Free-troposphere ozone and carbon monoxide over the North Atlantic for 2001-2011

In situ measurements of carbon monoxide (CO) and ozone (O3) at the Pico Mountain Observatory (PMO) located in the Azores, Portugal, are analyzed together with results from an atmospheric chemical transport model (GEOSChem) and satellite remote sensing data (AIRS (Atmospheric Infrared Sounder) for CO, and TES (Tropospheric Emission Spectrometer) for O 3) to examine the evolution of freetroposphere CO and O 3 over the North Atlantic for 2001– 2011. GEOS-Chem captured the seasonal cycles for CO and O3 well but significantly underestimated the mixing ratios of CO, particularly in spring. Statistically significant (using a significance level of 0.05) decreasing trends were found for both CO and O3 based on harmonic regression analysis of the measurement data. The best estimates of the possible trends for CO and O 3 measurements are −0.31± 0.30 (2-σ ) ppbv yr−1 and −0.21± 0.11 (2-σ ) ppbv yr−1, respectively. Similar decreasing trends for both species were obtained with GEOS-Chem simulation results. The most important factor contributing to the decreases in CO and O 3 at PMO over the past decade is the decline in anthropogenic emissions from North America, which more than compensate for the impacts from increasing Asian emissions. It is likely that climate change in the past decade has also affected the intercontinental transport of O 3.


Introduction
Carbon monoxide (CO) and ozone (O 3 ) are important atmospheric pollutants in the troposphere. The major sources of atmospheric CO include incomplete combustion of fossil fuels and biomass burning (Duncan et al., 2007;Holloway et al., 2000;Khalil and Rasmussen, 1994;Seinfeld and Pandis, 2006). CO is also formed through the oxidation of volatile organic compounds (VOCs) by the hydroxyl radical (OH). O 3 is formed in the troposphere through photochemical reactions in the presence of O 3 precursors, which include the nitrogen oxides (NO x ≡ NO + NO 2 ), CO and VOCs. The dominant loss pathway for tropospheric O 3 is photolysis followed by reaction with water vapor.
Atmospheric transport across the North Atlantic Ocean carries air pollutants from the continent of North America towards Europe. The atmospheric lifetime of CO ranges from weeks to months (Duncan et al., 2007), and the lifetime for O 3 varies from days to months in the troposphere (Law, 2010;. These relatively long lifetimes enable these species to undergo intercontinental transport. Thus, changes in emissions from source regions (such as North America) can have significant implications for atmospheric composition in downwind regions such as the North Atlantic or Europe. Li et al. (2002) found that 20 % Published by Copernicus Publications on behalf of the European Geosciences Union. Increases in anthropogenic emissions of NO x , CO and VOCs have led to increases in free-tropospheric O 3 since pre-industrial times  although several studies have reported divergent trends of tropospheric O 3 over different regions of the Northern Hemisphere (Brunke et al., 1998;Cooper et al., 2010;Fusco and Logan, 2003;Guicherit and Roemer, 2000;Jaffe et al., 2003;Lee et al., 1998;Oltmans et al., 2006;Parrish et al., 2010Parrish et al., , 2013. Over the past decade, emissions of O 3 precursors have declined significantly in North America and Europe (Hudman et al., 2009), while those in Asia have increased (Vingarzan, 2004). Figure 1 shows the trends of anthropogenic emissions of CO and NO x in the United States (US) from 2000 to 2011. US sources account for approximately 80 % of anthropogenic emissions from North America . These changes in emissions are expected to influence the freetroposphere CO and O 3 concentrations over the North Atlantic. The Pico Mountain Observatory located in the central North Atlantic region (details provided in the next section) is a unique site that can be used to measure species relevant to long-range transport and thereby examine the direct continental outflow from North America. In this study, we combine 10-year in situ measurements of CO and O 3 at this station with results from a chemical transport model (the GEOS-Chem chemical transport model) as well as satellite data to examine the seasonal variation and long-term trends in free-tropospheric CO and O 3 over the North Atlantic for the past decade.

