2.1. General Trends and Statistics
Figure 1 shows TGM concentrations measured at Patricia McInnes station from 21 October 2010 through 31 May 2013, inclusively. Summary statistics for these data are also listed in
Table 1. The average TGM concentration at Patricia McInnes station over the course of this study was 1.45 ± 0.18 ng∙m
−3, which is comparable to that measured at other stations in the province of Alberta (1.36–1.65 ng∙m
−3 [
7,
8]). The gap in data from 21 May 2011 through 20 October 2011 was due to instrument contamination as noted in
Section 3.3 below. There is a slight increasing trend in TGM concentrations measured at the Patricia McInnes station with a rate of 0.051 ± 0.003 ng∙m
−3∙y
−1 over the range of the study period as determined with linear regression. Note that at approximately 2.5 years in length, this period of record is not sufficiently long to definitively regard this as a true long-term trend, and this trend reported may only represent a medium-term fluctuation within a different longer-term trend. As seen in
Figure 1, there are several instances of elevated TGM concentrations. Most of these events coincide with forest fire smoke near the Patricia McInnes station as indicated by the shaded areas in
Figure 1. Conversely, not all cases where forest fire smoke was present led to increased TGM concentrations. This may be due to the distance between the forest fire and the monitoring station,
i.e., the age of the smoke. Data impacted by forest fire smoke is included in analyses described below except where noted.
Figure 1 indicates that, aside from periods impacted by forest fire smoke, high TGM concentrations are generally observed over short time scales, whereas low TGM concentrations are generally observed over longer time scales.
Table 1 shows the statistics for the data deemed to be measured during forest fire smoke events (N = 447) in relation to the data that was not impacted by forest fire smoke (N = 17,020).
Figure 1 also shows ambient TGM concentrations to have a seasonal dependence. This is further illustrated in
Figure 2, which shows the monthly averaged profile of TGM concentrations measured at Patricia McInnes station.
Figure 2 shows a maximum in the spring and a minimum in the fall for data not impacted by forest fire smoke (black data), a pattern that has been observed in other similar locations and linked to meteorological cycles [
8,
16,
17]. Forest fire smoke was only present during the months of May through September, inclusively, as shown in
Figure 2 (red data). In general, mean monthly TGM is significantly higher in these months (p < 0.0001; d.f. ≥ 654) when forest fire smoke was present than when it was not, with the exception of September (p = 0.16; d.f. = 423), according to t-tests of monthly mean values.
Figure 1.
Time series of hourly total gaseous mercury (TGM) concentrations at Patricia McInnes station. Grey shaded areas indicate when forest fire smoke was in the vicinity of the monitoring station. The de-trended/seasonally adjusted TGM time series is shown for reference. As described in the text, TGM concentration data was categorized as “high”, “average”, or “low”, with divisions at the 33rd- and 67th-percentiles in de-trended/seasonally adjusted TGM concentration.
Figure 1.
Time series of hourly total gaseous mercury (TGM) concentrations at Patricia McInnes station. Grey shaded areas indicate when forest fire smoke was in the vicinity of the monitoring station. The de-trended/seasonally adjusted TGM time series is shown for reference. As described in the text, TGM concentration data was categorized as “high”, “average”, or “low”, with divisions at the 33rd- and 67th-percentiles in de-trended/seasonally adjusted TGM concentration.
Table 1.
Summary statistics for Patricia McInnes station hourly TGM concentration data.
Table 1.
Summary statistics for Patricia McInnes station hourly TGM concentration data.
