UNMIX Methods Applied to Characterize Sources of Volatile Organic Compounds in Toronto, Ontario

UNMIX, a sensor modeling routine from the U.S. Environmental Protection Agency (EPA), was used to model volatile organic compound (VOC) receptors in four urban sites in Toronto, Ontario. VOC ambient concentration data acquired in 2000–2009 for 175 VOC species in four air quality monitoring stations were analyzed. UNMIX, by performing multiple modeling attempts upon varying VOC menus—while rejecting the results that were not reliable—allowed for discriminating sources by their most consistent chemical characteristics. The method assessed occurrences of VOCs in sources typical of the urban environment (traffic, evaporative emissions of fuels, banks of fugitive inert gases), industrial point sources (plastic-, polymer-, and metalworking manufactures), and in secondary sources (releases from water, sediments, and contaminated urban soil). The remote sensing and robust modeling used here produces chemical profiles of putative VOC sources that, if combined with known environmental fates of VOCs, can be used to assign physical sources’ shares of VOCs emissions into the atmosphere. This in turn provides a means of assessing the impact of environmental policies on one hand, and industrial activities on the other hand, on VOC air pollution.


Introduction
Principal human sources of volatile organic compounds (VOCs) include refining of fossil fuels, distribution of fuels, their combustion, and their usage in industries and consumer products. Most of the VOC species released into atmosphere contribute to ambient air pollution. Several are known carcinogens and many become toxic at higher concentrations or with prolonged exposure; see the toxicity list in reference [1]. References [2,3] provide toxicological and environmental-fate information for the compounds we investigate, with the most common pollutants having additional documentation.
Air pollution caused by urban traffic requires our constant attention: the ever-growing bulk of traffic may render ineffective the measures implemented with regard to limiting air pollution in large cities with heavy metals and other carcinogenic substances [24]. The concentrations of traffic pollutants form a recognizable hourly pattern, but the daily averages create a ubiquitous background of industrial air pollution. The determination of concentrations is complicated by atmospheric photochemistry when it comes to remote sensing of sources. Complex processes-chemical, physical, and biological-affect the environmental fate of VOCs. In particular, solar radiation converts the more reactive VOCs by oxygenation and through reactions with oxides of nitrogen to secondary VOCs, free radicals, ozone, and, subsequently, smog [25][26][27][28][29][30]. Secondary sources of ambient VOCs arise from contaminated water, sediment, landfill, contaminated urban soil, and waste-treatment facilities [31][32][33][34][35]. These secondary sources have the potential to contribute to the steady accumulation of substances and prolonged releases into the atmosphere [36,37]. Estimates of this pollution are compulsory for planning effective measures against the most widespread and persistent pollution.
The large number of organic substances contributing to air pollution is a challenge for air quality management, but it also allows for crafting intricate chemical fingerprints of putative sources by modeling ambient VOC multi-receptors operating in air quality monitoring stations. Some of these modeling methods consist of principal component techniques, mainly Positive Matrix Factorization (PMF) and Chemical Mass Balance (CMB) [38][39][40][41][42][43][44][45][46][47][48][49]. In such methods, the chemical characteristics of sources are constructed in the form of emission profiles representing the share each species has of the total concentration measured at a station. These emission profiles are assigned to the user-specified inventory of sources (in the CMB system), or to putative sources (in the PMF approach).
Complex atmospheric chemistry makes it a challenge to build a reliable receptor model [50][51][52][53]. The task is complicated not only by removal processes (photochemical degradation, aerosol formation), but also by intermedia transport and deposition [53][54][55][56][57][58]. This research addresses these issues by developing robust source discrimination routines searching for reliable information about VOC emissions, even if approximate or incomplete. This is achieved by applying the principal component methods implemented by the U.S. Environmental Protection Agency (EPA) in UNMIX [59,60]. UNMIX has the advantage of telling the user which modeling attempts cannot be successful, thus narrowing the output to reliable (even if partial in the sense of a limited VOC menu) receptor models. More extensive source profiles are built from vast libraries of restricted models created for the given receptor.

