follow), the regression equations included in the cumulative distribution plot in Figure 2 can be used to calculate the median value for each seasonal subset. This analysis yields values of 0.78, 0.65, 0.98, and 2.26 for [n-pentane]/[i-pentane] during fall, winter, spring and summer, respectively, which further supports the observation of increases in this isomer ratio during the summer. It needs to be emphasized that these interpretations are based upon a small fraction of the overall summer data set, and this subset is mostly representative of the higher concentration range during that time. However, this remarkable behavior of the pentane isomers in the data from Pico is in agreement with a similar behavior observed at other remote sites. Enhanced [n-pentane]/[i-pentane] ratios during the summer months and in low-concentration (well aged) samples were also evident in the Fraserdale data [Jobson et al., 1994a] as well as during ICARTT in the WP-3D data set, and in a 2006 data set from the Research Vessel Ronald H. Brown (D. Parrish, unpublished data, 2007).

[24] Next, we investigate if seasonal changes in the n-pentane/i-pentane emission ratio from anthropogenic sources play a role in the seasonal changes in the n-pentane/i-pentane concentration ratio seen at Pico. Boundary layer data from sites near anthropogenic source regions provide a good indication of this emission ratio. The [n-pentane]/[i-pentane] ratio in the year-round (2005) data set from Hohenpeissenberg, a rural area in southern Germany, averaged 0.58 ± 0.15 (median ± standard deviation) for January and 0.37 ± 0.10 for July [Plass-Duelmer and Berresheim, 2006; C. Plass-Duelmer and H. Berresheim, personal communication, 2007]. Similar behavior is seen in data from North American sites, e.g., measurements at four rural locations in the southeastern United States gave a ratio of 0.60 ± 0.05 for winter, and 0.43 ± 0.13 for summer (mean ± standard deviation of median seasonal data from four sites [Hagerman et al., 1997]), and the extensive NMHC data from Harvard Forest indicate a ratio of 0.45 in winter and approximately 0.35 in summer [Lee et al., 2006]. There are two striking features in the comparison of these emission ratios to the observed ambient concentration ratio at Pico. First, the absolute values of Pico [n-pentane]/[i-pentane] ratios are higher than in the source region data, and second, the seasonal tendency is reversed: while the urban/continental [n-pentane]/[i-pentane] data generally decrease toward the summer, an increase is seen at Pico and at the other above mentioned remote sites. Clearly, seasonal differences in emission ratios from anthropogenic sources cannot explain the magnitude and seasonal behavior of the observed [n-pentane]/[i-pentane] ratios at the remote sites.

[25] Also, photochemical processing of the two pentane isomers appears unlikely to play a role in the evolution of their concentration ratio. If OH oxidation dominates as is generally expected, a decrease in the [n-pentane]/[i-pentane] ratio with increasing transport time would be expected, because of the slightly higher n-pentane OH reaction rate constant (3.8 × 10−12 cm3 molecule−1 s−1 at 298 K) compared to i-pentane (3.6 × 10−12 cm3 molecule−1s−1) [Atkinson and Arey, 2003]. Reports in the literature have suggested significant roles for chlorine atoms and NO3 radicals in NMHC oxidation [Arsene et al., 2007; Penkett et al., 2007]. However, chlorine atoms, as OH, generally react with n-isomers more rapidly than branched isomers. Thus, any contribution from chlorine reactions would contribute to a further reduction of [n-pentane]/[i-pentane] during photochemical processing. In contrast, oxidation of NMHC by NO3 would shift the isomer ratio to larger values because the reaction of i-pentane is about two times faster [Atkinson and Arey, 2003]. However, a first-order estimate of the [NO3] needed to account for the observed shift in [n-pentane]/[i-pentane] resulted in unrealistically high NO3 levels. In summary, NMHC processing in the atmosphere does not support the larger [n-pentane]/[i-pentane] ratio observed at Pico.

