Comparing the impact of meteorological variability on surface ozone during the NEAQS (2002) and ICARTT (2004) field campaigns



[1] This paper investigates linkages between weather, climate, and air quality that contributed to the large difference in the number of ozone exceedances encountered in the northeastern United States (U.S.) during 2002 and 2004. Major air quality research field campaigns were conducted in the northeast during July and August of each year. Both the 2002 New England Air Quality Study (NEAQS-02) and the International Consortium for Atmospheric Research on Transport and Transformation 2004 field study (ICARTT-04) had research components focused on regional air quality. The primary environmental difference between the two field campaigns was the underlying climatic conditions. The July–August period in 2002 was much sunnier, warmer, and drier than normal. In contrast, the July–August period in 2004 was cloudier, cooler, and much wetter than normal. We conclude that these extreme climatic conditions were the underlying cause for the significant difference in the number of ozone exceedances that occurred during NEAQS-02 and ICARTT-04. We rule out the impact of other meteorological processes as the primary cause of this difference. We gauge horizontal transport using surface and upper air wind observations collected on Appledore Island (ADI), off the coast of New Hampshire and Maine, along with back trajectories based solely on wind observations collected by profiler networks deployed for each study. The wind conditions that favor pollutant transport to the northeast were more prevalent in 2004 than in 2002, yet the number of ozone exceedances in 2004 was more than a factor of three less than in 2002. Neither was daytime boundary layer mixing the cause for this discrepancy. Unlike other parts of the U.S. where poor air quality is generally associated with shallow boundary layers, in New England the boundary layers were deeper on high-ozone days than on clean days because the same sunny, warm, and dry conditions that favor boundary layer ozone production also produce deeper boundary layers. Both field campaigns were synoptically active. Lulls in synoptic activity explained most of the high-ozone events observed in 2002, whereas even an extended lull in synoptic activity during the summer of 2004 did not produce a single high-ozone day.

1. Introduction

[2] The 2002 New England Air Quality Study (NEAQS-02) was designed to provide a better understanding of (1) the role of long-range transport in shaping the air quality of New England, (2) the role of biogenic emissions in local and regional New England air quality, (3) the role of the land-sea breeze circulation in influencing New England air quality, (4) capabilities of current air quality forecasting systems, and (5) linkages between air quality and climate. This paper deals mainly with the last of these objectives. The NEAQS-02 intensive observing period was from 12 July to 10 August 2002. However, regional surface-based chemistry networks and meteorological networks deployed for the field study were in place for the period June–September 2002. Therefore we focus on observations collected during the entire months of July and August, when New England often experiences its worst air quality.

[3] The International Consortium for Atmospheric Research on Transport and Transformation (ICARTT) was formed to take advantage of the combined air quality research efforts taking place among several U.S., Canadian, and European agencies. A comprehensive ICARTT field study (ICARTT-04) was carried out primarily in July and August 2004. ICARTT-04 involved three broad areas of study: regional air quality, intercontinental transport, and radiation balance in the atmosphere. More than a dozen international aircraft and the NOAA Research Vessel Ronald H. Brown were deployed, in addition to a variety of surface chemistry and meteorological networks. More information on ICARTT-04 is provided by Fehsenfeld et al. [2006].

[4] A common component to both NEAQS-02 and ICARTT-04 was the air chemistry monitoring network operated by the Atmospheric Investigation, Regional Modeling, Analysis, and Prediction (AIRMAP) investigators at the University of New Hampshire. The primary mission of AIRMAP is to develop a detailed understanding of climate variability and the source of persistent air pollutants in New England through the collection and analysis of meteorological and chemical data from rural, maritime, and mountain sites in the northeast (Figure 1). One of the AIRMAP sites is located on Appledore Island, Maine (ADI), in the Isles of Shoals. Even though the chemistry measurements at ADI are made only during the summer when the island is inhabited, these measurements are extremely useful because under the proper meteorological conditions, polluted air masses that have been transported over the open water from major east coast metropolitan areas (e.g., Boston) can be sampled. In this paper we compare surface ozone measurements from ADI collected during NEAQS-02 and ICARTT-04. We use these measurements to represent broadly air quality conditions across the northeast during the two field studies.

Figure 1.

Map of New England showing the locations of the air quality observing stations in the Atmospheric Investigation, Regional Modeling, Analysis and Prediction (AIRMAP) network operated by the University of New Hampshire.

[5] The meteorological conditions that produce high-ozone events in New England are tightly coupled to the local, regional, and national climate. The relative frequency of air mass changes associated with passing cold fronts and the transport patterns that bring pollution from the industrial Midwest and the urban megalopolis along the east coast determine the regional air quality. Also, the polluted air that leaves New England can influence air quality in eastern Canada, as shown by Millet et al. [2006] and White et al. [2006], and ultimately in the North Atlantic and Europe. A better understanding of the linkages between climate and air pollution events will help scientists to predict more reliably the impact of anticipated changes in climate on future air quality.

[6] Over the 10-year period from 1996 to 2005, the average number of days per year that one or more of the five New England states exceeded the EPA's current National Ambient Air Quality Standard for ozone was 26.2. (For more information on the EPA's current National Ambient Air Quality Standards, see Within this period, the year with the maximum number of exceedance days (43) was 2002, whereas the year with the minimum number of exceedance days (13) was 2004 (see Figure 2). Therefore both the NEAQS-02 and ICARTT-04 ozone seasons were anomalous relative to the 10-year history. On the basis of NEASQS-02 data from Thompson Farm, New Hampshire, Griffin et al. [2004] concluded that the high ozone concentrations measured at Thompson Farm during 2002 were due to transport, not local production. With numerous sources of pollutants in Boston, New York, and Washington, D.C., to the south and southwest of New Hampshire, and more distant sources to the west, it stands to reason that if higher-ozone events are due to transport, not local production, high-ozone days in northern New England would be highly correlated with wind directions ranging from south to west, which, in fact has been shown by several researchers [e.g., Fischer et al., 2004; Griffin et al., 2004; Mao and Talbot, 2004]. Our analysis will show that low-level flow from the southwest quadrant was more predominant during ICARTT-04 than during NEAQS-02, despite the fact that NEAQS-02 experienced many more high-ozone events than ICARTT-04.

