TEMPORAL AND SPATIAL VARIABILITY OF TOTAL GASEOUS MERCURY IN CANADA: PRELIMINARY RESULTS FROM THE CANADIAN ATMOSPHERIC MERCURY MEASUREMENT NETWORK (CAMNet)

Markus Kellerhals*¹, Stephen Beauchamp², Wayne Belzer³, Pierrette Blanchard⁴, Frank Froude⁵, Bruno Harvey⁶, Karen McDonald¹, Martin Pilote⁶, Laurier Poissant⁶, Keith Puckett⁷, Bill Schroeder⁷, Alexandra Steffen⁷, Rob Tordon²

^{1,2,3,4,5,6,7}Environment Canada:

¹Prairie and Northern Region, 4999-98 Ave., Edmonton, AB, Canada, T6B 2X3, *(markus.kellerhals@ec.gc.ca)

²Atlantic Region, 45 Alderney Dr., Dartmouth, N.S., Canada, B2Y 2N6

³Pacific and Yukon Region, #700, 1200 W. 73rd Ave., Vancouver, B.C., Canada, V6P 6H9

⁴Ontario Region, 4905 Dufferin St., Downsview, Ont., Canada, M3H 5T4

⁵Centre for Atmospheric Research Experiments, Egbert, Ont., Canada, L0L 1N0

⁶Québec Region, 100 Alexis-Nihon suite 300, St-Laurent, Québec, Canada, H4M 2N8

⁷Meteorological Service of Canada, 4905 Dufferin St., Downsview, Ont., Canada, M3H 5T4

Abstract - Continuous measurements of total gaseous mercury (TGM) concentration were made at 10 rural sites in Canada. Mean TGM concentrations ranged among the sites from 1.3 to 1.9 ng/m³. The spatially averaged median concentration among all sites was 1.60 ± 0.15 ng/m³. Seasonal variations in TGM concentrations were observed at all sites. At most sites monthly median concentrations were highest in late winter and lowest in fall. Diurnal variations in TGM concentrations were also observed at most sites. The most common pattern of diurnal variability was minimum concentration just before sunrise and maximum concentration around noon. The diurnal variability was seasonally modulated, reaching maximum amplitude during spring or summer. Diurnal variability was affected by local wind patterns and mixed layer development.

INTRODUCTION

The Canadian Atmospheric Mercury Measurement Network (CAMNet) was established with the goal of providing accurate, long term measurements of TGM concentration across Canada. These measurements are being used to improve our understanding of the processes governing atmospheric concentrations of mercury, temporal and spatial variability of atmospheric mercury, and sources and sinks of atmospheric mercury. The measurements will also provide a high quality data set that may be used to validate global mercury models. This paper describes the network and summarizes results to date of the measurements.

EXPERIMENTAL

CAMNet currently consists of 11 monitoring sites (see Table 1) across Canada, chosen to represent major geographical and ecological regions. Each site was classified as either rural-remote or rural-affected. The latter category applies to sites that are expected to be significantly impacted by nearby anthropogenic mercury emissions.

Table 1: CAMNet sites. Results from the first ten sites were included in this study.

Station		Location		Classification	Site surroundings
Site and Code	Province	Lat.	Long.
Alert (ALT)	NU	82.5 N	62.3 W	rural-remote	arctic tundra, permafrost
Kejimkujik (KEJ)	NS	44.43 N	65.21 W	rural-remote	coniferous forest, sparsely settled
St. Andrews (STA)	NB	45.09 N	67.0 W	rural-remote	ocean, settled rural
Mingan (MIN)	PQ	50.27 N	64.23 W	rural-remote	boreal forest, ocean
St. Anicet (ANI)	PQ	45.12 N	74.28 W	rural-affected	settled rural, urban/industry influence
Point Petre (PPT)	ON	43.83 N	77.15 W	rural-affected	lake, deciduous forest, urban influence
Egbert (EGB)	ON	44.22 N	79.78 W	rural-affected	settled rural, urban influence
Burnt Island (BNT)	ON	45.8 N	82.95 W	rural-remote	lake, coniferous forest
Esther (EST)	AB	51.67 N	110.20 W	rural-remote	grassland, pasture, sparsely settled
Reifel Island (RFL)	BC	49.1 N	123.16 W	rural-affected	ocean, urban, settled rural
Kuujjuarapik (KUJ)	PQ	55.28 N	77.75 W	rural-affected	ocean, tundra, town

The standard operating procedures for the network are described in Steffen and Schroeder (1999) (available on request). Each site uses a Tekran 2537A analyzer to measure total gaseous mercury concentration. Each analyzer is calibrated daily, using an internal mercury source. The internal source emission rate is verified by quarterly manual calibrations. An audit (Poissant and Casimir, 1999) of the entire network is conducted annually.

