Trends in Passively-Measured Ozone, Nitrogen Dioxide and Sulfur Dioxide Concentrations in the Athabasca Oil Sands Region of Alberta, Canada

The Athabasca Oil Sands Region (AOSR) in northeastern Alberta, Canada has attracted much international attention in recent years due to the increased level of oil sands operations. A passive sampling program was initiated in 1999 to monitor ozone (O3), nitrogen dioxide (NO2) and sulfur dioxide (SO2) in the AOSR for the estimation of the exposure of the forest monitoring sites and the characterization of temporal trends. Since 1999, highest concentrations of O3 and NO2 occurred in April and winter, respectively. The observed spring O3 maximum is common in the northern hemisphere. The higher winter-time NO2 concentrations were due to low atmospheric mixing height, stable atmosphere, and higher emissions during winter. Sen-Theil trend analysis, a non-parametric analysis for temporal trending, determined that O3 concentrations from 2000 to 2009 did not change. NO2 concentrations increased slightly at three sites, and significantly increased at two sites closer to stationary and mobile sources. SO2 concentration was increasing at JP107 and was decreasing at JP101. SO2 concentrations did not increase at 4 other sites close to the major emissions. This suggests that SO2 emissions were likely stable. Spatial analysis was conducted to characterize the concentration distribution in the region. The O3 concentrations were low near the emission sources (9.4 km) likely due to local O3 titration. Highest NO2 and SO2 concentrations were measured near the main source area. Generally, passively measured monthly average concentrations of O3, NO2 and SO2 stabilized at 20, 48 and 48 km from the main source area suggesting NO2 and SO2 emission influences were limited to < 50 km away from the major sources. However, one site (JP107) located near the Athabasca River Valley, 94 km north of the main source area, had higher SO2 and NO2 concentrations. This could be attributed to influence of valley flow, and/or to additional sources added in the region since 2007.


