EFFECT OF SOURCES, AND METEOROLOGICAL CONDITIONS ON THE CONCENTRATION AND DISTRIBUTION OF GASEOUS ELEMENTAL MERCURY IN THE URBAN ATMOSPHERE

................................................................................................................... iii ACKNOWLEDGEMENTS ................................................................................................ iv TABLE OF CONTENTS ..................................................................................................... v LIST OF TABLES ........................................................................................................... vii LIST OF FIGURES......................................................................................................... viii LIST OF ABBREVIATIONS .............................................................................................. ix 1.0 INTRODUCTION .................................................................................................. 1 1.1 THE RESEARCH PROBLEM ............................................................................................................................ 1 1.2 MERCURY: PHYSICAL AND CHEMICAL PROPERTIES ............................................................................................. 2 1.3 SOURCES OF MERCURY IN THE ENVIRONMENT .................................................................................................. 3 1.3.1 Natural Sources ............................................................................................................................ 4 1.3.2 Anthropogenic Sources ................................................................................................................. 5 1.3.3 Legacy sources / reemissions ....................................................................................................... 6 1.4 MERCURY SPECIATION IN THE ATMOSPHERE .................................................................................................... 6 1.5 TRANSPORT .............................................................................................................................................. 7 1.6 TRANSFORMATIONS AND REMOVAL MECHANISMS ............................................................................................ 9 1.7 HEALTH EFFECTS ...................................................................................................................................... 10 1.8 GLOBAL MERCURY EMISSION AND TRENDS ..................................................................................................... 12 1.9 CANADA’S ROLE IN MERCURY RESEARCH AND EMISSION CONTROL ...................................................................... 13 1.9.1 Monitoring networks .................................................................................................................. 13 1.9.2 Latest Canadian emissions estimates ......................................................................................... 13 1.9.3 Health Canada guidelines for mercury exposure........................................................................ 14 1.9.3.1 Canadian Environmental Protection Act .......................................................................................... 15 1.9.3.2 Mercury legislation in Ontario and Canada ...................................................................................... 15 1.9.3.3 Products Containing Mercury Regulations ....................................................................................... 16 1.10 TRENDS IN ATMOSPHERIC MERCURY FROM NORTH AMERICAN CITIES ................................................................ 16 2.0 MATERIALS AND METHODS .............................................................................. 19 2.1 SITE AND EXPERIMENTAL SETUP .................................................................................................................. 19 2.2 INSTRUMENTATION .................................................................................................................................. 21 2.2.1 Mercury Vapour Analyzers ......................................................................................................... 21 2.2.2 Cleaning and storage of PTFE components ................................................................................ 21 2.2.3 Line setup & generation of zero air ............................................................................................ 22 2.2.4 Ambient line setup...................................................................................................................... 22 2.2.5 Calibration .................................................................................................................................. 23 2.2.6 Validation ................................................................................................................................... 23 2.2.7 Tekran model 2537A maintenance and operation ..................................................................... 24 2.3 WEATHER STATION SETUP .......................................................................................................................... 25 2.4 MATCHING THE ANALYZERS ........................................................................................................................ 25


iv `
ACKNOWLEDGEMENTS I would first like to thank my thesis advisor Dr. Julia Lu of the Faculty of Science at Ryerson University. She was supportive and available for consult whenever I ran into trouble regarding the research and writing, and she consistently allowed this paper to be my own work, but steered me in the right direction whenever she thought I needed it. Her patience, support and guidance are much appreciated.

The Research Problem
Toronto is the country's largest metropolis with an urban population of 2.6 million people. That number grows to 6 million when the surrounding metropolitan area is included (Statistics Canada, 2011). As the population of the city climbs, so too does population density.
Urban sprawl combined with the replacement of buildings with taller ones change the topography of the cityscape, which in turn may affect the air flow pattern in the city canopy hence the concentration and distribution of pollutants. This also means the volume of indoor air also increases. It is known that Hg concentrations in indoor air are higher than outdoor air. These factors together may influence the mercury concentration of outdoor ambient air by the movement of contaminated air from indoors to outside. Outdoor GEM concentrations are also influenced by expanding industries to meet increased demands from a growing population.
Fortunately technological advancements in power generation, waste management, materials manufacture, and chemical & ore processing have come a long way in mitigating Hg emissions.
Meteorology plays a key role in the distribution and cycling of mercury in urban, rural and remote settings although perhaps to different extents. In a rural or remote context, away from the effect of anthropogenic influence, meteorology might have a stronger effect compared to an urban environment because meteorology is the governing factor while in an urban environment, GEM concentration and distribution might be more heavily weighted towards point and regional sources. However, this is not to say that meteorology does not play a role in an urban environment. This study will attempt to characterize the effect of meteorological parameters on the concentration and distribution of elemental mercury within the urban atmosphere.
Policies to control the harmful effects of mercury emissions have been in place at the provincial and national level since 1999 and a measurable decrease in global and national atmospheric mercury levels have been independently verified by various groups, both governmental and academic agencies (NADP, ECCC, UNEP, and others (Butler et al., 2007;Temme et al., 2007;UNEP, 2013). These measurements, however, are usually made in rural and/or remote regions, far from point source emissions in order to obtain a better idea of the global background levels. In an urban atmosphere the background Hg levels are quickly surpassed by anthropogenic emissions from local and regional sources. It has been reported that indoor 2 ` environments might be contributing to the increase in atmospheric Hg (Cairns et al., 2011), and that the changing topography of the cityscape may influence the concentration and distribution of Hg within the city.
Mercury is a heavy metal naturally occurring in the environment. Its presence in the environment poses a complicated problem because of its persistent nature. Being an element, it does not break down or become benign after any amount of time. Once it enters into the environment, it can cycle indefinitely between air, soil and water. Mercury can travel long distances by air because of its long atmospheric residence time, deposit to far-off remote and isolated regions, and collect in bodies of water where it undergoes transformation to form more toxic Hg species. In preindustrial times mercury was not an environmental concern but since its widespread use during the industrial revolution and global transport due to its long atmospheric residence time, it has become a global problem that science and government have been trying to rectify since the 1950s.
In this experiment, outdoor ambient air is sampled from two locations on the Ryerson  Table 1 lists some properties relevant to mercury.   b (Sander, 2015) 1.3

Sources of mercury in the environment
The global atmospheric mercury budget is governed by anthropogenic emissions, natural emission sources, and re-emissions from aquatic and terrestrial surfaces otherwise known as legacy emissions. These each contribute approximately 30%, 10% and 60% of the mercury to the atmosphere annually (UNEP, 2013). It is important to understand these sources, quantify their emissions and build inventories of mercury polluters in order to gauge the ecological costs associated with mercury pollution. Once in the atmosphere, elemental mercury undergoes oxidative processes whereby it converts to an oxidized form (Hg 2+ ), usually taking the form of salts. The oxidized Hg species are much more soluble than the elemental form, allowing for easy removal from the atmosphere via precipitation. Hg in soils generally gets taken up by vegetation or re-emitted as mercury vapour. Hg in water generally gets sequestered in sediments where it 4 ` may remain for thousands of years, or be converted by bacteria into an organic form which bioaccumulates up the food chain. Expansion upon these concepts follows in the rest of the section.