Figure 2.

Number of days that a monitoring station in one of the New England states exceeded the 8-hour ozone standard, by year, from 1996 to 2005.

[7] In this paper we investigate this paradox by analyzing the meteorological factors that influence air quality and climate. We take advantage of mesoscale observations collected during research field campaigns in addition to local and regional climatic data. We analyze the wind fields that are used to initialize numerical weather models to evaluate large-scale transport. This paper does not investigate how the precursors and the chemical mechanisms responsible for producing ozone may have differed during the two field studies.

[8] Section 2 provides a summary of the ozone events that occurred during July–August 2002 and 2004 at ADI. Section 3 presents an overview of the synoptic and climatic conditions that were present during each field study period. Here we assess cold frontal passage frequency, temperature and precipitation anomalies in the northeast, differences in the amount of available sunshine, and climatic conditions across the contiguous U.S. These analyses place both the NEAQS-02 and ICARTT-04 ozone seasons into a climatological perspective, showing that, in addition to representing the extreme ends of the annual time series of ozone events measured over the past 10 years, both studies were conducted under climate extremes: the 2002 period was much warmer and drier than normal and the 2004 period was appreciably cooler and wetter than normal. Section 4 investigates horizontal transport and vertical mixing, two atmospheric processes that control the distribution of pollutants in the atmosphere. Here, in addition to assessing regional low-level transport through analyses of surface and upper air winds measured at ADI, we conduct a cluster analysis of back trajectories from ADI calculated using observations from the wind-profiler networks deployed for each field study. Diurnally averaged, convective boundary layer mixing depths from the two summers are compared and contrasted with the mixing depths observed on the high-ozone days. In section 5 we provide a summary and state our primary conclusions.

2. Appledore Island Ozone Measurements During the NEAQS-02 and ICARTT-04 Field Campaigns

[9] The manifold inlet for the ozone sensor on ADI was located at the top of the tower shown in Figure 3 (42.987°N, 69.335°W, 36.6 m above mean sea level). The ozone measurement technique used at ADI is the same as that used at Thompson Farm, another AIRMAP site [Mao and Talbot, 2004]. Daily maximum 8-hour average ozone concentrations measured at ADI during July and August 2002 and 2004 are shown in Figure 4. Marked on the bottom of Figure 4 are the times when cold fronts approaching from the west or northwest (as opposed to “backdoor” cold fronts that approach from the east or northeast) passed over ADI (see section 3.2 for further details). Dates when the EPA's 8-hour ozone standard was exceeded during July–August 2002 and 2004 are listed in Table 1. Ozone exceedances occurred at ADI on 17 days during July–August 2002, whereas only two exceedances occurred during the same 2 months in 2004. Ten-minute average ozone concentrations above 100 ppbv were observed on 206 occasions during July–August 2002 and only 22 times during the same period in 2004.

Figure 3.

Tower on Appledore Island, Maine, where the ozone sensor was deployed during the NEAQS and ICARTT field studies.

Figure 4.

Time series of daily maximum 8-hour average ozone concentration (ppbv) measured on Appledore Island during the NEAQS (2002) and ICARTT (2004) field campaigns. The current EPA National Ambient Air Quality Standard for ozone is indicated by the bold black line. The arrows on the bottom axis indicate the times (listed in Table 2) of eastward propagating cold frontal passages at the surface, as determined by analysis of NCEP/HPC daily weather maps and surface meterological data recorded on Appledore Island.

Table 1. Dates When the Daily Maximum 8-Hour Average Ozone Concentration Met or Exceeded the EPA 8-Hour Standard at Appledore Island, Maine, During July and August 2002 and 2004a
Maximum 8-Hour Average O3, ppbvNumber 10-Min O3 > 100 ppbvMaximum 10-Min O3, ppbvMaximum 8-Hour Average O3, ppbvNumber 10-Min O3 > 100 ppbvMaximum 10-Min O3, ppbv
  • a

    Number of 10-min ozone readings above 100 ppbv and maximum 10-min concentrations.

1 Jul81.77104.6   
2 Jul94.518127.8   
3 Jul90.35108.2   
4 Jul87.212110.6   
9 Jul90.92100.9   
13 Jul81.9095.1   
14 Jul85.92102.4   
22 Jul80.6098.0   
23 Jul91.425124.8   
30 Jul   91.522115.2
3 Aug   81.5096.7
4 Aug87.24100.7   
11 Aug83.7095.1   
12 Aug98.920118.8   
13 Aug124.945141.0   
14 Aug119.066150.9   
16 Aug82.0096.1   
18 Aug84.8091.3   
22 Aug82.8094.1   
Totals 206  22 

[10] Noting the dates of high-ozone events listed in Table 1, and then cross-referencing them to the dates of the cold front passages shown in Table 2, demonstrates how cold fronts often put an abrupt end to pollution episodes. Behind the fronts, sinking air originating from the middle and upper troposphere replaces the polluted air in the boundary layer and lower free troposphere. As evidenced by the number of cold front passages, both field studies were synoptically active. However, lulls in synoptic activity resulted in most of the high-ozone events observed in July–August 2002. In contrast, only one event occurred in July–August 2004, despite definitive breaks in synoptic activity. This issue will be discussed further in section 3.2.