INTRODUCTION
Oil sands are a mixture of clay, sand, water and bitumen.Oil sands deposits occur in the Athabasca, Peace River and Cold Lake regions of Alberta, Canada.Canada's Oil Sands deposits cover some 140,200 km 2 reported by Alberta Energy (2011).The AOSR in northeastern Alberta is the largest deposit, and it has attracted global attention due to environmental concerns over the rapid pace of industrial expansion related to bitumen extraction.The AOSR resource is estimated to contain some 170 billion barrels of recoverable oil.As of January 2011, Canada is third to Venezuela and Saudi Arabia in terms of global oil reserves (CAPP, 2011).In the AOSR, approximately 20% of the resource is situated close enough to the surface to be mined with shovels and trucks.The remainder of the bitumen lies deep below the surface and can only be recovered at this time by in-situ extraction methods such as Steam Assisted Gravity Drainage (SAGD).Oil production in 2010 was 1.5 million b/d (barrels of oil per day) with production expected to reach 3.5 million b/d by 2025 (CAPP, 2010).
In the AOSR there are varied emission sources including natural (e.g., forest fires and vegetation), oil sands operations (i.e., fixed, mobile, fugitive), and urban/transportation. Nitrogen dioxide (NO 2 ) and sulfur dioxide (SO 2 ) are the major criteria air contaminants emitted from the oil sands operation processes.According to the Canadian National Pollutant Release Inventory (NPRI), 2009 emissions from main stacks were 1.6, 79 and 307 tones/d for NH 3 , NO x (NO and NO 2 , expressed as NO 2 ) and SO 2 , respectively (http://www.ec.gc.ca/inrp-npri/).Facility reporting indicates stacks (flue-gas desulfurization) as the major SO 2 sources with NO x emission sources being split between stacks and area sources (i.e., mine-fleets) at ratios dependent upon process used.
The Wood Buffalo Environmental Association (WBEA; www.wbea.org) is a multistakeholder, not-for-profit organization created in 1997 and has been responsible for monitoring air quality in the AOSR since its inception.Continuous, time-integrated, and passive methods are used for assessing ambient air quality.Time-integrated methods include filter-based 24-hour PM 2.5 and PM 10 measurements for mass, ionic and metal species.The continuous monitors (84 analyzers at 15 stations) provide real time data but require ease of access and power.Outside of the Athabasca River Valley in the boreal forest, WBEA has used passive sampling techniques since 1999 to measure ambient concentrations of selected air pollutants.The advantages of passive samplers are low cost, no power requirement, and ease of maintenance.The disadvantages are low sensitivity, inability to resolve peak concentrations, possible interference from other species and significant influence of wind speed, temperature and relative humidity on passive sampler performance (Krupa and Legge, 2000).Passive samplers have been successfully used for monitoring large-scale, long term trends of air pollutants in remote areas (Krupa and Legge, 2000;Cox, 2003;Seethapathy et al., 2008;Bytnerowicz et al., 2010).The principle of passive samplers is that gas molecules are collected by diffusion using a collection medium coated with a chemical having specific affinity to the substance of interest.The sampling flow-rate of passive samplers is mainly controlled by pore size of the diffusion barrier, relative humidity, wind speed, and temperature.Many passive samplers have been developed for the pollutants of interest, including inorganic and organic compounds (Krupa and Legge, 2000;Cox, 2003;Namiesnik et al., 2005;Kot-Wasik et al., 2007;Levy et al., 2007;Seethapathy et al., 2008;Campos et al., 2010).The data collected by passive samplers can be used to estimate spatial distribution and temporal trends of the pollutants being measured to link with effects on forest health indicators and to estimate the deposition of atmospheric species of interest (Krupa et al., 2001;Ray, 2001;Cape et al., 2004;Delgado-Saborit and Esteve-Cano, 2008;Sicard et al., 2011).
To characterize SO 2 , NO 2 and O 3 concentrations for further analyses of sulfur/nitrogen deposition and photochemistry in the AOSR, passive sampling methodology has been applied to monitor these three species at remote forested sampling locations up to 150 km from oil sands processing facilities (upgrading facilities).The objectives of this study were (1) to characterize seasonal variation in concentrations and (2) to describe the temporal trend of ambient NO 2 , SO 2 , and O 3 concentrations in the AOSR.
The coordinates of a reference point (AMS 2), three community stations, and 10 remote sites with the corresponding distances to the reference point are listed in Table 1.AMS 2 is designated as industrial site and located in the middle of the industrial activities.AMS 1, AMS 6 and AMS 8 are located in residential areas of Fort McKay, Fort McMurray and Fort Chipewyan, respectively (Fig. 1).The populations of Fort McKay,Fort McMurray and Fort Chipewyan were 562,61,374,and 847 persons (respectively) in Canadian 2011 Census (http://www.statcan.gc.ca).AMS 1 is in close proximity to oil sands operations, and AMS 6 and AMS 8 are 34 km south and 186 km north from the reference point, respectively.Data from four of the 15 WBEA air monitoring stations (AMS 1, AMS 2, AMS 6, and AMS 8) with continuous monitors were used in this study.
AMS 2 was set as a reference point to characterize the relationship between concentrations of three species and distance since AMS 2 is in the geographic center of the two largest stationary emission sources.Passive samplers were collocated with continuous monitors at two community stations, AMS 1 and AMS 6, to evaluate the accuracy and precision of passive samplers for O 3 , NO 2 and SO 2 .AMS 8, a community station, was set as a background station for the O 3 concentration comparison from the continuous monitor.