Natural Sources
Mercury is a rare heavy metal found in the crust. It exists naturally in ores such as cinnabar and corderoite and is found is concentrated in geological structures called veins.
Mercury may naturally enter the environment by the weathering of rocks, volcanic eruptions, biomass burning / forest fires, and surface water-to-air transfers ( Figure 1). Settled mercury may also remobilize from soils, vegetation and surface waters and enter the atmosphere once again (Pirrone et al., 2009).
The world's supply of mercury comes from mining, or recovery as a byproduct from various industries, for example mining, smelting and production of other metals (Fe, Al, Cu, Pb, Zn), reprocessing of mine tailings, recycling from consumer products, and private stocks (such as mercury used in chlor-alkali plants) (UNEP, 2002). Typically Canada and the United States do not mine for Hg as a principal commodity but recover it through recycling programs or import it from mercury-producing nations such as China, Russia, Kyrgyzstan, Mexico and Peru (George, 2015;UNEP, 2002).
The United Nations Environment Programme (UNEP) estimates the largest mercury pollution sources on a global level are associated with ASGM and coal burning. Together they account for about 62% of the annual total anthropogenic emissions to air (62% of the estimated 1960 t in 2010) (UNEP, 2013). It is notable that ASGM is not a common practice in modern countries. Also some high-throughput power-producing countries are moving away from coal burning as a means of power generation in favour of renewables. Both practices are being phased out albeit by different motivating factors.
Mercury tends to be higher in an urban environment due in part to some of the abovementioned sources being located in and around cities. It has also been shown that indoor air is higher in mercury compared to outdoors because Hg is greatly diluted once it escapes outdoors (Cairns et al., 2011;Higueras et al., 2014). When mercury containing construction materials and consumer products are used indoors they produce vapours that can be inhaled (in the case of GEM) especially if those materials are disturbed (broken thermometer, thermostat, compact fluorescent light bulb) or absorbed through the skin (in the case of organic and inorganic mercury) by the application of skin-lightening creams, mascaras and eye-cleansing products (World Health Organization, 2011). This liberated mercury can eventually make its way back into the biogeochemical cycle by escaping to the atmosphere. For this reason, buildings are considered a point source of mercury in the context of urban environments.

Legacy sources / reemissions
Legacy emission sources refer to the reintroduction of previously liberated mercury to the biogeochemical cycle through natural or anthropogenic means. An example would be volatilization of Hg from contaminated soils by physical disturbance and/or solar radiation.
Another example would be volatilization of Hg from bodies of water. There are large uncertainties in the estimates for Hg contribution by reemission because the parameters that determine the rate of Hg exchange between air-sea, air-soil and air-vegetation are poorly understood (Pirrone et al., 2013). (1010-4070) tonnes of mercury was released to the atmosphere as a direct result of anthropogenic activity. This accounts for 30% of the total mercury released to the atmosphere that year. Natural sources are estimated to contribute 10% and "legacy emissions" are thought to contribute 60% i.e., mercury that has been recirculated though the atmosphere by re-emission from terrestrial surfaces and aquatic bodies (UNEP, 2013). Zhang et.al (2015) reports that two thirds of global Hg emissions to the atmosphere are due to reemissions based on the GEOS-Chem model, consistent with UNEP's estimate. The United Nations Environment Programme's 2013 report estimates natural and remissions account for between 35-70% of mercury deposition in most regions. Deposition is generally greater in low and mid-latitude regions due to more intense solar radiation, a higher concentration of oxidants, and higher volume of precipitation (Pirrone et al., 2013;UNEP, 2013).

Mercury speciation in the atmosphere
Mercury in the atmosphere is operationally defined as gaseous elemental mercury (GEM), gaseous oxidized mercury (GOM) and particulate-bound mercury (pHg). Total gaseous mercury (TGM) is the sum of the GEM and GOM species. In an urban environment, the distributions of these species are highly variable, especially near an emission source where local concentrations of any or all of the Hg species may be much higher than rural and remote background values.

`
GEM accounts for 95-98% of TGM in low-and mid-latitude rural areas while GOM and pHg make up the rest (Butler et al., 2007;Lin et al., 1999). GEM is relatively stable and quite a bit more volatile compared to its oxidized and particulate-bound counterparts. It also has a very slow deposition velocity, all factors which lead to its long residence time estimated between 6 months to 2 years (Lindberg et al., 2007;Schroeder et al., 1998). Elemental mercury also exchanges between air and surfaces due to its volatility. For the above reasons, GEM is capable of transport on a local, regional and global scale.
The background level of GEM in the northern hemisphere ranges from 1.5-1.7 ng/m 3 , and 1.1-1.3 ng/m 3 in the southern hemisphere. The reason for the difference in values is twofold; more anthropogenic mercury is produced in the northern hemisphere, particularly from Asia, and secondly it takes a long time for GEM to cross hemispheres due to the behaviour of wind patterns at the equator (Kock et al., 2005;Pirrone et al., 2013;Temme et al., 2004).
The Hg 0 to Hg 1+ and Hg 2+ reactions in the troposphere ( Table 2) are known to occur when GEM is in the presence of atmospheric oxidants and have been verified by reactor chamber experiments (Pirrone et al., 2009). The dominant oxidation species in the continental troposphere were thought to be ozone (O 3 ) and hydroxyl radical (·OH). A different trend is observed in the upper atmosphere and marine boundary layers where chlorine and bromine species (Cl, Cl 2 , Br, Br 2 , BrO) are the more dominant oxidants. Iodine and peroxide are also known to oxidize GEM (Lin et al., 2006). Due to the reactivity and high solubility of Hg 2+ compounds, they tend to quickly dissolve into water droplets and return to the ground in precipitation (Lindberg et al., 1998).
Particulate-associated mercury is the third type of atmospheric mercury studied in the science of the atmosphere. As the name implies this species is one that is governed by its association with small particles and aerosols. The particles can arise from natural and anthropogenic sources (e.g., volcanoes, biomass burning, sea aerosol, mineral dust, coal fired power plants, etc.) and are distributed mainly by wind.

Transport
The troposphere is the lowest layer of the atmosphere spanning 18 km in the tropics to as low as 7 km in polar regions. It is the thinnest layer of the atmosphere but it is also the most interesting as this is where most weather phenomena occur (Elliot, 1996). It's also the most 8 ` relevant to pollutant distribution because almost all airborne pollutants travel solely through this layer; rarely, if ever, crossing the tropopause and into the stratosphere. The troposphere serves as the main distribution mechanism by which GEM can travel worldwide. Evidence of global transport can be seen in remote regions where there are no point source emissions but background levels of TGM/GEM are above estimated pre-industrial levels (Pirrone et al., 2009).
There are a few mechanisms that assist in its transport and are discussed below.
Perhaps the most obvious mechanism in assisting pollutant transport is wind. The wind is responsible for the dispersion, mixing and dilution of pollutants through the atmosphere. Point and plume emissions are also influenced by wind patterns. Polluted air masses are able to be traced back to their sources using back-trajectory, dispersion models and meteorological data (Carslaw, 2006;Huang, 2012;Lindberg et al., 2007;Venter et al., 2015). This is particularly difficult in an urban environment however, due to turbulent wind patterns and irregularities in the topography, especially if the sampling sites are not high enough above the city canopy.
Particulate and GOM species are also transported through the atmosphere, albeit on a much shorter distance and timescale. GOM and pHg exists in the atmosphere on the order of hours to days and may travel tens to hundreds of kilometers . GOM is highly soluble and deposits to particles and surfaces or dissolves into water droplets. The captured mercury then deposits to surfaces and various environmental compartments, usually nearby the emission source. For this reason, GOM sources offer an immediate and acute risk to the surrounding environment compared to a GEM source whose effects threaten a long term chronic risk.
Aerosols play a role in the transport of mercury when particles are deposited to the ground or surface waters. Particle size is related to residence time in the atmosphere and therefore affects how far the particles can travel before settling. Ultrafine particles (<0.01 µm) and coarse particles (>20 µm) are removed within 24 hours by different mechanisms. The former are removed by phoretic diffusion to cloud particles and Brownian coagulation while the latter are removed by impaction and precipitation. Particles in the intermediate range (0.01 -10 µm) can remain in the lower troposphere for 1-2 weeks and much longer in the upper troposphere (Hobbs, 2000). Particulate mercury can be measured for a range of size fractions on the condition that it can be collected onto a filter (Pyta et al., 2016). ` Global distillation is the phenomenon whereby substances volatilize in warmer equatorial regions, travel by air and condense in colder, polar regions. This is sometimes referred to as the 'grasshopper effect' (Wania at al., 1995). This applies not only to mercury but any volatile (VOC) and semi-volatile (SVOC) organic compounds. It is due to this process that we observe an accumulation of mercury and POP's in remote places like the arctic where there are no significant pollution sources.