VOC Measurements and Data
In this work, ambient VOC concentrations-expressed in micrograms per cubic meter (µg/m 3 )-are examined. The concentration data are from the National Air Pollution Surveillance (NAPS) Canadian federal program [61]. NAPS sensors at air quality monitoring stations across Canada have performed every six days since 1989 hourly measurements of 175 species of VOCs, including hydrocarbons (from methane up to the C12 fraction), halogens, aldehydes, and ketones. The sensors operate based on a gas chromatography/flame ionization detector system (GC/FID) for quantification of light hydrocarbons in the C2 fraction, while a combined gas chromatography/mass selective detector (GC/MSD) system operating in selected ion monitoring (SIM) mode is used for quantification of the heavier hydrocarbons, in the C3 to C12 fractions. We use data obtained at four monitoring stations in Toronto, Ontario (see Figure 1); for sensor modeling purposes, we only utilize the daily averages of VOC concentration data acquired in years 2000-2009. Station #60413 is located in a park close to the international airport in Toronto. Stations #60418 and #60427 are situated in downtown Toronto. Station #60429 is located close to a large body of water (Lake Ontario) and nearby highway 401, farther from the industrial districts.
Many VOCs may have incomplete measurement data due to malfunctioning at one or more stations. For this research, species that missed more than 5% of their daily averages (within the study period) were removed from the affected station's data. The number of VOCs adequately measured and processed for each station is around 100. We also examined the VOC data acquired at the same NAPS station in 1991-1999 in order to observe general trends in the temporal variations of ambient VOC concentrations.

VOC Data Processing Method
The U.S. EPA makes available an application for processing VOC data, the UNMIX routine, through their website. UNMIX is a principal component method, but is based on geometrical analysis of the measurement dataset [59]. The routine performs linear modeling of many variables, with a few expressly constructed predictors named "source profiles". As usual, the profiles give chemical descriptions of putative sources or, precisely, apportionments of VOCs to those sources. UNMIX deduces the number of sources from the geometrical properties of the VOC data captured by a sensor. The application ensures that the source profiles represent physical sources and not only one of the large number of mathematically correct solutions to the modeling problem. Thus, UNMIX results are robust: the application fails to give results when a particular profiling cannot be made reliable.
The UNMIX routine does not constrain profiles to positive values; a profile, which is supposed to represent emissions of various gas-phase species from a source, is expected to be implicitly positive. However, persistent negative values appear in UNMIX models. Scrutiny of a large number of UNMIX models leads to a conclusion that the "negative emissions" hint at VOC depletion in the atmosphere and that the initial emissions from the "positive sources" are evaluated correctly (then their total percentage contribution may exceed 100%).
The success of the UNMIX profiling essentially depends on the collection of VOC species fed into the routine. Basically, no more than 30 VOCs can be fed into the routine at a time. Thus, one run can provide only a "partial reliable" profile. By combining results from many successful runs on various VOC menus-thus combining partial reliable profiles of diverse but limited VOC rangesreliable and exhaustive chemical descriptions of VOC sources can be produced. Thereby, modeling the 175-VOC multi-receptor required running UNMIX several hundred times in order to produce a sufficiently exhaustive library of reliable partial profiles. For a given monitor, up to 700 profiles were created to ensure that each correctly measured VOC was accounted for in a successful profiling.
When two partial profiles are defined using different lists of VOC species but characterize the same source, they can be spliced into one larger profile characterizing the source with VOCs from