Sampling and Analysis
The sampling period was monthly from April to September, and bimonthly from October to March.Due to lower concentrations, and due to access challenges, a bimonthly sampling schedule was adopted during the winter.This schedule resulted in nine sets of samples being collected each year at remote sites.Bimonthly concentrations were applied to each month directly.Duplicate samples were acquired at each remote site.The passive samplers were situated 2 m above the forest canopy on 18 m towers.To prevent the gravitational deposition of aerosols, the passive samplers were placed facedown under a shelter.
The all-season SO 2 passive sampling system (SPSS) (Tang et al., 1997), all-season NO 2 passive sampling system (NPSS) (Tang et al., 1999) and O 3 passive sampling system (OPSS) (Tang and Lau, 2000) were used to measure SO 2 , NO 2 and O 3 .As shown in Fig. 2, the OPSS comprises two parts: a Teflon film as a diffusion barrier and a filter coated with sodium nitrite as an O 3 collection medium, separated by an air gap as the diffusion area.NPSS and SPSS were similar to the OPSS, but used Teflon film as the diffusion barrier, and CHEMIX TM as the collection medium for NO 2 , and a sodium carbonate coated filter for SO 2 .SO 2 , O 3 and NO 2 samples were extracted and analyzed by ion chromatography (DX-120, Dionex Corp., US) for sulfate, nitrate and nitrite concentrations.Following collection, the passive sampler was stored in a sealed container delivered to a commercial laboratory and analyzed within 30 days.(Tang et al., 2000).
A SO 2 monitor (43A, Thermo Electron Instruments Inc), NO 2 monitor (42C, Thermo Environmental Instruments Inc), and O 3 monitor (49C, Thermo Environmental Instruments Inc) were employed to monitor real time SO 2 , NO 2 and O 3 concentrations at AMS 1 and AMS 6 for the evaluation of the passive samplers.The passive samplers at these stations were installed 1 m above the station roof, near the inlet for the gas analyzers.Triplicate passive samples of SO 2 , NO 2 and O 3 were collected monthly at the stations.Twelve sets of samples were collected each year from 2003 to 2008.

Data Analysis
Two models were used in the O 3 , NO 2 and SO 2 trending analyses: (1) simple linear regression (parametric method) for a preliminary trending analysis; and (2) Sen-Theil trend analysis (non-parametric method) for an advanced trending analysis.

Simple Linear Regression
The simple linear regression, as shown in Eqs.(4-6) (Weisberg, 2005), was used in the preliminary Exploratory Data Analysis (EDA).
where Y t is the value of the measured variable Y at time t, α and β are the regression intercept and slope respectively (ε represents the error term).Their estimators are computed as follows (bars represent corresponding means): n is the number of observations (n = 120).

Sen-Theil Trend Analysis
Sen-Theil trend analysis of time series was applied for advanced trending analysis.Sen-Theil trend analysis of time series does not assume any parametric properties, which are required for the standard simple linear regression trend analysis.It is based on a nonparametric concept calculating slopes between any pair of observed data.Considering a time series of observations x 1 , x 2 , x 3 , ..., x n (n = 120 in this study), the Sen-Theil algorithm computes all slopes of the form where x i is the pollutant concentration at time i = 1, 2, 3, ..., 120, i = 1 representing January 2000, and i = 120 representing December 2009.There are altogether  such pairs.The corresponding slopes are ordered according to their magnitude as follows: In order to get the lower limit S low and the upper limit S high of the corresponding (1 -α) × 100% confidence interval for Sen slope, we calculate where z α/2 is the corresponding critical point of the of the Gaussian distribution.Then we set (M 1 ) and S high = S (M 2 +1) (12)

Monthly Variation in Ambient Concentrations O 3 Concentration
Monthly O 3 concentrations at six remote sites are illustrated in Fig. 3.All sites clearly showed an annual cycle.The highest O 3 concentrations occurred in April, a spring maximum, with a small increase visible in November.The spring O 3 maximum is a common phenomenon at the northern hemisphere.Several hypotheses for the spring O 3 maximum are: (1) stratospheric intrusion (Levy et al., 1985;Logan, 1985;Dickerson et al., 1995;Moody et al., 1995; Oltmans et al., 2004); (2) wintertime accumulation of ozone at high latitudes (Liu et al., 1987;Honrath et al., 1996); and (3) phase overlap between stratospheric injection and tropospheric ozone production (Wang et al., 1998;Yienger et al., 1999).However, it should be noted that maximum 1-hr O 3 concentration was observed by the O 3 continuous monitor in summer, likely due to photochemical reaction.
Interestingly, the small increase observed in November has apparently become less pronounced in recent years.Angle and Sandhu (1986)

