Regional and Hemispheric Influences on Temporal Variability in Baseline Carbon Monoxide and Ozone over the Northeast US

Abstract

Interannual variability in baseline carbon monoxide (CO) and ozone (O₃), defined as mixing ratios under minimal influence of recent and local emissions, was studied for seven rural sites in the Northeast US over 2001 – 2010. Annual baseline CO exhibited statistically significant decreasing trends (−4.3 – −2.3 ppbv yr⁻¹), while baseline O₃ did not display trends at any site. In examining the data by season, wintertime and springtime baseline CO at the two highest sites (1.5 km and 2 km asl) did not experience significant trends. Decadal increasing trends (~2.55 ppbv yr⁻¹) were found in springtime and wintertime baseline O₃ in southern New Hampshire, which was associated with anthropogenic NO_x emission reductions from the urban corridor. Biomass burning emissions impacted summertime baseline CO with ~38% variability from wildfire emissions in Russia and ~22% from Canada at five sites and impacted baseline O₃ at the two high elevation sites only with ~27% variability from wildfires in both Russia and Canada. The Arctic Oscillation was negatively correlated with summertime baseline O₃, while the North Atlantic Oscillation was positively correlated with springtime baseline O₃. This study suggested that anthropogenic and biomass burning emissions, and meteorological conditions were important factors working together to determine baseline O₃ and CO in the Northeast U.S. during the 2000s.

1. Introduction

Carbon monoxide (CO) is a product of incomplete combustion (e.g. fossil fuel, biofuel, and biomass burning) and oxidation of hydrocarbon compounds (Worden et al., 2013). In the presence of nitrogen oxides (NO_x), oxidation of volatile organic compounds (VOCs) and CO can lead to the photochemical formation of ozone (O₃). CO is a major sink of hydroxyl radicals (OH) and O₃ is a precursor of OH, and hence they significantly affect the oxidizing capacity of the troposphere (Prinn, 2003). Tropospheric CO and O₃ are also serious and ubiquitous air pollutants, affecting human and natural ecosystem’s health (EPA, 2012).

CO has been used as a tracer of anthropogenic influence and fire emissions, due to its relatively unreactive chemical nature (Gratz et al., 2014; Price et al., 2004). A positive correlation between CO and O₃ has been identified in summer at various locations, attributed largely to the predominant influence of photochemical processes (Mao and Talbot, 2004; Hegarty et al., 2009). Therefore, many studies have used O₃-CO correlation to constrain O₃ sources and transport (e.g., Mao and Talbot, 2004; Hudman et al., 2009; Kim et al., 2013). The O₃-CO relationship was more complicated in other seasons due to the effects of stratospheric intrusion, dry deposition of O₃, or titration of O₃ by NO (Mao and Talbot, 2004; Voulgarakis et al., 2011; Kim et al., 2013).

The United States has made enormous efforts to control ambient mixing ratios of criteria pollutants since the 1970s (EPA, 2012). Nationally, annual second maximum 8 h average mixing ratios of CO decreased by 52%, and annual mean mixing ratios of nitrogen dioxide (NO₂) declined by 33% over 2001 – 2010 (EPA, 2012). Although anthropogenic emissions have decreased in Europe and North America, emissions in China and India have increased. Biomass burning emissions vary both spatially and temporally (Granier et al., 2011). The lifetime of CO and O₃ in the free troposphere is ~2 months and ~20 days, respectively (Price et al., 2004; Stevenson et al., 2006). Thus, transport of CO, other O₃ precursors, and O₃ from an upwind region as well as amounts produced in transit could affect downwind baseline CO and O₃ levels (Oltmans et al., 2008; Pollack et al., 2013). It remains unclear how opposing changes in emissions over North America, Europe and Asia have globally affected baseline CO and O₃, which are defined as mixing ratios of CO and O₃ under minimal influence of recent and local emissions (Chan and Vet, 2010; HTAP, 2010).

The terms “background” and “baseline” have often been used interchangeably, and a few studies discussed the difference (HTAP, 2010; Huang et al., 2015). The term “background” was used in modeling studies that estimated the atmospheric mixing ratio of a compound determined by natural sources only, while the term “baseline” was obtained from measurement records by removing data affected by local influences (Chan and Vet, 2010). Various methods have been utilized to diagnose baseline conditions, including using measurements at remote sites, analysis of the probability distribution of pollutants, correlations with reactive nitrogen oxides, or isentropic back-trajectories (e.g., Lin et al., 2000; Derwent et al., 2007; Wilson et al., 2012). Air masses with monthly to annual low percentile values of CO are commonly considered baseline air (e.g., Lin et al., 2000; Mao and Talbot, 2012). The monthly 10^th percentile value of afternoon CO was defined as baseline CO in this study for the lower elevation sites while the monthly 50^th percentile value of afternoon CO was used at high elevation sites. Baseline O₃ was then estimated as the monthly median of the afternoon O₃ data corresponding to CO mixing ratios below the baseline CO level (Lin et al., 2000). It should be noted that the monthly baseline CO and O₃ defined here excluded the influence of processes on hourly to daily time scales (DeBell et al., 2004; Honrath et al., 2004; Dutkiewicz et al., 2011b).

Trends in baseline O₃ have been investigated for northern hemispheric mid-latitude regions, such as North America, Europe, and Asia, and no consistent trends have been found (e.g., Chan, 2009; Cui et al., 2011; Logan et al., 2012; Oltmans et al., 2013; Wilson et al., 2012; Cooper et al., 2010; Xu et al., 2008). However, decreasing trends were found in baseline CO and CO concentrations in rural areas since the end of the 1980s (e.g., Hallock-Waters et al., 1999; Novelli et al., 2003; Duncan and Logan, 2008). Interpretation of long-term trends is difficult because interannual variability in emissions, climate, and photochemistry are intricately interwoven (Hess and Lamarque, 2007). CO is a long-lived tracer of combustion and using changes in CO, one can estimate continental influences on O₃ globally (Hudman et al., 2009). However, only a few studies (Kumar et al., 2013; Gratz et al., 2014) investigated the long-term trends in baseline CO and O₃ together. Kumar et al. (2013) found that trends of −0.31 and −0.21 ppbv yr⁻¹ for CO and O₃, respectively, at the Pico Mountain Observatory could be attributed to the North American anthropogenic emission declines over 2001 – 2010, and that climate change may have affected the intercontinental transport of O₃. Gratz et al. (2014) found that the springtime median mixing ratio of O₃ increased at a rate of 0.76 ppbv yr⁻¹ at the Mt. Bachelor Observatory over 2004 – 2013, while median CO decreased at a rate of −3.1 ppbv yr⁻¹. Overall, causes for temporal variability in both baseline O₃ and CO have not been adequately explained.

