Relationships between ozone and carbon monoxide at surface sites in the North Atlantic region

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL 103, NO Dll, PAGES 13,357-13,376, JUNE 20, 1998 Relationships between ozone and carbon monoxide at surface sites in the North Atlantic region D D Parrish, M Trainer, J S Holloway, J E Yee,2 M S Warshawsky3 and F C Fehsenfeld Aeronomy Laboratory, National Oceanic and Atmospheric Administration, Boulder, Colorado G L Forbes Atmospheric Environment Service, Sable Island, Nova Scotia, Canada J L Moody Department of Environmental Sciences, University of Virginia, Charlottesville Abstract As part of the North Atlantic Regional Experiment (NARE), measurements of 03 and CO at five surface sites were made from July 1991 to January 1995 The investigation of the variabilities and correlation of 03 and CO presented here indicates that the seasonal cycles of the medians and the means of 03 and CO are qualitatively similar to the cycles observed at other northern midlatitude sites The signature of 03 produced from anthropogenic precursors is clearest in the spring at the Azores and in the summer at Sable Island The influence of the natural stratospheric 03 source is apparent at Sable Island, particularly in the spring At all sites the variability of CO throughouthe year is dominated by episodes of pollution transport The slopes of the monthly O3-CO correlations in the summer in Atlantic Canada and the spring in the Azores are quite uniform at 03 to 04 However, individual pollution transport events often have larger (_<10) slopes, which indicate significantly different net 03 production efficiencies between episodes The average slope of 03 versus CO at Sable Island in the winter for moderate pollution transport events (CO _< 180 ppbv) is -028, which indicates that the titration of ambient 03 by emitted NO with little if any photochemical 03 production dominates the 03 chemistry over eastern North America in winter Diurnal cycles driven by photochemistry are observed in the summer for both 03 and CO at the Azores (net loss) and Sable Island (net production) These observations are consistent with the work of Duderstadt et al [this issue] who find positive net photochemical 03 production at Sable Island, and with the modeling of Atherton et al [1996] who find a region dominated by photochemical loss of 03 and CO in the central Atlantic 1 Introduction Industrialized countries on continental rims export anthropogenic pollution to the lower troposphere over the oceans The focus of the North Atlantic Regional Experiment (NARE) is to investigate the effect that this pollution export exerts upon the chemistry of the troposphere over the North Atlantic, especially with regard to the ozone budget This effect is particularly important in the North Atlantic because some of the world's largest sources are on the west in North America and on the east in Europe During the summer the Azores-Bermuda high-pressure system tends to circulate the Also at Cooperative Institute for Research in the Environmental Sciences, University of Colorado, Boulder 2Now at National Center for Atmospheric Research, Boulder, Colorado 3Now at Department of Chemistry, University of Colorado, Boulder Copyright 1998 by the American Geophysical Union Paper number 98JD ! /98/98 JD exported pollution from both of these regions around the North Atlantic basin A program to measure 0 3 and CO at surface sites in the North Atlantic has been a thrust of the NARE research Measurements of these two species are of particular interest because first, 0 3 is a critical atmospheric species that drives much of tropospheric photochemistry, has anthropogenic and natural sources, and is a sensitive marker for the photochemistry that has recently occurred in an air mass; second, CO provides a long-lived tracer of anthropogenic input, which helps to differentiate 0 3 sources; and third, both species can be easily and accurately measured by instrumentation that can be operated long-term in unattended locations An exploratory study was initiated in the summer of 1991 with measurements at three sites downwind of the eastern seaboard of the United States The 0 3 to CO ratios gave an estimate of the amount of 0 3 exported to the North Atlantic troposphere from North America [Parrish et al, 1993]; it was concluded that in the North Atlantic there was comparable input to the troposphere from stratospheric and anthropogenic sources during the summer Measurements were continued at 13,357

2 13,358 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE one of the sites through the winter; the strong, positive O3-CO correlation in summer disappeared and was replaced by a information for the interpretation of the history of sampled air masses One important thrust of the investigation will be an strong, negative correlation in the winter This negative attempt to delineate the seasons and regions in which correlation indicated removal of 0 3 by anthropogenic emissions in that season [Parrish, 1993] The purpose of this photochemical production of 0 3 from anthropogenic precursors dominates photochemical sinks paper is to examine the complete set of O3-CO measurements made during NARE by our laboratory at five different surface 2 Experimental Description stations Measurements at surface sites are useful but do have 2,1 Sites limitations The major concern is that the measurements can be confidently considered as representative of only the shallow surface layer of the troposphere Extensions to the lower troposphere as a whole must be made cautiously However, long-term, continuous surface measurements can provide climatologies of species including seasonal and diurnal cycles Infrequent but important events also can be observed Observations at remote, small island sites are likely to be particularly easily interpreted because first, the effects of the diurnal boundary layer evolution, which are strong at continental sites, are minimized; second, isolation from local sources is more likely; and third, anthropogenic pollution is likely to be strongly processed photochemically during transport to the site Table 1 summarizes the locations and measurement periods for the five sites, and Figure 1 shows their locations Measurements were made during only one summer intensive study period at three sites (Cape Race and Seal Island in 1991 and Chebogue Point in 1993) Over 4 years of data have been collected at Sable Island, and most of one annual cycle has been measured at the Azores The three sites (Seal Island, Sable Island, and Cape Race) chosen for the exploratory study during the summer of 1991 were spaced at approximately 500 km intervals from the East Coast of the United States in the direction of the prevailing winds For the summer 1993 intensive, Chebogue Point, which is on the mainland near Seal Island, was selected for ease of access The Azores is the one island site available in the central North Atlantic Our goal in this paper is to investigate the relationship between the variations of 0 3 and CO in the data sets collected from these five sites in order to learn as much as possible regarding the factors that determine the 03 budget in the North Atlantic region Generally, the important sources and sinks of 03 also affect the concentration of CO; thus the dependence of the concentration of 03 on that of CO provides useful Table 1 Measurement Sites Site Latitude, Longitude, Measurement øn øw Period Azores April 23, Jan 4, 1994 Cape Race July 13, Newfoundland Sept 16, 1991 Sable Island July 8, Nova Scotia Jan 16, 1995 Seal Island July 22, Nova Scotia Sept 22, 1991 Chebogue Point July 26, Nova Scotia Sept 3, 1993 Sites were selected to be as isolated from local influences as possible The instruments were placed in Canadian Coast Guard lighthouses at Seal Island and Cape Race, and in a Canadian Atmospheric Environment Service weather station on Sable Island During the summer intensive of 1993 measurements were carried out at Chebogue Point, a peninsula on the mainland Chebogue Point is in the vicinity of Seal Island, but is not as isolated However, the easier access to transportation and electrical power was essential for the much more extensive suite of instrumentation deployed for this study All of these Canadian sites were within 40 meters of sea level The Azores measurements were carried out on the highest point (elevation 100 km) of the island of Terceira, a volcanic peak on the island's western half The instruments were housed in a small building provided by the University of the Azores This site was nearly always within the marine boundary layer, and Peterson et al [this issue] discuss the diurnal evolution of this boundary layer Although generally quite isolated, each site was impacted by local pollution sources under specific conditions Significant influence from these sources was infrequent and identified by very high variability over short time periods Measurements from these data,periods have been removed from the data sets that we discuss More details of the Seal Island and Cape Race sites are given by Parrish et al [1993] Complete descriptions of the Chebogue Point, Sable Island and Azores sites are given by Fehsenfeld et al [1996] 22 Instrumentation Commercial instruments (Thermo Electron Corporation, Models 48 and 49), adapted for greater sensitivity and longterm unattended operation, were utilized to measure both CO and 03 Parrish et al [1993] give details of the setup for long-term unattended operation, and Parrish et al [1994] describe the modifications of the CO instrument to improve the precision of the measurements and its operation in the field The data were recorded as 1-min averages Longer-term (5-min, 12-min, and 1-hour) averages were calculated from the recorded data and are discussed here The precision of the 0 3 data are at least 1 ppbv for all averaging time periods The precision of the 1-hour CO data approaches 1 to 2 ppbv under the best circumstances [Parrish et al, 1994] and are correspondingly larger for the shorter averages as expected from the averaging statistics Early in the study the experimental precision for 1-hour averages was poorer (--6 ppbv) due to less frequent zeroing The accuracy of the 0 3 measurement is expected to be excellent, since the instrument is based on an absolute absorption measurement Before and after deploying an instrument in the field, it was run against a laboratory reference instrument to verify its performance The instruments at Sable Island (the one longterm site) were exchanged once per year except for the last 22 months of the study period No problems were ever found in