NO 2 Concentration
We measured higher NO 2 concentrations in the winter months (Fig. 4) due to a stable atmosphere, lower mixing height, slower photochemical reactions, and higher emissions.The examples resulting in higher NO x emissions in the winter include: (1) more NO x emit to atmosphere from both on-road and non-road vehicles at both idling (USEPA, 1998) and cold-start conditions; and (2) residential heating in the winter.At 57 degrees N latitude, day length is very short (circa 7 hours minimum), and temperature is extremely low in the winter months, and these factors necessarily result in slow photochemical reaction rates.Protelli (1977) and Davison et al. (1981) have estimated the mixing heights for Fort Smith, Stony Plain, and around the Mildred Lake area (near AMS 2).The mixing heights calculated by Davison et al. (1981) were based on the seasonal analysis of minisonde measurements, and the mean maximum afternoon mixing heights from Protelli (1977) were based on the analysis of radiosonde measurement.The mixing heights ranged from very low (208-270 m in January) to high (1,499-2,396 m in May and June) for Protelli's study, and 1,000 m in summer for Davison et al. (1981).In general, the mixing heights are well established to vary significantly from winter to summer.
Among the six sites, NO 2 concentrations were highest in the winter months at JP104 (median 6 ppb) and JP102 (median 4.13 ppb) as these sites are closer to two main emission sources.The NO 2 concentrations at AH7, AH3 and JP101 had a similar pattern, and concentrations ranged from 0.25 to 2.65 ppb.However, JP107, 94 km north of oil sands operations, but in the Athabasca Valley influence, also had higher NO concentrations during winter months.

SO 2 Concentration
Monthly SO 2 concentrations are shown in Fig. 5. Interestingly, three types of SO 2 concentration patterns were found.At JP104 and JP102, no clear trend was observed.The SO 2 concentrations at these two sites might be influenced by the emissions directly.At AH3, JP101 and JP107, SO 2 concentrations had a seasonal pattern which was low in summer and high in winter.The lower mixing height and stable atmosphere are the major reasons for high SO 2 concentrations in the winter months.In summer, the SO 2 reactions including dry deposition and heterogeneous reaction with aerosols to form sulfate (Seinfeld and Pandis, 2006) are usually faster due to more water surface area, higher concentrations of oxidants and higher temperature.The distance (or time) between the emissions and these three sites allowed SO 2 to undergo direct deposition, heterogeneous reaction, and within plume dilution.AH7 behaved as a transitional site between these two groups, sometimes impacted by emissions directly (i.e., May, Fig. 5(c)).
Overall, O 3 , and NO 2 monthly concentrations demonstrated the annual cycles in which O 3 and NO 2 concentrations reached their maximums in April and winter, respectively.SO 2 concentrations at sites (i.e., JP104 and JP102) close to emissions did not exhibit any seasonal patterns but the SO 2 concentration seasonal pattern was clear at sites (AH3, JP101 and JP107) 48 km away from the reference point.

Year Trend Analysis
The passive dataset comprises a monthly time series covering the time period from January 2000 to December 2009.It consists of 2,160 data points, including a few missing values due to sample loss in the field.Both simple linear regression model and Sen-Thiel trend analysis of time series were used for the 10 year trend analysis of O 3 , NO 2 and SO 2 data from JP104, JP102, AH7, AH3, JP101 and JP107.

Simple Linear Regression Model
The simple linear regression model used annual average concentrations of O 3 , NO 2 and SO 2 to examine the 10-year trend.During 2000 to 2009, all p-values for O 3 trend analyses (Fig. 6) at 6 sites were > 0.47 (Table 2) indicating no statistically significant change in O 3 during the period of this study.The highest O 3 concentrations occurred at all six sites in 2004 (Fig. 6).As displayed in Fig. 7, the peak annual average concentration was also observed by the O 3 continuous monitor at our background continuous monitoring station (AMS 8, located in Fort Chipewyan, Alberta) which is 180 km north from the reference point.Hence, the peak concentrations in 2004 might be influenced by the regional O 3 background concentration.Overall, the O 3 concentrations at JP104, JP102 and JP107 were lower than the concentrations at AH7, AH3 and JP101 (Fig. 6).JP104 and JP102 are close to large industrial NO x emissions.
The concentrations of NO 2 at four of the six sites increased slightly from 2000 to 2009 (Fig. 8).These sites were JP104 (p    2).JP104 and JP102 are 11.5 and 15.7 km away respectively from the reference point.Increased NO 2 concentrations may be related to increased oil sands activities over the past 10 years.The NO 2 concentrations at AH7 (p = 0.11) and JP101 (p = 0.73) exhibited no significant change over 10 years.
For the 10-year SO 2 trend analyses at the six sites (Fig. 9 and Table 2), the low R-square values (< 2.4 × 10 -1 ) with p-values (> 0.15) demonstrated no significant change in SO 2 concentrations during the past 10 years, as bitumen production increased.The intercepts (> 1.915), however, indicated that SO 2 concentrations at sites near the main industrial area (JP104, JP102 and AH7) were higher than those at other remote sites.SO 2 is generated from bitumen processing, notably combustion and/or coking processes as a part of the conversion of bitumen to synthetic crude oil.Sulfur reduction systems have been adopted to control sulfur emission from stacks.The use of sulfur reduction systems could be the key factor for no increase in SO 2 concentration level.