[26] Although anthropogenic emission sources are generally thought to dominate the atmospheric concentrations of the light alkanes, there are other sources that could possibly alter the relative ambient concentrations of these species, especially at the low concentrations characteristic of the Pico summertime data set. In particular, a stronger influence of biomass burning activities during the summer could shift the pentane isomer ratio more toward the n-isomer. During summer 2005 the geometric mean (with standard deviation) of [n-pentane]/[i-pentane] in 22 samples that had both isomers above the detection limit outside of fire events was 0.62 ± 0.33; this value is similar to the anthropogenic emission ratio. In contrast, in 12 samples that had both isomers above the detection limit during fire events, the ratio was 1.37 ± 0.55, more than twice as high. (For comparison, the [i-butane]/[n-butane] ratio outside of fire events was 0.37 ± 0.12 (n = 75) compared to 0.38 ± 7 (n = 25) during fire events.) The n-pentane/i-pentane emission factor ratio is about 2 for extratropical forest fires [Andreae and Merlet, 2001], which account for the primary impact at Pico [Val Martin et al., 2006; Lapina et al., 2006, 2008]. This ratio is 3–4 times higher than the reported emission ratios from urban areas. In contrast the i-butane/n-butane emission factor ratio from these fires is about 0.32, in much closer agreement with the emission ratio from anthropogenic sources. Thus, a strong influence of forest fire emissions in the summer is a plausible explanation for causing the summertime shift in the pentane isomer distribution, without a corresponding shift in the butane isomer ratio.

[27] Oceanic emissions of NMHC is another source that could conceivably affect [n-pentane]/[i-pentane]. Hopkins et al. [2002] measured NMHC concentrations in the Arctic marine boundary layer. In air masses that had been isolated from anthropogenic sources for extended periods, they found [n-pentane]/[i-pentane] ratios near 2. They attribute these relatively high ratios to the influence of emissions from the ocean. The butane isomer ratio was not markedly different from either anthropogenic emissions or forest fire emissions. It should be noted that Hopkins et al. [2002] did not directly measure NMHC fluxes from the ocean; they simply argued that oceanic emissions were the likely source in the isolated marine Arctic region. They also did not consider the possible effect of the transport of boreal forest fire emissions. As discussed above, the Pico site is not influenced by local boundary layer sources. However, there is the possibility of exchange of marine boundary layer air with the free troposphere over the days of transport upwind of Pico and it is possible that oceanic emissions have an influence in determining the elevated [n-pentane]/[i-pentane] ratios observed at Pico. Our current observations provide stronger support for the biomass source. However, as mentioned above only a small fraction of the summer NMHC measurements at Pico had both pentanes above the detection limit, so the preceding analysis is based on a small fraction of the measurements. At these low measured concentrations biases in the analyses may be possible. Subsequent measurements with higher sample volume collections and additional analytical tests were conducted at Pico in 2006 and are planned for 2008–2009. With the higher sensitivity expected in these new measurements a stronger data set is anticipated, which, together with FLEXPART transport studies, will be applied for a reevaluation of the seasonal trends in the [i-butane]/[n-butane] and [n-pentane]/[i-pentane] ratios with a focus of quantifying the relative contribution of anthropogenic, forest fire, and oceanic emissions.

3.4.2. Photochemical Processing as Indicated by NMHC Ratios

[28] By working with the ratio of pairs of hydrocarbons the influence of mixing on the composition of a given air parcel can be minimized and the influence of photochemical processing can be most clearly examined. Figure 5 investigates the seasonal dependence of NMHC oxidation by plotting [propane] and [n-butane] against ln([propane]/[ethane]) and ln([n-butane]/[ethane]). Smaller values of these ratios, along with smaller absolute NMHC mixing ratios, are seen in the summer, and larger values occur in the winter. A notable feature is that at a given [NMHCa]/[NMHCb] ratio, lower absolute mixing ratios of the faster reacting NMHC are observed during the earlier part of the year than during the later part. This hysteresis behavior arises from the different seasonal behavior of individual NMHC. If the seasonal cycle of all NMHC were in phase, then data in these plots would be expected to fall on one line throughout the year. The differences of the times for the seasonal maxima and minima between the NMHC plotted on the y axis and the ratio that is plotted on the x axis determines the spread of the data in the x-y domain. This analysis further illustrates the shift in the seasonal maxima and minima for individual NMHC (moving toward earlier in the year with decreasing atmospheric lifetime) as discussed above in section 3.1.

Figure 5.

Relationship of propane and n-butane as a function of the (top) ln([propane]/[ethane]) and (bottom) ln([n-butane]/[ethane]) with the color coding representing the time of the measurement.