3 , ß ß ß : -Chebogue cape Race Sable Islar d Seal Island : Azores ', ß ß ' , Figure 1 Map of the North Atlantic region Data are reported from the five indicated sites Sable Island 4OO 80 Seal Island I Chebogue Point Cape Race 3OO Azores - 60 O O 20O Plate 1 Bar plots indicating the distributions of measurements of 0 3 (red) and CO (blue) at each site for each summer that measurements were made The means (heavy solid lines) and medians (heavy dashed lines) are shown along with the central 67% (heavy shading); central 90% (light shading); and full range of the measurements At Sable Island and the Azores, data from June through August are included; for the other three sites, data from the complete periods listed in Table 1 are included The number of 1-hour average measurements included in each distribution is given on each bar

4 13,360 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE the performance of the 0 3 instruments The inlets were mounted on towers 5 to 30 m above the ground All calibrations and inlet line checks were done through valves mounted on the tower at the inlet Samples and calibrations passed through 47 mm OD Teflon filters (5-6 micron) Daily tests were performed to check for any inlet line losses of 03; none were found The accuracy of the CO instrument was monitored every 12 or 24 hours by a standard addition calibration process The accuracy is expected to be within 10%, which has been confirmed in an informal intercomparison [Fried et al, 1991] and by measurement of standards circulated in the Measurement of Air Pollution from Satellite study [Novelli et al, 1998] 23 Note on Correlation Procedure Much of the discussion that follows will be based upon linear correlations between 03 and CO The commonly used linear, least squares regression to obtain the function y = mx + b assumes that y is subject to error and x is not, and that any deviation from a straight line relationship is due to those errors in y However, we do not expect that an exact linear relationship exists between 03 and CO, and furthermore, the measurements of the concentrations of both are subject to error Hence a procedure, which we will call "two-sided," that allows for error in both variables is more appropriate [Draper and Smith, 1981, p 122] That procedure will be used exclusively here A weighting of the data by the inverse of the square of the approximate measurement precision will also be used; 1 for 03 and 1/9 for CO 3 Results and Discussion Plate I presents an overview of 0 3 and CO levels at five sites in the summer The sites are qualitatively ordered in increasing isolation from North American continental influence; overall, there is a noticeable decrease in the magnitude of the levels with degree of isolation The four summers at Sable Island show a significant interannual variability However, these systematic differences are small compared to the variability of the levels at any one site A large number of factors contribute to the magnitude and variability of surface levels of CO and 03 The present discussion will be organized around the effects of six: seasonal cycles of average levels, transport of anthropogenic pollution, degree of photochemical processing, diurnal cycles, vertical mixing of air from the free troposphere to the surface, and the mixing down to the surface of air with 03 of stratospheric origin These factors are not independent but do provide a structure for the discussion The goal will be to isolate the characteristic signature of each of these factors, to define as clearly as possible the physical processes driving each factor, and to compare the relative importance of the factors at the various sites The general approach will be to discuss the Sable Island data in most detail, since that site has the most extensive data record, and to contrast the behavior at the other sites The data sets discussed generally such as in Plate 1 will include all measurements made at the sites with one exception: the two periods of marked stratospheric influence at Sable Island, March 14-17, 1993 and 6 hours on May 15, 1993 The former episode will be discussed by J L Moody et al (manuscript in preparation, 1998), and the latter will be discussed as a separate episode in a section below 31 Seasonal Cycles Figure 2 presents the seasonal cycles of CO and 0 3 measured at Sable Island In broad features, the cycles of the median and mean for both CO (winter maxima and summer minima) and 03 (spring maxima) are generally similar in shape to those reported for other northern midlatitude sites [Oltmans, 1981; Oltrearis and Levy, 1992; Novelli et al, 1992] At Sable Island the 03 mean and median reach maxima of about 40 ppbv in April and are approximately flat at 30 to 34 ppbv from summer through winter The background 03 cycle reported for Niwot Ridge, Colorado [Parrish et al, 1986], another site at a similar latitude (4003 øn) that is isolated from local pollution sources, is similar in shape but about 10 ppbv higher in magnitude throughout the year, which may reflect the higher elevation (30 km) of Niwot Ridge The cycle observed at Bermuda JOltmarts and Levy, 1992], another North Atlantic site that lies a similar distance from the eastern US seaboard, provides an interesting contrast There the springtime peak is about 5 ppbv higher, and a summer minimum below 25 ppbv is observed Notably, the 03 levels observed at Bermuda are higher than at Sable Island from December through May At Sable Island the mean and median CO levels (141 and 136 ppbv annual average, respectively) are significantly higher throughout the year than observed at Niwot Ridge (123 ppbv annual mean) [Novelli et al, 1992] Thus the large CO sources in eastern North America clearly influence the Sable Island environment on an average and median, not just on an episodic, basis The magnitude of the annual cycle of the monthly means is also larger at Sable Island (63 ppbv) than at Niwot Ridge (48 ppbv) The monthly means are always higher at Sable Island, but the difference is smaller in the summer (8 ppbv) than in the winter (24 ppbv) In Figure 2 for all months the CO means exceed the medians; this is because the major excursions from the median are always in the positive direction The result is an approximate lognormal distribution as indicated by the approximately symmetric distributions on the log ordinate in the lower panel of Figure 2 For 03 a similar situation exists in the summer, but the means are below the medians in the winter This behavior reflects the seasonality of the sense of the major excursions of 0 3 from the medians: positive in summer as with CO, but negative in the winter The overall variability of both 0 3 and CO at all percentiles is greater in summer than in winter One feature of note is that the variability of 0 3 increases as spring advances, but the variability of CO remains small until the end of spring The higher variability of 03 in the spring may reflect the high but varying influence of the natural stratospheric 0 3 source, which would not significantly affect CO levels 32 O3-CO Correlations in Atlantic Canada: Signatures of Transport of Anthropogenic Pollution Figure 3 presents three example episodes of transport of high levels of anthropogenic pollution to Sable Island as indicated by the very high ( ppbv) maximum levels of CO The coherentransport of such highly polluted air masses over at least 1000 km is a notable indication of the stratification of the lower troposphere over this region of the North Atlantic These episodes represent three of the most

5 - -- PARRISH ET AL' RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 13, I I I I I I 1 I I I I - Sab Island O illlllllllllllllwwllll 2O -:: ' 95% mean 50% % t' 0% I I I I I I I I I I J F M A M J J A S O N D Month of Year Figure 2 Seasonal cycles of (top) 0 3 on a linear scale and (bottom) CO on a log scale at Sable Island The monthly means and cumulative probability percentiles for the hourly average data are indicated by the lines as denoted in the legend intense pollution episodes observed during the study (see maxima of CO and 03 in Figure 2) Importantly, the transport events occur during all seasons of the year; as indicated by the CO maxima in Plate 1 and Figure 2, the pollutant levels reach higher levels in the winter, but they still are substantial in the summer The example episodes in Figure 3 illustrate the seasonally varying correlative behavior between 03 and CO During the photochemically active period of the year from late spring to early fall, as represented by the late May example in the top panel, pollution transport episodes generally exhibit strong positive correlations (r 2 = 087 for this example) This correlation is attributed to photochemical production of 0 3 from anthropogenic precursors emitted from continental sources and transported over the North Atlantic [Parrish et al, 1993] CO is not only one of these precursors, but is also a quasi-conserved marker for anthropogenic pollutants in general, which leads to the high degree of correlation Wintertime episodes are marked by negative correlations of similar strength (r 2 = 090 for the example in Figure 3) The cause of this negative correlation is removal of 03 by reaction with anthropogenic emissions as discussed below During spring and fall there is no significant correlation as illustrated by the March episode Figure 3 (r 2 = 004) Episodes of pollution transport dominate the monthly variabilities of CO throughout the year, and of 03 in the winter and summer Figure 4 presents the correlations for all data collected over 4 years at Sable Island from one summer and one winter month The monthly correlations, while not quite as large as for the individual episodes of Figure 3, are strong The magnitude of r2 provides an approximate measure of the fraction of the variance of a dependent variable that can be