Sen-Thiel Trend Analysis of Time Series
The missing value pattern analysis revealed that the missing data patterns are conducive to imputation of missing data using standard statistical algorithms.The imputation method used fully conditional specification and predictive mean matching with regression model when required.There are no missing data after the imputation.
As shown in Figs. 3 to 5, the seasonal variations indicating strong periodicity can be clearly identified.In general, time series with a periodic component M(t) which has periodicity d, given by a monthly periodic/seasonal nature is expressed as a linear combination of sine and cosine functions, which is called as harmonic regression function (HRF) (Bloomfield, 2000).
where a j and b j are unknown parameters and λ j are fixed frequencies, each being an integer multiple of 2/d, and a 0 is a constant.Fig. 10 are the same time series (observed data) with the fitted harmonic regression function which captures seasonal periodicity for O 3 , NO 2 and SO 2 at JP107 and the corresponding harmonic trend functions (seasonality) are Eq.( 14), Eq. ( 15) and Eq. ( 16).The periodicity is very  Gilbert (1987) states that a standard regression test in such cases can be misleading if seasonal cycles are present, data are not normally distributed, and/or the data are serially correlated.Furthermore, Gilbert (1987) states that the slope computed using the simple linear regression model based on least squares can deviate greatly from the true slope if there are gross errors or outliers in the data.Hence, calculating Sen nonparametric slope and performing Sen-Theil test of trend significance (Sen, 1968;Hollander and Wolfe, 1973) are required to determine and verify the trending.Table 3 has listed the computation values: Sen slope (Sen) and the corresponding Sen-Theil 95% confidence interval (S low , S high ).Based on values of Sen, S low , and S high , SLOPE was determined and it states the result of the test of significance for the slope (existence of a trend).
It is quite interesting to compare results of the Sen-Theil tests with the linear regression results.There is a complete agreement in O 3 case indicating O 3 concentrations had no statistically significant differences in the 10 years.In case of NO 2 , there is a full agreement at JP107, JP104, JP101, and AH3 at which NO 2 concentrations were increasing at JP107, JP104 and AH3.In case of JP102, the linear regression model concludes that trend was increasing while Sen-Thiel test concludes perhaps increasing.In case of AH7, linear regression analysis concludes that there is no trend while Sen-Theil test concludes perhaps there might be trend.In case of SO 2 , linear regression concludes that there is no trend in all cases agreeing with Sen-Theil test conclusions except at JP107 where Sen-Theil test concludes that perhaps there might be an increase and at JP101 where Sen-Theil test concludes that there perhaps might be a decrease.Thus Sen-Theil analysis is in a very good agreement with the linear regression conclusions.However, while using linear regression may be questionable on the groups of theoretical assumptions, Sen-Theil analysis is in every respect fully statistically justified.Furthermore, Sen-Theil analysis yields slightly more refined conclusions indicating that linear regression analysis conclusions might be tenuous.
From 2000 to 2009, the O 3 and SO 2 concentrations did not exhibit any significant changes (increasing or decreasing).The NO 2 concentrations at sites, i.e., JP104, JP102, and AH3, near the major emission sources, have increased during the past 10 years.