The analyzer precision has been estimated to be better than 2% (Poissant, 2000). Comparison across the network also requires an estimate of inter-site variability, which is caused by two main factors: (1) inter-site variability of the difference between the nominal and actual emission rates of the internal mercury source, and (2) inter-site variability of the difference between the nominal and actual sample volumes. Based on the results of the last two audits a “network error” of ±5% was estimated.

All data were quality controlled using RDMQä, a quality control program that operates within the SASä statistical package. 1997 and 1998 data were used from all stations, except Esther and Reifel Island where 1999 data were also used because of the paucity of pre-1999 data from those sites. Data from each site, originally measured over 5, 15, or 30 minute intervals, were combined to produce time series of hourly average concentration for each site.

RESULTS AND DISCUSSION

Table 2: Statistical summary of TGM concentrations.

Station	N	Mean	Standard Deviation	Median	Minimum	Maximum
ALT	15021	1.55	0.39	1.59	<0.05	3.04
KEJ	15730	1.33	0.25	1.32	0.26	2.69
STA	15866	1.43	0.20	1.42	0.68	2.64
MIN	15530	1.62	0.22	1.67	0.36	3.07
ANI	16665	1.72	0.40	1.67	0.59	9.62
PPT	16390	1.90	0.43	1.83	0.97	8.50
EGB	14536	1.65	0.31	1.65	0.87	6.15
BNT	5229	1.58	0.23	1.56	0.65	3.04
EST	10766	1.69	0.19	1.69	1.11	4.57
RFL	6971	1.69	0.30	1.67	1.14	10.19

An overall average median concentration of 1.60 ± 0.15 ng/m³ for the ten sites was calculated by averaging together the ten site medians weighted by the number of months of monitoring. Table 2 provides a statistical summary of TGM results from each site. The four rural-affected sites had higher mean concentrations (1.65 to 1.90 ng/m³) than the six rural-remote sites (1.33 to 1.69 ng/m³). Maximum concentrations were substantially higher at the rural-affected sites (6.15 to 10.19 ng/m³) versus the remote sites (2.64 to 4.57 ng/m³). Concentrations were also more variable at the rural-affected sites than the remote sites. The one exception is Alert, a remote site that also had high variability. High variability at Alert was primarily due to springtime episodes of mercury depletion (Schroeder et al., 1998), during which TGM concentration dropped to very low levels. Besides Alert, relatively low minimum concentrations were also observed at the three easternmost sites (Mingan, Kejimkujik and St. Andrews). It has been hypothesized (Poissant, 2000) that reactive chlorine species in the marine air at Mingan might contribute to additional nocturnal depletion there. The north-south TGM gradient observed in Europe (Schmolke et al., 1999) was not observed in this study. Mean median concentration for the nine mid latitude sites (1.61±0.16 ng/m³) was not significantly different from the median concentration at the arctic site (1.59 ng/m³).

Figure 1: Probability distribution of TGM concentration at CAMNet sites.

Figure 1 is a box-whisker plot illustrating the probability distribution of TGM concentration at each station. The four rural-affected sites all had right-skewed probability distributions that approximate log-normality. A log-normal

distribution is the theoretical distribution of pollutant concentrations for a passive (no sources or sinks after being emitted) pollutant subject to successive random dilutions (Ott, 1995). Given the long atmospheric lifetime of mercury, its behavior might approximate a passive pollutant on the time scale of transport from regional or local sources. Four of the rural-remote sites (Esther, Burnt Island, St. Andrews and Kejimkujik) had roughly symmetrical probability distributions. At these remote sites multiple processes such as chemical depletion, surface deposition, volatilization from surrounding surfaces and occasional transport of mercury rich air from source regions may combine to create a roughly normal distribution. Two of the rural-remote sites (Alert, Mingan) had strongly left-skewed probability distributions. At Alert this distribution arises because TGM is fairly constant through fall and winter, and is then subject to large depletion events in spring. Mingan appeared to have a somewhat similar pattern of relatively constant levels interrupted by periods with significant depletion events, though at Mingan the depletions occur on a diurnal rather than seasonal cycle.

Figure 2: Seasonal cycle of TGM concentration.