6 13,362 PARRISH ET AL' RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 120 ' I ', 2, r = :%: aa% 4o Sable Island, 1992,!½"-,-, 0 I 5/22 5/23 5/ O {3' IO ' i, i, i, 500 2,iI'; ' r =090 ß ß ß ß (,-3 (, /'vz -oo - [, z '-6 o 2 ' O 0 _,,,,,,¾, %%,½, ½; ß / SableIsland, 1994 % / 1 O0, I I ':%½{ I 0 12/21 12/22 12/23 12/24 I ' I 4OO 80 2, ' r =004 ß 300 > 60 ß 0 o 200 "0 ( Sable Island, / I, I 0 3/5 3/6 3/7 Figure 3 Extreme examples of pollution transport events observed at Sable Island The data are 5-min averages of 0 3 (light symbols) and CO (dark symbols) The square of the correlation coefficient for the two species is indicated for the data from the time period included each plot related to a linear dependence upon the independent variable The average summertime slope of 03 to 04 is evidently [Draper and Smith, 1981, p 33]; thus, for each of these 2 quite consistent for the entire North Atlantic region, both at months, approximately two thirds of the variance in O 3 can be the surface and through the lower troposphere Anderson et al captured by a linear dependence upon CO Figure 5 summarizes [ 1993] report a slope of 04 within the mixed layer (altitude < the seasonal cycle of the O3-CO correlation at the Atlantic 16 km) over the Atlantic offshore from South Carolina to New Canada sites The slopes given in Figures 4 and 5 were derived Jersey Each of the three ground sites and the three aircraft from two-sided linear regressions, which give somewhat operated during the 1993 NARE summer intensive also found steeper slopes than we have reported previously [Parrish et al, similar slopes [Fehsenfeld et al, 1996] Even spring time 1993; Parrish, 1993], but generally they are of similar measurements at Bermuda [Dickerson et al, 1995] yield a very magnitude similar relationship: slope = 027 and r2 = 065 This

7 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 13, I I I I Sable Island ß / ß?o - / August 80 - _ ß ; " slope = 033 ß, r ß :- / ' 60 -: ' ' - _ "', - : December F xt slope = Oct Csebogue Poim Cape Jul Race Jun Aug ',?, r = Nov Jan Sable Island Dec o CO (ppbv) Figure 4 Correlation between 03 and CO for August (dark symbols) and December (light symbols) at Sable Island The data include all hourly averages from For each month the values are given for the square of the correlation coefficient and the slope from the two-sided linear regression The linear regression fits are shown for both months by the solid lines, and the linear regression to the December data at CO below 180 ppbv is shown by the dashed line consistency is perhaps remarkable, since 03 is produced from a mix of precursors, not simply directly from CO Evidently, there are no strong, overall spatial gradients in the mix of pollutants emitted over the eastern United States, and local variations in the pollution emissions are largely homogenized by mixing processes during the transport times to the sampling sites Despite the similarity of these overall O3-CO correlations, there are significant variations between the correlations at different sites and correlations for different episodes at the same site Parrish et al [1993, Figure 2] shows that the relationship between 03 and CO exhibited significant curvature for the particular episodes that impacted Seal Island in 1991; this curvature may indicate that the full photochemical production potential of the 03 precursor mix had not been reached by the time the air masses reached Seal Island However, this is not a general feature of all episodes in that region; no indication of such curvature was observed at the nearby Chebogue Point site in 1993 [see Roberts et al, 1996, Figure 3] Parrish et al [1993, Figure 2] also show that even though significantly lower levels of both 03 and CO were observed at Cape Race, the most remote Atlantic Canada site, the overall slope (033 from the two-sided fit) was quite similar to those at the other sites During the photochemically active season (May through September) of the 4 years of measurements, 38 different episodes (plus one major episode of stratospheric 0 3 transported to the surface, which is discussed in a separate section below) impacted Sable Island with at least 60 ppbv Figure 5 Relationship between the square of the correlation coefficient and the slope of the O3-CO correlations for the Atlantic Canada sites The results are presented as a function of month for Sable Island (dark open circles); the other sites include all measurement periods, which fell from mid-july to mid-september Of those 38 episodes, 36 had correlation coefficients between 03 and CO that were at least +075 One of the two with lower correlations had a negative correlation coefficient, and that high 03 is attributed to stratospheric influence as discussed below The other exhibited a low positive correlation that appeared to indicate different degrees of photochemical processing at different periods of the episode Figure 6 investigates the variation of the slopes for the 36 episodes of relatively high positive correlation The slopes are usually in the 03 to 04 range, particularly for the highest maximum 03 episodes (Figure 6, top) However, slopes of up to a factor of 3 larger are seen in some episodes, generally characterized by lower maximum 03 levels These episodes occur preferentially in the earlier portion of the summer period (Figure 6, bottom) In accord with this seasonal dependence, Figure 5 suggests that the overall monthly slopes are somewhat steeper at Sable Island in May-June compared to July-August-September The agreement between studies evidently results from the overall slope being most dependent upon the episodes with the highest 03 concentrations, and these particular events are characterized by slopes of about 03 to 04 The negative slope in winter implies that 03 is removed by reaction with the anthropogenic pollution marked by elevated CO levels The predominant reaction expected is 2 r NO+03-* NO 2+O 2 NO 2 can be lost essentially irreversibly during the daytime to nitric acid, NO 2 + OH -} HNO , and during the night by the reaction sequence

8 - 13,364 PARRISH ET AL' RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 12 I ' ' ' ' I I I I I ' ' 10 Sable Island 0 May- September o 08 0 o o 0 ( O0 0 o Q o o o o 0 0 ' 00,!, I I,,,, I,,,, I,,,, I,, O Maximum 03 (ppbv) 12 I I I ' ' ' ' I 10 o o Ooo 0 o CP o 0o o o O o 0 o o o o o 00 5,,,, I,,,, I,,,, I I Month of Year Figure 6 03 versus CO slope plotted as a function of (top) maximum 03 and (bottom) month of year for all episodes observed at Sable Islan during which 03 exceeded 60 ppbv and the linear correlation coefficient between 03 and CO was at least 075 The size of the symbols is proportional to the linear correlation coefficient on a scale of 06 to 10 NO 2+O 3-* NO 3+O 2 and other products [Wayne et al, 1991] All of these reactions of anthropogenic pollutants with 03 will irreversibly remove NO 2 +NO 3 -* N from the atmosphere and contribute to the observed negative correlation between 03 and CO in winter These followed by heterogeneous loss of N20 5 These reactions additional reactionsuggest that the negative slopeshould be representhe irreversible loss of 1 to molecules for even larger than -02 to -03 every NO molecule emitted The molar emission ratio of NO to The slopes indicated for the winter months in Figures 4 and CO over eastern North America is approximately are smaller than these values This may indicate that the [Saeger et al, 1989] Thus, if the above reactions proceed to removal of NO and NO 2 have not proceeded to completion by completion, a slope of about -019 to -028 is expected Other the time air masses are transported to Sable Island However, species from anthropogenic sources can also react with 03 Figure 4 shows that the correlation in December is dominated Most notably, the alkenes are emitted at about 30% the rate of by very high CO levels, and that a steeper slope is present at NO on a molar basis [Saeger et al, 1989] Other reaction lower CO levels For example, a two-sided linear regression to channels may also be important; for example, NO 3 reacts with the CO data below 180 ppbv yields a slope of (2c hydrocarbons to yield nitric acid, peroxyacetyl nitrate (PAN), confidence limit) with an r 2 of 041 This linear fit is indicated