VOC Data Processing Method
The U.S. EPA makes available an application for processing VOC data, the UNMIX routine, through their website. UNMIX is a principal component method, but is based on geometrical analysis of the measurement dataset [59]. The routine performs linear modeling of many variables, with a few expressly constructed predictors named "source profiles". As usual, the profiles give chemical descriptions of putative sources or, precisely, apportionments of VOCs to those sources. UNMIX deduces the number of sources from the geometrical properties of the VOC data captured by a sensor. The application ensures that the source profiles represent physical sources and not only one of the large number of mathematically correct solutions to the modeling problem. Thus, UNMIX results are robust: the application fails to give results when a particular profiling cannot be made reliable.
The UNMIX routine does not constrain profiles to positive values; a profile, which is supposed to represent emissions of various gas-phase species from a source, is expected to be implicitly positive. However, persistent negative values appear in UNMIX models. Scrutiny of a large number of UNMIX models leads to a conclusion that the "negative emissions" hint at VOC depletion in the atmosphere and that the initial emissions from the "positive sources" are evaluated correctly (then their total percentage contribution may exceed 100%).
The success of the UNMIX profiling essentially depends on the collection of VOC species fed into the routine. Basically, no more than 30 VOCs can be fed into the routine at a time. Thus, one run can provide only a "partial reliable" profile. By combining results from many successful runs on various VOC menus-thus combining partial reliable profiles of diverse but limited VOC ranges-reliable and exhaustive chemical descriptions of VOC sources can be produced. Thereby, modeling the 175-VOC multi-receptor required running UNMIX several hundred times in order to produce a sufficiently exhaustive library of reliable partial profiles. For a given monitor, up to 700 profiles were created to ensure that each correctly measured VOC was accounted for in a successful profiling.
When two partial profiles are defined using different lists of VOC species but characterize the same source, they can be spliced into one larger profile characterizing the source with VOCs from both lists. A measure of similarity between two profiles must be elaborated heuristically. Considering a profile a function on a domain of VOCs, the notion of distance between two profiles can be built involving the two profiles' values exhibited on the overlap of their domains. Also, as experimentation shows, the smaller the VOC concentration is, the less reliable its source apportionment. Therefore, a realistic measure of distance between profiles must also involve weighing with average VOC concentrations.
With such premises, two profiles could be considered describing the same source only if the overlap of their VOCs was large enough to cover at least half of the shorter profile. This usually signified 15 or more common VOCs. Next, the distance of the two profiles was defined as the normalized weighted sum of distances between the profiles' percentage values assumed on their common domain (on all shared VOCs). See ref. [62] for a more detailed description of the measure of distance between two profiles.
For each of the NAPS monitoring stations #60413, #60418, #60427, and #60429, several source profiles have been created from data captured at the station. Each profile was assembled by connecting several partial profiles produced by UNMIX, as described in the previous section. When searching for partial reliable profiles that extend a profile under construction, the possibility of forking-attaching partial profiles for two different sources-exists. This happens because the profiling fuzziness and error margins are often larger than separations-in any realistic metric-between the locally observed profiles of different distant sources. These bogus splices could be occasionally exposed by confronting models for different monitors. However, a fairly efficient method of eliminating the ambiguous connections consists in attempts to join together, repeatedly, different permutations of a collection of partial profiles-and getting a recurrent result. We report only the infrequent results of profiling attempts that were insensitive to permutations.
Actually, the results can be organized into two clearly distinguishable groups of profiles. The first group features long profiles of about 100 VOCs describing typical hydrocarbon urban air pollution, or the "inert background" of gases permeating the atmosphere, or else specific industrial emissions, showing several larger values and many small values implying an absence of emissions. Such a profile also has blank entries for VOCs not conceding to a reliable analysis of their sources.
The second group is made of up to several dozen shorter profiles (bringing out 30-50 VOCs) showing numerous negative values. Arguably, the negative values are attempts at describing the complicated dynamics of appearance and removal of VOCs in the urban atmosphere-the conspicuous processes of photochemistry and climate effects blurring the picture [66,67]. Here, as in the first group, the absence of a VOC in a source profile does not signify that the VOC is foreign to the source. It rather signifies that, for some VOC menus, UNMIX was unable to quantitatively evaluate the emissions from "positive" putative sources and the absorptions by "negative" alleged sources. An analytical overview of the short profiles is not given; at this stage of research they elude analysis.
Negative and higher than 100% percentage points also appear in the profiles we describe here in some detail. For the most prominent VOCs, if they possess values, possibly negative, reliably assigned to all the sources "visible" at a station, the values add up to roughly 100%. This signifies that UNMIX was able to estimate the depletion of some VOCs travelling from their sources to the sensor.
When describing a compound marking a point source or characterizing a distributive source, we accompany the compound's name with its Compound Identification Number (CID number). This can be used to access information about the compound on the Internet; see PubChem portal in the United States' National Center for Biotechnology Information [68]. The Chemical Abstracts Service (CAS) registry number can be used as well to access the compound's data via chemistry portals. Also, references may provide information about the compound's production or its industrial applications. We tabulate this information about compounds; it has the potential for providing leads to point sources of the ambient VOCs in Toronto.