Although a number of studies have been conducted to understand the distributions of surface CO and O₃ in the Northeast US and their controlling mechanisms (e.g. Bae et al., 2011; DeBell et al., 2004; Mao and Talbot, 2004; Schwab et al., 2009), little work has been done on baseline CO and O₃ together using long-term measurement data for the region, in particular using the data from the seven sites in this study. Here, we aimed to examine the trends in baseline CO and O₃ as well as interannual and seasonal variation at seven rural sites in the Northeast U.S. over 2001 – 2010 and investigate their potential association with emissions, dynamics, and O₃ photochemistry.

2. Methods and data

2.1. Measurement data

Our study used long-term observations at seven rural sites in the Northeast US. Five are located in rural New Hampshire (NH) and two are in rural New York (NY) State. These sites are 100 – 200 km away from major sources (Table 1 and Fig. 1). The Appledore Island (AI) site is located in the marine boundary layer, ~11 km offshore in the Gulf of Maine, while the Thompson Farm (TF) site is located in an open lot surrounded by agricultural fields and mixed vegetation, ~21 km away from the coastline. The Castle Spring (CS) and Pack Monadnock (PM) sites are ~75 km to the northwest and west of TF, respectively. These four sites are on the eastside of the Appalachian Mountain Range. The Mountain Washington Observatory (MWO) and Whiteface Mountain (WFM) sites are located on the summit of the Appalachians, ~2 km and 1.5 km a.s.l., respectively. The Pinnacle State Park (PSP) site is located on the west side of the Appalachians.

Table 1.

Ground stations with geographical coordinates and measurement periods.

Site	Latitude	Longitude	Elevation	Measurement Period (CO)	Measurement Period (O3)	Time Resolution (min)
Appledore Island (AI)*	42.97°N	70.62°W	18 m	Jul, 2001–Jul, 2011	Jul, 2002–Mar, 2012	1
Thompson Farm (TF)	43.11°N	70.95°W	23m	Apr, 2001–Jul, 2011	Apr, 2001–Aug, 2010	1
Mt. Washington (MWO)	44.27°N	71.30°W	1917m	Apr, 2001–Apr, 2009	Apr, 2001–May, 2010	1
Castle Spring (CS)	43.75°N	71.35°W	396m	Apr, 2001–Jun, 2008	Apr, 2001–May, 2008	1
Pack Monadnock (PM)	42.86°N	71.88°W	698m	Jun, 2004–Jul, 2011	Jul, 2004–Oct, 2008	1
Whiteface Mountain (WFM)	44.40°N	73.90°W	1484 m	Jan, 1996–Dec, 2010	Jan, 1996–Dec, 2010	60
Pinnacle State Park (PSP)	42.09°N	77.21°W	504 m	Jan, 1997–Dec, 2010	Jan, 1997–Dec, 2010	60

^*CO at AI was measured seasonally from May to September over 2001 – 2006 and year-round over 2007 – 2011, and O₃ was measured seasonally from May to September over 2002 – 2007 and year-round over 2008 – 2011.

Note: The time in all of the datasets was expressed in coordinated universal time (UTC), i.e. local time +5 h for non-daylight saving time and +4 h for daylight saving time (March–November).

Figure 1.

Map of the Northeast U.S. The seven measurement sites used in the study are marked with black dots.

Measurements of CO, O₃, wind direction, wind speed, and relative humidity at AI, CS, MWO, PM, and TF were conducted continuously by the University of New Hampshire AIRMAP Observing Network (http://www.eos.unh.edu/observatories/). All measurements have undergone rigorous quality controls as described in Section S1 in the Supplement. Measurements at PSP and WFM were operated by the Atmospheric Sciences Research Center at State University of New York at Albany (Section S1). Hourly data of solar radiation flux at the Harvard Forest was downloaded from http://harvardforest.fas.harvard.edu/.

2.2. Quantification of baseline CO and O₃

The local afternoon time window (18:00 – 24:00 UTC) was selected to avoid including the data representing nighttime depletion of O₃ due to dry deposition and titration (Talbot et al., 2005), and the afternoon well-mixed planetary boundary layer (PBL) best includes the influences of local and distant sources. The monthly 10^th percentile mixing ratio of CO at AI, CS, PM, TF, and PSP was used to represent baseline CO levels. As MWO and WFM are located atop the mountains, they are far less impacted by local anthropogenic emissions (Dutkiewicz et al., 2011a). Therefore, monthly median values of CO were selected at MWO and WFM to represent baseline levels there. To determine baseline O₃ levels for individual sites, we first created a subset of O₃ data corresponding to CO mixing ratios below monthly 10^th percentile values at AI, CS, PM, TF, PSP and monthly median values at MWO and WFM. The monthly median value of this subset at each site was then defined as the site’s baseline O₃. As the datasets used for this study started in 1997 at PSP and 1996 at WFM, comparison in baseline CO and O₃ were made between the time period of 2001 – 2010 and the years prior to that at these two sites.

2.3. Global datasets of climate indices, wildfires, meteorology, and tropospheric O₃

The North Atlantic Oscillation (NAO) index is a measure of the intensity of NAO, which is defined based on the leading empirical orthogonal function of the normalized sea level pressure difference between the subtropical high and the subpolar low (Barnston and Livezey, 1987). The Arctic Oscillation (AO) index was obtained by projecting the daily 1000 hPa geopotential height anomalies poleward of 20°N onto the loading pattern of the AO (Thompson and Wallace, 2000). The monthly climate index values of NAO and AO used in this study were obtained from the Climate Prediction Center of the National Centers for Environmental Prediction (NCEP) (http://www.cpc.ncep.noaa.gov/data/teledoc/telecontents.shtml).

The Global Fire Emission Data (GFED) combines satellite information of fire activities and vegetation productivities and contains gridded monthly burned area and fire emissions with 0.5°×0.5° horizontal resolution since 1997. GFED 3 (http://www.globalfiredata.org/) was used in this study to estimate the biomass burning emissions of CO over Russia, Canada, California, and Alaska. Monthly mean global CO columns with 1°× 1° resolution obtained from the Measurements of Pollution in the Troposphere (MOPITT) instrument on the satellite Terra (https://www2.acd.ucar.edu/mopitt/) were used over grids containing Russia, Canada, Alaska, and California, when wildfire CO emissions in these grids calculated from GFED were larger than 1 g m⁻²month⁻¹.