Concentrations of SO 2 , NO 2 and O 3 in the AOSR
In 2005, three sites (JP205, JP201, and JP213) were added to the passive sampling network.In 2007, site JP212 was established to further characterize NO 2 , O 3 and SO 2 concentrations in air near the major facilities.The annual average concentrations at six sites from 2001 to 2003, and at 10 sites from 2008 to 2009 were used to evaluate concentration changes over time.Three sites, JP205, JP210 and JP213, are situated in northeast, southeast and east of the oil sands operations, and are all around 100 km away from the reference point.The lowest O 3 concentration was observed at JP205.
The highest O 3 concentrations were observed at site JP213 located in Saskatchewan.Most of the time, the monthly O 3 concentrations at JP213 were higher than those at other sites.Monthly concentrations were highly correlated at sites JP213 and JP205 (R = 0.957) and sites JP213 and JP210 (R = 0.956).

NO 2 Concentration
The NO 2 concentrations were lower and less variable at a distance of 48 km (Fig. 11(b)).The NO 2 concentrations from 2008 to 2009 were higher than those from 2001 to 2003.The NO 2 concentration decreased significantly within 30 km of the reference point.The highest NO 2 concentration (5.7 ppb) was at JP104, 2.7 times the concentration at 30 km in 2009.The NO 2 concentrations (4.16 ppb) at 9.4 km, site JP212, were lower than those (5.70 ppb) at 11.5 km, JP104.Site JP212 is close to the major emission sources of NO x and NO is the major form of NO x emitted.The O 3 titration results in elevated NO 2 concentration and lower O 3 concentrations.In addition, there was another industrial NO x emission source close to JP104 site which might cause the elevated NO 2 concentration.However, higher NO 2 concentration and lower O 3 concentration were observed at JP107, 94 km from the reference point, for both 2008 and 2009.

SO 2 Concentration
The monthly average SO 2 concentrations (> 1.7 ppb) were higher within 28 km of the reference point.The SO 2 concentrations became quite stable and were low at a distance of 48 km.Lower SO 2 concentrations could result from emissions dilution, deposition and atmospheric chemical reactions of SO 2 .The highest SO 2 concentration (2.6 ppb in 2003 and 2.2 ppb in 2009) was at 11.5 km (JP104).
In 2009, JP107 had slightly higher SO 2 concentration than in 2001 to 2003.As shown in Fig. 5(f), SO 2 concentrations were low in summer and the emissions from forest fire might not be the reason to explain the elevated SO 2 concentration.Besides, the occurrence of fires in 2008 and 2009 was low.In general, the SO 2 concentrations in the past 10 years at all sites were very similar year to year.
In summary, O 3 concentrations were low near the emission sources, while SO 2 and NO 2 concentrations were higher near the reference point.The highest NO 2 concentration was observed at 11.5 km, JP104.Elevated NO 2 and SO 2 concentrations were observed at JP107 in2008 and 2009.The possible explanations are the influence of the Athabasca Valley flow and the new industrial emission near JP107.Further investigation should be conducted to verify the main source.

Comparison of Passive and Active Measurements
Collocated passive and active measurements were obtained at AMS 1 and AMS 6 (n = 138 for O 3 , n = 139 for NO 2 , and n = 142 for SO 2 ).

O 3 Measurement
The O 3 concentrations from the OPSS and the continuous monitor are shown in Fig. 12(a).The O 3 monthly average concentrations from the OPSS ranged from 10.0 to 43.3 ppb, and hourly concentrations from the continuous monitor ranged from 9.6 to 36.9 ppb.Most data points are within the ± 30% range of O 3 concentrations from continuous monitor.
Descriptive statistics are listed in Table 4.The median of O 3 concentrations from the OPSS, 22.0 ppb, was slightly higher than that from the continuous monitor, 19.5 ppb.The Pearson correlation coefficient was 0.81 which implies the O 3 concentrations from two measurements were correlated.However, the p-value was < 0.01 indicating the O 3 concentrations from two measurements were significantly different.

NO 2 Measurement
Monthly average NO 2 collocated concentrations exhibited a wider range from 1.23 to 15.50 ppb.The NO 2 concentrations from the NPSS were lower than those from the continuous monitors as shown in Fig. 12(b).Most data points from the NPSS were between -60% and 30% of NO 2 concentrations from the continuous monitor.The deviation increased when NO 2 concentrations were higher.
The median NO 2 concentration from the NPSS, 3.67 ppb, was lower than hourly concentration from the NO 2 continuous monitor, 4.85 ppb.The Person correlation coefficient was 0.90 and the p-value was < 0.01 which indicates that the means between two measurements were statistically significantly different.