Dataset | Date Range | N | Mean (ng∙m−3) | Median (ng∙m−3) | Standard Deviation (ng∙m−3) | Minimum (ng∙m−3) | Maximum (ng∙m−3) |
---|
All data | 21 October 2010–31 May 2013 | 17467 | 1.45 | 1.46 | 0.18 | 0.64 | 4.43 |
Excluding smoke | 17020 | 1.45 | 1.46 | 0.17 | 0.64 | 3.05 |
Only smoke | 447 | 1.73 | 1.69 | 0.34 | 1.08 | 4.43 |
Figure 3 shows the diel profile of mean TGM concentrations for each season (spring: 21 March–20 June; summer: 21 June–20 September; fall: 21 September–20 December; winter: 21 December–20 March). As observed in other studies [
16,
18], minimum and maximum TGM concentrations are observed in the morning and in the afternoon-evening, respectively. Seasonal diel profiles of TGM concentrations qualitatively resemble those of the mean temperatures and ozone concentrations at Patricia McInnes (
Figure 3) and highlight the relationship between these variables under average conditions. This relationship is also identified in the principal component analyses described in
Section 2.2.
Figure 2.
Monthly averaged TGM concentrations measured at Patricia McInnes station. Red data are impacted by forest fire smoke, whereas black data are not.
Figure 2.
Monthly averaged TGM concentrations measured at Patricia McInnes station. Red data are impacted by forest fire smoke, whereas black data are not.
Figure 3.
Seasonal diel profiles of mean TGM concentrations at Patricia McInnes station after removing data impacted by forest fire smoke. Seasonal diel profiles of mean temperatures and ozone concentrations at Patricia McInnes are also shown for reference Vertical bars indicate ± 2 standard deviations of TGM concentration data at each hour.
Figure 3.
Seasonal diel profiles of mean TGM concentrations at Patricia McInnes station after removing data impacted by forest fire smoke. Seasonal diel profiles of mean temperatures and ozone concentrations at Patricia McInnes are also shown for reference Vertical bars indicate ± 2 standard deviations of TGM concentration data at each hour.
2.2. Principal Component Analysis
In addition to TGM concentration measurements, other air quality and meteorological measurements were considered, as described in
Section 3.2. Air quality parameters include nitrous oxide (NO), nitrogen dioxide (NO
2), oxides of nitrogen (NO
X), ozone (O
3), fine particulate matter (PM
2.5), sulphur dioxide (SO
2), total hydrocarbon (THC), total reduced sulphur (TRS), ammonia (NH
3), and carbon monoxide (CO). Meteorological parameters include temperature (T), relative humidity (RH), wind speed and direction, solar radiation, and snow depth. Principal component analysis (PCA) was performed to identify correlations between parameters and serves as a means for reducing the number of parameters in the dataset into a handful of important factors. PCA was similarly applied in previous work where TGM concentrations were measured alongside a variety of ancillary variables to quantify relationships within the dataset and identify common factors that may influence TGM concentrations [
19]. Three data subsets were established in order to better understand the factors affecting TGM and other air quality and meteorological parameters. The data subsets are grouped by rank according to “high”, “average”, and “low” TGM concentrations (
i.e., three bins divided at the 33rd- and 67th-percentiles). In order to avoid biasing data subsets with predominately spring values in the “high” subset or fall values in the “low” subset, for example, the TGM concentration data was de-trended and seasonally adjusted only to
categorize any given TGM concentration as “high”, “average”, or “low”, but the
original TGM concentration values were used as input for PCA. De-trending and seasonal adjustment was carried out with additive decomposition with a linear trend component and 4th-order polynomial fit to monthly averaged data (similar to that shown in
Figure 2) as a seasonal filter [
20,
21]. The de-trended and seasonally adjusted time series is shown in
Figure 1 for reference. PCA results for each data subset, “high”, “average”, and “low” TGM concentrations, are shown in
Table 2,
Table 3 and
Table 4, respectively. For comparison, PCA results for the full, combined dataset are shown in
Table 5. These tables show the rotated factor loadings, the amount of variability explained by each factor and the eigenvalues. Only factors with eigenvalues greater than 1 are deemed significant and are interpreted below [
22]. Parameters are an important contribution to a given factor when the absolute value of the corresponding rotated factor loading is ≥ 0.5 (bold values in
Table 2 through
Table 4) [
22]. The meaning of each factor was inferred based on the groupings of parameters that load heaviest on that factor (e.g., NO
X is interpreted as anthropogenic combustion processes, and PM
2.5, CO, and NH
3 together are interpreted as forest fire smoke).