Motor Vehicle Exhaust and Automotive Evaporative Emissions in Toronto, Ontario
Pollution associated with urban traffic forms an intense background to the pollution from industrial sources and from the usage of industrial goods. All C4-C8 aliphatic and aromatic alkanes (and some alkenes, even an alkyne, the 1-propyne) appear in Toronto's ambient air and have associations with traffic. Traffic source scarcely emits any aldehydes or ketones. Halogenated species are rare, while C2 to C12 hydrocarbons-aliphatic and aromatic-appear often. Actually, the traffic source is made of two sources: "exhaust"-which includes fuel combustion products and unburned hydrocarbons-and "evaporative emissions", which include evaporation from moving and parked vehicles, from refueling operations, and losses from the usage of Liquefied Petroleum Gases (LPG): propane, butane, propylene, and butadiene (1,3-butadiene).
In Toronto, the two sources are clearly visible; UNMIX consistently produces two distinct profiles. This is compatible with reports from many other places affected by dense traffic [67,[69][70][71][72][73][74][75]; however, the UNMIX traffic profiles have many blanks (uncertain, but not null, values), particularly when the monitoring station is situated far away from the intense urban traffic. Indeed, the traffic sources appear to be strongly affected by factors such as weather and landscape. Nonetheless, evaporations and exhaust gases coexist in the atmosphere, but their contributions to air pollution in the presence of petrochemicals or sources of natural gas are relatively small. In fact, after petrochemicals and natural gas, evaporative emissions may be the chief environmental concern for VOCs in many large cities [41,42,47,62]. The city of Toronto does not seem affected by petrochemicals or natural gas mining; still the concentrations of evaporation and exhaust gases appear lower than those reported elsewhere. The presence of LPG is significant; indeed, the "evaporative" profile may be distinguished from the "exhaust" profile just by the strong presence of LPG. Also, the profile of the evaporative emissions mostly consists of light hydrocarbons (C4 to C6) and it contributes more than exhaust to urban VOC pollution now that vehicle exhaust systems have improved [76][77][78][79][80].
Among the four constituents of BTEX (benzene, toluene, ethylbenzene, and xylenes), benzene is not markedly present in exhaust or in vehicular evaporations in Toronto, which apparently is due to the significantly increased usage of reformulated gasoline in Canada and the USA [64,81,82]. The three other BTEX components abound, particularly toluene showing the highest concentration in the air among all measured VOC species; in the urban air in Toronto it is about three times more concentrated than benzene (see also [83][84][85]). There also appears to be an omnipresent and strongly correlated pair: o-xylene and ethylbenzene (cf. [86]).
The BTEX constituents are also used extensively in manufacturing processes; in particular, ethylbenzene is used in the production of an important HPV chemical-styrene, which makes ethylbenzene a common pollutant [87,88]. However, apportionment of BTEX amid point sources could not be assessed with our method of remote sensing and modeling.
The traffic sources contribute to the city-wide presence of straight-chain alkanes, from ethane and propane up to undecane and dodecane. The presence of LPG in the traffic profiles indicates an "evaporative" source. Here, the "evaporative" concentrations of the alkenes decrease when the carbon number increases. The "exhaust" source is more noticeable for heavier hydrocarbons, in fractions C8-C10 (cf. [89]). Table 1 describes the traffic sources by their typical VOC menus.