9 PARRISH ET AL' RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 13,365 by the short dashed line in Figure 4 Thus agreement is found between the observed slopes and the expected signature of the removal of 03 by reaction with anthropogenic emissions of NO and alkenes Smaller 03 versus CO slopes are found at high CO levels because there is simply more anthropogenic NO emission than required to nearly completely remove 0 3, as occurred in the episode illustrated in the middle panel of Figure 3 The 0 3 that is present in the higher-pollution episodes and the corresponding smaller negative 0 3 versus CO slopes may indicate that significant photochemical production of 0 3 has occurred in the pollution plumes, even in the wintertime, or that the products HNO 3, N20 5, PAN, etc have not been efficiently removed by heterogeneous processes, and in daylight, photolysis of NO 3 and NO2 regenerates a fraction of the 03 During the spring and fall it appears that an approximate balance between chemical removal and photochemical production of 03 occurs This balance accounts for the disappearance of the correlation between 03 and CO in these seasons The March episode in Figure 3 clearly illustrates that balance In the less polluted edges of the plume, the levels of 0 3 were higher than the approximately 40 ppbv observed before and after the episode; but in the center of the episode, the 03 levels were lower than outside the plume Evidently, enough photochemical production occurred in the edges to lead to a net increase in 03, while in the center there was still a net decrease The episode provides a larger scale analogy of the similar pattern often observed in power plant plumes [see, eg, Gillani and Pleim, 1996] where edges of the plume show net production with net loss of 03 in the center of the plume 33 Contribution of Degree of Photochemical Processing to O3-CO Correlations Ozone and CO are simultaneously removed by photochemical processes in regic ns isolated from sources of nitrogen oxides: 03 + sunlight - O(1D) + 02 O(1D) + H20-2OH OH+CO+O 2 - CO 2 +HO 2 If the levels of NO are low, then the hydroperoxy radicals form hydrogen peroxide, and the overall result is the loss of 0 3 and CO The remote marine boundary layer in low-latitude regions is generally a region of low NO and high sunlight and water vapor, and thus rapid photochemical loss The levels of both 03 and CO tend be lower in these regions than in the northern midlatitude air usually sampled at the Atlantic Canada sites However, on some occasions, subtropical marine boundary layer air is transported to the Atlantic Canada sites, either by the circulation around the Bermuda high or in that associated with hurricanes The contrast of the northern midlatitude air with air transported from lower latitudes thus contributes to the observed O3-CO correlations Figure 7 shows one striking example of these contrasting air masses In August 1991, Bob, a category 3 hurricane, moved up the Atlantic seaboard of North America to the west of the measurement sites As a tropical storm, it then passed to the north of the three sites and moved generally east across the North Atlantic At each site, air with very low mixing ratios of 03 and CO was observed in association with the passage of O 5O 0 0 8/19 8/20 8/21 8/ ' ' ' ' I Sable Island August, i,, I ' ' o 150 loo 0,,,, I,,,, I,,,, I, I I I CO (ppbv) Figure 7 (top) Time series for 0 3 (light symbols) and CO (dark symbols) recorded at Sable Island during the time period of the passage of Hurricane Bob (bottom) Data from this period (open symbols) are highlighted on the August O3-CO correlation from Figure 4 (light symbols) the storm The upper panel of Figure 7 shows the time series observed at Sable Island An episode of polluted air transport was terminated on August 18 by the inflow of photochemically processed marine air with low levels of both CO (often below 70 ppbv) and 03 (often below 20 ppbv) At the time of this dramatic drop in levels, the center of the storm was located approximately 250 km south of Cape Hatteras, North Carolina By the end of the data period shown, Bob had moved to 2900 km east of Sable Island after having passed to the north The Seal Island and Cape Race data show generally similar temporal evolution In particular, each of the three sites exhibits the initial displacement of polluted air by the marine air, and the displacement occurred at progressively later times: 1000, 1600, and 19:00 UTC on August 18 at Seal Island, Sable Island, and Cape Race, respectively Thus air from low latitudes was transported into a vast area of the midlatitude North Atlantic by the circulation around the hurricane Figure 8 shows insentropic back trajectory calculations [Merrill et al, 1986] for part of the period of the passage of the hurricane The trajectories are consistent with the displacement of continentally influenced air by low-latitude 50

10 13,366 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE I i i i i i i i i i i i i i i '_, ----,r!,, I --L /', I-- i i i Figure 8 Isentropic back trajectory calculations for air masses arriving at the indicated times at about 950 mbar above Sable Island for part of the period shown in Figure 7 Each arrowhead indicates 1 day travel distance All of the trajectories remained near the surface (<800 mbar) for the duration of the calculation marine air Between 1200 and 2400 UTC on August 18 the of polluted air; the other two are likely due to the inflow of air transport of continental air was terminated by transport of air with a small admixture of stratospheric air that accounts for the that had spent at least the trajectory period at tropical or peaks in Os The data from these peaks tend to define the upper subtropicalatitudes and altitudes below 800 mbar edge of the O s versus CO scatter in the lower panel of Figure 7 The lower panel of Figure 7 shows where the Sable Island At Sable Island the degree of variability of CO below the data from the hurricane passage period fall on the correlation median is likely a good indication of the frequency of transport plot for August from Figure 4 They constitute a significant of air with low-latitude marine character to the site The lower fraction of the data that lie below and to the left of the bulk of panel of Figure 2 indicates that this variability is greatest in the data that represent relatively clean, midlatitude air Thus the summer and declineslowly through the fall These are the the contrast of low-latitude marine air with midlatitude air seasons when the Bermuda-Azores high is established and contributesignificantly to the degree of correlation between when hurricane season occurs This variability is lowest in the CO and 03 Such contrasts occur throughout the year winter and spring, when zonal flow dominates During the period of generally low CO and 03 levels in The contribution of variations in the degree of Figure 7, three relative peaks are observed in the 0 3 levels photochemical processing to the Os-CO correlations is more Only one of these (about 1200 UTC on August 21) is significant in the Azores due to their more isolated location associated with a coincident peak in CO, which is the signature and concomitant reduction in the influence of transport of

11 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 13, O 6O Azores Spring " slope = 042 ' 2 r = 059 Azores Spring 5O 4O I,,,, I I,,, 0 5/1/93 5/6/93 5/11/93 5/16/93 80 t ' I I I I ' ', Azores --] 200 summer I- % ::, ; -I 15o O ::: ß,' ß 0 ' 0 7/6/93 7/11/93 7/16/93 7/21/93 80 J J I, ' ' ' ' I t 250 Azores ' "2 -I 60 Fall %'i':, I 200 ½ -,: I- ' '"" G '"""' ': ' ;:' - "/--"'"g"j 2O 6O 4O 20 Azores July, August slope = 042 r ee e eee ee j?:?' ' e e e e ß ß -,,,, I,,,, I I,,,, I,,,, O CO (ppbv) 0 I,,,, I,,,, I,,,, I / 0 10/21/93 10/26/93 10/31/93 11/5/93 11/10/93 Figure 9 Example 3-week time series of 0 3 (light symbols) and CO (dark symbols) for three seasons at the Azores site The data are 1-hour averages Figure 10 Correlation between 03 and CO for the (top) spring and (bottom) summer for the Azores site The data include all hourly averages recorded in April and May, and July and August, 1991 The square of the correlation coefficient and the slope from the two-sided linear regression are indicated for each site polluted air Figure 9 shows 3 week time periods from three successive seasons to contrast the 03 and CO behavior at this site with that at Sable Island (Figure 3) Correlation plots for spring and summer periods are shown in Figure 10 The highest 03 and CO levels were observed in the spring, and the spring Azores correlation is remarkably similar to the summer monthly correlations at Sable Island Thus a strong signature of transport of anthropogenic pollution is evident during the spring season at the Azores In the summer this transport signature is largely absent; the high, correlated 03 and CO episodes are absent A significant, but much smaller correlation is present (Figure 10, bottom) The levels of 0 3 and CO are low, and the correlation likely arises from the contrast of air transported from northern latitudes versus that transported from tropical regions, with the concomitant contrast of degree of photochemical processing Figure 11 summarizes the correlations seen at the Azores as a function of season and compares them to those at Sable Island Positive correlations are seen throughout the year; significant episodes of negative correlations between 0 3 and CO have not been observed in the Azores The highest correlations are observed in the spring and the early summer when they are comparable to and perhaps stronger than at Sable Island In Figure 11 there is some indication of stronger correlation returning in the fall; this is primarily due to the one episode of quite high CO levels shown in the lower panel of Figure 9 that had a small coincident ozone peak Throughout the year, there is a small but significant positive correlation that is likely due to the contrast between air masses with different degrees of photochemical aging An isentropic trajectory analysis has been performed for the transport of air to the Azores sampling site for two