Table 2.
Principal component analysis (PCA) rotated factor loadings for the highest 1/3rd of de-trended/seasonally adjusted TGM concentrations (i.e., “high” TGM concentrations). Factor loadings with an absolute value greater than or equal to 0.5 are bolded. The shaded row highlights TGM loadings on each factor.
Table 2.
Principal component analysis (PCA) rotated factor loadings for the highest 1/3rd of de-trended/seasonally adjusted TGM concentrations (i.e., “high” TGM concentrations). Factor loadings with an absolute value greater than or equal to 0.5 are bolded. The shaded row highlights TGM loadings on each factor.
Parameter | Combustion Processes | Diel Trending | Forest Fire Smoke | Snow Depth | Industrial Sulphur |
---|
NO | 0.93 | 0.04 | 0.03 | 0.06 | 0.01 |
NO2 | 0.86 | 0.28 | 0.06 | 0.16 | 0.14 |
NOX | 0.96 | 0.17 | 0.05 | 0.12 | 0.08 |
THC | 0.60 | 0.04 | 0.25 | −0.45 | 0.32 |
O3 | −0.55 | −0.70 | 0.01 | 0.07 | −0.10 |
Temperature | −0.23 | −0.68 | 0.19 | −0.48 | 0.01 |
RH | 0.14 | 0.86 | 0 | 0.12 | 0.11 |
Solar Radiation | −0.03 | −0.78 | −0.03 | −0.12 | 0.09 |
TGM | −0.05 | −0.49 | 0.59 | 0.28 | 0.17 |
PM2.5 | 0.06 | −0.05 | 0.87 | −0.11 | 0.17 |
NH3 | 0.02 | 0.04 | 0.74 | 0.01 | −0.06 |
CO | 0.11 | 0 | 0.79 | −0.10 | −0.03 |
Snow Depth | 0.22 | 0.35 | −0.01 | 0.81 | 0.08 |
SO2 | 0.05 | −0.05 | −0.06 | 0.22 | 0.83 |
TRS | 0.25 | 0.10 | 0.19 | −0.35 | 0.70 |
Variability (%) | 22.7 | 18.6 | 16.1 | 9.5 | 9.4 |
Cumulative Variability (%) | 22.7 | 41.3 | 57.5 | 67.0 | 76.4 |
Eigenvalues | 4.62 | 2.99 | 1.60 | 1.13 | 1.11 |
Table 3.
PCA rotated factor loadings for the middle 1/3rd of de-trended/seasonally adjusted TGM concentrations (i.e., “average” TGM concentrations). Factor loadings with an absolute value greater than or equal to 0.5 are bolded. The shaded row highlights TGM loadings on each factor.
Table 3.
PCA rotated factor loadings for the middle 1/3rd of de-trended/seasonally adjusted TGM concentrations (i.e., “average” TGM concentrations). Factor loadings with an absolute value greater than or equal to 0.5 are bolded. The shaded row highlights TGM loadings on each factor.