Table 2. Date and Approximate Time of Eastward Propagating Cold Frontal Passages at Appledore Island, Maine, During 2002 and 2004 From the End of June Until the End of August, the Corresponding Lapsed Time (Δt) in Hours Between Frontal Passages, and the Dates of Ozone Exceedance Days From Table 1 That Occurred Prior to the Frontal Passage
DateUTCΔt Since Last Cold Front, hoursOzone Exceedance Days Prior to FrontDateUTCΔt Since Last Cold Front, hoursOzone Exceedance Days Prior to Front
29 Jun0100  29 Jun0700  
5 Jul02001451–4 Jul2 Jul090074 
10 Jul07001259 Jul6 Jul080095 
15 Jul230013613 and 14 Jul9 Jul070071 
19 Jul220095 15 Jul0900146 
23 Jul20009422 and 23 Jul19 Jul2200109 
30 Jul2000168 23 Jul210095 
6 Aug02001504 Aug29 Jul0800131 
19 Aug000031011–14, 16, and 18 Aug1 Aug22006230 Jul
23 Aug070010322 Aug4 Aug0100513 Aug
26 Aug230088 14 Aug1000249 
    21 Aug1900177 
    24 Aug030056 
    31 Aug0900174 
  equation image = 141.4   equation image = 114.6 

3. Synoptic and Climatological Overview

3.1. Geostrophic Wind Analyses

[11] The large-scale conditions responsible for producing high-ozone events at ADI during NEAQS-02 and ICARTT-04 were gauged by constructing NCEP–NCAR composite synoptic analyses for the two study periods: July–August 2002 and July–August 2004. While composites for several different variables and for two significant pressure levels (925 mbar and 500 mbar) were produced, for brevity we present here only the 925-mbar geopotential height fields (Figure 5), which are representative of low-level pollutant transport for each season. The lower-troposphere geostrophic winds inferred from the height fields were on average out of the northwest during NEAQS-02. In contrast, the geostrophic winds were out of the west-southwest during ICARTT-04. As will be shown later, high-ozone events in both field campaigns occurred when the surface winds measured at ADI ranged from south-southeasterly to southwesterly. This implies that the large-scale transport, as reflected by the geostrophic winds, was more in line with producing high-ozone events at ADI in 2004 than it was in 2002. Figure 5 also indicates that the July–August weather in the northeast was characterized in the mean by a 925-mbar ridge, whereas the same period in 2004 exhibited a more persistent trough. The local transport that contributed to the high-ozone events at ADI will be investigated in section 4. Because the number of events observed in 2004 was so small compared to 2002, factors other than large-scale horizontal transport must have contributed to the increased number of events in 2002.

Figure 5.

Composite synoptic-scale mean 925-mbar geopotential height analyses (m) derived from the NCEP-NCAR daily reanalysis global gridded data set for (top) July–August 2002 and (bottom) July–August 2004.

3.2. Cold Fronts

[12] Tables 1 and 2 show how eastward propagating cold fronts were important for curtailing high-ozone events at ADI during NEAQS-02. In this section we compare the frequency of cold fronts passing through ADI during NEAQS-02 and ICARTT-04 to investigate whether synoptic weather activity contributed to the large difference in the number of high-ozone events observed at ADI during the two field campaigns.

[13] Table 2 gives the dates and approximate times of eastward propagating cold frontal passages at ADI for July–August 2002 and 2004. Two data sets were used to estimate the frontal passage times. First, daily synoptic weather maps from the NCEP/HPC Daily Weather Map Project ( for 2004 maps and for 2002 maps) were used to determine the date when a frontal passage occurred. Two-minute resolution surface meteorological data (pressure, temperature, relative humidity, wind speed, wind direction, solar and net radiation, and precipitation) collected on ADI by NOAA/ESRL then were used to pinpoint the closest hour to the frontal passage, as manifested at the surface. Table 2 also lists the lapsed time between successive cold fronts. Because of the higher frequency of cold fronts in July and August 2004, the average time between cold fronts was shorter by more than a day as compared to the same period in 2002.

[14] Although the shorter time between cold fronts may have contributed to the low number of high-ozone events observed at ADI during ICARTT-04, by shortening the time allowed for the buildup of pollutants in the major metropolitan areas along the east coast, it is unlikely that it is the sole contributing factor. For example, both field studies experienced an extended lull in synoptic activity during August that lasted for more than ten days. During 2002, the lull period was 6–18 August, while in 2004 the lull period was 4–14 August (Table 2). The synoptically stagnant period in 2002 caused the worst air quality episode of either field campaign, but the stagnant period in 2004 did not result in a single high-ozone event at ADI. The 2004 lull period, and the months of July and August in general, were characterized by a 500-mbar low over eastern Canada which kept New England's weather unsettled. The 2002 lull period was dominated by high pressure off the east coast of the U.S. These synoptic patterns were reflected in the surface temperature recorded during the two lull periods. During the 2002 lull period, the average high temperature in Boston was 89.3°F (8 of the 13 days from 6 to 18 August 2002 registered high temperatures of 90°F or above). During the 2004 period (4–14 August) the average high temperature in Boston was 74.5°F, which was 4.1°F below normal for the period and nearly 15°F cooler than the 2002 period. Thus, in addition to differences in synoptic activity, these results suggest that the climatic conditions were dissimilar between the 2 field study years. This hypothesis will be explored further in the following section.

3.3. Bimonthly (July–August) Climate Summaries

[15] We first show how the two field study periods fit into a climatological context by examining the monthly rankings for temperature and precipitation (Figures 6 and 8, respectively). Figure 6 shows how the northeast and close to the entire nation in general were warmer, if not significantly warmer, than normal during July and August 2002. During August 2002, when ADI experienced its worst air quality episode of the two field campaigns, eight northeastern and mid-Atlantic states (New Hampshire, Vermont, New York, Massachusetts, Connecticut, Rhode Island, New Jersey, Delaware, and Maryland) each experienced one of their top ten warmest Augusts on record. The Ohio River Valley states also were warmer than normal. This means that the conditions favorable for daytime ozone production existed not only locally near ADI but also throughout the northeast and in regions upwind of there. In contrast, the combined temperature ranks for July and August 2004 indicate that the 2004 period was slightly cooler than normal, for the most part, in the northeast and much cooler than normal to the west, especially over the Ohio River Valley and the lower and upper Midwest.

Figure 6.

Statewide ranks for monthly average temperature observed in (top) July and (bottom) August during (left) 2002 and (right) 2004.