Inert Gases and Stable Background
Great restrictions have been imposed since 1980 in the use of chlorofluorocarbons [20,105-108]freon-11, freon-12, freon-113, freon-114-as refrigerants and propellants. They were replaced by the less ozone-depleting hydrochlorofluorocarbon, freon-22 (CID 6372). Despite the restrictions, the freons still persist in the urban environment in two distinct constituents of ambient air, forming a "stable background" of the atmospheric pollution. These are extremely slow reacting gases (the half-lives of the reactions with hydroxyl radicals are measured in years). The stable background also features halogens (chloroform, chloromethane, bromomethane, dibromomethane, and dichloromethane) and the ozone-depleting carbontetrachloride; these are slowly reacting gases (the half-lives of the reactions with hydroxyl radicals are measured in months). Besides this, many more reactive hydrocarbons exist in the stable background; it appears that they persist in the atmosphere in equilibrium between the atmospheric degradation and the renewal processes associated with slow evaporation from contaminated waters and soils.
UNMIX consistently creates two profiles featuring freons and other inert gases. The first, well-defined "inert" profile describes a uniformly distributed source featuring the inert gases and many light hydrocarbons, including ethane, ethylene, and propane; see Table 2. The second, "stable" profile features besides the inert compounds numerous alkanes of industrial provenance, sometimes including heavier hydrocarbons such as dodecane. Traces of 2,2-dimethylpropane (isomer of pentane), 2,2-dimethylpentane (isomer of heptane), 2,4-/2,5-dimethylhexane (octanes), and cis-1,3-dimethylcyclohexane (a cyclohexane) appear to persist in the "stable" background as sequels of the fugitive releases of pentanes, heptanes, octanes, and cyclic hydrocarbons from petroleum industries and fuels. (Still, alkene, like 1-Methylcyclohexene, has no discernible source.) The "stable" profile is poorly defined and less distributed in the environment: it adopts local pollution features (see Table 3; the table does not show station #60427, where profiling of the "stable" source was unreliable).  Table 4 describes the inert gases persisting in the inert and stable background air pollution (the hydrocarbons occurring in the inert/stable background are described in the "traffic" Table 1 and/or the "industrial" Table 5).

Isobutane
Fugitive releases from industrial processes, but primarily production and usage of high-octane gasoline (see Table 1) Refrigeration systems, propellant for aerosol cans and in foam production (as an ozone friendly gas) Reacts with hydroxyl and nitrate radicals; half-lives are a week and half a year, resp. 6360 [95] 2-methyl-2-butene Nearby point sources, but mostly traffic (see Table 1) Intermediate in chemical industry Indirect photolysis mediated by hydroxyl radicals and ozone, with a half-live at most some hours 10553 [96] Pentane Evaporations from solvent and refrigerant blends, manufacturing of petroleum (see Table 1) General laboratory solvent, polymerization reactions, primary blowing agent in the production of polystyrene foam, refrigerant Reacts with photochemically produced hydroxyl radicals, half-life 3 to 4 days. [31,95,98] m and p-Xylene Releases from printing, rubber, plastic, and leather industries Precursor to terephthalic acid and dimethyl terephthalate, monomer in polyethylene terephthalate (PET) Reacts with hydroxyl radicals, degrades by half in less than 20 h.

[97]
Propylene Fugitive releases from ethylene production , petrochemical industry, and numerous polymer, plastics, and metalworking industries Chemical and plastic products: detergents, automotive brake fluids, fibers, polyurethane for foams, films" ABS resins, automotive trim parts, also production of polypropylene, acrylonitrile, propylene glycols, cumene, butyraldehydes, and acrolein Degrades in the atmosphere by reaction with hydroxyl radicals with a half-life of 15 h. 8252 [120,121] Cyclohexane Escapes from production of nylon precursors, but mainly exhaust and evaporations from fuels (see Table 1)

Solvent in chemical industry , intermediate in production of adipic acid and caprolactam
Degrades in the atmosphere by reaction with hydroxyl radicals, half-life of about 2 days. 8078 [93,97,98,122,123] Cyclopentane Escapes from the manufacture of synthetic resins, rubber adhesives, polyurethane insulating foam, pharmaceuticals but mainly evaporations from fuels (see Table 1) Domestic appliances (replacing freon-11), advanced lubricants (extremely low volatility), pharmaceutical products.