Parameter | Combustion Processes & NH3 | Diel Trending | Industrial Sulphur & PM | Temperature & Snow Depth | CO |
---|
NO | 0.89 | 0.07 | 0.14 | 0.07 | 0.10 |
NO2 | 0.70 | 0.29 | 0.36 | 0.30 | 0.20 |
NOX | 0.87 | 0.20 | 0.28 | 0.20 | 0.17 |
NH3 | 0.64 | −0.11 | −0.25 | −0.11 | −0.35 |
THC | 0.49 | 0.39 | 0.42 | −0.30 | 0.07 |
TGM | −0.06 | −0.76 | 0.05 | 0.28 | 0.05 |
O3 | −0.48 | −0.70 | −0.27 | −0.07 | −0.18 |
Temperature | −0.19 | −0.60 | 0 | −0.63 | 0.18 |
RH | 0.09 | 0.81 | 0.03 | 0.20 | −0.01 |
Solar Radiation | 0.02 | −0.70 | 0.05 | −0.18 | 0.01 |
PM2.5 | 0.29 | −0.13 | 0.55 | 0.02 | 0.34 |
SO2 | −0.03 | −0.15 | 0.71 | 0.27 | −0.19 |
TRS | 0.19 | 0.17 | 0.69 | −0.16 | −0.07 |
Snow Depth | 0.11 | 0.08 | 0.04 | 0.91 | 0.09 |
CO | 0.09 | −0.04 | −0.10 | 0.01 | 0.86 |
Variability (%) | 20.8 | 19.5 | 12.3 | 11.5 | 7.8 |
Cumulative Variability (%) | 20.8 | 40.2 | 52.5 | 64.0 | 71.8 |
Eigenvalues | 4.80 | 2.16 | 1.52 | 1.23 | 1.06 |
Table 4.
PCA rotated factor loadings for lowest 1/3rd of de-trended/seasonally adjusted TGM concentrations (i.e., “low” TGM concentrations). Factor loadings with an absolute value greater than or equal to 0.5 are bolded. The shaded row highlights TGM loadings on each factor.
Table 4.
PCA rotated factor loadings for lowest 1/3rd of de-trended/seasonally adjusted TGM concentrations (i.e., “low” TGM concentrations). Factor loadings with an absolute value greater than or equal to 0.5 are bolded. The shaded row highlights TGM loadings on each factor.
Parameter | Combustion Processes | Diel Trending | Temperature & Snow Depth | Industrial Sulphur & NH3 |
---|
NO | 0.85 | 0.06 | 0.16 | 0.08 |
NO2 | 0.74 | 0.17 | 0.51 | −0.07 |
NOX | 0.88 | 0.12 | 0.34 | 0.01 |
PM2.5 | 0.59 | −0.12 | 0.07 | −0.31 |
THC | 0.68 | 0.42 | −0.16 | −0.1 |
O3 | −0.54 | −0.71 | −0.06 | 0 |
TGM | −0.02 | −0.72 | 0.05 | 0.02 |
RH | 0.03 | 0.83 | −0.03 | 0.02 |
Solar Radiation | 0.05 | −0.64 | −0.34 | −0.02 |
Temperature | −0.14 | −0.26 | −0.87 | 0.03 |
Snow Depth | 0.07 | −0.14 | 0.9 | −0.06 |
SO2 | 0.16 | −0.17 | 0.28 | −0.61 |
TRS | 0.49 | 0.11 | −0.07 | −0.51 |
NH3 | 0.21 | −0.07 | 0.08 | 0.61 |
CO | 0.41 | −0.10 | −0.10 | 0.20 |
Variability (%) | 24.5 | 16.7 | 14.9 | 7.8 |
Cumulative Variability (%) | 24.5 | 41.2 | 56.1 | 63.9 |
Eigenvalues | 4.61 | 2.14 | 1.74 | 1.10 |
Focusing on the factors in relation to TGM,
Table 2 correlates “high” TGM concentrations with parameters that are associated with forest fire smoke, namely: PM
2.5, CO, and NH
3 [
23]. In other words, and not surprisingly, many of the high values in TGM are often a result of forest fire smoke impinging on the station. The next important factor for TGM during “high” TGM concentration observations relates to diel trending that strongly links ozone, temperature, RH, and solar radiation. TGM also shows minimal loading on the factor attributed to snow depth for the “high” TGM concentration data subset, which is not surprising given the strong association between “high” TGM concentration observations and forest fire smoke during the summer when there is no snow cover. PCA results for the “average” TGM concentration data subset (
Table 3) show a shift of TGM correlation to the diel trending factor shared with ozone, temperature, RH, and solar radiation. The next important factor for TGM during “average” TGM concentration observations relates to temperature and snow depth. Similarly, PCA results for the “low” TGM concentration data subset (
Table 4) also correlate TGM with diel trending ozone, RH, and solar radiation (but not temperature). Note that, in general, temperature is not as strongly associated with diel trending factors in all the PCA result tables, likely due to the wide variability of temperature in this region. Interestingly, for “average” and “low” TGM concentration subsets, TGM is effectively unrelated to pollutants mainly associated with local anthropogenic sources (
i.e., NO, NO
2, NO
X, SO
2, TRS). Likewise, “high” TGM concentration data is very minimally, and essentially insignificantly, associated with SO
2 and TRS. Considering the PCA results from the complete dataset in
Table 5, the overall ranking of importance for factors driving TGM variability is: diel trending > forest fire smoke > temperature and snow depth > industrial sulphur > combustion processes. Thus,
Table 5 further indicates the lack of association between industrial pollutants and TGM. This is despite the fact that the station is within an urban area and near the industrialized center of the Canadian oil sands region.