[16] The amount of available sunshine is important to regional air quality because the photochemistry that produces ozone from its precursors during the day requires sunlight. We gauged the amount of available sunshine by comparing the solar irradiance measured with pyranometers at each of the NOAA profiler sites during the two field campaigns. During the 2002 study, LICOR model L1200 pyranometers were deployed. During the 2004 study, Kipp & Zonen model CM3 pyranometers were deployed. The limited spectral response of the LICOR sensor suggests that it may overestimate solar irradiance in hazy or partly cloudy conditions [Zamora et al., 2003]. However, we are not aware of published results that quantify and compare the performance of the two different sensors under a variety of sky conditions. For brevity, here we show the results from only the Concord, New Hampshire, site, which is located approximately 75 km to the northwest of ADI. The results from ADI revealed statistics very similar to those obtained from Concord, but there were nine days with missing data from the ADI 2004 data record. The solar irradiance data from the pyranometers were recorded with 2-min averaging periods. For the comparison we summed the 2-min averages over the course of each day and compared the difference of these daily sums between the 2 years (Figure 7). The weather in July and August 2002 was sunnier, on average, than the same period in 2004, which helps explain why there were more high-ozone events in 2002 and also warmer temperatures.

Figure 7.

Daily difference in accumulated solar irradiance (W m−2) measured at NOAA's research field station located in Concord, New Hampshire, during July–August 2002 and 2004.

[17] Precipitation statistics (Figure 8) support the hypothesis that the local and regional seasonal climatic conditions were, in general, more favorable for ozone production during NEAQS-02 than during ICARTT-04. July and August 2002 were drier than normal over much of the nation except for the lower and upper Midwest. During August 2002 the driest conditions, compared to climatology, were recorded over the northeast and southwest. In fact, Maine experienced a record-breaking drought for the month of August. In contrast, 2004 was an abnormally wet year over much of the northeast. July 2004 set monthly rainfall records in New York and Pennsylvania (see also Table 3). The wet conditions that dominated in 2004 resulted in reducing the amount of available sunshine and, therefore, help explain the paucity of high-ozone events experienced at ADI during the ICARTT-04 campaign. The impact of increased seasonal precipitation on biogenic emissions was not considered.

Figure 8.

Statewide ranks for monthly accumulated precipitation observed in (top) July and (bottom) August during (left) 2002 and (right) 2004.

Table 3. Bimonthly (July–August) Temperature and Rainfall Climatologies for Six Eastern National Weather Service (NWS) Airport Observing Stations During 2002 and 2004a
NWS Observing StationAverage Maximum Temperature, °FDeparture From Normal, °FNumber of Days With Maximum Temperature Above 90°FNumber of Days With Rainfall ≥0.01 inchesNumber of Days With Rainfall ≥0.1 inchesTotal Rainfall, inchesTotal Rainfall, % of Normal
  • a

    Stations are listed from north to south.

Portland, ME80.1575.55+2.10−2.509023259164.619.9772157
Concord, NH84.3080.00+2.45−1.8520214257174.208.5164129
Boston, MA83.6078.65+2.45−2.5022316218163.558.2555128
New York, NY86.4081.85+2.50−2.0527412208156.1513.0572154
Philadelphia, PA88.7083.50+3.95−1.2532512217154.5912.0856147
Washington, DC88.7584.30+2.10−2.3534417259155.557.5176102

[18] We next examine bimonthly (July–August) climatic data for six coastal metropolitan areas stretching from Portland, Maine, to Washington, D.C. The data are summarized in Table 3. The seasonal climatic conditions in all six of these coastal cities were more favorable for ozone production in 2002 than in 2004. All six cities experienced warmer than normal conditions in 2002 and cooler than normal conditions in 2004. A particularly astonishing statistic was the number of days when the maximum daytime temperature met or exceeded 90°F, which was roughly an order of magnitude larger in 2002 than in 2004 for all cities examined. In fact, the maximum temperature in Boston met or exceeded 90°F on 14 out of the 17 ozone exceedance days observed at ADI in July–August 2002. With regard to precipitation, the number of days with significant rainfall (≥0.1 inch) in 2004 was about double what it was in 2002. Note, in particular, the large July–August 2004 rainfall totals in New York City and Philadelphia.

3.4. Discussion

[19] The synoptic and climatic data presented in this section present a somewhat inconsistent picture of why ADI experienced nearly a factor of ten more ozone exceedances during NEAQS-02 than during ICARTT-04. The mean geostrophic wind direction present during each study suggested that transport of air pollution from the Ohio Valley and east coast to ADI was more likely in July–August 2004 than it was in July–August 2002. However, cold fronts more frequently swept away pollution in the northeast during the 2004 period than during the same period in 2002. In addition, July–August 2002 experienced much sunnier, warmer, and drier conditions than did the same period in 2004. Therefore we conclude that the seasonal departures from the local and regional climate norms and the frequency of cold frontal passages played more important roles than horizontal transport in controlling the relative number of high-ozone events experienced at ADI during NEAQS-02 and ICARTT-04.

[20] Perhaps more importantly, with respect to generalizing the research results from either study, both the NEAQS-02 and ICARTT-04 field campaigns were conducted during summer months characterized by climate extremes. In 2002, the main anomaly was surface temperature, with very warm conditions experienced in the northeast, especially during August. In 2004, the main anomaly in the northeast was precipitation, with greatly increased rainfall accumulations compared with climatology, especially during July. These climatic features must be considered when air quality managers use results from one or both of these field campaigns to help form pollution control strategies.