Measurements at the Pico Mountain Observatory
The Pico Mountain Observatory (PMO, formerly called PICO-NARE) is located on the summit of Pico Mountain on Pico Island in the Azores, Portugal (38 • 28 N, 28 • 24 W), at an altitude of 2225 m . This station is well above the Marine Boundary Layer (MBL) during summertime and ideal for examining the atmospheric composition of the lower free troposphere (FT). The island population is low (around 15 000) and is concentrated near sea level, which results in the site having negligible anthropogenic influence on the free-troposphere composition, although upslope flows have been found to occur, resulting in the transport of low-altitude air to the mountaintop (Kleissl et al., 2007). However, the occurrence of such flow is not very frequent and even when it occurs it was found that the air did not originate from the surface (Kleissl et al., 2007).
The station is also frequently impacted by export of North American pollution during summertime and outflow from Arctic and sub-Arctic regions, resulting in transport of biomass burning emissions from Canada, Alaska and Siberia Val Martin et al., 2008b). Val  reported that North American boreal wildfires contributed significantly to enhancements in CO and O 3 background mixing ratios during the summer of 2004.  also observed frequent enhancements in CO levels above the marine background levels during the summertime in 2001-2003 which they attributed to North American pollution outflow or long-range transport of biomass burning emissions. High levels of both CO and O 3 were observed during the periods of biomass burning.
CO and O 3 have been measured at PMO using instruments described by Honrath et al. (2004). We use hourly CO data covering 2001-2010 and O 3 data spanning 2001-2011 with most of the available data spanning May-September. Considering the diurnal variations of CO and O 3 , we have applied a 24 h filter to avoid unexpected bias -i.e., only days with full 24 h data availability were used to calculate the daily average mixing ratios of CO and O 3 . The application of this filter resulted in the inclusion of approximately 76 % (87 %) of the total CO (O 3 ) measurements in the final analysis.

AIRS/AQUA data for CO
The Atmospheric Infrared Sounder (AIRS) was launched onboard the AQUA satellite on 4 May 2002. It is a polarorbiting nadir-viewing thermal IR sounder with cloud clearing capability and retrieves CO at 4.7 microns with 70 % daily global coverage -100 % between 45 • and 80 • longitude (McMillan et al., 2005;Yurganov et al., 2008). We use AIRS level 3 version 5 monthly data obtained from Giovanni, Goddard Earth Sciences Data and Information Services Center (GES DISC). Data are available at http://disc. sci.gsfc.nasa.gov/giovanni#instances (Acker and Leptoukh, 2007). Level 3 datasets have undergone rigorous processing and are available with no subsequent processing required. As done in prior studies, AIRS measurements with more than 0.5 degrees of freedom  were included in the analysis. Previous comparisons of AIRS with in situ measurements reveal a positive bias of ∼10 % in the Northern Hemisphere Kopacz et al., 2010). Retrievals for the area matching the same horizontal grid as GEOS-Chem over PMO (32.5 • to 27.5 • W and 36 • to 40 • N) centered at 802 hPa (roughly 2.2 km, the same elevation as PMO) are used in this study. Since both the measurement data at PMO and GEOS-Chem use both day and night values, day and night retrievals are used for AIRS as well.

TES/AURA data for O 3
The Tropospheric Emission Spectrometer (TES) was launched in July 2004 aboard the Earth Observing System (EOS) Aura satellite. Data were obtained from Giovanni online data system, developed and maintained by the NASA GES DISC as mentioned above (Sect. 2.2.1). TES is nadir viewing in a polar-orbiting Sun-synchronous orbit on the same track as AIRS/AQUA with a local crossing time of 01:45 and 13:45 . We use available data for the period covering 2005-2010 in this study. The data for 2008-2009 were not available. The level three data used were pre-processed and all negative values were filtered out. Previous comparisons with in situ measurements show that TES has a positive bias of 5.3 ppbv for O 3 at 500 hPa .