Table 5.
PCA rotated factor loadings for the full, combined dataset. Factor loadings with an absolute value greater than or equal to 0.5 are bolded. The shaded row highlights TGM loadings on each factor.
Table 5.
PCA rotated factor loadings for the full, combined dataset. Factor loadings with an absolute value greater than or equal to 0.5 are bolded. The shaded row highlights TGM loadings on each factor.
Parameter | Combustion Processes | Diel Trending | Forest Fire Smoke | Temperature & Snow Depth | Industrial Sulphur |
---|
NO | 0.92 | 0 | 0.07 | 0.06 | 0.01 |
NO2 | 0.83 | 0.20 | 0.09 | 0.29 | 0.17 |
NOX | 0.96 | 0.10 | 0.08 | 0.19 | 0.09 |
THC | 0.62 | 0.27 | 0.15 | −0.39 | 0.28 |
O3 | −0.56 | −0.70 | −0.01 | 0.01 | −0.08 |
TGM | −0.12 | −0.65 | 0.37 | 0.24 | 0.14 |
RH | 0.10 | 0.84 | 0.02 | 0.12 | 0.07 |
Solar Radiation | 0.01 | −0.73 | −0.07 | −0.25 | 0.03 |
Temperature | −0.26 | −0.54 | 0.15 | −0.64 | −0.01 |
PM2.5 | 0.08 | −0.07 | 0.82 | −0.06 | −0.31 |
CO | 0.12 | −0.06 | 0.71 | −0.03 | −0.06 |
NH3 | 0.03 | 0.04 | 0.73 | 0 | −0.04 |
Snow Depth | 0.17 | −0.11 | 0 | 0.89 | 0.06 |
SO2 | 0.04 | −0.11 | −0.06 | 0.24 | 0.80 |
TRS | 0.27 | 0.11 | 0.16 | −0.21 | 0.73 |
Variability (%) | 22.5 | 17.4 | 13.0 | 11.4 | 9.2 |
Cumulative Variability (%) | 22.5 | 39.9 | 52.9 | 64.3 | 73.5 |
Eigenvalues | 4.53 | 2.48 | 1.50 | 1.42 | 1.10 |
2.3. Directional and Back Trajectory Analyses
Although the PCA results suggest a lack of localized industrial source influences on TGM concentrations, the directionality of TGM concentrations was explored and is plotted as a pollutant rose alongside a wind rose in
Figure 4. Note that the pollution rose in
Figure 4b excludes both data observed in calm wind conditions (where wind speed < 5 km∙h
−1) and data impacted by forest fire smoke. Although the pollution rose, plotted as bars on the black axis in
Figure 4b, at first appears to suggest possible point-sources to the north-northwest (
i.e., oil sands development) and southeast (
i.e., urban sources within Fort McMurray), a comparison to the wind rose (
Figure 4a) indicates similar north-northwest and southeast dominant wind direction. The predominant winds at the station follow river valleys from the north-northwest and the southeast, hence TGM is simply observed from wherever the wind originates with minimal increased weighting toward any direction of a point-source. This is further illustrated with the directionally dependent TGM concentration percentiles, plotted as lines on the red axis in
Figure 4b, which do not indicate any significant directional dependence to TGM concentrations. Thus, both the PCA results and
Figure 4 suggest a limited role of local point-source influences on ambient TGM concentrations. These results suggest that the TGM observed at Patricia McInnes is predominately due to broad area sources (
i.e., surface flux) and long-range transport.