Transformations and removal mechanisms
In the atmospheric environment mercury transformations fall into two main categories: gas-phase oxidation ( Table 2) and aqueous phase oxidation which take place in cloud droplets.
The atmospheric oxidation of gaseous elemental mercury is important in the context of its biogeochemical cycle. Generally, GEM compared to other mercury species is not easily oxidized and is relatively volatile. Oxidation of GEM purely in the gas-phase by atmospheric oxidants is difficult to explain mechanistically and the kinetic and thermodynamic data for reactions between GEM with atmospheric oxidants do not agree. It was thought that the main gas-phase oxidants of Hg 0 are ozone and hydroxyl radicals however, in a re-evaluation of empirical data and mercury models, bromine has been shown to be the dominant atmospheric oxidant (Holmes et al., 2006).
Overnight reactions of Hg 0 with nitrate (NO 3 -) in the atmosphere have also been quantified.
Experiments mostly done in reaction chambers attempt to quantify rate laws pertaining to the reaction of Hg 0 and atmospheric oxidants, radicals, and halogens.
The product of the first 3 reactions in Table 2 yields HgO (s) which precipitates as a small particle. The pHg exists in the fine particulate matter fraction (<2.5 µm) with a low deposition velocity of <0.5 cm/s and low concentration typically <10 pg/m 3 . The slow deposition velocity indicates that it is not a significant component of dry deposition but the HgO (s) particles are susceptible to scavenging by larger particles, aerosols, black carbon, and water vapour / cloud droplets. Because of this, the residence time of particulate-bound mercury is on the order of days to weeks and has been shown to negatively correlate with rain.  ( In the arctic springtime for 3 months after polar sunrise, field measurements revealed episodic drop-offs in elemental mercury that were highly correlated with ozone depletion (Barrie et al., 1998). Ozone was later found to be depleted by reaction with high levels of Br and BrO during this time (Barrie et al., 1998). BrO was also found to oxidize Hg 0 and is the cause for the removal of Hg 0 from the atmosphere by forming HgO (s) . The transformation of GEM to HgO (s) and the subsequent deposition is the mechanism that governs this removal process. It is not entirely certain where the Br is coming from, but leading hypotheses include sea-salt aerosols, sea-salt deposits on snow and new sea-ice surfaces. The settled mercury can volatilize back into the atmosphere driven by more intense sunlight and higher temperatures during warmer periods.
The last type of transformation is classified as biological, in that a living organism is required to metabolize, catalyze or otherwise facilitate the transformation. As mercury enters an aquatic ecosystem and moves through the water column or deposits to the sediment, sulphur-(Desufovibrio) and iron-reducing (Geobacter) bacteria methylate Hg 2+ through the methyl transferase metabolic pathway. The methylmercury [CH 3 Hg] + produced by this pathway is much more harmful than its precursors and is prone to bioaccumulate in animal tissue to levels high enough to cause damage at the top of the food chain Merritt et al., 2009).

Health effects
The common exposures routes for mercury compounds are different for each species.
Elemental mercury has a high absorption of around 70-80% when inhaled as a vapour but negligible for ingestion (0.01%) as a liquid. It leaves the body through urine and feces. Inorganic 11 ` mercury salt absorption varies based on its form and host conditions. It is absorbed most aggressively through the stomach and intestines and leaves the body through urine and feces over the course of weeks to months. Organic mercury has a very high bioavailability and almost completely absorbs through the GI tract when ingested. Dermal absorption is the most severe for this class, especially concerning dimethylmercury where a few drops applied to bare skin can have lethal effects on the human body in less than 1 year. Mercury in this form readily accumulates in organs and crosses the blood-brain barrier. Mercury in this form is slowly excreted in feces over several months (Agency for Toxic Substances and Disease Registry, 1999).
Once elemental mercury vapour or inorganic mercury enters the bloodstream, it is rapidly oxidized by red blood cells via the hydrogen peroxide-catalase pathway to Hg 2+ . Oxidation also occurs by this pathway in tissues such as the brain and liver. Organic mercury is also thought to be converted to Hg 2+ by the hypothesized reaction between alkyl mercury and hydroxyl radicals produced by cytochrome P-450 reductase (Agency for Toxic Substances and Disease Registry, 1999).
It should be emphasized that people do not often find themselves at risk of dangerous levels of exposure to Hg 0 or Hg 2+ . Methylmercury found in fish from contaminated water poses a much larger risk because consumption of contaminated fish leads to accumulation of MeHg in the body. There is virtually no way of knowing beforehand if fish is contaminated and the effects of MeHg build-up in the body won't manifest in symptoms until significant damage has been done.
Lethal dose of organic mercury in humans is difficult to assess because they rely on case studies that are few and far between. The EPA estimates the lethal dose of MeHg is 10-60 mg/kg bodyweight. Victims affected by Minamata disease, aka mercury poisoning from contaminated fish, died from pneumonia, non-ischemic heart attack and non-inflammatory CNS disease (Agency for Toxic Substances and Disease Registry, 1999). Divalent mercury acts on the body by binding to thiol (-SH) groups, such as the one on cysteine. Typically cysteine is associated with protein folding by means of a disulphide bridge. The interaction of Hg 2+ with this amino acid may disrupt the folding of a protein and this can have a wide variety of unintended effects such as inactivation of enzymes, disrupted transport processes, changes in membrane permeability by the formation of thiolates (a salt with structure R-S -M + ), increased oxidative stress, disruption of microtubule formation, increased permeability of blood-` brain barrier, disruption of protein synthesis, disruption of DNA replication and DNA polymerase, impairment of synaptic transmission, impairment of immune response, disruption in calcium homeostasis, and other detrimental health effects (Agency for Toxic Substances and Disease Registry, 1999). Mercury is allowed to cross the blood-brain barrier by mimicking methionine after binding cysteine through the thiol (-SH) group.