Model description and simulations
GEOS-Chem is a global three-dimensional model of tropospheric chemistry driven by assimilated meteorological observations from the Goddard Earth Observing System (GEOS) of the NASA Global Modeling Assimilation Office (Bey et al., 2001;Evans and Jacob, 2005;Fairlie et al., 2007;Huang et al., 2013;Johnson et al., 2010;Martin et al., 2002;Park et al., 2004;Wu et al., 2007;Yu et al., 2010). The model has fully coupled O 3 -NO x -VOC aerosol chemistry and has been extensively evaluated and applied to a wide range of research topics related to atmospheric chemistry and air quality (e.g., Duncan et al., 2007;Evans and Jacob, 2005;González Abad et al., 2011;Hudman et al., 2007Hudman et al., , 2008Martin et al., 2002;Park et al., 2004;Wu et al., 2008aWu et al., , b, 2012. There have been different versions of the GEOS meteorology, including GEOS-3, GEOS-4 and GEOS-5, with each spanning different periods. Emission inventories in GEOS-Chem cover various sources, including anthropogenic, biomass burning, biofuel and biogenic emissions. Anthropogenic emissions fol-low the Emissions Database for Global Atmospheric Research (EDGAR) global inventory (Olivier and Berdowski, 2001) and are updated with regional inventories, including the US EPA National Emissions Inventory (EPA/NEI05 and EPA/NEI99), the Environment Canada National Pollutant Release (CAC) Inventory, the European Monitoring and Evaluation Programme (EMEP) inventory, the Streets emission inventory in Asia (Streets et al., 2003(Streets et al., , 2006Zhang et al., 2009) and the Big Bend Regional Aerosol and Visibility Observational (BRAVO) study emissions inventory for Mexico and some neighboring US states. The biomass burning emissions follow either the Global Fire Emissions Database (GFED) v2 (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) (van der Werf et al., 2006) or GFEDv3 (1997GFEDv3 ( -2010 monthly inventories (van der Werf et al., 2010) and biogenic VOC emissions are taken from the Model of Emissions of Gases and Aerosols from Nature (MEGAN) global inventory. Emissions from other natural sources (e.g., lightning, volcanoes) are also included. For chemistry, either the Kinetic PreProcessor (KPP) (Eller et al., 2009;Sandu and Sander, 2006) or the Sparse-Matrix Vectorized Gear 2 (SMVGEAR 2) (Eller et al., 2009) chemistry solver can be used. The Linearized Ozone (LINOZ) mechanism (McLinden et al., 2000) is used for the stratospheretroposphere exchange of O 3 .
In addition to the standard full chemistry simulations with GEOS-Chem, we also carry out special tagged simulations (tagged CO and tagged O 3 simulations) to better identify the contributions to CO and O 3 at PMO from various source regions. A brief description of the simulations used in this study and the archived data is provided in the following sections.

Full chemistry simulations used
The full chemistry simulations using the standard version of GEOS-Chem with normal emissions are referred to here as FCNE simulations. We used the 4 • × 5 • (latitude × longitude) horizontal resolution and a one-year spinup for all the FCNE simulations. GEOS-4 meteorological fields were used for 2001-2004, whereas those from GEOS-5 covered the period from 2005 onwards. Regional emission inventories for anthropogenic (fossil fuel and biofuel) emissions were used over Canada (CAC inventory), Mexico (BRAVO), Asia (Streets), USA (NEI2005 and EPA/NEI99) and Europe (EMEP). The other regions were covered by the EDGAR global anthropogenic emissions inventory. Biogenic VOCs in the model include emissions of isoprene, methyl butenol and seven monoterpene compounds following the MEGAN scheme (Guenther et al., 2006). For biomass burning emissions, the GFEDv2 monthly inventory was used (as the updated GFEDv3 inventory was not available with the model version (v8-03-01) used). Model-simulated CO and O 3 mixing ratios for the grid box covering PMO were archived with a 4 h temporal resolution. These 4 h instantaneous mixing ratios were used to compute the daily averages for both the species. Table 1 summarizes the anthropogenic and biomass burning CO emissions in GEOS-Chem for various regions of the Northern Hemisphere for 2001-2009. The emission inventories indicate significant declines in anthropogenic CO for North America and Europe but increases from Asia during this time period.
A sensitivity run was also performed to assess the impact of North American fossil fuel combustion on CO and O 3 at PMO by excluding those emissions. This run had no fossil fuel emissions for the region covering the US, Canada and Mexico (15 • N to 88 • N (latitude) and 50 • W to 165 • W (longitude)).