Figure 4.
(a) Wind rose and (b) TGM concentration rose for Patricia McInnes station. Data corresponding to calm winds (wind speed < 5 km∙h−1) are not shown here. The TGM rose excludes data impacted by forest fire smoke. Frequency of concentration observations are plotted on the black scale and percentiles in TGM concentrations are plotted on the red radial axis in panel (b).
Figure 4.
(a) Wind rose and (b) TGM concentration rose for Patricia McInnes station. Data corresponding to calm winds (wind speed < 5 km∙h−1) are not shown here. The TGM rose excludes data impacted by forest fire smoke. Frequency of concentration observations are plotted on the black scale and percentiles in TGM concentrations are plotted on the red radial axis in panel (b).
Since the PCA results and
Figure 4 suggest that local conditions were not sufficient to explain all observations of TGM concentrations, the role of long-range transport was assessed for the observations at Patricia McInnes. To investigate the role of long-range transport, back trajectories were calculated with the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model [
24,
25]. Back trajectory analysis has been identified as an important tool for assessing the history of air masses influencing concentration measurements of atmospheric species [
26]. The 72-hour HYSPLIT back trajectories from Patricia McInnes station were obtained for all observations of hourly TGM concentrations. Data observed with forest fire smoke were excluded from this analysis. The coordinates of the starting point of each back trajectory were plotted in ArcMap10 as points in
Figure 5a where the colour of each point represents the de-trended/seasonally adjusted TGM concentration observed with each back trajectory (as defined in
Section 2.2 and
Figure 1). De-trended/seasonally adjusted TGM concentrations were used rather than actual TGM concentrations to gain insight into which regions were most frequently associated with
unseasonably high or low TGM concentrations for any given time of year, rather than the times of year when higher or lower TGM concentrations are expected (as shown in
Figure 2). Due to the presence of many overlapping points,
Figure 5b shows the same back trajectory starting points, but illustrated as point density (see
Section 3.5 for further details on point density mapping).
Figure 5b indicates that most 72-hour back trajectory starting points were located in northern Alberta, northern Saskatchewan, northern British Columbia, and southern Northwest Territories where the highest point density is mapped in red. Only a limited number of back trajectories started outside western Canada, where the lowest point density is mapped in blue in
Figure 5b; grey regions in
Figure 5b indicate where no back trajectory starting points exist within a 700 km radius. Subsequently, a boundary condition to this back trajectory analysis was set to exclude regions with less than 1% of the total point density (
i.e., fewer than 170 back trajectory starting points within a 700 km radius neighborhood), represented as the dashed lines in
Figure 5a–c. Regions outside this dashed line were deemed to have an insufficient number of back trajectory starting points to draw any useful conclusions. Finally,
Figure 5c shows the point density of back trajectory starting points weighted with de-trended/seasonally adjusted TGM concentrations and normalized by the point density shown in
Figure 5b, resulting in a normalized de-trended/seasonally adjusted TGM concentration map (see
Section 3.5). Red and blue areas in
Figure 5c correspond to areas leading to the highest and lowest normalized de-trended/seasonally adjusted TGM concentration observations, respectively. The grey areas in
Figure 5c are outside the boundary limits of the analysis due to an insufficient number of back trajectory starting points.