12 _ - 13,368 PARRISH ET AL' RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 10 ' I examined to determine if the respective cycles are consistent Confidence limits (1 a) can be assigned to the means of lhe 1- hour averages, that is the standard deviation of the averages divided by the square root of the number of averages collected For quantitative comparisons, sine functions have been fit 05 to the cycles of the means and the medians of each data set The period of the sine was fixed at 24 hours, and the amplitude Azores and phase of the function were fit by a nonlinear, least squares 4/93-1/94 procedure The resulting functions are illustrated in Figure 12 and the amplitudes (expressed as peak to peak) and phases 00 (expressed as local solar time of diurnal maximum) are summarized in Table 2 At each site the phase and amplitude of the cycles determined from the means and medians were consistent within their respective confidence limits, so the averages of these parameters for the two cycles at each site are -05 tabulated with corresponding 2 a confidence limits The Sable Island poorer measurement precision for CO and the smaller relative 7/91-1/95 magnitude of the CO diurnal cycles account for the relatively larger confidence limits on the CO cycle parameters The functional fits well represent the data; at nearly every point -10 I, I I they fall within the 1 confidence limits of the means Table 2 indicates that statistically robust diurnal cycles are Month of Year present in each of the five data sets' the amplitude of each diurnal cycle is larger than its 2 confidence limit by at least a Figure 11 Seasonal variation of the correlation coefficient factor of 2 The phases of the cycles are consistent with between O3 and CO at the Azores (dark symbols) compared to photochemical causes; the diurnal maximum of each cycle that at Sable Island (light symbols) The Azores data include occurs near either sunrise (indicating dominance of daytime April 1993 through January 1994, and Sable Island data destruction) or near sunset (indicating dominance of daytime include July 1991 through January 1995 production) Figure 12 and Table 2 indicate daytime decreases of both CO and 03 at the Azores, and daytime increases of both species at Sable Island; this feature is also consistent with representative months, May and August of 1992 and 1993, in photochemistry as the cause of these cycles The remote an effort to determine the cause of the very pronounced marine boundary layer is generally considered to be a region of difference in the levels of O 3 and CO between the two seasons photochemical destruction of 03 Diurnal 03 cycles with illustrated in Figure 10 The primary difference in transport daytime decreases, similar to that in Figure 12 a for the Azores, patterns is a greater influence of descending air from northerly have been reported for a number of sites [see, eg, Oltmans latitudes during May and a greater influence of southerly and Levy, 1994] In the eastern Atlantic in the vicinity of the latitudes during August Both patterns suggest a potential Azores the Atlantic Stratocumulus Transition Experiment influence from upwind continental regions However, the (ASTEX) study used Lagrangian experiments and modeling trajectories do not resolve the boundary layer structure We studies to estimate a diurnally averaged ozone decay rate of hypothesize that the boundary layer, where the measurements about 011 day -1 in the marine boundary layer [Paluch et al, were made, is more isolated from transport layers aloft in the 1995] C is also expected to be photochemically removed in summer than in the spring This is consistent with remote regions' a typically cited summertime lifetime is 30 observations of Oltrnans et al [1996] who have reported days, which corresponds to a diurnally averaged CO decay rate ozonesonde data from the Azores during spring and summer; of about 0033 day -1, which in turn implies a diurnally the average O 3 gradient observed in the lower 4 km of the averaged OH concentration of 16 x 106 cm -3 troposphere was significantly steeper in August (=35 ppbv) To use these decay rates to derive diurnal cycles, the than in May (=25 ppbv) photochemistry of the marine boundary layer and transport to the boundary layer is treated very simply Photochemical 34 Diurnal Cycles of O: and CO destruction (or production) is assumed to vary sinusoidally with a noontime peak during the day and to be zero at night Diurnal cycles in 0 3 and CO levels can be caused by The integral over the daytime hours then must equal the photochemical and/or transport processes that vary in diurnally averaged decay (or production) rate A transporterm strength between day and night Figure 12 shows the diurnal is required to balance that sink (or source) in order to derive a 03 and CO cycles extracted from the summertime measurements repetitive diurnal cycle; that transport term is assumed to be from the two more remote island sites, and Table 2 summarizes constant over the 24 hour period and to be due to slow vertical the diurnal cycles for all five sites Given the large variability transport from (or to) the free troposphere or horizontal of the 03 and CO levels in the summer that are largely driven advection within the boundary layer This model easily yields by synoptic transport, the relatively small diurnal cycles are an approximate analytical solution For the decay rates cited not obvious from simple inspection of time series Therefore above, diurnal cycles with peak-to-peak amplitudes of 17 and the data sets were considered as a whole as a function of the 18 ppbv for 0 3 and CO, respectively, and morning maxima at time of day Both the means and medians of the 1-hour 0715 are predicted These predictions are in excellent averages collected in each hour of the diurnal cycle were agreement with the values from Table 2: amplitudes of 16 +

13 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 13, I I I ' 36 I a Azores 1993 Sable Island O % 24 I I I I I I I I I I T o T T itti :00 06'00 12:00 18'00 00: :00 06:00 12:00 18:00 00'00 Local Solar Time Local Solar Time Figure 12 Summertime diurnal cycles of 03 and CO at the Azores All hourly averages collected in the months of June, July, and August regardless of year are included The data points indicate the average and its confidence limit for all hourly averages falling in a 1-hour segment of the diurnal cycle; the heavy dark line indicates a sine function fit to the averages The thick light curve indicates the median of the hourly averages, and the thin dark curve indicates a sine function fit to the medians The local solar time is calculated from UTC corrected to the longitude of the site 03 and ppbv for 03 and CO, respectively, and morning maxima at 0610 _ and 0820 _+ 0210, respectively Thus the diurnal evolution of these two species is quantitatively consistent with the photochemical removal of both of these species expected in the Azores region Even though the observed diurnal cycles are well matched by the expected photochemical destruction of 03 and CO, a contribution to the diurnal cycle by local effects on the Azores could be possible Peterson et al [this issue] report a diurnal cycle for the oxides of nitrogen that indicates enhanced Table 2 Parameters of Summertime Diurnal 0 3 and CO Cycles Site Species Median* Mean* Amplitude* of Diurnal Cycle Timer of Diurnal Peak Seal Island CO Chebogue Point CO Sable Island CO Cape Race CO Azores CO The indicated uncertainties are 2 o confidence limits *Units: ppbv, amplitude is expressed as peak to peak flocal solar time _

14 13,370 PARRISH ET AL' RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE transport of local emissions to the measurement site during the day This leads to a diurnal cycle of the median and mean NOy levels of about 01 and 02 ppbv, respectively We have no information concerning the ratio of CO to NOy emissions the Azores, but if we assume a ratio of about 5, which is the approximate eastern North American value [Parrish et al, 1991], then a diurnal cycle of 05 to 1 ppbv with an afternoon maximum would be expected for CO This cycle is the opposite phase to that observed; if this transport is important for CO, then the photochemical removal must be faster than assumed above to counter the local effects Similarly, local effects might be expected for 03 as well The transport patterns bringing local emissions to the site could also bring air that has been more affected by surface deposition of 03 or marine boundary layer processing Such transport could be partially responsible for the observed daytime 03 decrease Photochemical destruction of 0 3 within the marine boundary layer would be expected to maintain a vertical gradient in 0 3 between the boundary layer and the free troposphere The loss of 0 3 is approximately 3 ppbv d -1, which is derived from the product of the O 3 decay rate (011 d-1 [Paluch et al, 1995]) and the average 03 level (278 ppbv from Table 2) Peterson et al [this issue] present a simple model for vertical transport into the marine boundary layer in the Azores region They suggest a characteristic venting time of about 5 days and a boundary layer thickness of about 2 km A gradient of 15 ppbv 0 3 would be required for this venting rate to balance the O 3 loss The ozonesonde measurements from the Azores of Oftmarts et al [1996] indicate approximately 32 ppbv in the lower 2 km of the troposphere and 48 ppbv from 2 to 4 km Thus the observed 03 gradient is consistent with that required to balance the photochemical destruction implied by the observed diurnal cycle The Sable Island data in Figure 12 b and Table 2 present a contrasting picture to the Azores data The cycles are of similar magnitude, but of opposite phase with afternoon peaks These cycles, if due to photochemical processes, indicate that the Sable Island region is one of photochemical production of both 0 3 and CO Duderstadt et al [this issue] report that there is indeed net photochemical production, on average of approximately 36 ppbv 03 per day, at Sable Island This rate of production, combined with a balancing constant transport term in the simple marine boundary layer treatment outlined above, implies a diurnal cycle of 20 ppbv peak-to-peak amplitude with a maximum at 1645, again in excellent agreement with the observed diurnal cycle parameters of with a maximum at Duderstadt et al [this issue] do not report the rate of production or loss of CO, but their Table 3 indicates that photochemical production and variation in the transport of the polluted air masses However, loss of CO may be in approximate balance at Sable Island if all we can compare the observed cycles with those modeled by Lin of the carbon in the hydrocarbons that are photochemically et al [this issue] They find at Chebogue Point 03 and CO oxidized is converted to CO Linet al [this issue] model the cycles with amplitudes of 74 and 77 ppbv, respectively, each temporal behavior of 03 and CO for the entire month of August with a peak at about 1700 The measured cycle amplitudes (see 1993' their average diurnal cycles indicate photochemical Table 2) at Chebogue Point, and Seal Island as well, are in production of both O 3 and CO with peak-to-peak amplitudes of excellent agreement with those modeled However, the timing 22 and 30 ppbv, again in excellent agreement with the of the peaks differ significantly between 03 and CO and parameters in Table 2 between sites; this is an indication that the observed diurnal The shape of the 03 cycle from Sable Island in Figure 12 b cycles have contributions from predominant arrival times of differs systematically from the sine function in that the polluted air masses as well as photochemical processes increase appears to occur primarily before noon This is In summary, the observed diurnal cycles of O 3 and CO at the exactly the behavior expected for photochemical production of 03 Wang et al [1996] report that NO, the limiting reactant in photochemical 03 formation, exhibits an unsymmetric diurnal cycle, with a strong peak in the morning and dropping off in the afternoon Hence the strong 03 rise in the morning is consistent with photochemical production rates dependent upon this NO cycle The variability of the photochemical environment in the summer at Sable Island can be investigated to some extent, Duderstadt et al [this issue] emphasize that the photochemical evolution of ozone varies from large net production to smaller net loss depending upon degree of continental influence Figure 4 suggests that we can use the measured 03 level to distinguish the photochemical realm As a rough approximation, we take 0 3 below 35 ppbv as air characteristic of the remote marine boundary layer, and 03 above 35 ppbv as air more strongly influenced by continental sources with a concomitant effect upon the photochemistry Figure 13 presents the CO diurnal cycles for these two 03 regimes; the parameters of the diurnal cycles are included in the figure In the lower 0 3 regime the CO cycle is consistent with that observed at the Azores, while in the higher 0 3 regime, net photochemical production of CO is indicated These results support the conclusion that the observed CO and 03 diurnal cycles arise from net photochemical production or destruction of the species The magnitude of the diurnal cycle of CO in the higher 0 3 regime is surprisingly high and can only be supported by a rich hydrocarbon photochemistry Measurements of formaldehyde in the region indicate this is possible Formaldehyde is the primary photochemical precursor of CO Lee et al [1996] report formaldehyde measurements generally between 15 and 5 ppbv in the vicinity of Sable Island during a pronounced pollutant transport event Formaldehyde has a noontime loss rate through photolysis and reaction with OH in moderately polluted conditions of about 04 hour -1 [Finlayson-Pitts and Pitts, 1986] Formaldehyde levels of 2 ppbv would give a diurnal cycle for CO of 34 peak-to-peak amplitude when included in the simple marine boundary layer treatment without CO removal However, it would be attenuated by simultaneous photochemical removal of CO for a net cycle of about 18 ppbv Thus the diurnal cycle of CO can be explained by solely photochemical effects only if the average formaldehyde concentration in the polluted air is greater than 2 ppbv or there are other significant photochemical sources of CO The 0 3 and CO cycles observed at Seal Island, Chebogue Point, and Cape Race are less easily interpreted, since these sites are impacted by coastal effects Also, the Seal Island and Chebogue Point sites are closest to the large continental sources and may be more strongly influenced by any diurnal Azores and Sable Island are entirely consistent with photochemical processes The western Atlantic marine boundary layer, as represented by Sable Island, is dominated