Monthly wind, geopotential height, temperature, relative humidity, and potential vorticity (PV) (http://rda.ucar.edu/) with 2.5°×2.5° resolution from the NCEP National Center for Atmospheric Research (NCAR) were used for meteorological conditions and for identifying stratospheric intrusion. The dataset representing O₃ of stratospheric origin (Liu et al. 2013, ftp://es-ee.tor.ec.gc.ca/pub/ftpdt/Stratospheric%20Climatology/), was also used to verify the contribution of stratospheric O₃ to the two mountain sites and the decadal trends there. This dataset included monthly amounts of stratospheric O₃ from the surface to 26 km altitude with 5°× 5°× 1 km spatial resolution from the 1960s to the 2000s.

2.4. Mid-latitude cyclone identification and tracking

Many algorithms have been developed since the 1970s to identify mid-latitude cyclones (Hu et al., 2004; Murazaki and Hess, 2006; Racherla and Adams, 2008). The algorithm developed by Bauer and Del Genio (2006) was adopted in this study to track the sea level pressure minima with the following two steps. First, the local minimum was searched for by a 2 grids × 2 grids matrix. Second, the local (within 720 km) minimum was searched for in the next time step, assuming that a cyclone cannot move faster than 120 km h⁻¹, the same criterion used by Bauer and Del Genio (2006). If more than one local minimum was found, the point with the lowest sea level pressure was designated as the center of a cyclone. Two more criteria were applied, its duration > 24 h and central pressure ≤ 1020 hPa. Long-term cyclone frequency statistics were calculated for the Northeast US (37.5 – 47.5°N, 67.5 – 82.5°W). Six hourly mean sea level pressure data with a spatial resolution of 2.5°×2.5° from NCEP-DOE Reanalysis 2 (http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis2.html) were used in this analysis.

2.5. Statistical methods

The open-air package in the statistical programming language R 3.0.2 was used to determine the statistical significance of a rate of change. Trends in baseline CO and O₃ in ppbv yr⁻¹ were reported using SenTheil slopes from non-parametric Mann–Kendall analysis with 90% confidence intervals. Pearson correlation was computed to determine the relation between variables (e.g. baseline CO, baseline O₃, NAO index, relative humidity). The Student t test was conducted to verify statistical significance (α = 0.10).

To quantify the contribution of biomass burning emissions over an area to a location of interest, we applied the linear regression model (Wotawa et al., 2001; Jaffe et al., 2004; Lu et al., 2016) using:

$$\text{CO}={\text{a}}_0+{\text{a}}_{1}E+\epsilon$$ (1)

where CO is the mixing ratio of baseline CO at each site, E the total CO column over the area, a₀ the intercept value, a₁ the slope value, and ε the error. The combined effect of biomass burning emissions over two areas was computed using:

$$\text{CO}={\text{b}}_0+{\text{b}}_1{E}_1+{\text{b}}_2{E}_2+\epsilon$$ (2)

where b₀, b₁, b₂, and ε are regression parameters, and E₁ and E₂ the total CO column over the two areas.

3. Characteristics of temporal variations in Baseline CO and O₃

Time series of baseline CO and O₃ both showed annual maxima in spring and minima in fall during the entire study period at most sites, except baseline CO at WFM over 2001 – 2010 (Fig. 2). Baseline CO showed averaged annual maxima ranging from 149 to 189 ppbv uniformly in March and minima ranging from 103 to 142 ppbv in varying months between August and October at AI, CS, MWO, PM, TF, and PSP from 2001 to 2010 (Figs. 2a-b). Baseline O₃ at all sites displayed very similar averaged annual maxima, ranging from 47 to 51 ppbv, in April and averaged annual minima varying between 28 and 37 ppbv, in August – October over 2001 – 2010 (Figs. 2c-d). Baseline CO at WFM had an average annual maximum (171 ppbv) in February and minimum (104 ppbv) in July over 1996 – 2000 (Fig. 2b). Note that, over 2001 – 2010, the pattern of annual cycles at WFM was the opposite with annual maxima averaged 144 ppbv in July and minima 103 ppbv in December (Fig. 2b), which was also found in Dutkiewicz et al. (2011a). The change of the pattern of annual cycles occurred in 2000, when the annual averaged mixing ratio was extremely low (70 ppbv) and no distinct annual cycle was found in this year. The unusual annual cycle of baseline CO at WFM might be caused by combined effects of mountain meteorology, reduced nationwide emissions, and less exposure to reduced anthropogenic emissions in Montreal (Section S2; Dutkiewicz et al. 2011a).

Figure 2.

Monthly baseline CO (ppbv) at (a) AI, CS, PM, TF, and PSP, and (b) MWO and WFM. Monthly baseline O₃ (ppbv) at (c) AI, CS, PM TF, and PSP, and (d) MWO and WFM.

The Mann-Kendall test suggested that baseline CO decreased significantly at rates ranging from −2.5 to −4.3 ppbv yr⁻¹ (p < 0.01) at MWO, PM, TF, PSP, and WFM over 2001 – 2010 (Table 2). Unlike for baseline CO, the Mann–Kendall test suggested no significant trends in baseline O₃ during the time period of 2001 – 2010 at all sites and the time period prior to 2001 at PSP and WFM (Table 2). Baseline CO at CS exhibited an increasing trend of 2.8 ppbv yr⁻¹ over April 2001 – June 2008. Prior to May 2003, the mixing ratio of baseline CO at CS was similar to that at TF. After May 2003, baseline CO at CS was ~30 ppbv higher than that at TF (Fig. 2a), resulting in the overall increasing trend. No possible causes were found for such increases.

Table 2.

Trends (ppbv yr⁻¹) of baseline CO and O₃ in spring, summer, fall, and winter. p-values are in the parentheses. Boldfaced numbers indicate p-value < 0.10.