Tagged CO simulation
The GEOS-Chem tagged CO simulation was used to determine the contribution to CO mixing ratios at PMO from various geographical regions (e.g., Asia, North America) and sources (e.g., biomass burning, biofuels). It is one of the several offline simulations included in GEOS-Chem and consists of source/region-specific CO tracers. The CO sources accounted for include fossil fuels, biofuels, biomass burning, anthropogenic and biogenic VOCs. A detailed description can be found in Duncan et al. (2007). The reaction of CO with OH is the only sink considered. We used v9-01-02 of the model with the 4 • × 5 • (latitude × longitude) horizontal resolution. For this simulation, emission inventories included to account for anthropogenic CO production were the global EDGAR inventory, CAC inventory over Canada, BRAVO over Mexico, EPA/NEI99 over North America, Streets over southeastern Asia and the RETRO (for global anthropogenic VOCs) inventory. For biomass burning emissions, the monthly GFEDv3 emissions (available with the version used) were used. The OH concentrations used were those archived from a separate full chemistry simulation. Restart files for the tagged CO simulation were generated by an eleven-month (

Tagged O 3 simulation
The tagged O 3 simulation was used to attribute the O 3 mixing ratios at PMO to O 3 production from various regions. It uses chemical production/loss rates of O 3 archived from the FCNE simulation. We used version 09-01-02 of GEOS-Chem (horizontal resolution 4 • × 5 • (latitude × longitude)) for this simulation, and the same version was also used to archive the O 3 production/loss rates from 2000 to 2011.
Restart files were generated in the same manner as for the tagged CO simulation. The time series of 4-hourly instantaneous values of all the tracers was archived from 2001 to 2011 for the location and altitude corresponding to PMO. The daily averages for all tracers were computed using these 4-hourly instantaneous values.

GEOS-Chem model evaluation
GEOS-Chem output for CO and O 3 mixing ratios was compared with the in situ measurement data for the two species at PMO as well as satellite retrievals from AIRS/AQUA (for CO) and TES/AURA (for O 3 ). Figure 2 shows the time series of 24 h average O 3 and CO mixing ratios at PMO from September 2000 to August 2010 along with the 24 h averages computed from GEOS-Chem time series output for this period. For CO, GEOS-Chem was found to underestimate the observations, with the magnitude during summer being 24.7 ± 5.2 ppbv (1-σ ). The significant underestimate of CO by GEOS-Chem possibly reflects some low biases in the CO emission inventories used in the model. The model underestimates of CO have also been reported by previous studies (Bey et al., 2001;Duncan and Logan, 2008;Duncan et al., 2007;Fiore et al., 2002;Heald et al., 2003;Kopacz et al., 2010;Val Martin et al., 2008a;Zhang et al., 2008). For O 3 , the GEOS-Chem performance was much better. The model was 6.7 ± 2.6 ppbv higher and showed the least agreement in winter and best agreement in summer and fall. The seasonal variation of CO and O 3 at PMO was further examined in Fig. 3 with box plots for GEOS-Chem model simulations (24 h averages) compared with ground-based in situ measurements and satellite data. The data used in Fig. 3 cover the period of September 2004 to August 2005 as this period has the best data availability for the measurements. Both species show maxima in the springtime and minima in the summertime with consistency between the model results and measurement data. Similar seasonal variations have also been reported for these species at other remote sites such as Mace Head (Derwent et al., 1998) and mountaintop stations such as Mt. Cimone (Bonasoni et al., 2000) and for CO in a previous study at the PMO (Val Martin et al., 2008a). The AIRS data further confirm the model underestimate of CO which is particularly large during late spring-early summer. We also compared model results with observational data on the interannual variability of O 3 at PMO (Fig. 4). The period of May-August was chosen for best data coverage, and average O 3 mixing ratio for this period was calculated for each year. Some years without sufficient observational data coverage were excluded from the comparison. We find that besides the period of 2004-2006, the interannual variability of O 3 at PMO reflected from the GEOS-Chem model results generally agrees well with the observational data (Fig. 4).