4. Regional Transport and Vertical Mixing

4.1. Wind Profiler Network

[21] During NEAQS-02 and ICARTT-04, NOAA/ESRL added to the small number of existing wind profilers in the northeast to build wind profiler networks capable of documenting the regional transport of pollutants. The wind profiler network available for each of the field campaigns is illustrated in Figure 9 and listed in Table 4. The configurations of the 2002 and 2004 profiler networks differed because the underlying objectives of the two field programs were different. NEAQS-02 focused on air pollution generated within and upwind of New England; thus there was a greater emphasis on inland wind profilers for that campaign. While both field programs had a research component on regional air quality, ICARTT-04 more heavily emphasized intercontinental transport, so the ICARTT-04 profiler network was designed with more coastal profilers to help characterize outflow of pollution from the eastern U.S. In addition, results from the NEAQS-02 study suggested that the profilers in New York State were too far north to document pollution transport from the Ohio River Valley. Therefore the ICARTT-04 profiler network included a site further to the south and west at Pittsburgh, Pennsylvania. Locating a profiler at ADI suited the objectives of both field campaigns and provided us with the opportunity afterward to study how the local wind conditions were related to the two disparate summers.

Figure 9.

Map showing the locations of the wind profilers available for the NEAQS-02 and ICARTT-04 field campaigns and the wind profiler trajectory tool.

Table 4. NOAA and Cooperative Agency Wind Profilers That Provided Data for the NOAA Wind Profiler Trajectory Tool in Either or Both of the NEAQS (2002) or ICARTT (2004) Field Campaigns
Map NameLocationLatitude, degLongitude, degElevation, mAvailabilitySponsor
ADIAppledore Is., ME42.99−70.625yesyesNOAA
BHBBar Harbor, ME44.44−68.364 yesNOAA
CCDConcord, NH43.21−71.52104yesyesNOAA
CBEChebogue Pt., NS43.75−66.1215 yesNOAA
RHBGulf of Mainevariousvarious5yesyesNOAA
DALLunenburg, NS44.40−64.3030 yesEnvironment Canada
RUTNew Brunswick, NJ40.50−74.4510yesyesRutgers University, NJ DEP
OREOrange, MA42.57−72.29169yes NOAA
PEAPease Tradeport, NH43.09−70.8330yesyesNOAA
PSPPinnacle SP, NY42.09−77.21587yes NOAA
PITPittsburgh, PA40.48−80.26335 yesNOAA
PYMPlymouth, MA41.91−70.7346yesyesNOAA
SCHSchenectady, NY42.85−73.93115yes NOAA
STOStorrs, CT41.80−72.22198 yesNOAA

[22] The type of ultra-high-frequency (UHF) wind profilers used in these studies is described by Carter et al. [1995]. Wind profilers are Doppler radars that do not require a hard target to receive a signal. In the clear atmosphere, wind profilers receive backscattered signals from refractive index turbulence. Winds are retrieved by transmitting pulses of electromagnetic radiation to form a vertical beam and at least two orthogonal off-vertical beams, a technique that is commonly referred to as Doppler beam swinging (DBS). The wind speed and direction are calculated by combining the Doppler-shifted signals retrieved from each beam direction. The return signals are also sampled at discrete intervals called “range gates.” This sampling strategy results in a height-resolved wind profile. All wind profiles measured over a particular sampling period (hourly, in this case) are averaged together using a consensus routine that aids in removing outliers.

[23] The hourly winds measured during NEAQS-02 and ICARTT-04 were transmitted from the profiler sites to a data hub in Boulder, Colorado, using conventional telephone lines where available or, in remote areas, via satellite communications. After each experiment, quality control procedures were applied to the wind profiler data sets using the Earth System Research Laboratory (ESRL) Physical Sciences Division wind editor, which includes the time-height continuity method developed by Weber et al. [1993].

[24] At each of the NOAA land-based profiler sites, a meteorological tower was deployed to measure surface winds and other relevant meteorological variables. Measuring the surface winds is important because the lowest altitude measured by the wind profilers is usually about 120 m above ground level (AGL). NOAA/ESRL measured surface winds on ADI at an altitude of 10 m AGL using an RM Young model 5103 wind monitor. The signals were acquired with a Campbell CR23X data logger and averaged to 2-min resolution. The final data products consist of vector- and scalar-averaged wind speed, wind direction, and peak gust. The archived wind profiles and surface meteorology data from each field experiment are publicly available on the internet.

4.2. Local Wind Conditions

[25] To compare how the surface winds at ADI behaved on high-ozone days as compared to the rest of the field campaigns, we generated wind rose diagrams. Figure 10 shows wind roses for the July–August period of both field campaigns and for the highest-ozone (an exceedance day with at least one 10-min ozone measurement >100 ppbv) days in both seasons. During both NEAQS-02 and ICARTT-04, the highest-ozone days occurred when the wind directions fell primarily within a sector from 150° to 240°. For the 2002 highest-ozone days, 71% of the surface flow fell within this sector. For the single 2004 highest-ozone day, 82% of the surface flow fell within this sector. Given this distribution, one might have expected July–August 2004, when 51% of the surface flow fell within this sector, to produce a similar or greater number of high-ozone days than July–August 2002, when only 45% of the surface flow fell within the sector. In reality, the 2004 period produced only one day with ozone >100 ppbv, whereas the 2002 period produced 11 days.

Figure 10.

Wind rose diagrams of surface winds measured at Appledore Island, Maine (ADI), during (a) July–August 2002, (b) July–August 2004, (c) the 11 episode days during July–August 2002 with high ozone (one or more 10-min average samples ≥100 ppbv) measured at ADI, and (d) the one episode day during July–August 2004 with high ozone measured at ADI. The gray-shaded numbers give the percentage of time that the wind direction was within each of the 30° sectors.

[26] Because of the strong vertical wind shear present in the North Atlantic marine boundary layer [Millet et al., 2006], we were motivated to look at the winds above the surface using the ADI wind profiler. Figure 11 shows diurnally averaged wind profiles for the July–August period of both field campaigns. The color background indicates the wind persistence, defined as the vector wind speed divided by the scalar wind speed. The overall lower-tropospheric wind patterns measured by the wind profiler agree with the 925-mbar geopotential height analyses; that is, the flow was decidedly more northwesterly in 2002 than in 2004. Both years exhibited a diurnal variation in wind direction in the lowest ∼500 m consistent with a land-sea breeze circulation. This pattern was more persistent in 2004 than in 2002. The sea breeze pattern at ADI produces low-level southwesterly to southeasterly flow during the afternoon and evening hours (local time). These wind directions make it more likely for ADI to be impacted by pollution transported from the Boston metropolitan area either directly from the southwest or following a more circuitous route over the southern Gulf of Maine.