		Spring		Summer		Fall		Winter		Annual
Site	Period	CO	O3	CO	O3	CO	O3	CO	O3	CO	O3
AI	2002–2010			0.8(0.66)	−3.1(0.07)
CS	2001–2008	3.4(0.06)	0.9(0.65)	2.4(0.19)	−2.9(0.14)	1.1(0.57)	1.5(0.45)	6.1(<0.01)	0.4(0.86)	2.8 (<0.01)	0.8 (0.39)
MWO	2001–2009	−13.2(0.51)	−0.7(0.71)	−4.5(0.01)	−4.7(0.01)	−4.4(0.01)	−0.9(0.64)	−1.7(0.36)	0.1(0.98)	−2.3 (<0.01)	−0.7 (0.42)
PM	2005–2010	−6.5(<0.01)	−1.9(0.39)	−5.5(<0.01)	−3.5(0.14)	−4.2(0.05)	−3.4(0.11)	−5.5(0.01)	0.1(1.00)	−3.5 (<0.01)	−0.8 (0.43)
TF	2001–2010	−3.7(0.02)	2.4(0.10)	−4.5(<0.01)	−0.1(0.94)	−3.2(0.04)	0.2(0.90)	−4.8(<0.01)	2.7(0.10)	−2.5 (<0.01)	0.8 (0.29)
PSP*	1997–2000	-	-	-	-	-	-	-	-	−3.1(0.01)	−1.2(0.32)
PSP	2001–2010	−4.5(<0.01)	1.3(0.43)	−4.3(<0.01)	−0.8(0.56)	−4.2(0.01)	−1.9(0.23)	−3.9(0.02)	−0.7(0.68)	−4.3 (<0.01)	−0.7 (0.40)
WFM	1996–2000	−7.7(<0.01)	1.3(0.62)	−8.8(<0.01)	−0.3(0.92)	−5.7(0.02)	−0.8(0.77)	0.2(2.38)	0.7(0.84)	−5.6(<0.01)	0.3(0.80)
WFM	2001–2010	−0.5(0.78)	0.4(0.83)	−1.9(0.23)	−4.7(<0.01)	−6.4(<0.01)	0.5(0.76)	−2.1(0.21)	−1.3(0.45)	−2.8 (<0.01)	−0.9 (0.27)

^*Trends at PSP were not computed for each season over 1997 – 2000, as only four-year datasets were available.

Trends in seasonal baseline CO and O₃ were also computed and were found to be variable at each site (Table 2). In spring and winter, baseline CO at PM, TF, and PSP decreased significantly at rates of −6.5 – −3.7 ppbv yr⁻¹ over 2001 – 2010 while no significant decreasing trends at the two highest sites MWO and WFM (Table 2). In spring and winter, TF was the only site where baseline O₃ increased significantly at rates of 2.4 ppbv yr⁻¹ and 2.7 ppbv yr⁻¹, respectively, while other sites showed no trends. Figure 3 further underscored large difference in variability of baseline CO and O₃ between sites in spring and summer. Emphasis was placed on spring and summer, when strong intercontinental transport and photochemistry involving O₃ and CO (Mao and Talbot, 2004) as well as exceedances of NAAQS tended to occur. In summer 2003, baseline CO and O₃ reached the decadal maxima at most sites, except baseline O₃ at PSP (Fig. 3c-d). Springtime baseline CO at TF also showed the largest value (170 ppbv) in 2003. In summer 2004, the second largest baseline CO mixing ratio of the decade (2001 – 2010) was found at TF and WFM. Both baseline CO and O₃ were the lowest of the decade (2001 – 2010) at AI, PM, TF, and PSP in summer 2009.

Figure 3.

Springtime averaged baseline (a) CO and (b) O₃ and summertime averaged baseline (c) CO and (d) O₃ at each site.

4. Discussion

According to the findings in the previous section, the following questions need to be addressed:

• 1)What factors drove the interannual variations of baseline CO and O₃? Did their influences vary from spring to summer? Which factor(s) was (were) dominant in each season?
• 2)Why were there decreasing trends in baseline CO and yet no trends in baseline O₃?
• 3)Why did baseline CO at WMO and WFM exhibit no trends in spring and winter? and
• 4)Why did baseline O₃ at TF show significant increasing trends in spring and winter?

The above questions were addressed in terms of changing anthropogenic and biomass burning emissions, stratospheric intrusion, meteorological conditions, and their geographical locations. Since the causes for the unusually high values at CS after May 2003 are unknown, baseline CO at CS was not included in the following discussion.

4.1. Impact of wildfires in summer

Studies (Hecobian et al., 2011; Oltmans et al., 2010) suggested that biomass burning effluents from Russia and Canada flowed into North America. In addition, California and Alaska were two US states with considerable fire emissions of CO, which reportedly impacted the air quality over North America (McKendry et al., 2011; Real et al., 2007). Fire emissions of CO in summer were estimated using the GFED dataset and MOPITT retrievals (Figs. 4a and b). The GFED data suggested that massive wildfires occurred in Russia in 1998, 2002, 2003, and 2008 with annual CO emissions of 64.0, 42.1, 71.1, and 35.8 Tg, respectively. The top three annual fire emissions of CO in Canada were 23.3 Tg in 1998, 17.4 Tg in 2004 and 18.3 Tg in 2010. In Alaska, the largest fires occurred in 2004 with 13.1 Tg CO emitted, while in California the largest were 1.3 Tg in 2008. Over 2001 – 2010, the total CO emissions from wildfires in Russia, Canada, Alaska and California varied from 19.9 to 84.3 Tg, with the lowest and the highest in 2007 and 2003, respectively.

Figure 4.

(a) CO emissions from biomass burning based on GFED dataset. (b) Total CO columns based on MOPITT retrievals over Russia (black), Alaska (red), Canada (blue), and California (magenta).

Monthly MOPITT total CO columns were found to be significantly correlated with monthly GFED fire emissions of CO for the four areas. The correlation coefficients were 0.89, 0.81, 0.81, and 0.84 (p < 0.01 for the four values) for Russia, Canada, California, and Alaska, respectively, suggesting that the variability in total column CO over those areas was dominated by that of fire emissions. Based on the linear regression model described in Section 2.5, it was found that the contributions of fire emissions from Russia and Canada to the variability of summertime baseline CO at the 5 sites averaged between 38% and 22%, respectively (Table 3), and their combined contribution averaged 41%. Contributions from Alaska and California were negligible at these five sites. Wotawa et al. (2001) used the same method and found that 14% of the CO variability in the extra-tropical Northern Hemisphere in the 1990s can be explained by boreal forest area burnt in North America, 53% by area burnt in Russia, and 63% by combination of both. Globally, there is approximately 1 billion ha closed forest in the boreal region, about two thirds of which is situated in Russia (Harden et al., 2000). CO emissions from wildfires in Russia and Canada contributed 49.5 and 29.6%, respectively, to the total CO emissions from wildfires in Northern Hemispheric midlatitudes (30 – 90°N). Understandably, baseline CO was well correlated with wildfires in Russia and Canada at most sites, except at PSP.

Table 3.

The contributions, in r², of CO emissions from wildfires over Russia, Canada, Alaska, and California to variation (r²) in baseline CO at each site. The combined effect of wildfire emissions over Russia and Canada was also computed. Boldfaced numbers indicate p-value < 0.10.