Global mercury emission and trends
There does not exist a single unified global monitoring network for mercury but rather numerous regional and national programs and long term monitoring sites. Since the early-to-mid 1990s large decreases in atmospheric mercury have been observed in both the Northern and Southern hemispheres from long-term monitoring stations in Cape Point, South Africa (Venter et al., 2015), Mace Head, Ireland (Ebinghaus et al., 2002), Neumayer, Antarctica (Temme et al., 2003), Mauna Loa, Hawaii (Krnavek et al., 2010) and from cruise ship campaigns over the Atlantic (Soerensen et al., 2012). At the same time, over this period, governments have been heavily regulating mercury emissions with an emphasis on power generation and industrial sources.
The mercury research community spends a great deal of effort to model mercury transportation, transformation and fate in the environment in an attempt to understand the variables and to what extent they affect the ecosystem, be it on a local, regional or global scale. In constructing these models all the steps need to be understood and quantified and herein lies the problem. Uptake rates by aquatic organisms, parameterization of rates of exchange between natural reservoirs, coordinated measurements of tropospheric distribution of GEM, and more accurate and extensive inventories are just some of the missing pieces needed to build better models (UNEP, 2013).
Even with the missing pieces, there are reasonable models that can estimate global emission trends using the current knowledge on mercury movement and transformation, national mercury budgets, natural, anthropogenic, and legacy emissions, and coordinated data from the various networks (section 1.7). Neglecting to account for the missing pieces means the models are over-simplified, especially if incorrect or inaccurate assumptions are made, or if the models are built using inaccurate or incomplete data. Zhang et al. (2015) compiled TGM data from various monitoring networks over 20 years and concluded that observed concentrations of Hg 0 have decreased ~1-2% y -1 since 1990, but global emission inventories, whose data are used to construct 13 ` the models, project flat or increasing emissions over the same period. They resolve this inconsistency by identifying flaws with the current Hg inventories including (i) failure to account for release of Hg from commercial products, (ii) biased estimates of ASGM emissions, and (iii) failure to properly account for changes in Hg speciation in CFPPs after implementation of emission controls for SO 2 and NO x . Their revised model was better able to predict the observed trends in the free troposphere in North America and Eastern Europe but had difficulty with high northern latitude regions, likely due to more complex interactions involving bromine radicals and AMDEs. The Air Pollution Emission Inventory Report tracks 17 air pollutants from sources at the national and provincial level. This information is obtained by a reporting system with which companies producing these pollutants must comply. From 1990 to 2014, the APEI reported total Hg releases to air have decreased an average of 90% over all sectors. This breaks down to a 95% reduction from the industrial sector, 74% reduction from open sources, and a 62% reduction from the non-industrial sector. The largest emission reductions are attributed to changes in ore processing and shutting down of coal-fired power plants (Environment and Climate Change Canada, 2014).

Health Canada guidelines for mercury exposure
Health Canada has set guidelines on exposure to mercury by fish, air, soil, drinking water, and biological materials. In the gas phase alkyl, aryl and elemental/inorganic forms are each given different recommended threshold exposure values ( Table 3). The threat of mercury lies in the transport of mercury compounds from the atmosphere to bodies of water and wetlands and the transformation of these compounds from a relatively benign form into a bioaccumulative and highly poisonous organic form. Aquatic life exposed to 15 ` even low levels of mercury can experience accumulation up the food chain, affecting piscivorous predators, disrupting ecosystem dynamics and posing a threat to human populations that subsist on fish.
Canada Prohibits the manufacture and import of products that contain mercury and/or any mercury compounds.

Canadian Environmental Protection Act
The Canadian Environmental Protection Act (CEPA) came about as the result of an effort to develop a systematic approach for dealing with toxic substances. Prior to 1999, existing legislation was inadequate; a more comprehensive approach was needed to manage the full lifecycle of toxics (cradle-to-grave philosophy). CEPA provided a framework for the management and control of toxic substances at each stage of their cycle: development, manufacture, import, transport, distribution, use, storage, and disposal. CEPA granted federal authority to control ocean dumping through a permit system, regulate waste handling and disposal practices, and establish environmental guidelines & codes of practice (Environment and Climate Change Canada, 2009).

Mercury legislation in Ontario and Canada
Most people living in developed countries are not at risk of high levels of mercury exposure by air because the concentration required to cause harm is quite high, and practices and 16 ` regulations have been put in place that reduce and/or outright ban mercury in the indoor environment ( Table 4).
The Canada-wide emissions for coal-fired electric power generation aims for a 60% reduction in emissions from CFPPs from 1998 to 2010 across Canada. Not all provinces shared the same burden of cutting their emissions based on their individual situation, however, Ontario committed to gradually phase out their 4 CFPPs. In 2014, they met their goal after closing the Nanticoke Generating Station. The Ending Coal for Cleaner Air act was then passed to ensure that no coal would ever be burned in the province starting in 2015.

Products Containing Mercury Regulations
This set of regulations applies to the prohibition of the manufacturing and import of products containing mercury or any of its compounds. Exceptions are made for products at the end of their useful life and intended for recycling or for products that have no technical or economically viable alternative. Examples of some exceptions include medical and research applications and dental amalgam.

Trends in Atmospheric Mercury from North American Cities
Atmospheric mercury monitoring studies have been intermittently conducted in Toronto, Canada. From June 2001-February 2002 found GEM over sidewalk and tree canopy peaked in the fall and was lowest in the winter. Diurnal patterns were found to vary be season. Data from 2004 (Song et al., 2009) found GEM/GOM/pHg to be lowest in fall and highest during the summer months. Both studies also find GEM to be higher overnight compared to daytime hours.
Monitoring studies from other North American cities ( Table 5) demonstrate similar patterns and trends in an urban environment. Two study locations in Michigan compared the GEM from urban Detroit with the nearby rural town of Dexter.  found that GEM was higher at the urban location compared to the rural location due to local and regional sources.
GEM was also found to be higher overnight between the hours of 10 pm and 4 am, agreeing with studies from Toronto . TGM was measured in Houston, the most populated city in Texas, from August 2011 to October 2012. Lan et al. (2014) found the same diurnal variation and seasonal trends observed at other North American cities with a periodic night-time maximum and ` daytime minimum. Maximum TGM measurements were observed in the summer while the lowest were recorded in the winter. It is expected that climate and population density play a role in the increased levels of TGM in an urban environment due to their proximity to industry and power generation facilities. It was shown that the increased ambient Hg in Houston during the warmer months correlated well (0.735, p<0.005) to an increase in energy consumption.
A study by Wang et al. (2013) in Rochester, NY compared a 2-year data record of mercury species in ambient air with a period of time following the decommissioning of a CFPP 12 km to the northwest. Following the closing of the power plant, they observed a median decrease of 12% in GEM overall and a 25% reduction in median GEM from the direction of the power plant. They also found the largest decrease in median mercury species in winds associated with the northwest direction (Wang et al., 2013). A similar study in the Lake Wabamun area of Alberta, west of Edmonton, measured TGM, SO 2 , and NO x from a site within 500 km of a chain of four CFPPs.
Pollution data were analyzed for episodic events of high TGM. High levels of TGM correlated with both NO x and SO 2 and were associated with wind from the direction of the power plants and from the city of Edmonton; northwest and east respectively (Mazur et al., 2009). Both studies made use of tracer gases SO 2 and NO x to attribute high levels of TGM to wind in the direction of the CFPPs, and observed a significant decrease in TGM concentration after the closing of the power plant.