15 _ PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 13, Sable Island -- - _ 03 > 35 ppbv _ () ½) -- _ " "-',½½'" '""'"'"'" amplitude: 44_ Sable Island 03 < 35 ppbv amplitude: peak: 8:40 Z'00 00:00 06:00 2:00 8:00 00:00 Local Solar Time Figure 13 Summertime diurnal cycles of CO at Sable Island for the two indicated ranges of 03 The format is generally the same as Figure 12 Note the scale change between the two panels by photochemical formation of both species, while the marine the lower troposphere in the Atlantic Canada region during boundary layer in the central Atlantic, as represented by the summer [Angevine et al, 1996] In general, polluted air is Azores, is dominated by photochemical destruction These transported in highly stratified layers The episodes of conclusions are supported by other studies in both the Azores polluted air observed at the western Atlantic surface sites region [Paluch et al, 1995] and at Sable Island [Linet al, this described in this paper were restricted to very shallow layers issue; Duderstadt et al, this issue] In general, these diurnal (=200 m) Vertical profiles during such episodes are shown by cycles of 03 and CO are in phase; hence, they contribute to Angevinet al [1996, Figure 6] and Daum et al [1996, Figure some extent to the O3-CO correlations However, the 4] At other times a surface site was confined in a shallow magnitude of the cycles are so much smaller than the observed layer of relatively clean air while highly polluted air was variability in each species that this contribution to the overall transported in a confined layer in the lower 1 to 2 km of the correlation is minimal The importance of the determination troposphere Examples of such elevated plumes are shown by of these diurnal cycles is the information that they convey Angevinet al [1996, Figure 6] and Daum et al [1996, Figure concerning the photochemistry dominating in the respective 6] At the Azores a similar situation existed in the summer; the areas of the North Atlantic region air sampled at the 1 km elevation site primarily was representative of the marine boundary layer [Peterson al, this issue] and isolated from the free troposphere Up to this 35 Transport of Free Troposphere Air to the point, we have discussed the variability of CO and 03 that was Surface due to processes occurring within these isolated layers Vertical stability is a major feature that characterizes the Vertical transport has not been considered, except perhaps as a meteorology and associated transport of chemical species in constant, slow mechanism to transport CO and/or 03 into or

16 13,372 PARRISH ET AL' RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE out of the surface layer to balance the photochemical production or destruction implied by the diurnal cycles discussed above significantly correlated with the 0 3 deviations Although the correlation was very weak (r = -02), it was as high as could be expected given the precision of the dew point instrument and Figure 14 shows two episodes that we interpret as polluted air from transport layers above the surface layer being directly mixed down to the sampling sites The characteristic that defines these episodes is relatively high variability of the 1- min 0 3 averages At nearly all times that variability was below 1 ppbv, as shown on each side of both episodes in Figure 14, but during these episodes the variability was markedly increased This increased variability is attributed to that the 03 and dew point instruments had different time delays that could not be corrected, since data were collected only as 1 min averages Thus the Chebogue Point episode is consistent with drier, 0 3 rich air being mixed to the surface At the Azores site there were no other data to suppport the identification of vertical transport The Azores episode in Figure 14 represents the single example of significant transport of anthropogenic pollution small-scale turbulence associated with the vertical transport to the site during the summer of 1993 In addition, two because we can identify no other mechanism that could cause short-term (--1 min), positive deviations of 03 correlated with CO At the Chebogue Point site the dew point also shows somewhat larger variability (standard deviation approximately 025 K) during the episode in Figure 14 than outside the episode (standard deviation approximately 01 K, which is the episodes that brought very high levels of natural 03 to the surface at Sable Island were also marked by large 0 3 variability; one of these episodes is discussed in more detail in the next section These four episodes indicate that even though a high degree of stability usually confines surface sites in the North Atlantic instrument precision) The dew point deviations were to sampling air characteristic of only a shallow, near-surface 8O 250 6O Azores ; t ' 2OO 150 o o 4O 100 2O (x5) - 5O, :', i :,,,,, I I / I I I I I I I I I I I I 1 '1 / 12'00 00'00 12' O 25O 6O Point ' ' 2OO t %%, " _ "i i, n ; 2O ' '00 8/1/93 2:00 00'00 2'00 8/2/93 Figure 14 Two episodes of mixing of anthropogenic pollution down to the surface from transport layers aloft Shown are time series for 0 3 (thick light line) and CO (dark line) recorded at (top) the Azores and (bottom) Chebogue Point The data are 12-min averages; the standard deviations (multiplied by a factor of 5) of the 1-min averages that compose the 12-min 03 averages are given by the thin light line