Trend analysis methodology
Two-sided parametric hypothesis testing (null hypothesis of no trend and alternate hypothesis of non-zero trend) was used to analyze the trends in CO and O 3 mixing ratios at PMO for 2001-2011. We employed a multiple regression model that is similar to the additive model of a time series and con-tains sinusoidal functions to represent the characteristic seasonal variations of both CO and O 3 . Following Helsel and Hirsch (2002), we express the model for daily average values of CO or O 3 as C t = a 0 + a 1 t + a 2 sin 2π t 365 + a 3 cos 2π t 365 where C t is the daily average mixing ratio of the species at a time t (days since a reference time); a 0 , a 1 , a 2 , and a 3 are regression coefficients; and ε t is residual from the model. The sinusoidal functions use a time period of one year to reflect the seasonal variations of CO and O 3 . The linear term (a 1 t) with time was used to represent the long-term trend in CO or O 3 .
For all datasets consisting of daily averages, a five-day centered simple moving average scheme was applied to the data before fitting the regression model. Coefficients in the regression equation were determined by the least squares method, and the statistical significance of the t statistic corresponding to the slope term was examined by comparing the probability of observing a t statistic of same or greater magnitude with the null hypothesis being true (p value) with the significance level α (0.05 used for this study).

Trends of CO and O 3 at PMO
The regression model in Eq. (1) was applied to the time series of CO and O 3 (for both the in situ measurements and GEOS-Chem FCNE simulation output) at PMO. Table 2 lists the trends for both the measurement data and GEOS-Chem output along with the associated p values. Figure 5