Site and Experimental Setup
In this experiment, outdoor ambient air is sampled from two locations on the Ryerson Loggers) equipped with sensors to measure meteorological parameters (wind speed, direction, gust speed, precipitation, temperature, relative humidity, solar radiation, barometric pressure).
Both the mercury analyzers and weather stations record a measurement every 5 minutes.  In addition to air sampling from the two sites, the experimental area also includes a large skyscraper located to the northwest of KHN (265-295°) and was constructed in the years 2010-2014 ( Table 6).

Mercury Vapour Analyzers
The Hg vapour analyzers (model 2537A, Tekran Inc., Toronto, Canada) used in this experiment employ a dual-channel design allowing for continuous monitoring of GEM. Ambient air is sampled at a rate of 1 L min -1 for 5 minutes while filtered for particulate matter and passed through a quartz tube containing gold-coated beads. The gold serves as a necessary preconcentration step because the instrument would not be able to detect Hg if the measurements were taken in real time due to the real time concentration of GEM in air being lower than the detection limit of the analyzer. The gold cartridge is heated under argon atmosphere to ≈500°C for a few seconds to desorb mercury. Argon serves as a carrier gas to move the analyte through the sample line to a quartz cuvette. A mercury vapour lamp produces collimated UV light at 253.7 nm which collides with mercury atoms in the cuvette and excites an electron in the outer shell (5d 10 ( 1 S)6s 2  5d 10 ( 1 S)6s6p). As the excited electron relaxes back to its ground state, the atom emits energy as a photon of fluorescent light (Ball, 2006). Fluorescent radiation from the electron is detected by a photomultiplier tube. A series of dynodes amplify the intensity of the ejected photons until an anode measures the current produced by the cascade of electrons. The voltage produced is then proportional to the flux of photons from the photocathode and corresponds to a concentration (Ball, 2006).

Cleaning and storage of PTFE components
Before the monitoring, the PTFE components (e.g. unions, tees, tubing, filter housings, fasteners) were soaked in 1:3 HNO 3 overnight to dissolve any elemental mercury or other metal contaminants that may have been present. Nitric acid also dissolved oxidized mercury (Hg 2+ ).

Hg 0 (s) + HNO 3 (aq) -> Hg(NO 3 ) 2 (aq)
The components were then rinsed 5 times with distilled water and 5 times with NANOpure water. Components were left to dry in a fume hood and placed in a plastic zip-lock bag and double bagged for storage. Subsequent handling of the cleaned components was done while wearing nitrile gloves.

Line setup & generation of zero air
Zero-air, that is, air free of mercury, moisture and interfering absorbing species is required for the operation of the analyzer. Zero air was made from the compressed air tap in the lab following the scheme in Figure 3. Air from the tap was first run through a filter to remove the oil produced from the compressor. The air was then put through a column of soda lime (266434, Sigma Aldrich) to remove the CO 2 and other acidic gases, and a column of a silica gel desiccant (SiO 2 ) (85815, Sigma Aldrich) to remove moisture. The pressure was reduced to 20 psi before entering the first of two Hg scrubbers. A particulate filter was attached to the second Hg scrubber to catch any particles that may have leaked from upstream. The pressure was reduced one last time before entering the analyzer to 0.5 psi with an inline regulator. PTFE carrier lines were used throughout.

Calibration
Manual mercury vapour injections were made to the analyzers to calibrate the machines every ~14 days and to validate the calibration after it has been performed. Injections require a gas tight syringe with digital readout (1710N, 25 µL or 100 µL syringe, Hamilton Company), Tekran model 2505 mercury calibration unit and zero-air.
Air that is saturated with mercury vapour follows an empirical relationship governed by temperature. First discovered by Dumarey, the equation calculating the mass concentration of mercury as a vapour at a given temperature is given below (Dumarey et al., 2010). Although the formula was never formally published, it was alluded to in Dumarey et al. (1985). As long as the temperature can be carefully controlled and the container is air-tight, a saturated mercury vapour source can be used as a primary standard for calibrating this instrument. The equation below describes the mass concentration of mercury vapour in a sealed container The calibration unit was set to at least 2°C below the room temperature and allowed to equilibrate overnight. Before making an injection, the syringe needle was left inserted into the calibration unit and the plunger fully extended for 24 hours. When the syringe was not in use, it was left inserted in the calibration unit.

Validation
Following a calibration operation on the 2573A analyzer, more injections were made to verify the accuracy. The model 2537A was set to sample zero air in order to clean the instrument and obtain a baseline reading of zero. Then a series of injections was made at around one third to one half of the calibration volume (e.g. if 30 µL was used to calibrate, 10-15 µL was used to verify).
Three sets of injections were made according to the scheme below to avoid memory effect and contamination (i.e., avoid 6 back-to-back injections).
The model 2537A produced [Hg] = 0.000 ng/m 3 (no peak) when no injection was made. If peaks were observed, zero air was continuously run until no peaks were observed. Small peaks 24 ` <0.100 ng/m 3 following an injection were due to residual mercury either still adsorbed to the cartridge or the septa. When this happened, zero air was continuously drawn through the cartridge and the cartridge heated until no peaks were observed.
The calibration was deemed valid if the difference between the measured concentration and expected concentration was less than 5% and if the difference between cartridges was less than 5%. The analyzer was used for monitoring only after it had passed this test. If it failed, one or both of the cartridges were recalibrated.

Tekran model 2537A maintenance and operation
During the sampling period the machines and data are checked on for general maintenance and data quality. Weekly maintenance for the analyzers include checking for correct time and date, drifting or low baseline voltage, lamp performance indicated by a high baseline deviation (consistently <0.150V), peak status = OK, sample volume = 5.00 ± 0.01 L, argon tank > 300 PSI, and front panel LED indicators.
Biweekly maintenance included calibration and validation, injection port septa changed, outdoor sample inlet and machine inlet filters changed, syringe and needle inspected for damage.
Monthly maintenance included inspecting the nichrome heaters, measuring the UV lamp voltage, syringe maintenance included cleaning the syringe barrel, plunger and needle with nitric acid, DI water and methanol. Other components were maintained or replaced as required (UV lamp, Hg adsorbent canisters, 0.2 µm front-end particulate filters, nichrome heaters), usually on an annual or semi-annual basis. The gold traps were conditioned if they were suspected to be underperforming by heating in ambient air and purged to remove impurities or washing with nitric acid, DI water and methanol. The cuvette was inspected and cleaned by the same procedure if there was reason to suspect a problem.
Each day's values were collected and tabulated for percent difference between cartridges.
If they deviated by more than 5%, they were investigated for possible errors (heater malfunction, improper heating time, reduced cartridge efficiency, leaks in the sample line) and the values were either discarded or corrected to align the offending cartridge with the correct one. This was determined from the calibration data and by the standard operating procedure that assumed the 25 ` higher cartridge (following a valid calibration) is the correct one when compared with an underperforming cartridge.

Weather station setup
A weather station was setup at each sampling location to monitor meteorological parameters. The station consisted of a tripod and data logger with modular sensors. The tripod was affixed with sensors to measure solar radiation, precipitation (rain), wind direction, wind speed, gust speed, barometric pressure, temperature and relative humidity. Each of the sensors continuously monitored these parameters for the duration of the study. A solar panel recharged an onboard battery each day.
The weather station data logger was launched (initiated) by USB connection to a laptop containing HOBOware software. The logger uploaded the sensor data to the manufacturer's server provided there is an active internet connection, otherwise data was manually downloaded from the onboard computer via USB connection. Data were downloaded once a week and inspected for sensor damage or malfunction.