[29] The distribution of NMHC in a double natural logarithm plot of [n-butane]/[ethane] versus [propane]/[ethane] is particularly useful to investigate photochemical processing and mixing that occurs during atmospheric transport from source regions to remote sites [Rudolph and Johnen, 1990; McKeen and Liu, 1993]. Photochemical processing of NMHC in the atmosphere is expected to follow first-order kinetics, in which case atmospheric NMHC mixing ratios will decrease exponentially with the extent of photochemical processing. Thus, when the logarithms of these NMHC ratios are plotted, larger values indicate a lesser extent of photochemical aging, and smaller values indicate a greater extent of aging (this is discussed in more detail in the companion paper [Honrath et al., 2008]). The Pico data shown in Figure 6 fall between two theoretical limits. The steeper “kinetic” line with a slope of 2.61 indicates the evolution of isolated air parcels from the assumed source region emission ratios of these compounds (indicated by the black diamond) when subjected to oxidation by OH radicals. Here, we used the emission ratios of 0.63 for [propane]/[ethane] and 0.35 for [n-butane]/[ethane] [Parrish et al., 2007], which was derived from the mean of the Goldstein et al. [1995], and Swanson et al. [2003] results. Applied rate constants were k = 0.18 × 10−12, 0.89 × 10−12 and 2.05 × 10−12 cm3 molecule−1 s−1 for ethane, propane and n-butane, respectively (Atkinson and Arey [2003], with T = 273 K). The less steep “dilution” line indicates the effect of dilution of an air parcel starting at the same emission ratio when mixing with “background” air that has aged to the point that there are negligible concentrations of the more reactive NMHC (propane, n-butane) in the numerators of the ratios.

Figure 6.

Relationship between the natural logarithms of [n-butane]/[ethane] versus [propane]/[ethane] for all fall 2004 to summer 2005 Pico NMHC data with the seasonal dependency illustrated by the color bar. The solid black diamond marks the assumed source emission ratio. The data are bound by two lines representing the behavior of NMHC pairs assuming a sole dependency on OH oxidation (kinetic slope) and sole dependency on mixing, where background mixing ratios were assumed negligible for propane and n-butane. Four two-sided regression lines for the four seasonal subsets were added in colors corresponding to the times of year indicated by the color bar. Also added are source emission ratios from other studies, with (1) Boston, New York City, ratioed to acetylene, 5 July to 12 August 2004 [Warneke et al., 2007]; (2) Boston, New York City, ratioed to acetylene, 12 July to 10 August 2002 [Warneke et al., 2007]; (3) Boston, New York City, ratioed to CO, 5 July to 12 August 2004 [Warneke et al., 2007]; (4) Los Angeles, April–May 2002 [Warneke et al., 2007]; (5) Boston, New York City, August 2003 [Warneke et al., 2007]; (6) average of 39 U.S. cities, June–September 1984–1986 [Seila et al., 1989; Warneke et al., 2007]; (7) Harvard Forest, summer, August 1992 to July 1994 [Goldstein et al., 1995]; (8) Harvard Forest, August 1992 to July 1994, winter [Goldstein et al., 1995]; (9) source emission estimates derived by ratioing the slope of the fall accumulation rates observed at Summit, Greenland, June 1997–1998 [Swanson et al., 2003]; and (10) mean ratio seen in January 2005 flask samples from the NOAA Cooperate Flask Network for samples collected at 30–60°N (J. Pollmann, unpublished results, 2006).

[30] The seasonal differences in the degree of NMHC oxidation are clearly visible in these plots. During winter, most data have larger ratios, fall closer to the assumed emission ratios, and are less variable. This behavior indicates the lower degree of photochemical processing in that season. In contrast, in spring and summer the lower [NMHC]/[ethane] ratios are indicative of the higher degree of NMHC oxidation that occurred during transport in these seasons, and the data are more scattered. The two-sided regression analysis of the four seasonal subsets of these data resulted in slopes (with 95% confidence interval and (R2)) of 1.65 ± 0.16 (0.82), 1.57 ± 0.07 (0.91), 1.57 ± 0.11 (0.72), and 1.46 ± 0.18 (0.55) for the fall, winter, spring, and summer, respectively. The regression through all data yielded 1.60 ± 0.04 (0.89). All of these slopes fall within the range of slopes (1.44–1.78) from the eleven data sets summarized by Parrish et al. [2007].