Fig. 3. Seasonal variation of CO (top) and O 3 (bottom) at PMO (September 2004-August 2005) as indicated by GEOS-Chem model simulations (yellow), in situ measurements (dark brown) and satellite observations (AIRS for CO and TES for O 3 ) (light blue).
The thick (thin) bars represent the 67 % (90 %) confidence intervals. The mean and median are represented with red and green dots, respectively.The statistics are based on daily (24 h) averages. shows the regression fits for the CO and O 3 measurements, and Figure 6 contains the fits for the modeled (GEOS-Chem FCNE simulation) CO and O 3 . The observations show statistically significant decreasing trends for both the species over the study period, with CO showing a decrease of GEOS-Chem also shows statistically significant decreasing trends for both CO (−0.34 ± 0.08 (2-σ ) ppbv yr −1 , p value < 0.001) and O 3 (−0.53 ± 0.04 (2-σ ) ppbv yr −1 , p value < 0.001). These decreasing trends are in agreement with those obtained from observation data at other stations in the North Atlantic such as Tudor Hill, Bermuda (TH), and Mace Head, Ireland (MH) (data available at http://ds.data.jma. go.jp/gmd/wdcgg/cgi-bin/wdcgg/catalogue.cgi). For TH, the harmonic regression analysis of O 3 mixing ratios from 2003 to 2011 yields a statistically significant decreasing trend (p value < 0.001) of 1.1 ± 0.11 (2-σ ) ppbv yr −1 , whereas CO at MH (2001-2011) shows a statistically significant decreasing trend (−1.4 ± 0.20 (2-σ ) ppbv yr −1 , p value < 0.001) as well. However, measurements at Izana, Tenerife (IT) (data available at http://ds.data.jma.go.jp/ gmd/wdcgg/cgi-bin/wdcgg/catalogue.cgi), show an increasing trend of 0.10 ± 0.07 (2-σ ) ppbv yr −1 for O 3 over the study period. This opposite trend at IT could indicate the fact that it is further south and less affected by the continental outflow from North America. Previous studies such as Lelieveld et al. (2004) and Derwent et al. (2007) have reported upward trends in O 3 over the North Atlantic for the periods 1977-2002 and 1987-2007, respectively. The discrepancy appears One of the major factors that could lead to a decrease in CO and O 3 mixing ratios over the North Atlantic is the decline in anthropogenic emissions from North America (of which 80 % of emissions are from the US) in the past decade. The trends for US anthropogenic emissions of CO and NO x are shown in Fig. 1. In order to investigate this hypothesis, we first carried out sensitivity runs by turning off North American fossil fuel emissions in the model (Sect. 3.1) to derive the contribution of North American anthropogenic emissions to CO and O 3 at PMO. Our results show that the anthropogenic emissions from North America enhance the annual means of CO and O 3 at PMO by 8.6 and 4.0 ppbv yr −1 , respectively (Table 3). Also, Figure 1 shows that the US CO emissions have declined by about 1/3, from around 105 Tg in 2000 to 70 Tg in 2011. Assuming a linear correspondence between the US emissions and their contribution to CO mixing ratios at PMO, this would amount to a decrease of approximately 2.9 ppbv in the CO at the station (over the past decade). This change is similar to that obtained from the harmonic regression analysis of CO measurements at PMO, suggesting that the observed possible decline in the CO at the station cor- relates particularly well with the US emission declines. For CO, the largest contribution from North American emissions was found in the winter and spring, which reflects stronger atmospheric transport and longer chemical lifetimes of CO (due to low levels of OH) during those seasons, as previously noted by Val Martin et al. (2008a). In contrast, for O 3 , the least North American contribution was found during winter, reflecting the NO x -saturated regime where O 3 production is more sensitive to the photochemically generated hydrogen oxide radicals (HO x ) (Jacob et al., 1995;Kleinman et al., 1995). In order to better understand and quantify the effects associated with the decreases in North American emissions, we also carried out tagged CO and O 3 simulations with the GEOS-Chem model. Results of the trend analysis for the tagged CO simulation are summarized in Table 4. The contribution of CO produced from anthropogenic fossil fuel emissions in the US to CO at PMO shows a statistically significant (p value < 0.001) decrease over the study period. The contribution of biomass burning emissions in Europe also shows a statistically significant (p value < 0.001) decrease. On the other hand, contribution of fossil fuel emissions in Asia shows a statistically significant (p value < 0.001) increase. These results indicate that reductions in CO from North America and Europe could be the primary reasons for the possible decrease in CO at PMO, which outweigh the effects of increasing CO from Asia in the past decade.  Table 5 summarizes the results of the statistical analysis for the tagged O 3 simulation. Statistically significant (p value < 0.001) decreasing trends are obtained for the contributions to O 3 at PMO from O 3 production in the boundary layer over various regions, including North America, Europe and Asia. The contribution of O 3 production in the middle troposphere (MT) (boundary layer top -350 hPa) also shows a decreasing trend (p value < 0.001) over the study period. On the other hand, an increasing trend (p value < 0.001) was identified for the contribution from O 3 produced in the upper troposphere (UT) (350 hPa -tropopause).
The decreasing trends for contributions to O 3 at PMO from North America and Europe are consistent with the declines in anthropogenic O 3 precursor emissions in these regions. On the other hand, the decrease in the contribution of O 3 produced in Asia (where anthropogenic precursor emissions have been increasing) may reflect the impacts from climate change in the past decade. Except for regions with strong emissions of O 3 precursors, climate change generally decreases surface O 3 due to enhanced O 3 destruction associated with higher water vapor concentration in the atmosphere (e.g., Johnson et al., 1999;Wu et al., 2008b). With reduced chemical lifetime of O 3 , this effect is particularly important for remote regions and the long-range transport of atmospheric O 3 . Analysis of the atmospheric water vapor concentrations in the past decade shows significant increases (up to 20 %) over the North Atlantic for the study period which contributes to a shorter O 3 lifetime during transport. The increase in the contribution of O 3 produced in the UT could be due to an increase in the NO x production from lightning. Harmonic regression analysis of global lightning flashes archived from GEOS-Chem for 2001-2011 shows a statistically significant increase over this period, which points to an increase in the NO x produced. The uppertroposphere O 3 is highly sensitive to NO x emissions from lightning, with the O 3 production efficiency (referenced to NO x emissions) being about six times higher than that for NO x emissions in surface air .
The statistically significant decreasing trend for the contribution of O 3 production in the North Atlantic largely reflects the decline in the anthropogenic emissions from North America. When comparing the contributions from various regions (North America, Europe and Asia) in Table 5, we can see that the reductions in anthropogenic emissions of O 3 precursors are the most important factor contributing to the possible decrease in O 3 mixing ratios observed at PMO.