Matching the analyzers
Part of the quality control process involved ensuring the performance of the two analyzers was within the tolerable variation range (5%) throughout the sampling campaign. In order to do this, both machines were placed side-by-side, calibrated and allowed to run while sampling outdoor air overnight from the same inlet. Their performance was verified by calculating the percent difference in observed values between the machines. Before the campaign started, the analyzers showed a disagreement of 4.03% (Figure 4). Following the campaign the difference was measured at 4.77% ( Figure 5). The difference between analyzer measurements was quantified to eliminate the question of machine bias in measurements at either sampling site. It had to be reasonably certain that under identical sampling conditions, both analyzers produced the same data from which the same conclusions could be drawn. Otherwise one could not reliably compare measurements at two sites given that one or both machines are not working properly.
The data selected for analyzer comparison were chosen for a period of relative stability in the baseline because these are representative of the majority of data collected. Machine performance also tended to deviate more greatly at higher concentrations ie. >5% difference at values >3 ng/m 3 . Analyzer difference is 4.03% for n=176. Analyzer difference is 4.77% for n=176.

Data handling
A systematic approach was implemented to process the large amount of data collected from both the weather station and mercury analyzer. Figure 6 describes the general process of collecting, formatting and applying the quality control procedure. The mercury vapour analyzers were required to be connected to a computer for the duration of the data collection period. The manufacturer supplied software automatically generated a text file every day the analyzer was active. Every couple of days the text files were copied to a USB, backed up onto another PC, and put through the data handling procedure. The text file from the analyzer contained the following parameters: date, time, status, active cartridge, user pre-set flag condition, sampling time, volume of air sampled, baseline voltage, baseline voltage deviation, max voltage, peak area, and concentration (ng/m 3 ) (see Appendix A for example).
The raw text files were backed up and manually transferred to a spreadsheet containing a single month's worth of data. Only the relevant data were copied to the spreadsheet while metadata, column headings and calibration summaries were omitted. If the mercury analyzer was stopped for any reason during the sample period, blank space was left in the spreadsheet for the

Weather station data
The files from the weather station were written to a .csv file and downloaded from HOBOlink.com or transferred from the data logger via USB. These data were then manually transferred into a spreadsheet containing a single month's worth of data. Metadata was omitted and blank space was left in the spreadsheet if any downtime was experienced during the month.

Combining Hg & weather data
The data from the weather station must be lined up and matched with the corresponding mercury data. A script in R was written and used to do this (see Appendix B for script). The monthly weather and mercury data are loaded into the R workspace, their columns renamed and then their data frames are fused together. Both the weather and Hg excel files have the same number of cells because they contain the same number of measurements in a given month at the same time resolution. Redundant and unnecessary columns are removed and the combined weather/Hg data frame was exported and saved as a .csv.

Post processing & quality control
Data quality control involved the identification, removal, correction or deletion of erroneous data. The combined Hg/weather files were submitted to an R script for data scrubbing and quality control (see Appendix B for script). Flagging criteria were coded to emulate a modified version of Environment and Climate Change Canada's research data management quality (RDMQ) for GEM. Table 7 presents a modified version of the RDMQ found in Steffen et al. (2012). Data were removed if they violated a condition in the table and in the odd cases where Hg and wind direction are not both present simultaneously i.e. during analyzer maintenance. Observations for weather and Hg were made on a 5 minute interval and each variable was oriented as a column header. GEM values were retained only if the corresponding weather data was also present. GEM values with a missing case for wind direction, wind speed, solar radiation or pressure were excluded. Wind speeds below 0.5 m/s were set to zero to approximate calm conditions and wind direction is split into 36 sectors for polar plots and 16 sectors for the wind rose. GEM below the detection limit 0.1 ng/m 3 were also removed from the data set. Flagging criteria for instrument error were coded to emulate those outlined in Environment and Climate Change Canada's research data management quality (RDMQ) for GEM. Data processing and quality control of the collected data were managed using a script written in R; an open-source programming language and statistical and graphics software environment. The data was analyzed and images produced using the OpenAir package, an add-in useful for analyzing air pollution data, ` developed at King's College and University of Leeds. It is suited for handling meteorological parameters and pollution data. Images can then be produced from the input data and interpreted to give insight into temporal and spatial distributions, anomalous peak events and overall trends (Carslaw et al., 2012).

Data from CAMNet
The Canadian atmospheric mercury measurement network was started in 1994 and has expanded to include 15 sites over 22 years at rural and remote sites across Canada. Most of these sites are classified as rural, rural-affected (as in affected by nearby urban or industrial emissions), or remote. The objective of the network is to further the understanding of the transport, transformation and removal mechanisms of atmospheric mercury and to identify spatiotemporal trends, provide data to validate models and provide data to develop risk assessments (Puckett et al., 2010). Some of the CAMNet sites have since been shut down and only 8 are currently in operation.

`
The CAMNet data is used here to make a comparison between 2004 data from sites in the network with that from KHN collected by Song et. al (2004) in the same year. The latest datasets from those same sites will also be used to compare to the 2015/2016 data collected in this study.
Unfortunately 2010 is the latest available year for data from these sites. Environment Canada has been contacted in an attempt to acquire the 2014 or 2015 data but they are not willing to make the data public before performing their QA/QC.

Statistical Analysis
Normality was tested using Shapiro-Wilk test, qq plots and histograms. Data was determined not to be normally distributed. The data looks more like a log-normal distribution but does not fit into that distribution either. The non-parametric versions of a t-test are Kolmogorov-Smirnov test and Mann-Whitney U. Both were used to determine if mercury measurements between the two sites came from the same statistical population. The difference between the two is that in KS the test compares the cumulative distribution of two continuous variables and work out a p-value based on the largest discrepancy between the two distributions. MWU first ranks the data and works out a p-value based on the difference in mean ranks of the two groups.
Both are used to confirm each other's' conclusion. Rank refers to sorting a numerical variable according to its magnitude, for example the number series [9, 3.14, 5, 0.77] has the ranking [4, 2,

3, 1].
A Kruskal-Wallis test is used when performing analysis of variance when assumptions for normality are violated. This test does not assume the data belong to any specific distribution. KW test is sometimes called 'one-way ANOVA on ranks'. Kruskal-Wallis tests the null hypothesis that all samples are drawn from identical populations (i.e. have the same distribution). It does not give pairwise comparisons between groups. This test was used on seasonal data at a site.
Spearman's rank correlation coefficient (ρ) is used to determine if there are any correlations between GEM received at the sampling sites and the meteorological parameters simultaneously measured. Spearman's rho (ρ) is the non-parametric version of Pearson's productmoment (r). Spearman's rho is used because data from both sites are non-normally distributed.
Spearman's correlation was run between the weather variables and Hg for each subset to see if any correlations existed between events of high Hg and any meteorological parameters. ` After running the QC scripts over the data and combining it into spreadsheets, they were separated by [Hg] cut-offs based on percentiles of 90%, 99% and 99.9%. Correlations were done for both sites for each subset (90th, 99th and 99.9th percentiles). The whole dataset for each site was also treated this way (ie. with no cut-off values) in order to compare them to the peak events.