To investigate seasonal variability of TGM concentration, median monthly concentrations were calculated for each site and normalized to a common annual average. The resulting normalized seasonal cycle for the nine mid-latitude sites is shown in Figure 2. The seasonal cycle at Alert, distinctly different from the mid-latitude sites, is shown separately in Figure 2. Overall highest mid-latitude concentrations were observed in winter and spring, and lowest concentrations in summer and fall. The seasonal pattern was relatively consistent between sites in fall and winter (s between 0.07 and 0.13 ng/m³), while in spring and summer there was greater inter-site variability (s between 0.13 and 0.18 ng/m³). At two sites highest monthly median concentrations were actually observed in the late spring or summer months. The magnitude of the mean seasonal cycle was 0.29 ng/m³, or 18% of the network median concentration. Seasonal cycles in TGM observed elsewhere are somewhat similar. For example, measurements of TGM at Wank Summit in Germany over seven years (Slemr and Scheel, 1998) found highest concentrations in February-April and lowest concentrations in December-January.

Figure 3: Mean hourly TGM concentrations at CAMNet sites.

Diurnal variations in TGM concentration have been reported in several studies (Schroeder and Munthe, 1998). The mean diurnal variations in TGM concentrations for each site are shown in Figure 3. All sites showed evidence of diurnal variability. The majority of sites experienced, on average, a diel cycle of maximum concentrations near solar noon and minimum concentrations just before sunrise. A major exception to this pattern was the Reifel Island site, which experienced maximum concentrations in the early morning and minimum concentrations in the evening which are likely a reflection of local land/sea breeze dynamics.

At six of the sites there was significant correlation between the amount of nocturnal TGM depletion and the day-night temperature difference. Six sites also showed significant anti-correlation of nighttime wind speed and nocturnal TGM depletion. Both these results are expected if the degree of nocturnal depletion is dependent on the strength of the nocturnal inversion; strong day-night temperature differences being associated with strong nocturnal inversions, and strong nighttime winds tending to weaken the nocturnal inversion (Oke, 1978)

Diurnal variability was seasonally modulated at all sites. In general the greatest mean variability was observed in the summer months. Figure 4 shows the seasonal variation of the diel cycle at Esther, a fairly typical case. At Esther the amplitude of the mean diel cycle during summer was approximately 10% of the mean TGM concentration, while in winter the cycle was largely absent. Other sites showed an even more dramatic seasonal effect. For example Point Petre changed from a diurnal minimum in spring to a diurnal maximum in summer. There are several possible reasons for a stronger diel cycle in summer including: stronger diurnal variations in insolation, greater day/night temperature difference, greater role of uptake/emission by vegetation, increased volatilization from nearby water bodies, increased effect of nearby water bodies on local wind patterns and on the evolution of the mixed layer, and temperature dependent chemistry. Further site-specific analyses are being undertaken to elucidate under what conditions these various processes are active and important.

Figure 4: Seasonal modulation of diel cycle of TGM concentration at Esther

REFERENCES

Oke, T.R., (1978). Boundary Layer Climates, Methuen, London.

Ott, W.R., (1995). Environmental Statistics and Data Analysis. CRC Press, Boca Raton.

Poissant, L., (2000). Total gaseous mercury in Québec (Canada) in 1998. Science of the Total Environment, (in press).

Poissant, L. and Casimir, A., (1999). Audit Protocol for Total Gaseous Mercury Measurements (Version 4 /March 1999), Environment Canada, Montreal.

Schmolke, S.R, Schroeder, W.H., Kock, H.H., Schneeberger, D., Munthe, J., and Ebinghaus, R., (1999). Simultaneous measurements of total gaseous mercury at four sites on a 800 km transect: Spatial distribution and short-time variability of total gaseous mercury over central Europe. Atmospheric Environment, 33, 1725-1733.

Schroeder, W.H., Anlauf, K.G., Berg, T., (1998). Arctic springtime depletion of mercury. Nature, 394, 331

Schroeder, W.H. and Munthe J., (1998). Atmospheric mercury - an overview. Atmospheric Environment, 32, 809-822.

Slemr, F. and Scheel, H.E., (1998). Trends in atmospheric mercury concentrations at the summit of the Wank Mountain, southern Germany. Atmospheric Environment, 32, 845-853.

Steffen, A. and Schroeder, W.H., (1999). Standard Operating Procedures Manual Procedure for Total Gaseous Mercury Measurements - Canadian Atmospheric Mercury Measurement Network (CAMNet), Environment Canada, Toronto.