Figure 11.

Diurnally averaged wind profiles for (a) July–August 2002 and (b) July–August 2004 measured by the Doppler wind profiler at ADI, Maine. The wind barb convention follows: Each half-barb accounts for a 2.5 m s−1 increase in wind speed, each full barb accounts for 5 m s−1, and the single half-barb covers winds speeds from 1.2 to 3.7 m s−1. The color background indicates the wind persistence (%), which is defined as the vector wind speed divided by the scalar wind speed. Altitudes where data coverage fell below 50% because of reduced signal are indicated by open circles.

4.3. Regional Transport

[27] The results of the previous section identified patterns in the local winds observed at ADI when high-ozone events were in progress. In order to characterize and compare transport during the two field studies, we examine back trajectories. Many forms of model-based trajectories exist, but for this study we chose to use back trajectories based solely on the wind observations from the profiler networks and from ocean buoy and Coastal-Marine Automated Network (C-MAN) stations deployed over the Gulf of Maine and along the New England coastline. This observationally based trajectory tool is described by White et al. [2006].

[28] Numerous studies linking transport patterns to chemistry measurements using cluster analysis of back trajectories have been published [e.g., Moody and Galloway, 1988; Fernau and Samson, 1990a, 1990b; Brankov et al., 1998; Cape et al., 2000; Eneroth et al., 2005; Abdalmogith and Harrison, 2005]. All of these studies used either gridded upper air measurements for back trajectory models or weather forecast model output to calculate the back trajectories. Since the upper air observations that were assimilated into these models were obtained only every 12 hours, the accuracy of the trajectories was largely dependent on the skill of the model forecast. The spatial resolution of the upper air measurements was also sparse, with few or no measurements over water.

[29] Positive aspects of using model output for the calculation of back trajectories (as seen in many of the above-referenced papers) include having access to a long, continuous record of gridded data, which allows for the assessment of annual and seasonal trends, when smaller-scale wind features are not as important, and the incorporation of pressure measurements to account for vertical motion (although Cape et al. [2000] found that the horizontal transport far outweighed the contribution of vertical transport in their cluster analysis results for Mace Head, Ireland). The cluster analysis results from these papers revealed preferred transport patterns for measured increases in the atmospheric constituents of interest, e.g., ozone or carbon dioxide.

[30] Since the measurements used for the radar wind profiler back trajectories were hourly averaged winds, the temporal resolution for these back trajectories was a vast improvement over using standard upper air measurements. The timescale of interest was 24 hours or less, making the smaller network appropriate for our analysis. These back trajectories were particularly useful for characterizing transport to ADI and in the northeast, in general, because of the complex mesoscale meteorology associated with the coastline and frequent changes in synoptic-scale weather, as indicated in Table 2.

[31] Twenty-four-hour back trajectories were completed with ADI as the endpoint for July and August of 2002 and 2004. The end times were on the hour, every hour, between 0000 UTC 2 July and 2300 UTC 31 August (1464 hours total). Trajectories were calculated for five different height levels: surface (buoys only), 0–250 m ASL (lowest profiler gates only), 250–500 m ASL, 500–1000 m ASL, and 1000–1500 m ASL.

[32] Each back trajectory comprises 24 latitude-longitude pairs indicating the calculated position of the air parcel every hour from 1 hour to 24 hours before the trajectory end time. Cluster analysis was performed on these latitude-longitude pairs. In the nonhierarchical cluster analysis method used, the cluster analysis algorithm sorted the 1464 trajectories, grouping like trajectories together in one of the 12 clusters, as described by Darby [2005].

[33] During the cluster analysis process, statistics were calculated to determine if the number of clusters chosen was reasonable. In this case 12 was a good number, as indicated by the differences in the standard deviation of the cluster members within each cluster versus the standard deviation among the clusters. With 12 clusters, the among-cluster standard deviation was more than twice the size of the within-cluster standard deviation. This indicates that there was much smaller variability among the members of the clusters, with substantial differences among the clusters, which is the goal of a successful cluster analysis. Using this method to assess cluster analysis success, fewer clusters could have been used in the present analysis (a minimum of 8), but we wanted the enhanced information garnered from 12 clusters.

[34] Some clusters had very few trajectories assigned to them (e.g., fewer than ten). In these cases, there were just a few consecutive hours out of the whole 2-month period that had this particular transport pattern. Many clusters had 100–350 trajectories assigned to them. The dominant transport patterns for each summer were determined by examining the percentage of trajectories assigned to each cluster.

[35] Figure 12 shows the cluster analysis results for 2002 and 2004. Each trajectory plotted is an average of all trajectories assigned to the cluster. Thus any small-scale features that may have been detected in any of the trajectories have been smoothed out. At the start point of each cluster-averaged trajectory is a single-letter identifier and the percentage of trajectories assigned to the cluster. The two most commonly occurring clusters are labeled with blue labels. The two clusters most likely to be coincident with the highest hourly averaged ozone concentrations at ADI are labeled in red.

Figure 12.

Cluster analysis of 24-hour back trajectories computed using the NOAA wind profiler trajectory tool for each hour (left) during July and August 2002 and (right) during July and August 2004. The end point for the back trajectories is Appledore Island, Maine. The dots indicate mean hourly trajectory positions in each of the 12 clusters. The two most commonly occurring clusters are labeled in blue. The two clusters most often coincident with high ozone concentrations are labeled in red.