[31] Seasonally varying emission ratios may make a significant contribution to the apparent difference in the linear fits in Figure 6. A number of available determinations of regional emission ratios are included in Figure 6. The fit to the wintertime Pico data agrees well with several of these determinations. However, data from a number of other studies deviate to a greater extent, with the earlier data from Seila et al. [1989] falling markedly above the Pico regression lines, and most of the recent midsummer data from the Boston area [Warneke et al., 2007] falling below it. It is notable that the Pico summer measurements deviated the most from the assumed source emission ratio, and that the summer regression line most closely approaches the summertime emission ratios of Warneke et al. [2007] The better agreement between the Pico summer data and the Warneke et al. [2007] summer emission ratios suggest that Pico data may, to some degree, reflect emission ratio changes in NMHC seen in the northeastern U.S.; for example, a lower summertime [n-butane]/[ethane] emission ratio is also supported by the recent analysis of Lee et al. [2006].

3.5. Relationship Between NMHC Processing and Ozone

[32] In general, understanding the temporal variability of tropospheric ozone at any particular location is complex because several processes can have significant impacts, and these impacts vary strongly on different time scales. In situ photochemical production and destruction proceed at rates that vary with the ambient levels of ozone precursors, and variables such as sunlight and water vapor concentration. Surface deposition and destruction by reaction with local emissions of NO or reactive NMHC can drastically reduce near-surface ozone concentrations at rates that vary with the flux of local emissions and the structure and evolution of the planetary boundary layer. Transport of ozone from the stratosphere or from upwind regions of strong photochemical production can greatly increase ozone concentrations. However, if the effects of local influences (surface deposition, reaction with emissions) and transport of stratospheric ozone can be eliminated, then concurrent measurements of NMHC ratios and ozone provide information on the net effect of photochemical ozone production or loss; Parrish et al. [2004] show that the relative change of ozone with NMHC ratio evolution is an indication of ozone production or loss during long-range transport.

[33] The Pico Mountain site is a good location to isolate the effects of the regional photochemical production and destruction in the central North Atlantic. Previous discussion in this paper and by Kleissl et al. [2007] have shown that air sampled at the site is generally characteristic of the lower free troposphere, with negligible impacts from local processes such as surface deposition, destruction by reaction with local emissions, or local emissions in most cases. The varying influence of the transport of stratospheric ozone often dominates the variability of ozone in free tropospheric data sets. Since ozone from the stratosphere has a steep, negative correlation with CO [see, e.g., Danielsen et al., 1987], the influence of stratospheric ozone transport can be evaluated from the correlation of ozone with CO. Honrath et al. [2004] discuss the strong, positive ozone-CO correlation observed at Pico Mountain; Figure 7 of Honrath et al. [2004], which presents these correlations, shows only a few scattered points with relatively high ozone at low CO that may be due to stratospheric influence. Thus, consistent with other Pico analyses [Lapina et al., 2006; Owen et al., 2006; Val Martin et al., 2006], Honrath et al. [2004] concluded that events of elevated ozone were dominated by periods of long-range transport from North American source regions.

[34] Parrish et al. [1992, 2004] demonstrate that NMHC aging correlates with increasing ozone when photochemical ozone production dominates, and correlates with decreasing ozone when ozone destruction processes dominate. These previous studies focused upon marine boundary layer observations to avoid the confounding influence of stratospheric ozone. Given the low stratospheric influence at Pico the same analysis can be directly applied to the Pico data. Figure 7 shows the dependence of ozone concentrations on the natural logarithm of [propane]/[ethane] as the indicator of the photochemical processing. During fall and winter both the [propane]/[ethane] ratio and the ozone concentrations are relatively constant with no clear dependence on the NMHC ratios. In spring and summer ozone has a higher variability, both in the higher and lower range of photochemical processing, and an increasing correlation between the degree of NMHC processing (as indicated by the [propane]/[ethane] ratio) and the ozone concentrations is observed. Higher ozone levels were more consistently observed in air that had relatively “fresh” photochemical signatures (i.e., ln [propane]/[ethane] > −2.5), and overall lower ozone levels were seen in more processed air (i.e., ln [propane]/[ethane] < −2.5).