Conclusions and discussion
This study analyzes in situ measurements of CO and O 3 at the Pico mountaintop observatory and output from GEOS-Chem model simulations to determine the possible trends in CO and O 3 over the North Atlantic region for the past decade. The GEOS-Chem model performance is also evaluated by comparing simulation results against measurements at PMO and satellite observations. The model reproduces the seasonal variations of CO and O 3 at PMO reasonably well although it significantly underestimates CO at this remote site.
Sensitivity studies show that North American fossil fuel emissions contribute to CO and O 3 at the Pico site by 8.6 and 4.0 ppbv, respectively. This contribution to the CO mixing ratios is greatest during winter and spring, reflecting the longer CO lifetime and stronger transport during those seasons. In contrast, the contribution to O 3 is found to be smallest during winter, reflecting the low O 3 production and low sensitivity to NO x emissions during wintertime.
Harmonic regression analysis of the observed CO and O 3 mixing ratios at PMO for 2001-2011 shows statistically significant decreases for both species. The best estimates of the possible trends for the CO and O 3 measurements are −0.31 ± 0.30 (2-σ ) ppbv yr −1 and −0.21 ± 0.11 (2-σ ) ppbv yr −1 , respectively. GEOS-Chem model simulations for the past decade also yield decreasing trends for CO and O 3 at PMO. We conducted tagged simulations for CO and O 3 , respectively, with GEOS-Chem to attribute these decreases. Contribution to CO at PMO from fossil fuel emissions in the United States shows a statistically significant decreasing trend (−0.27 ± 0.05 (2-σ ) ppbv yr −1 ), while the contribution from fossil fuel emissions in Asia shows a statistically significant increasing trend (+0.35 ± 0.03 (2σ ) ppbv yr −1 ). In addition, biomass burning emissions in Europe show statistically significant decreasing contribution (−0.3 ± 0.03 (2-σ ) ppbv yr −1 ). These results indicate that, even though the increasing CO emissions from Asia in the past decade have led to increasing enhancement in CO over the North Atlantic, the decreases in CO from North America and Europe more than compensate for these increases and result in an overall decreasing trend in CO at PMO. The changes in O 3 at PMO are attributed to O 3 production in various regions (e.g., North America, North Atlantic, Europe and Asia) through tagged O 3 simulations with the GEOS-Chem model. Contributions from O 3 produced in the lower and middle troposphere also show statistically significant decreasing trends. The decreases in the contribution of O 3 from North America and Europe as well as the North Atlantic are consistent with the decreases in anthropogenic emissions of O 3 precursors in North America and Europe in the past decade. While the anthropogenic emissions in Asia have increased in the past decade, the decreasing trend for contribution to PMO O 3 from Asia possibly reflects the impacts of climate change which can decrease the atmospheric lifetime of O 3 and therefore suppress its long-range transport. Further study is required to examine the potential impacts of climate change on intercontinental transport. Overall, the emission reduction of O 3 precursors in North America in the past decade is the most important factor contributing to the possible decreasing trends of CO and O 3 at PMO.