[36] In the case of 2002, clusters E and G represent the two most common transport pathways for the period, indicating that westerly and northwesterly flow was quite common at low levels during summer 2002. Transport from the northwest quadrant occurred during ∼32% of the time analyzed. Flow from the southwest quadrant, however, was also common, with transport from this quadrant occurring ∼36% of the time, demonstrating a slightly different picture from that portrayed by the 925-mbar geopotential heights analysis (Figure 5, top). The two clusters representing back trajectories that were most likely associated with high ozone at ADI in 2002 were clusters B and D. Table 5, which shows the number of times the end time of a trajectory assigned to one of these clusters was coincident with hourly averaged ozone >70, 80, 90, and 100 ppbv at ADI, indicates that the back trajectories assigned to Cluster B were the most likely to be coincident with high ozone at ADI during 2002. Cluster B was composed of back trajectories with light winds from the southwest. Figure 13 shows all trajectories assigned to Cluster B, indicating the representativeness of the cluster analysis results shown in Figure 12, and emphasizes that the trajectories assigned to this cluster had similar wind speeds. Cluster D, the cluster second most likely to be associated with high ozone at ADI, represents back trajectories that had even lighter winds than those assigned to Cluster B, with a shift from northwesterly to southwesterly flow. Approximately the final 6 hours of transport was from the southwest. Combined, back trajectories assigned to either clusters B or D were associated with hourly averaged ozone concentrations > 100 ppbv 25 times.

Figure 13.

The 201 individual hourly back trajectory members composing cluster B in Figure 12.

Table 5. Number of Occurrences When the End Time of a Trajectory Member From a Particular Cluster Coincided With Hourly Averaged Ozone Greater Than 70, 80, 90, or 100 ppbv
YearCluster>70 ppbv>80 ppbv>90 ppbv>100 ppbv
2002B (see Figure 12, left)64/20145/20128/20115/201
2002D (see Figure 12, left)36/12733/12719/12710/127
2004N (see Figure 12, right)24/14715/1475/1473/147
2004M (see Figure 12, right)10/1693/1691/1690/169

[37] For 2004 the two clusters representing the most commonly occurring back trajectories were clusters P and T. Both of these clusters represent very light wind cases, one from southwest of ADI, the other from northeast of ADI. In a marked difference from NEAQS-02, transport from the southwest quadrant occurred ∼56% of the time, while transport from the northwest quadrant occurred only a mere 10% of the time. These results closely tie in with the 925-mbar geopotential heights analysis for ICARTT-04 (Figure 5, bottom). The two clusters representing back trajectories that were most likely to have their end times coincident with higher ozone at ADI were clusters M and N. Both of these indicated transport from the Boston area, as did the higher-ozone clusters from 2002. However, as seen in Table 5, the number of times ozone exceeded 70, 80, 90, or 100 ppbv with these back trajectories was far fewer than the data indicate for 2002, with only 3 hours with hourly averaged ozone concentrations > 100 ppbv. These two clusters combined occurred about the same amount of time as clusters B and D in 2002 (22% in 2004 vs. 23% in 2002). Clearly, factors other than horizontal transport were involved in determining the number of high-ozone events experienced at ADI during 2002 and 2004.

4.4. Boundary Layer Depth

[38] Other than horizontal transport, vertical mixing is perhaps the most critical process controlling the concentration of pollutants in the boundary layer. The depth of the convective boundary layer (CBL) determines the vertical extent over which mixing occurs in the atmosphere during the daytime. Wind profilers are capable of detecting the depth of the CBL because there is an enhancement in radar backscattered energy at the capping inversion, where mean gradients in temperature and humidity occur. We chose to use the Concord, New Hampshire, wind profiler to measure the depth of the CBL over land, which is more representative of the vertical mixing that occurs where the pollution sources exist than the depth of the boundary layer over the water or near the coastline. Boundary layer depths were estimated by visual inspection of the high-time-resolution (approximately 90 s) radar reflectivity vertical profiles [Bianco and Wilczak, 2002; Angevine et al., 1994].

[39] Figure 14 compares hourly averaged daytime boundary layer depths measured at Concord, New Hampshire, for the July–August 2002 and 2004 periods and for the highest-ozone days observed during each period. The boundary layers were significantly deeper during NEAQS-02 (red lines) than during ICARTT-04 (blue lines). One may be led to believe that the shallower boundary layers observed in 2004 would lead to more high-ozone events because of the decreased atmospheric volume containing the pollutants. However, the boundary layer depths during the highest-ozone days in both years were even deeper than the means for the two seasons. This is in contrast to the southeastern U.S. and central valleys of California, where pollution events are often associated with shallower boundary layers caused by strong subsidence inversions [Banta et al., 1998; Seaman et al., 1995]. In the northeast, the conditions that are conducive to producing high ozone, i.e., warm and dry with abundant sunshine, are the same conditions that produce deeper boundary layers.

Figure 14.

Diurnally averaged convective boundary layer (CBL) heights measured at NOAA's field research station located in Concord, New Hampshire, for the periods indicated in the key, with episode days as defined in Figure 10. Hourly CBL heights were estimated by visual inspection of radar reflectivity profiles measured by the wind profiler located at Concord, New Hampshire. For the bimonthly curves, the error bars enclose two standard deviations.

[40] This point is illustrated further by Figure 15, which compares the sensitivity of the ozone concentration observed near the profilers deployed in the northeast for the ICARTT-04 study and the observed boundary layer depths estimated from those profilers with the same variables forecast by three different coupled meteorology/air chemistry models. A description of each model is given by McKeen et al. [2005]. The daily maximum hourly boundary layer depths measured by the profilers and forecast by each of the numerical models were compared to the daily maximum 8-hour average ozone concentrations. Least squares linear regression fits to the comparison data are shown in Figure 15. All three models show the same trend as the observations: first, there is a weak correlation between boundary layer depth and ozone concentration, and second, deeper boundary layers tend to be associated with higher ozone concentrations. Given the observational and numerical modeling evidence, we conclude that the deeper boundary layers observed during NEAQS-02 did not cause more high-ozone events than during ICARTT-04, but rather they were coincident with the warm, dry, sunny days that resulted in producing more ozone.

Figure 15.