Figure 7.

Ozone in relation to the natural logarithm of [propane]/[ethane]. The two lines indicate the linear least squares fit regression lines to the natural logarithm-transformed NMHC ratios for spring and summer, with the end points of both lines spanning the range of observed x values. The slopes of these regression lines with their 95% confidence limits and correlation coefficients are 0.72 ± 0.07 and R2 = 0.43 for spring and 1.02 ± 0.12 and R2 = 0.22 for summer, respectively.

[35] The short-term variability of this behavior further illustrates how, in addition to the seasonal dependency, changes in transport time influence the ozone chemistry-NMHC relationship. In Figure 8 the seasonal record of ln([propane]/[ethane]) is plotted, with the data points being color coded according to the ozone mixing ratio observed at the time of measurement. This analysis shows the dynamic range of photochemical processing at a given time of year (on the order of 1–3 unit differences in ln([propane]/[ethane]) and it reemphasizes how ozone variability and the dynamic range of NMHC ratios increases from winter to spring-summer. It furthermore shows that lowest ozone levels are observed during events when air that is well aged (as indicated by low ln ([propane]/[ethane]) values) is transported to the site, whereas ozone levels are higher when fresher air (indicating faster transport) is encountered. Significant changes in the degree of NMHC processing and ozone are observed on 1–5 day time scales. These analyses indicate that in spring and summer the highest ozone concentrations were observed in air masses transported relatively rapidly from continental source regions to Pico, and lower concentrations were observed in air masses that had been processed for longer times. Thus, the photochemical environment of the lower free troposphere over the central North Atlantic during spring and summer is primarily one of net ozone destruction.

Figure 8.

Ozone measured at Pico Mountain (indicated by the color coding) as a function of ln([propane]/[ethane]) and time of year during 2004–2005. The insert shows a blowup of the spring 2005 period which was used for the case transport study presented in the companion paper by Honrath et al. [2008].

[36] The regression lines for the spring and summer included in Figure 7 were calculated by using a linear, least squares weighted regression algorithm that allows for uncertainties in both the x and y variables. Each variable was weighted by 1/σ2, where σ is the estimated uncertainty in each measurement, as obtained from the specified precision and accuracy of the analysis for the NMHC (discussed in section 2.2), and using the greater of 1 ppbv or the standard deviation of the ozone measurements over the NMHC sampling period. The positive slope values for the spring, and even more so during summer, reflect the degree of the photochemical destruction of ozone. Even though these dependencies account for a relatively small fraction of the variance of ozone in spring and summer (approximately equal to the R2 values given in Figure 7), the relationship is highly significant (as indicated by the 95% confidence limits of the slopes given in Figure 7). The same analysis was also conducted on a subset of these data that excluded periods with suspected upslope conditions according to the criteria given by Kleissl et al. [2007]; this analysis yielded regression line slopes for the ozone-ln([propane]/[ethane]) relationships that were within 5% of those in Figure 7 and not statistically different.

[37] During the summer of 2004 the NASA DC-8 aircraft conducted flights over the western North Atlantic Ocean as part of the INTEX-A field study [Singh et al., 2006]. Data collected over the North Atlantic during those flights have isolated the vertical distribution of ozone chemical sources and sinks, and were used to conduct box model calculations initialized with observed concentrations of measured species (J. R. Olson and J. W. Crawford, private communication, 2007). These calculations found that net ozone destruction dominated in the lower free troposphere, while net production characterized the upper troposphere. Thus, the model calculations based upon the INTEX aircraft measurements are consistent with the present analysis based on the lower free troposphere monitoring at the Pico Mountain station.

[38] Table 2 compares the springtime and summertime slope of the ozone-ln([propane]/[ethane]) relationship found at Pico with those reported from the northern, temperate Pacific marine boundary layer. On the basis of the earlier data included in Table 2, Parrish et al. [2004] argued that the recent Pacific studies (ITCT-2K2, TRACE-P) indicated only weak net ozone destruction (small positive slopes) in the more remote Pacific marine boundary layer, or no evidence for net ozone destruction (PHOBEA). This weak photochemical destruction is in sharp contrast with the much stronger photochemical destruction indicated by a study at Point Arena from nearly two decades earlier. The exception to this picture is the strong photochemical production (large negative slope) in the PEM West-B study, which investigated the western North Pacific region of strong outflow of ozone precursor emissions from Asia. Comparison of these results from the Pacific region to those from Pico suggest that, at the present time, springtime and summertime photochemistry more effectively destroys ozone in lower troposphere over the central North Atlantic than in the springtime marine boundary layer of the central North Pacific.