	Russia		Canada		Alaska		California		Combined
	r2	p	r2	p	r2	p	r2	p	r2	p
AI	0.39	0.01	0.12	0.15	0.13	0.19	0.12	0.21	0.41	0.02
CS	0.41	0.01	0.17	0.09	0.06	0.38	<0.01	0.92	0.41	0.02
MWO	0.41	0.01	0.13	0.15	0.01	0.77	<0.01	0.88	0.43	0.02
TF	0.64	0.01	0.40	0.05	0.01	0.80	0.03	0.52	0.65	0.01
PSP	0.11	0.18	0.15	0.11	0.09	0.27	<0.01	0.93	0.16	0.27
WFM	0.32	0.01	0.32	0.01	<0.01	0.90	0.01	0.69	0.38	0.03
Mean	0.38		0.22		<0.05		<0.03		0.41

Note: PM was not included due to insufficient data

The second largest summertime baseline CO mixing ratio of the decade (2001 – 2010) was found in summer 2004 at TF and WFM (Fig. 3c). During this summer, Alaska saw its decadal maximum (13.1 Tg) fire emissions of CO, Canada its second largest (17.4 Tg) of the decade, whereas Russia had below average emissions (Fig. 4a). The circulation pattern over Alaska and Canada appeared to be anomalous in summer 2004, with the stronger than normal high pressure system (Fig. 5a–c) leading to the driest (82% relative humidity) and warmest (12°C) summer of the decade (Fig. 5d). Such weather conditions are conducive to the occurrence of wildfires. In summer 2004, CO emissions from wildfires in Russian, Canada, and Alaska contributed 16%, 48.5% and 36.5%, respectively, to the Northern Hemispheric total, a stark contrast to 49.5%, 10.6%, and 29.6% on decadal average. Moreover, the predominant contributions from Canadian and Alaskan fire emissions in summer 2004 appeared to be closely linked to the unusual transport to the Northeast US via the anomalously strong northeasterly component of the high pressure system over Alaska and southwestern Canada (Fig. 5c). The combination of these two factors resulted in efficient transport of massive fire emissions of CO from Alaska and Canada.

Figure 5.

Geopotential height at the 850 hPa pressure level during summer in (a) 2001–2010 (b) 2004. (c) The difference of geopotential height at 850 hPa between summer 2004 and the 10-year average. (d) The annual surface temperature and relative humidity over Alaska and southwestern Canada (55°N – 70°N, 110°W – 160°W) over summer 2001 – 2010. Red stars indicated the area of the study sites. (Source: NCEP/NCAR reanalysis)

Since wildfires provide a substantial source of NO_x and VOCs, O₃ is expected to form in fire plumes. Corresponding to the largest fire emissions in summer 2003 in Russia (Fig. 4a), baseline CO in summer at all sites reached the decadal maxima, and baseline O₃ was the highest of all summers at AI, CS, MWO, TF, and WFM (Fig. 3c-d). Previous studies also found enhanced tropospheric O₃ levels over North America in summer 2003 and spring 2008, which was attributed to biomass burning in Siberia (Oltmans et al., 2010; Jaffe et al., 2004). The enhanced O₃ from Canadian boreal forest fires was reported to vary from being insignificant to substantial (Alvarado et al., 2010; Mckeen et al., 2002). Linear regression analysis suggested that contributions of fire emissions from Russia and Canada to the variability of summertime baseline O₃ at MWO were 27% (p = 0.07) over 2001 – 2010. No significant contribution to summertime baseline O₃ was found at other sites. On a decadal scale, wildfires appeared to have a significant impact on baseline O₃ at high elevated sites only, where the effect of continental to intercontinental transport was more important (Jonson et al., 2010).

Mann-Kendall trend analysis indicated that CO emissions from wildfires in Russia decreased at a rate of −0.51 Tg yr⁻¹ (p = 0.10) and no trends were found in other regions. Hence, the decreasing trend of biomass burning emissions in Russia likely contributed to the decreasing trends in summertime baseline CO and O₃ in the Northeast U.S. (Table 2).

4.2. Meteorological conditions

4.2.1. Impact of cyclone activity and AO in summer

Midlatitudinal cyclone frequency ranks highly among meteorological variables that can impact regional air quality. It affects not only boundary layer ventilation, humidity, solar radiation, and temperature but also general circulations of the regional atmosphere (Leibensperger et al., 2008). Time series of summertime counts of cyclones in the Northeast US showed strong interannual variability (Fig. 6a). Our calculated numbers of cyclones were consistent with the results for the same years in Leibensperger et al. (2008) and Bauer and Del Genio (2006). The counts of cyclones in 2003, 2006, 2008, 2009, and 2010 were greater than 12, the average of summer 2001–2010. Summer 2009 experienced the largest number of cyclones (20) passing the Northeast US over 2001 – 2010.

Figure 6.

(a) Counts of cyclones in the Northeast U.S. (black) and the AO index (blue) in summer. (b) Geopotential height at 500 hPa from the NCEP/NCAR reanalysis data during summer 2001 – 2010. (c) The difference of 500 hPa geopotential height between years with strong (2003, 2006, 2008, 2009, and 2010) and weak (2001, 2002, 2004, 2005, and 2007) cyclone activities. (d) The difference between geopotential height at 500 hPa in summer 2009 and the 10-year means. (e) The difference between sea level pressure in summer 2009 and the 10-year means. (f) Time series of summertime baseline CO (black) and O₃ (red) averaged over all seven sites, and CO emissions (blue) from wildfires in Russia and Canada. Dashed lines indicate the 10-year means. Red stars indicate the area of the study sites.

In summer, midlatitudinal cyclones tend to move around the 500 hPa vortex, which is located over the cold Arctic Ocean with broadly symmetric flow around it (Serreze et al., 2007). On the North American side, high latitudinal flow on the 500 hPa pressure level has a southerly component, which tends to steer systems away from the Arctic Ocean (Fig. 6b). Composite analyses of years of strong (2003, 2006, 2008, 2009, and 2010) vs. weak (2001, 2002, 2004, 2005, and 2007) cyclone activities revealed distinct differences in regional to large scale circulation (Fig. 6c). Results showed a pronounced positive difference of ~35 gpm centered over Baffin Island (north of the Northeast US) and a negative difference of ~25 gpm centered over the Northeast US (Fig. 6c). This difference was related to the negative phase of AO (Fig. 6a), when Arctic lows and westerlies are weaker, leading to more frequent cold-air outbreaks down to Eurasia and the US (Hess and Lamarque, 2007) and ultimately low baseline CO and O₃ there.