Annual Trends
It can be seen from The KHN site showed larger variations in GEM. A reason for the discrepancy in variability between the sampling sites is its proximity to an Hg source. Suppose a ground level source was emitting GEM to the atmosphere, be it anthropogenic or reemissions from pre-existing sources, an air parcel containing GEM would be less evenly mixed at low elevation compared to one at high elevation because of a dilution effect that would no doubt take place on its way to the higher site. Support for the variability at KHN is given by the proximity of the inlet to rooftop exhaust vents from chemistry and biology labs. Building and room use have also been shown to influence indoor air concentrations of GEM for example in chemistry labs GEM is higher than other indoor environments (Cairns et al., 2011;Cizdziel et. al., 2011). The indoor-to-outdoor movement of air through ventilation or permeation through cracks and seals may explain the source and be the cause of the variability at the KHN site.
The similar general response to GEM in the atmosphere can be seen in Figure 9 a-c where the hourly, daily and weekly values track to each other but remain separated by ≈0.2 to 0.4 KHN JOR 35 ` ng/m 3 . The arithmetic average for annual GEM measured at KHN and JOR are 1.78 ± 0.89 ng/m 3 and 1.46 ± 0.54 ng/m 3 respectively ( Table 8). The solid line through each plot represents the arithmetic mean value and the shaded region is the 95% confidence interval.  Ar. mean is arithmetic mean and Geo.mean is geometric mean.

Seasonal Trends
On the seasonal level, GEM was highest in the summer at both sites but lowest in winter at KHN (1.55 ng/m 3 ) and lowest in autumn followed closely by winter (1.28 ng/m 3 , 1.30 ng/m 3 ) at JOR (Figure 10). The seasonal mean and median values largely agree at both sites in that they follow a similar pattern (e.g., summer had the highest mean and median GEM, winter mean and medians are lower than spring and summer).
Spring has a lower median and a smaller range compared to summer for both sites. Winter has the smallest range of measurements for all seasons at both sites i.e. smallest standard deviation which agrees with previous studies in Toronto (Song et al., 2009). The values measured in autumn are not consistent between the sites (Figure 10). Autumn values were second highest at KHN but lowest at JOR. At the KHN site, there is no statistical difference between winter and spring and no statistical difference between summer and fall. All the seasonal measurements at JOR are statistically different from each other (see Appendix C). Ignoring autumn, the seasonal comparison between sites have nearly identical trends. The seasonal difference in autumn is due to the missing data from JOR at the beginning of the sampling campaign as can be seen when

Weekly Trend
As shown in Figure 9 a) and c) over the course of a week, GEM peaked on Thursday and decreased over the rest of the week with a sharper decrease over the weekend. The weekend minimums could be attributed to less activity from regional industrial polluters since some may operate at lower capacity or shut down all together over the weekend. Within the city there is a slowdown in construction due to noise bylaws that are in effect on the weekend; construction is permitted on Saturday from 9 am -5 pm and not on Sunday (with exceptions). The decrease over the weekend is especially noticeable at the Jorgenson site where weekend values drop off to ~1.39 ng/m 3 and remain until Monday. One might expect the mercury to quickly rise on Monday and stay relatively constant until dropping off during the weekend. Instead GEM picks up on Tuesday, peaks on Thursday and declines the rest of the week. The boxplot in Figure 11 more clearly represents the smaller variability in measurements on the weekend compared to week days. The 100 th percentiles, excluding outliers, also follow the same trend as the mean and medians with lower values observed over the weekend.

Diurnal Trend
The average GEM at both sites varied diurnally with an evening minimum (~6 pm) and morning maximum (~6 am) (Figure 9 b). This may be influenced by the atmospheric boundary layer; the layer of atmosphere between the ground and the inversion layer. As solar radiation heats the ground, air rises until its density matches that of the surrounding air and mixes. The daytime height is ~1 km and may drop to a few hundred meters at night due to the temperature difference between the ground and air. This is because air cools as it rises, however the rate of cooling changes based on moisture saturation, temperature, and expansion of air with elevation change. At night, the rate of change of air temperature, or lapse rate, is slower compared to the day causing less frequent and weaker turbulent fluctuations of air masses moving past each other.
This leads to a deeper atmospheric boundary layer in the day and a shallower one at night (Colbert et al., 2008;Hemond at. al., 2006;Schneider, 1996). The lower rate of cooling is directly related to higher mercury within the urban canopy because the cooler lapse rate at night slows the upward dispersion of air and, with it, pollutants. In this study, mercury climbs overnight and ` peaks at both sites at or just before dawn. At this point the boundary layer is the lowest and has accumulated the most GEM. After dawn the boundary layer expands and dilutes the pollutants upward.
This pattern of higher night time values has been observed in other urban monitoring studies where a low nocturnal boundary layer allows mercury to accumulate overnight to reach a daily maximum before sunrise, followed by a rapid decline into the midday and afternoon. This scenario applies to an urban setting where local and regional anthropogenic sources dominate over background sources and reemissions (Lan et al., 2014;Song et al., 2009).

Spatial Distribution
Polar plots represent Hg data along with wind speed and wind direction. The data can also subsetted to plot specific time periods (e.g., seasonally, weekly, daily, hourly, diurnally).
Information is represented on 3 axes where the wind speed is indicated by dashed grey circular lines, wind direction is grouped into 10° increments and correspond to wind direction much like a wind rose, and the colour of a cell correseponds to the legend, indicating the value of that cell.
The value contained in the cell can be calculated for mean, median, max, min and weighted mean.
Cells are plotted with a minimum bin size of 1. A "bin" is defined as a region which corresponds to a particular wind direction and speed (Carslaw, 2015).

Annual -mean of GEM by site
During 2015/2016, JOR has the higher wind speed as indicated by the larger range in wind speed measurements ( Figure 12). KHN rarely passed 5 m/s while JOR regularly does, and on average, higher mercury concentrations are measured under conditions of slow wind speeds (1-2 m/s) for both sites. This is observed on the plot at the inner-most rings where the warm colours (red, orange, yellow) are seen.
Higher wind speeds are expected at JOR because the building is higher and therefore subject to stronger winds. This might affect the motion of polluted air plumes passing over the sampling inlet. The higher elevation might also play into lower concentrations if Hg sources were at ground level due to a dilution effect. The higher values observed at low wind speed seems apparent because of a similar reasoning. Under strong wind conditions any pollutants would be dispersed more aggressively over a larger volume compared to under weaker wind speeds. In the 40 ` case that the GEM in the air is homogenously mixed, wind speed should not have any effect on concentration.

Annual -weighted mean of GEM by site
The Figure 13 plot is coloured according to the percent contribution of Hg to the site.
These plots do not give insight to the concentrations observed from any direction, rather it tells us where most of the mercury is coming from in terms of percentage by mass.  The pollutant release inventory for Canada does not account for illegal sources or multiple sources that may cumulatively contribute a significant amount of Hg to the atmosphere. For example crematoriums are estimated to contribute <1% of the global mercury burden via incineration of dental amalgam (UNEP, 2013) but no single crematorium produces enough Hg to be obligated to report their emissions. If there were many crematoria in a small area (there are ~12 or so within the boundary of the map in Figure 14) they might contribute a substantial amount of Hg to the surrounding environment.