Figure 5c shows that the lowest normalized de-trended/seasonally adjusted TGM concentration observations primarily arrive via arctic regions of northern Canada. This region has minimal land disturbance or industry, so lower TGM concentration observations from air passing through this region are not surprising. Note that this back trajectory analysis cannot clearly distinguish between lower TGM concentrations originating from arctic regions as a result of arctic atmospheric mercury depletion events (AMDEs) [
1,
3,
5,
27,
28,
29,
30,
31] as opposed to as a result of generally cleaner air (
i.e., air with lower TGM concentrations). However, given that during the study period, arctic air parcels are often associated with unseasonably lower TGM concentrations, this possibility is explored further in
Section 2.4 below.
Figure 5.
HYSPLIT 72-hour back trajectory analysis. Panel (a) shows all back trajectory starting points colour-coded to the corresponding de-trended/seasonally adjusted TGM concentrations; panel (b) shows the point density of back trajectory starting points; panel (c) shows the normalized de-trended/seasonally adjusted TGM concentrations as described in the text. Dashed lines indicate the boundary limits of the analysis as described in the text; grey areas contained insufficient or no back trajectory starting points. Fort McMurray is indicated in northern Alberta with the cross-hair symbol. Data observed with forest fire smoke were excluded from this analysis.
Figure 5.
HYSPLIT 72-hour back trajectory analysis. Panel (a) shows all back trajectory starting points colour-coded to the corresponding de-trended/seasonally adjusted TGM concentrations; panel (b) shows the point density of back trajectory starting points; panel (c) shows the normalized de-trended/seasonally adjusted TGM concentrations as described in the text. Dashed lines indicate the boundary limits of the analysis as described in the text; grey areas contained insufficient or no back trajectory starting points. Fort McMurray is indicated in northern Alberta with the cross-hair symbol. Data observed with forest fire smoke were excluded from this analysis.
Higher normalized de-trended/seasonally adjusted TGM concentration observations can be traced back to regions further southeast and west from the monitoring station, suggesting the role of transport from more populated areas of North America and Asian influenced trans-Pacific transport. That said, high TGM concentration may be introduced into an air parcel anywhere along a back trajectory, and the intention of this analysis is not to identify sources of atmospheric mercury, but rather to highlight the importance of long-range transport when dealing with atmospheric mercury as a global pollutant. Indeed, it is unlikely there are any significant sources of atmospheric mercury in the areas shaded red in
Figure 5c, but only that the air has passed through those regions on its way to the monitoring station where higher atmospheric mercury concentrations were observed. Note that
Figure 5c uses de-trended/seasonally adjusted TGM concentrations; thus, red or blue regions in this map indicate the starting points for the unseasonably high or low TGM concentrations for any given time of year and cannot be related to the seasonal trend in TGM concentrations illustrated in
Figure 2. In other words,
Figure 5c does not indicate whether back trajectories in the spring or fall tend to originate from regions shaded red or blue, respectively. The intermediate regions in
Figure 5c between extremes do not show any consistently high or low values as exhibited in the red or blue regions, but rather a relatively similar number of high and low values (as can be seen in
Figure 5a). The normalized de-trended/seasonally adjusted TGM concentrations approach 0 ng∙m
−3 in these regions. For example, normalized de-trended/seasonally adjusted TGM concentrations are approximately 0 ng∙m
−3∙km
−2 in the region near the sampling station because the high and low de-trended/seasonally adjusted TGM concentrations associated with these starting points cancel each other in the weighted point density calculations.