A case in point was summer 2009 with the largest cyclone count (20) and the strongest negative AO phase (−0.92) of the decade (Fig. 6a), and baseline CO and O₃ were the lowest of the decade at AI, PM, TF, and PSP. Consistent with earlier results, negative anomalies of 500 hPa geopotential height up to ~−60 gpm were found over the North American continent and positive anomalies up to ~65 gpm centered near the pole (Fig. 6d). The sea level pressure field (Fig. 6e) featured a pronounced mean low over southern Canada, and streamlines suggested an unusually strong northeasterly component. Indeed, the frequency distribution of wind direction at each site (not shown) suggested more frequent occurrence of northeasterly wind (22.5–112.5°) during 2009, with 21% at PM, 9% at MWO, 42% at TF, 11% at PSP, and 13% at WFM, compared to 10%, 5%, 12%, 6%, 8% averaged respectively during other summers. Decadal trends were computed for summertime cyclones and AO index, and no significant trends were found.

It should be noted that of all the sites, significant negative correlation was found between counts of cyclones and baseline O₃ only at PSP (r = −0.56, p = 0.05). PSP was the only site that did not seem to be significantly affected by the Russian and Canadian wildfire emissions during the 2000s (Section 4.1). Compared to other sites, PSP had the lowest wind speed of 0.47 m s⁻¹ (Fig. 7a), which appeared to be consistent with the position of PSP relative to the pressure systems year-round (Fig. 7b–e). The climatological seasonal composites of sea level pressure suggest that PSP is located either on the periphery of the subtropical high in summer-fall and on the periphery and near the axis of the U.S. east coast trough in fall-winter, where wind tends to be the weakest. In comparison, other sites are either located at the top of the boundary layer, and/or tend to be positioned within the U.S. east coast trough, more directly under the influence of the westerly wind often facilitating global transport. This very dynamic characteristic could cast a predominant influence of synoptic systems over the site, e.g., the Bermuda High and cold frontal passages. As commonly known, high mixing ratios of O₃ in the Northeast often occurred under summertime stagnant, clear sky conditions associated with the Bermuda High (Logan, 1989; Hegarty et al., 2007), while low O₃ was often linked to cold fronts which sweep out polluted air leaving much cooler and cleaner air in the Northeast (Leibensperger et al., 2008). Conceivably, with more frequent cyclones passing the Northeast US, lower concentrations of baseline CO/O₃ would be expected, and the predominant effect of such synoptic systems could lead to anticorrelation between the baseline CO/O₃ levels and cyclone activities. However, significant correlation was not found between baseline CO and counts of cyclones at PSP. This indicates that, for CO at PSP, global transport could be as important as regional transport facilitated by processes such as cyclones, whereas O₃ has a shorter lifetime and it was thus more sensitive to synoptic systems. The unique characteristics of baseline CO at PSP warrant further study.

Figure 7.

(a) Annual surface wind speed with yearly variation at each site over summer 2001 – 2010. Northeast U.S. sea surface pressure (hPa) in (b) spring, (c) summer, (d) fall, and (e) winter averaged over 1980 – 2010. Red stars indicate the location of PSP.

4.2.2. Impact of NAO in spring (March and April)

Wildfires in March and April were scarce, with mean CO emissions of 1.78 Tg in Russia and 0.004 Tg in Canada over 2001 – 2010, negligible compared to emissions during the fire season (May – September). To focus on the impact of large-scale circulation on baseline CO and O₃ in spring, the May data were excluded to avoid the effect of biomass burning. Springtime baseline O₃ at each site showed strong and consistent interannual variation up to 10 ppbv, and site average baseline O₃ mixing ratios exceeded the decadal average 46.5 ppbv in 2001, 2003, 2005, 2008, 2010, and was below average in 2002, 2004, 2006, 2007, 2009 (Fig. 8a).

Figure 8.

(a) Baseline O₃ and the NAO index averaged in March and April. The thick orange line indicates the baseline O₃ averaged over the seven sites and the thick dark blue line indicates the mean value 46.5 ppbv over 2001 – 2010. (b) Averaged daytime (18:00 – 24:00 UT) relative humidity at each site and daily maximum solar radiation flux at the Harvard Forest in March and April. (c) Averaged wind speed (> 2 m s⁻¹) from the west (247.5° – 337.5°) and the NAO index in March and April.

The difference of 850 hPa geopotential height between the lower and higher O₃ years is shown in Figure 9a. There was a pronounced difference up to 40 gpm in the Bermuda/Azores high and ~−40 gpm in the Icelandic low, resulting in stronger gradient flow between the two pressure systems indicative of the positive phase of NAO. Over 2001 – 2010, the NAO index was significantly positive in 2002, 2004, 2007, 2009 and negative in 2001, 2005, 2008, 2010, which corresponded mostly to the years of below and above the decadal average baseline O₃, respectively. Significant negative correlation was found between the NAO index and baseline O₃ at each site (Table 4). No significant trend was found in the NAO index over 2001 – 2010.

Figure 9.

The difference of (a) geopotential height (m) and streamlines at 850-hPa and (b) PV (10⁻⁹ m² s⁻¹ kg) at 350 K between the low O₃ years (2002, 2004, 2006, 2007, and 2009) and high O₃ years (2001, 2003, 2005, 2008, and 2010). The red star or box indicated the study region.

Table 4.

Correlation coefficient (r) and p-value between the pairs of variables in March and April over 2001 – 2010. Boldfaced numbers indicate p-value < 0.10.

Pair of Variables	CS	MWO	PM	TF	PSP	WFM	Harvard Forest
NAO index vs Baseline O3	−0.75 (0.03)	−0.68 (0.03)	−0.81 (0.03)	−0.81 (<0.01)	−0.58 (0.06)	−0.51 (0.10)	-
NAO index vs Baseline O3*	-	-	-	-	−0.49 (0.04)	−0.48 (0.04)	-
NAO index vs Baseline CO	−0.24 (0.30)	−0.51 (0.12)	−0.06 (0.46)	0.30 (0.22)	−0.14 (0.36)	−0.16 (0.34)	-
Relative humidity vs NAO index	0.85 (0.02)	-	-	0.63 (0.08)	0.23 (0.26)	0.40 (0.13)	-
Solar radiation flux vs NAO index	-	-	-	-	-	-	−0.64 (0.04)
Surface wind speed vs NAO index	0.68 (0.06)	0.76 (0.02)	-	0.57 (0.09)	0.44 (0.12)	-	-

^*The correlation between baseline O₃ and NAO index was computed at PSP over 1997 – 2010 and WFM over 1996 – 2010.