Polymer, Plastics, and Metalworking Industries
The city of Toronto and its surrounding area produces more than half of Canada's manufactured goods. The polymer and plastics industries, which mostly consume petrochemical products, release many of the VOCs already seen in traffic profiles. For example, the ethylbenzene present in the traffic pollution is an HPV alkylbenzene important in the petrochemical industry [6], but its derivative, styrene (also an HPV), is characteristic of the polymer and plastics industries [88,89].
Another abundant species is isoprene (yearly emissions of isoprene by vegetation are about equivalent to total emitted methane and accounts for~1/3 of all hydrocarbons released into the atmosphere [116]); however, in urban air, anthropogenic isoprene from industrial sources and fossil fuels may dominate [117,118].
Many species usually reported as solvents and intermediates can provide clues leading to local industrial sources. The following compounds allow for making informed guesses about their provenance when considering apportionments of putative industrial sources versus dominant traffic sources.
Tables 6-9 describe point sources "visible" at different stations in Toronto. The visibility strongly depends on the distance between the source and the station. A clearly visible source has a well-defined profile (a limited number of blanks) on the whole large menu of about 100 VOCs that usually appear. For large number of VOCs the profile is close to 0, pointing to an absence of emissions of VOCs from the source; the tables may not show the absent VOCs.
The point-source profiles are labeled A, B, C... and are shown in bold-faced columns, together with the background traffic pollution, i.e., with the distributed-source profiles "Evaporation" and "Exhaust" detected for the same monitoring station. The tables also show average ambient concentrations (in µg/m 3 ) of the VOCs.
There are point sources that can be seen from two monitoring stations. The more distant station provides a rather incomplete profile. In Table 6, source A is detected at stations #60413 (good visibility) and #60418 (bad visibility). At station #60427 source A is not detectable (neither at station #60429, not shown). It can be seen that the definition of a traffic profile is clearer when point sources are not in sight. Table 7 shows point sources B and C also detected at station #60413; they are not visible at other stations.
At station #60418, there are two point sources clearly visible and one many-VOC hardly visible and therefore poorly defined, arguably more distant, source. The two well-defined sources are shown in Table 8.
At station #60427 two many-VOC sources can be detected and partially defined; judging by their VOC menu, they are industrial sources and one of them has been seen from station #60418. The other seems to also appear at station #60429. However, the partial definitions do not yield unambiguous conclusions. By contrast, the traffic profiles seen at station #60427 are very well defined and catch almost all of the ambient hydrocarbon load. Station #60429 appears to feebly "see" an industrial source already detected at station #60427. The feebly defined profiles from stations #60418, #60427, and #60429 are shown in Table 9.
Station #60429 also detected a strong concentration of styrene (about three times stronger than elsewhere); all of the styrene was assigned to one very well-defined source (not tabulated), which also was emitting trace amounts of 1-hexene and n-propylbenzene-and nothing else.

Food Industries
There are more than 500 food companies in Toronto; however, the flavoring and fungicidal agents cis-3-methyl-2-pentene (CID 643935) and cis-4-methyl-2-pentene (CID 5326159) only occur in some profiles in trace quantities. The occurrences of dichloromethane and trichloroethylene in the profiles indicate industrial usage of the agents as solvents rather than in flavoring (trichloroethylene was banned from the food and pharmaceutical industries in most of the world due to concerns about its toxicity; see [129]). Also, there are no discernible sources of aldehydes or ketones; these compounds are almost undetectable in Toronto's air, unlike in Montreal, a large metropolis in the neighboring province of Quebec, which accommodates the forestry, lumber, and pulp industries [62].