Sensitivity of maximum 8-hour ozone concentration to daily maximum boundary layer depth as observed by four of the ICARTT-04 wind profilers and EPA AIRNow ozone sensors in the vicinity of the wind profilers (black line) and as predicted by three different coupled meteorology/air chemistry model configurations listed in the key (colored lines). The four wind profilers used in this analysis were located at Bar Harbor, Maine; Concord, New Hampshire; Pittsburgh, Pennsylvania; and Plymouth, Massachusetts. Least squares linear regression fits to the comparison data are shown.

5. Summary and Conclusion

[41] This paper examined several meteorological processes that could have potentially influenced the number of high-ozone events observed during NEAQS-02 and ICARTT-04. A large difference in the number of ozone exceedances was observed during the two field campaigns. During NEAQS-02 there were 17 days when daily maximum 8-hour average ozone concentration at ADI exceeded 80 ppbv. During ICARTT-04 there were only two days. More generally, during the 10-year period 1996–2005, the New England states experienced the greatest number of ozone exceedances in 2002 (43) and the least number of exceedances in 2004 (13).

[42] To accomplish our objectives, we took advantage of special and operational observing systems, networks, and algorithms made available for these studies, including the University of New Hampshire's AIRMAP network, The Environmental Protection Agency's AIRNow Network, NOAA and cooperative agency wind profiler networks, NOAA's instrumented research vessel Ronald H. Brown, NOAA and cooperative agency ocean buoy and C-MAN surface observing stations in the Gulf of Maine and the surrounding vicinity, and the NOAA wind-profiler trajectory tool. In addition, we used the composite synoptic reanalysis tool developed by NOAA/ESRL Physical Sciences Division (a portion of which was formerly NOAA's Climate Diagnostics Center), daily weather maps from NOAA/Hydrometerological Prediction Center's Daily Weather Map Project, and climate station data and monthly state climate rank maps from NOAA's National Climatic Data Center. We also used the output of coupled meteorology/air chemistry models to evaluate the relationship between vertical mixing and ozone.

[43] Composite synoptic 925-mbar geopotential height fields hinted that the mean pollution transport pathway for producing high-ozone events in the northeast was more prevalent in 2004 than in 2002. This was confirmed using surface and upper air wind measurements from ADI as well as cluster analysis of surface and upper air back trajectories calculated from the wind-profiler networks and ocean buoy/C-MAN station networks.

[44] Both field study seasons were synoptically active. The average time between cold fronts was more than a day shorter in 2004 than it was in 2002. However, both field seasons experienced an extended lull in synoptic activity. In 2002, this lull resulted in the worst air quality event of either field season, whereas during the 2004 synoptic lull, there was not a single high-ozone day observed at ADI.

[45] Daytime vertical mixing was evaluated by comparing the CBL depths measured by the wind profiler at Concord, New Hampshire, during the two field campaigns. During NEAQS-02, the CBLs were deeper than during ICARTT-04 (peak afternoon values differed by 223 m), but both observations and numerical model results from the northeast indicated that highest-ozone events were associated with deeper boundary layers because the warm, dry, sunny conditions that favor local ozone production also produce deeper boundary layers. We concluded that the deeper CBLs observed during the 2002 study did not cause the greater number of observed high-ozone events, but rather were the result of the coincident meteorological conditions that produced these events.

[46] We also investigated the local, regional, and national climatic conditions that were present during July–August 2002 and 2004. Here it was found that both field campaigns were conducted under climate extremes. NEAQS-02 was conducted under anomalously hot and dry conditions. During August 2002, which produced the worst air quality episode measured at ADI during either field campaign (Figure 4 and Table 1), eight northeastern and mid-Atlantic states experienced one of their top ten warmest Augusts on record (Figure 6). The state of Maine experienced its driest August on record (Figure 8). Major coastal cities stretching from Portland, Maine, to Washington, D.C., were all substantially warmer and drier than normal (Table 3). These conditions were consistent with the mean ridge pattern identified in the composite synoptic 925-mbar geopotential height fields. In contrast, the 2004 study period was cool, especially upstream of the northeast, and extremely wet. Most of the eastern three quarters of the nation was cooler, if not much cooler, than normal (Figure 6), and record or near-record precipitation occurred over the entire northeast, with New York and Pennsylvania each receiving record statewide precipitation during the month of July (Figure 8). These conditions were consistent with the mean trough pattern identified in the 925-mbar composite synoptic geopotential height fields. The same trends held true for the bimonthly mean climatic data observed at the six coastal cities (Table 3).

[47] We conclude that the local, regional, and national climatic conditions were the primary cause for the large difference in the number of ozone exceedances observed at ADI during NEAQS-02 and ICARTT-04. One could argue that conducting these field campaigns under two diametrically opposed sets of extreme climatic conditions (warm and dry versus cool and wet) was fortunate because it will allow scientists to study the impact of a possible climate change scenario on regional air quality. However, the observed meteorological variability means that care should be taken when interpreting research results and policy-relevant findings derived from either study. From another viewpoint, the accidental scheduling of these air quality field studies during climate extremes points to the value of extending, when the cost is not prohibitive, research field campaigns to cover multiple, successive seasons of interest. This approach has recently been undertaken with summertime measurements conducted in Mexico for the North American Monsoon Experiment and wintertime measurements in the western U.S. for NOAA's Hydrometerological Testbed.


[48] We thank the dedicated engineers and technical staff responsible for installing, operating, and maintaining the wind profiler and surface meteorology networks used in this study. We also thank Tom Shyka with the Gulf of Maine Ocean Observing System for providing buoy and C-MAN station data for the NOAA wind profiler trajectory tool. The ozone data from Appledore Island were provided by the University of New Hampshire's AIRMAP Program, which is supported by NOAA's Office of Oceanic and Atmospheric Research. Climate station data and climate maps were provided by NOAA's National Climatic Data Center. We gratefully acknowledge Cathy Smith of NOAA's Earth System Research Laboratory, and additional colleagues at the former NOAA-CIRES Climate Diagnostic Center, for developing the NCEP–NCAR reanalysis products and making them available at The critical reviews written by Sara Michelson, Klaus Weickmann, and three anonymous reviewers helped improve this manuscript substantially. This research was sponsored by the NOAA Health of the Atmosphere Program.