17 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 13,373 layer, there are episodes of pronounced vertical transport when the surface receives polluted air mixed down from transport layers aloft This feature is particularly important for the detailed analyses of surface measurements The meteorological conditions that caused the two episodes in Figure 14 would certainly be of interest, but the detailed meteorological data necessary for such analysis are unavailable 36 Transport to the Surface of Air With 0 3 of Stratospheric Origin Two dramatic but rare episodes of air with marked stratospheric character were observed at Sable Island The first episode occurred in association with a major storm system that ravaged the eastern United States; on March 14-17, 1993, 03 remained above 90 ppbv at Sable Island for 71 consecutive hours and reached a maximum 1-hour average of 167 ppbv The observations and the associated meteorology of this episode will be discussed in a separate paper (J L Moody et al, manuscript in preparation, 1998) The second episode, illustrated in Figure 15, which occurred on May 15, 1993, was much shorter in duration (4 hours of 0 3 above 90 ppbv), but reached an even higher 1-hour average of 185 ppbv Similar high 03 events observed at surface sites have been reported occasionally [see, eg,chung and Dann, 1985; Davies and Schuepbach, 1994; and references therein] Tropopause fold events mix stratospheric air into the troposphere [see, eg, Holton et al, 1995] One signature of such air is very high 03 and low CO levels For example, Danielsen et al [1987] report extrema of 03 and CO levels of 320 and 30 ppbv, respectively, in a troposphere fold sampled at an altitude of 88 km over the southwestern United States from the NASA Convair 990 aircraft This point is plotted as the upper left terminus of the mixing curves shown in the lower panel of Figure 15 The interpretation of the Sable Island 03 episode in Figure 15 is complicated due to mixing of a variety of air masses In the lower panel, three dashed lines indicate the possible O3-CO levels that could result from the mixing of tropopause fold air into lower troposphere air The different mixing lines correspond to three different boundary layer air masses observed at Sable Island as labeled in the upper panel The onset of the high 03 event displaced air mass C, but the peak 0 3 lies near the B air mass mixing line Evidently, the tropopause fold air was mixed with moderately polluted, lower troposphere air similar to air mass B during transport to the site The position of the peak 03 on the mixing line suggests it represented about a half:half mixture of tropopause fold air and this polluted lower tropospheric air This high 03 air was in turn displaced by air similar to air mass A This rather complicated scenario is supported by the condensation nuclei (CN) data shown in Figure 15 CN are good markers for polluted air, and the CN levels are elevated in both air mass B and the high 03 event, so the scenario is at least plausible It should be noted, however, that the CN measurements are not necessarily definitive since the measurements are most sensitive to the freshest pollution, and local pollution sporadically affected the CN measurements during this whole period We have attempted to edit out the local influence by removing the data with very large CN levels and high variability and believe the CN data presented in Figure 15 do indeed qualitatively representhe regional pollution level The mixing of air with such high 03 to the surface occurs only under rare conditions when first, the high 03 air is present above the marine boundary layer, and second, the boundary layer becomes unstable and allows mixing to the surface J L Moody et al (manuscript in preparation, 1998) discuss the meteorological situations that lead to these conditions in greater detail For this May, event, the presence of turbulent mixing into the boundary layer is indicated by the relatively high variability of the 0 3 levels, which was identified in the preceding section as a marker of vertical transport to the surface The average standard deviation of the 1-min averages that constitute 10-min averages was 54 ppbv, a value significantly larger than those for the vertical transport events shown in Figure 14 Although extremely high 0 3 events are very rare, more modest increases in 03 with little change in CO are common Examples have been pointed out previously in Figure 7 Another is apparent in Figure 15 at about 1700 UTC on May 16, 1993 Section 32 discusses 38 episodes when the 03 at Sable Island rose above 60 ppbv; this May 16, 1993 event is the only one without a positive correlation with CO In the upper panel of Figure 15 the decrease in CO that accompanies the increase in 03 is small but significant The correlation between 0 3 and CO for the 11 hours including that event is -045, and the corresponding points are those paralleling the "A" mixing line in the lower panel This event represents natural 0 3 mixed into the marine boundary layer at another location and transported to Sable Island because the short-term variability of 03 was quite small; the average standard deviation of the 1-min averages composing the 10-min averages was 053 ppbv One other significant example of natural 03 is apparent in Figure 2 of Parrish et al [1993] at Cape Race; these are the points at a constant CO level of about 100 ppbv with 03 increasing to about 55 ppbv Detailed examination of time series of data indicates that events with 0 3 increasing by _< 20 ppbv and no significant change in CO are common (perhaps approximately weekly in spring and summer) Such events contribute to the scatter of points about the overall O3-CO correlations such as shown in Figures 4 and 10 4 Summary and Conclusions The measurements of 0 3 and CO at the five surface sites over a greater than 4 year period allow certain conclusions to be drawn in the context of comparison with other studies: At Sable Island and the Azores the seasonal cycles of the medians and the means of 03 and CO are qualitatively similar to the cycles observed at other northern midlatitude sites: spring time peaks and late summer minima Figure 16 presents the seasonal cycle of the monthly 03 means at Sable Island and the Azores contrasted with the seasonal cycle measured at Bermuda by S J Oltmans of Climate Monitoring and Diagnostics Laboratory (CMDL), National Oceanic and Atmospheric Administration (NOAA) At Sable Island the summer 0 3 minimum is absent; the 0 3 levels are approximately flat from summer through winter This is ascribed to the transport of 03 from anthropogenic sources on North America; the maximum amount of 03 is produced in the photochemically active summer when the transport is preferentially to the northeast toward Sable Island The NARE 1993 summer intensive found the most discernable contribution of anthropogenic air was associated with warm

18 13,374 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE O so 14o :00 00:00 00:00 00:00 00:00 5/14/93 5/15/93 5/16/93 5/17/93 5/18/ popause Fold 25O ,,, I I,,,CI A,,, I,, CO (ppbv) Figure 15 Episode of mixing of ozone from a tropopause fold event down to the surface at Sable Island The upper panel shows time series of 1-hour averages for 03 (light line), CO (thick dark line), and condensation nuclei (CN), multiplied by a factor of 20 (thin dark line) The lower panel shows mixing curves (dashed lines) of tropopause fold air [Danielsen et al, 1987] with boundary layer air from the four periods labeled by the respective letters in the upper panel The data marked with circles in the upper panel are the circles included in the lower panel sector flow ahead of cold fronts or troughs, which tended to push polluted air out in pulses The relative summertime 03 levels at the three sites in Figure 16 also reflect the varying effect of photochemicaloss of 03 The more southerly sites are more greatly affected by the transport of air from low-latitude marine regions that are characterized by low 03 levels Figure 16 indicates the highest monthly average 03 levels occur in the springtime at all three sites At least three causes have been proposed for this annual maximum First, it has been attributed to the annual maximum in natural ozone input to the troposphere from the stratosphere by Oltrnans and Levy [1992] and Moody et al, [1995] Second, Dickerson et al [1995] argue that half or more of the air associated with the springtime high 0 3 episodes at Bermuda is of polluted continental boundary layer origin and that substantial amounts of anthropogenic 03, both transported from the continent and produced during transport, contribute to the springtime maximum Third, photochemical production of 03 in the spring from pollutants that build up at high northern latitudes during the winter may also contribute [see, eg, Honrath et al, 1996 and references therein] The third of these processes is

19 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE 13, I I I , 35 o o I I I Azores 4/93-3/94 Bermuda 10,;88 I2/9S Sable Island 7/91-1/95 I I I I I I I I Month of Year Figure 16 Seasonal cycles of the monthly mean 03 levels at three sites in the North Atlantic: Sable Island (solid, dark line), Azores (dashed, dark line), and Bermuda (solid, light line) The Bermuda data are from S J Oltmans, CMDL, NOAA The spans of the data records from each site are indicated in the figure There are significant gaps in the Azores data; when measurements were available for less than half of a month, the data were combined with an adjacent month and plotted at the mean of the time period expected to be stronger at northern latitudes, adjacent to the wintertime reservoir of anthropogenic pollution, while the continental contribution is expected to be stronger at southern latitudes, since the transport from North America to the North Atlantic is preferentially to the southeast in the spring Thus the higher springtime peaks observed at Bermuda (32øN) and the Azores (39øN) than at Sable Island (44øN) suggest that the first two of the proposed causes must dominate, at least at the two more southerly stations The relative importance of these two causes at Bermuda is open to question and is the subject of current research [Moody et al, 1995; Dickerson et al, 1995] The importance of a significant role of anthropogenic pollution at the Azores is consistent with the strong anthropogenic signature seen in the O3-CO correlations in the spring at that site The influence of the natural stratospheric source is apparent at Sable Island The springtime 03 median peak (Figure 2) occurs in April when there is little correlation with CO (Figure 5) and the minimum in the variability in CO (Figure 2) indicates that episodes of pollution transport to Sable Island are at a minimum in frequency Figure 2 shows that the 0 3 variability increases at Sable Island in the spring, which may indicate the increasing frequency of stratospherically influenced air masses In the Atlantic Canada region the variabilities of CO throughouthe year and of ozone in the summer and winter are dominated by episodes of pollution transport from North America At the Azores, pollution transport episodes dominated the variabilities of CO in the spring and fall and of 0 3 in the spring only The slopes of the monthly O3-CO correlations in the summer in Atlantic Canada and the spring in the Azores are quite uniform at 03 to 04 Similar slopes have been reported! I from Bermuda in the spring [Dickerson et al, 1995] and off the central eastern US coast [Anderson et al, 1993] in summer However, individual pollution transport events often have significantly larger (<10) slopes, which indicate significantly different net 03 production efficiencies between episodes The average slope of 03 versus CO at Sable Island in the winter for moderate pollution transport events (CO < 180 ppbv) is -028 This slope agrees well with that expected for the reaction of anthropogenic NO with 03 to yield N20 5 followed by heterogeneous removal of N20 5 Thus this indicates that this reaction dominates the odd-nitrogen chemistry over eastern North America in winter Diurnal cycles driven by photochemistry are observed in the summer for both 03 and CO at the Azores and Sable Island These cycles indicate for both species net photochemicaloss at the Azores and net photochemical production at Sable Island, on average At Sable Island the net photochemical tendency of CO is loss in clean air (0 3 < 35 ppbv) and production in more polluted air (03 > 35 ppbv) These conclusions are consistent with those of Duderstadt et al [this issue] who find positive net photochemical 0 3 production at Sable Island, but that the value varies from small negative values to large positive values depending upon air mass history They are also consistent with the modeling of Atherton et al [1996] who find a region dominated by photochemical loss of 03 and CO in the central Atlantic Acknowledgments The authors would like to thank J T Merrill of the University of Rhode Island for providing trajectory calculations and fruitful discussions, M C Peterson and R E Honrath of Michigan Technological University for very useful suggestions, the staff at the Canadian AES weather station at Sable Island for their excellent logistical support, V H Forjaz and F Rocha of the University of the Azores for providing the field site and site support in the Azores, and S J Oltmans for providing unpublished 03 data from Bermuda This research was partly funded by the NOAA Climate and Global Change Program References Anderson, B E, G L Gregory, J D W Barrick, J JE Collins, G W Sachse, D Bagwell, M C Shipham, J D Bradshaw, and S T Sandholm, The impact of US continental outflow on ozone and aerosol distributions over the North Atlantic, J Geophys Res, 98, 23,477-23,489, 1993 Angevine, W M, M Trainer, S A McKeen, and C M Berkowitz, Mesoscale meteorology of the New England coast, Gulf of Maine, and Nova Scotia: Overview, J Geophys Res, 101, 28,893-28,901, 1996 Atherton, C S, S Sillman, and J Walton, Three-dimensional global modeling of the transport and photochemistry over the North Atlantic Ocean, J Geophys Res, 101, 29,289-29,304, 1996 Chung, Y S, and T Dann, Observations of stratospheric ozone at the ground level in Regina, Canada, Atrnos Environ, 19, , 1985 Danielsen, E F, R S Hipskind, S E Gaines, G W Sachse, G L Gregory, and G F Hill, Three-dimensional analysis of potential vorticity associated with tropopause folds and observed variations of ozone and carbon monoxide, J Geophys Res, 92, , 1987 Daum, P H, L I Kleinman, L Newman, W T Luke, J Weinstein-Lloyd, C M Berkowitz, and K M Busness, Chemical and physical properties of plumes of anthropogenic pollutants transported over the North Atlantic during the North Atlantic Regional Experiment, J Geophys Res, 101, 29,029-29,042, 1996 Davies, T D, and E Schuepbach, Episodes of high ozone concentrations at the Earth's surface resulting from transport down from the upper troposphere/lower stratosphere: A review and case studies, Atrnos Environ, 28, 53-68, 1994 Dickerson, R R, B G Doddridge, P Kelley, and K P Rhoads, Large-scale pollution of the atmosphere over the remote Atlantic Ocean: Evidence from Bermuda, J Geophys Res, 100, , 1995