SO 2 Measurement
The monthly average SO 2 concentrations were generally low and the concentrations ranged from 0.32 to 3.66 ppb.The SO 2 concentrations from the SPSS were generally higher than those from the continuous monitors within the range of -30% to 40% as displayed in Fig. 12(c).
The median SO 2 concentration from the SPSS, 1.27 ppb, was slightly higher than the mean hourly concentration from the SO 2 continuous monitor.The Pearson correlation coefficient is 0.76 and the p-value of t-Test is lower than 0.01 indicating the mean SO 2 concentrations from two measurements were statistically significantly different.
In general, the NPSS underestimated the NO 2 concentration and the OPSS and SPSS overestimated the O 3 and SO 2 concentrations slightly.Significant difference between continuous measurement and passive measurement is likely due to inability of passive samplers to react to short-exposures, and higher concentrations (Krupa and Legge, 2000).In other words, the concentrations from the passive samplers cannot be applied as the concentrations from the continuous measurement.However, high correlation coefficients between two measurements still demonstrate the passive measurement can be effectively used to determine the long-term trend.

CONCLUSIONS
The monthly O 3 concentrations showed the expected spring O 3 maximum which is a common phenomenon at the northern hemisphere.The NO 2 concentrations were high in the winter months due to the stable atmosphere, lower mixing height, and higher emissions.The SO 2 concentrations did not have a clear seasonal pattern near the major emission sources and the SO 2 concentrations had a similar seasonal pattern with NO 2 concentrations at distant sites (50 km away from the reference point).
Trending analysis using linear regression and nonparametric Sen-Theil testing agreed very well and both analyses showed no significant change for O 3 concentrations during the past 10 years at all six sites.Sen-Theil analysis concluded that NO 2 concentrations at the locations near the major emission sources were increasing.
The concentrations of O 3 , NO 2 and SO 2 became less variable at 20, 48 and 48 km, respectively from the reference point (AMS 2).The highest concentrations of NO 2 and SO 2 were at 11.5 km from the reference point and they were 5.7 ppb for NO 2 and 2.2 ppb for SO 2 in 2009.JP107 may have received contributions from the influence of valley flow, new industrial or natural emission sources since 2007 resulting in higher NO 2 and SO 2 concentrations and lower O 3 concentration.
Continuous and passive measurement techniques have different purposes and uses.The data from passive measurement can be compared for general trending, but should not be compared on an absolute basis.

ACKNOWLEDGEMENT
The author greatly appreciates the thorough statistical analysis completed by Dr. Milo Nosal, University of Calgary.I also appreciate the insightful interpretation of this analysis provided by Dr. Nosal.The author would like to thank Dr. Ken Foster, Dr. Allan Legge, Dr. Kevin Percy and Lori Adamache their valuable comments and suggestions to improve the quality of the paper.
This study was funded by WBEA.The content and opinions expressed by the author in this study do not necessarily reflect the views of the WBEA or of the WBEA membership.
The spatial variation of O 3 concentrations is shown inFig.11(a).The multi-year O 3 concentrations were relatively low (17.4-21.5 ppb) at 9 km from AMS 2. The likely explanation is localized O 3 titration.Higher NO concentrations occur near major emission sources, and oxidation of NO by O 3 to form NO 2 would be expected.The O 3 concentration was stable at around 20 km, at 25 ppb in 2009.
also reported on O 3 concentrations in Birch Mountains and at Bitumount from 1977 to 1980, when only two oil sands operations were in production.They found that monthly O 3 concentrations exhibited a bimodal distribution in the Birch Mountains, approximately 100 km NE of Fort McMurray, but not in Bitumount closer to the oil sands development near the Athabasca River Valley.It is unlikely the oil sands activities caused the observed secondary November O 3 peak.

Table 2 .
Simple linear regression analysis results for O 3 , NO 2 and SO 2 passive concentrations.

Table 4 .
Descriptive statistics for O 3 , NO 2 and SO 2 concentrations (ppb) from passive samplers and continuous monitors.