The negative correlation between NAO and baseline O₃ could be a result of multiple factors, such as reduced solar flux, enhanced continental export of O₃ produced in North America, and less frequent/weak stratospheric intrusion. Significant correlation was found between relative humidity and the NAO index at sites near the Northeast US coast (CS and TF), while the correlation was weaker at inland, elevated sites (PSP and WFM) (Table 4 and Fig. 8b). During a positive NAO year, the mean North Atlantic storm track parallels the eastern North American coastline before extending northeastward to near Iceland (Rogers, 1997). This storm track and its associated moisture transport and convergence could lead to relatively wet conditions near the eastern US coast and conversely dry conditions during a negative NAO year (Archambault et al., 2008; Hurrell, 1995). Significant negative correlation was found between NAO index and solar radiation flux at the Harvard Forest (Table 4), which was only ~45 km southwest from PM. No significant correlation was found in other seasons. This indicates more cloudiness, consequently reduced solar radiation flux near the surface and subsequently less O₃ production.

North American continental export facilitated by NAO could also affect baseline O₃ over the Northeast US. Annual wind speed from the west-northwest (247.5 – 337.5°, the prevailing winds at each site) was calculated at the study sites (Fig. 8c). Positive correlation was found between surface wind and NAO index at MWO and TF (Table 4). During a positive NAO phase, the anticyclonic circulation off the U.S. east coast and the cyclonic circulation across the North Atlantic were amplified with a northward shift (Rogers, 1997; Eckhardt et al, 2004) and hence stronger surface wind could be found near 50°N across the North Atlantic basin (Hess and Lamarque, 2007). The O₃ produced over the Northeast US was less likely accumulated in the region and was more likely transported faster off the continent and across the Atlantic Ocean. This speculation was consistent with the positive anomalies of O₃ observed over northwestern Europe (Christoudias et al., 2012).

Stratospheric intrusions and NAO could be linked, as the U.S. east coast trough induces descending air on its tailing side, which can cause tropopause folding with stratospheric air mixing downward into the troposphere. One of the physical characteristics of stratospheric air is high values of PV. The difference of PV between positive NAO years and negative NAO years exhibited negative anomalies ~ −0.6 × 10⁻⁹ m²s⁻¹kg over the Northeast US (Fig. 9b), suggesting that positive NAO was related to less stratospheric intrusion over the Northeast US. Negative correlation, although insignificant, was found in PV and NAO index over the Northeast US (r = −0.31, p= 20). This is consistent with lower baseline O₃ levels during positive NAO springs, which was further verified using the stratospheric O₃ dataset developed by Liu et al. (2013). Liu et al. (2013) showed that stratospheric O₃ was hardly detected in the lowest two layers (i.e., 0.5 and 1.5 km) in April. In March, ~40 – 60 ppbv of stratospheric O₃ reached the lowest layer 0.5 km in our study area in 2004 and 2006–2008 and reached the 1.5 km layer in 2001–2008. The stratospheric contribution to the 0.5 km layer in March 2008 was the largest of all Marchs of the decade, when NAO was negative (Fig. 8a)

Negative correlation, although insignificant, was also found between baseline CO and the NAO index at most of our study sites (Table 4). North American continental export could also impact the variation of baseline CO, while this impact could be confounded by other factors, e.g. stratospheric intrusion. Specifically, during positive NAO years, stronger continental outflow could lead to a decrease in baseline CO, while less stratospheric intrusion could lead to less dilution of surface CO and thus increase baseline CO levels. Further research is warranted to fully understand the relationship between baseline CO and NAO.

4.3. Combined effects of biomass burning and cyclone activity in summer

In summer 2009, the Northeast US was under the influence of frequent cold frontal passages associated with the largest number of cyclones passing through the region and was consequently exposed most frequently to air masses of Arctic origin. Moreover, summer 2009 saw the lowest fire emissions of CO (~11.9 Tg) of the decade from Russia and Canada (Fig. 6f). Hence, the lowest fire emissions of CO and the most frequent cyclone activities were likely two dominant factors leading to the decadal lowest summertime baseline CO and O₃ in 2009.

However, in summer 2003, AO was negative and 15 cyclones passed the region (Fig. 6a), greater than the decadal mean (12), and yet baseline CO and O₃ at the sites reached the decadal maxima (Figs. 3c, d and 6f). According to the analysis above, baseline CO and baseline O₃ were expected to be lower during this summer than the decadal average as a result of above-average passages of cyclones. However, the decadal maximum fire emissions of CO in Russia and Canada in summer 2003 (Section 4.1) counteracted the effect of the cyclone activity. Another interesting example was summer 2007 which had the decadal minimum of cyclone activity (Fig. 6a), and the total CO emissions from wildfires in Russia and Canada were 13.6 Tg, the second lowest of the decade following summer 2009 (Fig. 6f). As a result, the site-average baseline CO and O₃ levels in summer 2007 were below the decadal means. Therefore, the effect of biomass burning may dominate over that of AO and cyclone activity during some summers, while the two worked in concert during others.

4.4. Anthropogenic Emissions

4.4.1. Impacts on baseline CO

Global anthropogenic CO emissions showed a slight decrease of ~1% yr⁻¹ over 1990 – 2010 (Granier et al., 2011). Consistent with the global emission decline, a decreasing trend was found in northern hemispheric CO using a trajectory-mapped MOZAIC-IAGOS CO dataset (Osman et al., 2016). In the US, total anthropogenic CO emissions declined at a rate of −3% yr⁻¹ over 2000 – 2010, while increasing trends were found in India (~1.5% yr⁻¹) and China (~3% yr⁻¹) (Granier et al., 2011). It was likely that global CO emission decline with significant U.S. anthropogenic emission reductions could have contributed to the significant decreasing trends in baseline CO identified at our sea level to low elevation sites (i.e., 18 – 698 m). However, for the two highest sites MWO and WFM, it was hypothesized that their overall insignificant trends in baseline CO in spring and winter 2001 – 2010 (Table 2) resulted from the combined effect of decreasing US emissions and increasing Asian emissions.

Ten years of wintertime 36-hour backward trajectories (Fig. S1) suggested that nearly half of the time the air masses arriving at MWO and WFM originated from and passed through major source regions below their respective elevations, which placed the two sites under the influence of domestic anthropogenic emission reductions at least half of the time. On the other hand, the impact of continental to intercontinental transport, which has been suggested to be occurring most strongly in late winter to early spring in the middle to upper troposphere (~700 hPa – tropopause, Zhang et al. (2008)), could be just as important as, perhaps at times more important than, regional transport to MWO and WFM. This is because these are the highest sites situated close to the top of the daytime convective boundary layer, which are more likely impacted by free tropospheric air compared to low elevations sites. This was supported by the finding from Jonson et al. (2010)’s multi-model study of the long-range transport of O₃ and its precursors that impact of intercontinental transport from Asia was stronger in the free troposphere (~ 800 hPa) than what was calculated for U.S. surface sites by Fiore et al. (2009) and Reidmiller et al. (2009). Transpacific transport of Asian pollution was found to be particularly strong in spring and winter due to the combined effect of efficient ventilation of the Asian boundary layer via midlatitudinal cyclones and convection, long lifetime of CO, and/or strong springtime biomass burning emissions in southeastern Asia, which weakened in summer (Liang et al., 2004).