20 13,376 PARRISH ET AL: RELATIONSHIPS BETWEEN OZONE AND CARBON MONOXIDE Draper, N R, and H Smith, Applied Regression Analysis, 2nd ed, John Wiley, New York, 1981 Duderstadt, K A, et al, Photochemical production and loss rates of ozone at Sable Island, Nova Scotia during the North Atlantic Regional Experiment 1993 summer intensive, J Geophys Res, this issue Fehsenfeld, F C, P Daum, W R Leaitch, M Trainer, D D Parrish, and G Hiibler, Transport and processing of 03 and 03 precursors over the North Atlantic: An overview of the 1993 North Atlantic Regional Experiment (NARE) summer intensive, J Geophys Res, 101, 28,877-28,891, 1996 Finlayson-Pitts, B J and J N Pitts Jr, Atmospheric Chemistry: Fundamentals and Experimental Techniques, John Wiley, New York, 1986 Fried, A, B Henry, D D Parrish, J R Carpenter, and MP Buhr, Intercomparison of tunable diode laser and gas filter correlation measurements of ambient carbon monoxide, Atmos Environ, Part A, 25, , 1991 Gillani, N V, and J E Pleim, Sub-grid-scale features of anthropogenic emissions of NOx and VOC in the context of regional Eulerian models, Atmos Environ, 30, , 1996 Holton, J R, P H Haynes, M E Mcintyre, AR Douglass, R B Rood, and L Pfister, Stratosphere-tropospherexchange, Rev Geophys, 33, , 1995 Honrath, R E, A J Hamlin, and J T Merrill, Transport of ozone precursors from the Arctic troposphere to the North Atlantic region, J Geophys Res, 101, 29,335-29,351, 1996 Lee, Y-N, X Zhou, W R Leaitch, and C M Banic, An aircraft measurement technique for formaldehyde and soluble carbonyl compounds, J Geophys Res, 101, 29,075-29,080, 1996 Lin, X, M Trainer, and E-Y Hsie, A Modeling study of tropospheric species during NARE, J Geophys Res, this issue Merrill, JT, R Bleck, and D Boudra, Techniques for Lagrangian trajectory analysis in isentropic coordinates, Mon Weather Rev, 114, , 1986 Moody, J L, S J Oltmans, H Levy II, and J T Merrill, Transport climatology of tropospheric ozone: Bermuda, , J Geophys Res, 100, , 1995 Novelli, P C, L P Steele, and P P Tans, Mixing ratios of carbon monoxide in the troposphere, J Geophys Res, 97, 20,731-20,750, 1992 Novetli, P C,et at, An internally consistent set of globally distributed atmospheric carbon monoxide mixing ratios developed using results from an intercomparison of measurements, J Geophys Res, in press, 1998 Oltmans, S J, Surface ozone measurements in clean air, J Geophys Res, 86, , 1981 Oltmans, S J, and H L Levy II, Seasonal cycle of surface ozone over the western North Atlantic, Nature, 358, , 1992 Oltmans, S J, and H L Levy II, Surface ozone measurements from a global network, Atmos Environ, 28, 9-24, 1994 Oltmans, S J, et al, Summer and spring ozone profiles over the North Atlantic from ozonesonde measurements, J Geophys Res, 101, 29,179-29,200, 1996 Paluch, I R, S McKeen, D H Lenschow, R D Schillawski, and G L Kok, Evolution of the subtropical marine boundary layer: Photochemical ozone loss, J Atmos Sci, 52, , 1995 Parrish, D D, Carbon monoxide and light alkanes as tropospheric tracers of anthropogenic ozone, in The Tropospheric Chemistry of Ozone in the Polar Regions, edited by H Niki and K H Becker, pp , Springer-Verlag, New York, 1993 Parrish, D D, D W Fahey, E J Williams, SC Liu, M Trainer, P C Murphy, D L Albritton, and F C Fehsenfeld, Background ozone and anthropogenic ozone enhancement at Niwot Ridge, Colorado, J Atmos Chem, 4, 63-80, 1986 Parrish, D D, M Trainer, MP Buhr, B A Watkins, and F C Fehsenfeld, Carbon monoxide concentrations and their relation to concentrations of total reactive oxidized nitrogen at two rural US sites, J Geophys Res, 96, , 1991 Parrish, D D, J S Holloway, M Trainer, P C Murphy, G L Forbes, and F C Fehsenfeld, Export of North American ozone pollution to the North Atlantic Ocean, Science, 259, , 1993 Parrish, D D, J S Holloway, and F C Fehsenfeld, Routine, continuous measurement of carbon monoxide with parts-per-billion precision, Environ Sci Technol, 28, , 1994 Peterson, M C, R E Honrath, D D Parrish, and S J Oltmans, Measurements of nitrogen oxides and a simple model of NO 2, fate in the remote North Atlantic marine atmosphere, J Geophys nes, this issue Roberts, J R, et at, Episodic removal of NOy species from the marine boundary layer over the North Atlantic, J Geophys Res, 101, 28,947-28,960, 1996 Saeger, M, et al, The 1985 NAPAP emissions inventory (version 2): Development of the annual data and modelers' tapes, Rep EPA- 600/ a, US Environ Protect Agency, Washington DC, 1989 Wang, T, et al, Ground-based measurements of NO x and total reactive oxidized nitrogen (NO y) at Sable Island, Nova Scotia, during the NARE 1993 summer intensive, J Geophys Res, 101, 28,991-29,004, 1996 Wayne, R P, et al, The nitrate radical: Physics, chemistry, and the atmosphere, Atmos Environ, Part A, 25, 1-206, 1991 FC Fehsenfeld, JS Holloway, DD Parrish, M Trainer, MS Warshawsky, and JE Yee, NOAA, R/E/AL7, 325 Broadway, Boulder, CO ( parrish alnoaagov) GL Forbes, Atmospheric Environment Service-Sable Island, 1496 Bedford Highway, Bedford, Nova Scotia, B4A 1E5, Canada ( gforbes foxnstnca) JL Moody, Department of Environmental Sciences, University of Virginia, Charlottesville, VA ( moody virginia edu) (Received April 4, 1997; revised January 27, 1998; accepted January 28, 1998)