2.4. Mercury Chemistry Implications
In this dataset, low concentrations of TGM are often concurrently observed with low concentrations of ozone and higher RH, as indicated with the PCA results above. Furthermore, unseasonably low TGM concentrations are often traced back to arctic regions as illustrated in
Figure 5c. These associations suggest the possible role of atmospheric mercury chemistry, particularly oxidation of GEM to form GOM [
3,
5,
27,
32]. GOM can then undergo deposition faster than GEM, thereby reducing the ambient TGM concentration in an area. Previous studies have provided detailed discussions on this type of mercury chemistry, most notably in terms of arctic AMDEs during spring-time polar sunrises [
1,
3,
5,
27,
28,
29,
30,
31]. AMDEs are associated with reduced GEM concentrations, reduced ozone concentrations, and increasing solar radiation. The PCA results above, however, show a positive correlation between TGM concentrations and solar radiation (
i.e., reduced TGM concentrations with decreasing solar radiation), which suggests this reduction in TGM concentration is not likely a locally occurring process from oxidation by ozone. Since, in this dataset, lower TGM concentrations are associated with lower ozone and lower levels of solar radiation, the link between reduced TGM concentrations and reduced ozone concentrations is more likely an association with cleaner air passing over the monitoring station rather than a result of local GEM oxidation.
To further support this argument,
Figure 6 shows the diel and monthly fractional distributions of observations of “high”, “average”, and “low” TGM concentrations, with divisions at the 67th- and 33rd-percentiles of de-trended/seasonally adjusted TGM concentrations (described above in
Section 2.2 and illustrated in
Figure 1) excluding data affected by forest fire smoke.
Figure 6a shows the “low” TGM concentration observations typically occurred in the early morning or overnight while “high” TGM concentration observations typically occurred in the mid-day. The frequency of “average” TGM concentration observations is consistent throughout the day as indicated in
Figure 6a. This is not to say that “average” TGM concentration observations do not exhibit a diel trend, but rather that “average” TGM concentrations can be observed with the same frequency at any time of day. In contrast,
Figure 6b shows that the “high”, “average”, and “low” TGM concentration subsets do not have any clear seasonal trend.
Figure 6b emphasizes that “low” TGM concentration observations are distributed more or less equally over the course of the year, suggesting “low” TGM concentrations originating from arctic regions are more likely due to generally cleaner air than to AMDEs. Had the “low” TGM concentration observations been a result of local/regional AMDEs, one would expect to see the highest frequency of “low” TGM concentration observations during daylight hours. Likewise, had the “low” TGM concentrations been a result of arctic air transported into the region after arctic AMDEs, one would expect to see the highest frequency of “low” TGM concentration observations during the spring. Note that
Figure 3 also indicates that lower TGM concentration observations are not particularly common in the spring.
Figure 6.
(a) Diel and (b) monthly fractional distribution of “high”, “average”, and “low” de-trended/seasonally adjusted TGM concentrations as defined in the text. These plots exclude data affected by forest fire smoke.
Figure 6.
(a) Diel and (b) monthly fractional distribution of “high”, “average”, and “low” de-trended/seasonally adjusted TGM concentrations as defined in the text. These plots exclude data affected by forest fire smoke.
Figure 6a indicates again that, in general, local daytime oxidation by ozone is an unviable reaction mechanism for TGM depletion observed at Patricia McInnes. The diel trend in
Figure 6a also supports conclusions that TGM is driven by diel trends affecting underlying surface-air flux and long-range transport of global atmospheric mercury. In other words, a diel trend is superimposed on the TGM concentration time series that undergoes lower frequency oscillations as air with higher or lower TGM concentrations move over the monitoring station. The back trajectory analysis above also supports this conclusion, in that the lowest TGM values came with air that has passed through arctic regions of northern Canada, and are thus presumably cleaner than air originating from more populated regions. GEM oxidation chemistry may still be occurring locally, but it is difficult to speculate to what extent it occurs, if at all, without speciated mercury measurements providing GOM concentration data. Also, this dataset cannot provide any insight to the role of particulate bound mercury in the area. Mercury speciation measurements will form a future component of the Joint Canada-Alberta Implementation Plan for Oil Sands Monitoring, and that future work may provide further information to more definitively understand any local mercury chemistry taking place in the region with associated effects on local mercury deposition.