Table 2. Ozone–NMHC Oxidation Relationship as Derived From a Plot of the Ozone Mixing Ratio Against ln([propane]/[ethane]) and a Two-Sided Linear Regression From Data in the Marine Troposphere in Spring at Northern Temperate Latitudes
  • a

    Includes marine boundary layer measurements only [Parrish et al., 2004].

  • b

    Lower free troposphere measurements (this work).

Point Arena, Californiaa24 Apr to 9 May 19840.86 ± 0.10
PEM West-B: Asian outflow in western N. Pacifica8 Feb to 14 Mar 1994–0.39 ± 0.11
ITCT-2K2: eastern N. Pacifica22 Apr to 19 May 20020.19 ± 0.06
PHOBEA: eastern N. PacificaMar–May 1997–1999, 2001–2002–0.03 ± 0.08
TRACE-P: west to central N. Pacifica26 Feb to 10 Apr 20010.19 ± 0.04
Pico Mountainb20 Mar to 20 Jun 20050.72 ± 0.07
 21 Jun to 21 Sep 20051.02 ± 0.12

4. Summary and Conclusions

[39] This is the first data set that characterizes NMHC concentrations downwind of North America through a complete seasonal cycle. Concentrations of NMHC at Pico Mountain are lower than at remote, higher northern latitude sites, but higher than at MLO. The observed NMHC levels at Pico reflect the increased influence of the adjacent continents on air composition in the central Atlantic region in comparison to the northern mid-Pacific (MLO), as well as the station's latitude and elevation above sea level. Analyses of the NMHC variability-lifetime relationship provided evidence for the remote character of the Pico site and the lack of significant local influences on NMHC concentrations, with the exception of isoprene. These analyses are in accord with previous conclusions that air encountered at Pico reflects the composition and chemistry of the lower free troposphere above the North Atlantic Ocean. Subsets of seasonal data showed similar behavior in NMHC variability as a function of OH lifetimes. This finding suggests similarity in NMHC transport regimes and source regions throughout the year. Ratios of butane isomers behave as expected from OH chemistry.

[40] NMHC remained elevated in air masses that had been influenced by either anthropogenic emissions or injections from biomass burning after time scales of 1–2 weeks during their transport from source regions to Pico (see more discussion on this topic by Honrath et al. [2008]). Increases in the [n-pentane]/[i-pentane] ratio during the summer, and in particular in identified biomass burning plumes, further reemphasize the influence of boreal biomass fires on atmospheric composition and chemistry in the North Atlantic region. It is possible that oceanic pentane emissions further contribute to their concentrations at Pico. Future analyses will more fully investigate the biomass burning and possible oceanic emission influences.

[41] Springtime and summertime ozone levels at a given [propane]/[ethane] ratio showed a higher variability, indicating, more extensive photochemical processing and variable ozone chemistry than during winter and fall. During spring and summer, average ozone decreased with increased photochemical processing. This relationship implies that during the spring and summer the photochemical environment of the lower free troposphere over the central North Atlantic is characterized by net ozone destruction during transport of air masses to Pico. This behavior is in contrast to the springtime North Pacific, where the photochemical processing is closer to ozone neutral; that is, it neither produces nor destroys ozone.


[42] We thank P. Goldan, NOAA Earth System Research Laboratory, Boulder, CO, for the reference analysis of the primary NMHC standard prior and after its use at Pico; M. Dziobak and M. Val Martin, Michigan Technological University, for GC instrument maintenance tasks at Pico; and T. Jobson, Washington State University, for making available the Fraserdale data. The anonymous reviewers provided valuable comments that helped to further develop discussions in this manuscript. This research was funded by a grant from the NOAA Office of Global Programs (award NA03OAR4310072). R.E.H. and R.C.O. acknowledge support from NOAA grant NA03OAR4310002 and National Science Foundation grant ATM-0535486 and ATM-0720955.