Trends in CO were also computed for Niwot Ridge, Colorado (3526 m) and Mauna Loa, Hawaii (3402 m), two long-term remote monitoring locations from NOAA’s Earth System Research Laboratory Global Monitoring Division (Section S3). Consistent with our study, no trends were found in winter and spring while significant decreasing trends in summer and fall at both sites. This further supports the hypothesis that in winter and spring high elevation sites in the Northeast U.S. were impacted by long-range transport of increasing Asian emissions.

4.4.2. Impacts on baseline O₃

Significant decreasing trends were found in year-round baseline CO, while no trends in year-round baseline O₃ at any site. This indicates that changes in baseline CO did not seem to affect year-round baseline O₃, and other factors, e.g., opposite trends in emissions of different ozone precursors and varying meteorological conditions, could have been more dominant. In examining baseline O₃ by season, it was found that TF was the only site with significant increasing trends in spring and winter at a rate of 2.4 ppbv yr⁻¹ and 2.7 ppbv yr⁻¹, respectively, which were not associated with trends in meteorological conditions such as NAO as discussed in Section 4.2.2. This points to the possible effect of anthropogenic emissions.

US Emissions of NO_x were reduced by 48% over 1990 – 2010, largely due to control of emissions from power plants and mobile sources (Xing et al., 2013). The Northeast U.S. Urban Corridor, extending from Washington D.C. in the south to Boston in the north, was dominated by mobile combustion emissions of NO_x. Annual mixing ratios of NO₂ in New York City decreased at a rate of −0.3 ppbv yr⁻¹ over 1980 – 2007 (Buckley and Mitchell, 2011). In winter and early spring with weakened photochemical production, decreased NO_x emissions in urban areas could cause less loss of O₃ via titration by NO (Liu et al., 1987; Jonson et al., 2006), resulting in enhanced O₃ mixing ratios in urban plumes (Wilson et al., 2012). From measurements at our sites, data points of O₃ corresponding to winds from the urban corridor were examined. It was found that the 10^th percentile mixing ratio of O₃ at TF in air masses from the urban corridor (157.5 – 202.5°) had been increasing at a rate of 1.81 ppbv yr⁻¹ (p = 0.05) in spring and 1.52 ppbv yr⁻¹(p <0.01) in winter (Figs. 10a and b). No trends were found at CS, MWO, PSP, and WFM under the influence of Northeast Urban Corridor air masses. Trends at PM were not computed, as wind direction data at this site were only available over 2007 – 2010. This strongly suggests that decreased NO_x emissions in the urban corridor likely had a significant impact on springtime and wintertime baseline O₃ at TF. We also investigated impacts of stratospheric intrusion, long-range transport, and NO_x emission reductions from power plants (Section S4). None of them turned out to be the dominate factor contributing to the increasing springtime and wintertime baseline O₃ at TF.

Figure 10.

Seasonal 10^th percentile mixing ratios of O₃ with wind from the directions aligned with the urban corridor in (a) winter and (b) spring. Specifically, the wind directions selected for AI: 157.5° - 202.5°; CS: 157.5° - 202.5°; MWO: 157.5° - 202.5°; PM: 112.5° - 157.5°; TF: 157.5° - 202.5°；WFM: 112.5° - 157.5°；PSP: 67.5° - 112.5°.

5. Summary

Our study suggested that over 2001 – 2010 baseline CO at most sites (MWO, PM, TF, PSP, and WFM) decreased significantly at a rate between −4.3 and −2.3 ppbv yr⁻¹ (p<0.01), which was associated with global anthropogenic CO emission reductions and decreasing biomass burning emissions in Russia. It was found that biomass burning impacted summertime baseline CO at all sites, except at PSP, and baseline O₃ only at high elevation sites. In summer, ~38% of baseline CO variability was attributed to CO emissions from forest fires in Russia and ~22% to emissions from forest fires in Canada. Baseline CO at MWO and WFM did not exhibit a significant trend in spring and winter, hypothesized to result from the combined effect of decreasing domestic emissions and increasing Asian emissions. Different from baseline CO, no trends were found in year-round baseline O₃ at all sites. The increasing trends in springtime and wintertime baseline O₃ at TF was most likely caused by reduced NO_x emissions over the US Northeast Urban Corridor. Summertime baseline CO and O₃ was also related to the AO. When the AO was in negative phase, more frequent cyclone activities brought cleaner Arctic air to midlatitudes lowering CO and O₃ at our sites. Significant negative correlation was found between baseline O₃ and the NAO index in spring, potentially due to NAO-related changes in solar flux, stratospheric intrusion, and continental export.

The U.S. EPA lowered the NAAQS for ground-level O₃ to 70 ppbv to improve protection of public health and welfare on 1 October 2015 (http://www3.epa.gov/ozonepollution). As the O₃ NAAQS are set closer to background levels, states will face ever increasing challenges with regard to fulfilling their obligation for NAAQS attainment. This study reinforced that, in addition to domestic emission controls, intercontinental transport of anthropogenic and wildfire emissions together with meteorological conditions should also be considered when designing encompassing, cost-effective emission control strategies that account for impacts of regional to global emissions of multiple pollutants (e.g., CO, CH₄, NO_x, and NMHCs). In addition, the relationships between baseline O₃/CO and various factors (e.g. NO_x emission controls, biomass burning emissions, NAO, and AO) examined in this study can also be used as reference point for evaluating global/regional air quality modeling systems that are used in air quality management applications. Different from most studies that focused on episodic influence of wildfires (e.g., DeBell et al., 2004; Dutkiewicz et al., 2011b; Liu et al., 2005), this was the first measurement-based study to examine the impact of wildfires on baseline CO and O₃ over time periods of ten years in the 2000s. Moreover, this study found that trace gases over the upwind areas could change in response to varying intensity of NAO. One limitation of this study is that it was based on ten-years of observations, and hence it was unlikely to predict the potential longer-term changes in natural emissions and AO/NAO signals as well as their impacts on baseline O₃. Also, while data analysis studies provide measurement-based estimates of contributions of various processes to baseline CO and O₃, they may be limited in clearly separating such contributions. Future research that integrates measurements and model simulations is warranted to further address the issues identified in this work on climatological time scales.