Peak Event Analysis
The set of gaseous elemental mercury measurements for each site were analyzed for correlations to meteorological variables. The measurements obtained at both sites were found to be non-normally distributed ( Figure 17) and tests for normality are presented in Appendix C.
Spearman's rank correlations were used instead of Pearson's product-moment.  Table 9 and Table 10). Spearman's rho (ρ) describes the association between two variables. It should be interpreted just like Pearson's r. The correlation coefficient may take on a value between -1 and 1, with -1 indicating a perfectly negative correlation, +1 indicating a perfectly positive correlation, and 0 indicating no correlation at all. The correlation between two variables are deemed significant if p < 0.01. A 2tailed significance is chosen because there is no a priori reason to expect a 1-tailed result.
At the KHN site, temperature is positively correlated with GEM for the whole set (ρ=0.265) but the trend disappears at higher GEM values e.g. 90th percentile (ρ = 0.090), 99th percentile (ρ = -0.118), and 99.9th percentile (ρ = -0.078). As a general trend, temperature and solar radiation have been observed to positively correlate with GEM in some urban and rural studies (Blanchard et al., 2002). Both of these variables accelerate the volatilization of GEM from the ground level at times of peak incident radiation (Poissant et al., 1998). Because of the lack of KHN JOR 46 ` correlation in this study, this suggests that surface reemissions are not a major contributor of GEM to the urban environment. For wind speed at both sites there is a weakly negative correlation indicating higher GEM readings are more prevalent at lower wind speeds. This is supported by the polar plot from Figure 12 where the average GEM is higher for low wind speed.
It would be expected that at high wind speeds GEM would be more vigorously dispersed compared to low wind speeds. Indeed this is what is observed when looking at the whole data set. However, when GEM is subsetted for 90 to 99.9 percentile values, wind speeds do not appear to have any association with the measurements.
The correlation coefficient for Spearman's rank, ρ, is weak and inconsistent for all of the subsetted percentile cut-offs for all variables at both sites, indicating that either the meteorological parameters measured have only a small or negligible influence or that other factors like anthropogenic pollution dominate the GEM in Toronto so much so that it buries the effect of meteorological influence. Other studies have also reported no clear correlation between meteorology and GEM but still find empirical trends between the two (Nair et al., 2012;Shon et al., 2008). `

49
In addition to meteorological parameters, many studies used 'tracer gases' or 'copollutants' to gauge the relationship between criteria pollutants (PM, SO 2 , NO X , O 3 , CO) and mercury. In Blanchard's (2002) study found TGM at Egbert was strongly correlated to SO 2 (R 2 = 0.85 for logarithmic regression) in the winter, which led them to conclude TGM and SO 2 came from the same source; plumes of air from urban center where coal combustion was practiced for domestic heating. In Houston, highly concentrated plumes of THg were found to be related to combustion tracers (CO, CO 2 , NO X ) and high spikes in CO, CO 2 , NO X , SO 2 , CH 4 were usually found to be coincident with high TGM events. It was not determined if combustion tracers and atmospheric mercury are directly related or related by another or some combination of other mechanisms, for example planetary boundary layer height or vertical mixing rates (Lan et al., 2014). In Detroit from 2002-2004 mercury emissions, meteorological variables, and various pollutants were measured over 4 separate sampling campaigns. PCA revealed GEM was associated with a factor that has high loadings for CO and NO X (Lynam et al., 2006). The authors suggest these are indicative of a combustion source and may be related to motor vehicles. The authors also point to Liang et al. (1996) who first measured mercury in gasoline and petroleum products and found diesel, heating oil and gasoline can contain up to 2.97, 0.59 and 3.20 ng/g respectively when extracted with BrCl. However US EPA National Emissions Inventory estimates <1 Mg of Hg is released by mobile sources per year which is less than 1% of the national annual output. Carpi and Chen (2002) presented data from Manhattan in 2000 and also pointed out that traffic may be a cause of elevated mercury in the city but could not associate spikes in TGM values with rush hour spikes motor vehicle traffic. Since Toronto has fewer people now than Manhattan did in 2000, and since motor vehicles have been evaluated to contribute so little Hg to the environment, it stands to reason that they are not a significant source of GEM in this study.

Identifying peak events using maxima polar plots
Another novel way to look at peak events is to plot the max values obtained for each combination of wind speed and direction. Figure 18  was registered at the sampling site as coming from the north. The wind rose plot in Figure 16 suggests that very little wind comes from the south. It seems more likely that local sources are the cause of peak mercury events from directions that are not related to large regional emissions.

Diurnal maxima and weighted average GEM by site and daylight
When sites are plotted for maxima separated diurnally in Figure 19, max values most often occur and are higher during the day. They also tend to not come from directions of regional sources.
This narrows down the sources to daytime GEM pollutants within the city.  night (6 pm -6 am).

Weighted mean GEM by site and daylight
The day/night splitting of the weighted average site data in Figure 20 show a markedly higher GEM contribution to the KHN site at night while the JOR site is relatively uniform regardless of time of day. In contrast to Figure 19 which suggests local sources are   For instance TGM reached maxima in the spring and winter, minima in the early fall (Blanchard et al., 2002;Temme et al., 2007). Sites in close proximity to urban centers or those receiving air masses from the direction of an urban center measured Hg to be higher than other rural sites with the exception of Alert due to different circumstances regarding arctic atmospheric and ocean chemistry. All the studies measured some statistically significant decreases in TGM over the sampling campaign for most, if not all sites. However the rate of decrease varied seasonally and mid-latitude sites showed faster decreases in median TGM compared to arctic sites (Cole et al., 2013). to the site is largely the same over both sampling periods i.e. a high proportion of GEM is observed at low wind speeds. The plots even suggest that under 0.5 m/s winds from the east in 2004 are proportionally higher than from the same direction after the building was built. Suppose Aura building did contribute significant GEM to the sampling site, the polar plot in Figure 15 would show higher proportions of GEM along the 265-295° directions for all wind speeds. This is not to say that the Aura building might not contributing Hg to the outdoor air, just that it was not determined to be a significant or measureable source in this experiment. It is also noteworthy to keep in mind that Hg in indoor air is directly related to Hg sources within the space, ventilation, and regulations pertaining to building materials.

Effect of Aura building
Given that indoor air has been reported to have higher GEM than outdoor air, it was hypothesized that a higher proportion of GEM might come from the direction of the building or that GEM concentrations from that direction might be higher than others. Subsetting the data based on direction revealed very different GEM distributions between sampling campaigns as shown in Figure 22 a)   were so high and have since decreased, and because the Aura building was built during this period of mercury policy reform, it is very hard to tell from the data that was gathered if Aura even produced any mercury that could be measured at a site downwind. If the building does produce mercury, which we can expect to be true because generally indoor Hg levels are higher than outdoor, we were just not able to measure it because it got lost to the background GEM in the city. Huigeras et al. (2014) reported a similar phenomenon whereby high GEM measured in a herbarium tended to quickly disperse and dilute to much lower values. Since urban background levels are already very low and indoor air, while higher than outdoor air, is not orders of magnitude higher, it is reasonable to assume the indoor-to-outdoor exchange of mercury would be minimal and hard to measure at a site 300 m away in an urban atmosphere where there is much turbulence, pollutant dispersion and dilution as is the case with KHN and Aura. Prolonged periods of elevated mercury levels are thought to be associated with regional sources and meteorological influence while short-lived spikes in GEM, especially when they appear simultaneously at both sites are thought to be associated with local sources. These sources, if they exist, are not registered to the NPRI and warrant further investigation.

iii) Kolmogorov-Smirnov test
The KS test is applied to two non-parametric samples to see if they came from the same distribution. It takes the 2nd data set and tests it against the 1st. No assumptions are made on the distribution except that it's non-normal and deals with a continuous variable.