Secondary Sources
Persistent air pollution from aromatic and chlorinated hydrocarbons arises due to volatilization from contaminated soil, sediment, and water (cf. refs. [31][32][33][34][35][36][37]). Formerly (before 2000), in the Great Lakes areas in Canada and the United States, volatilizations from sediment, contaminated soils, and water surface contained significant concentrations of aromatic and chlorinated hydrocarbons; see refs. [130][131][132][133]. Research concentrated on the deposition of these pollutants into lakes and on the prevention of ecological disasters; chemical waste dump leachates and direct manufacturing effluents were reported to be the major source of chlorobenzenes [134,135], particularly in the Great Lakes of North America. Their biodegradation in water and soil is slow; volatilization is the most important environmental fate process and the main mechanism of their transport from soil and water into air.
Volatilization of deposited materials (see e.g., [56,136]), the most persistent input to air pollution, does not appear as releases from discernible sources. The volatilizations contribute to the "inert" and "stable" sources (see Tables 2 and 3)-probably as remnants of their presently reduced, but constant, release from wastewater, sediment, and lake surfaces.
The dimethylbutanes (described in Table 1) also exist in the stable background, supposedly because of their potential to adsorb to soils, suspended solids, and sediments and then slowly volatilize. Also, dichlorobenzenes (1,2-, 1,3-, 1,4-) can appear in the stable background as uniform releases from contaminated sediments and soils (through industrial and municipal effluents); in Toronto, only 1,4-dichlorobenzene was detected.

Conclusions
Our results may provide information on identifying and resolving pollution sources. Many less prominent species captured at monitoring stations may be considered characteristic of specific point sources, while the prominent ones describe distributed background sources and secondary sources, and their VOC loadings.
These results endorse the finding concerning the trends in urban VOC pollution over time. Biogenic VOCs, and their photochemistry and natural cycles, have no role in degrading the quality of urban air. The pollution associated with traffic and transportation becomes dominant; here, the evaporative emissions of a large variety of fuel hydrocarbons, including unsaturated hydrocarbons, both aliphatic and aromatic, overwhelm the pollution caused by exhaust hydrocarbons. However, in Toronto, the ambient concentrations of hydrocarbons strongly diminish over time; growing urban traffic produces less exhaust pollution due to reformulated gasoline, better combustion technology, and governmental regulations targeting fossil fuel usage and combustion.
Halocarbons (freons) still persist in the atmosphere, accompanied by the abundant and relatively stable chlorinated hydrocarbons. The ambient concentration of the stable and/or abundant gases strongly depends on the prevailing winds; this may be the cause of negative values of apportionments for the "Inert" or "Stable" source. Indeed, removal by the wind of some locally occurring compounds may be correlated with arrivals of the compound released by a remote source located in the upwind direction (cf. [146,147]).
Governmental restrictions on water chlorination result in a very limited presence of trihalomethanes in the Toronto atmosphere (see [148])-excepting persistent but moderated amounts of chloroform in the inert/stable background of urban pollution.
The highly hazardous pollutants, trichlorobenzenes, were not detected in Toronto. They have a relatively short half-life in the atmosphere. Their possible sources-volatilization from sediments and soils contaminated by the petrochemical industry-are apparently absent. Based on our analysis, it would seem as though the clean-air bylaws prevented their release into industrial waste streams.
Dichlorobenzenes (1,2-, 1,3-, 1,4-) are widely used in manufacturing polymers, degreasing engines, and metal cleaning; however, they are hard to assign to a specific source. In Toronto, they appear in the stable background as uniform releases from contaminated sediments and soils.
Chlorobenzene-a massively used solvent in degreasing automobile parts-is not detectable in Toronto, since, arguably, American production declined considerably in the second half of the last century [131]. Chloroethane-used in leaded gasoline, but banned in Canada since 1980-disappeared from traffic sources; its counterpart in unleaded gasoline, methyl tert-butyl ether (MTBE), is also absent, which indicates a broad reduction in exhaust emissions [149].
For all the "industrial" VOCs, volatilization from soils and waters receiving industrial seepage is an important environmental fate process. It runs differently for different substances, and the dynamics of removal also vary. This confuses simultaneity of incidence of VOCs coming from the same source. Thereby, industries shall be distinguished mainly by their chemical menus rather than by the temporal patterns of their emissions. Still, UNMIX mainly associates with one source the VOCs that occasionally vanish simultaneously from the air. Then, irregular volatilizations may incapacitate modeling for many VOC menus.
The UNMIX results suggest that without taking into consideration atmospheric photochemistry (cf. [150]), weather (cf. [53,147]), and the mechanisms of secondary formation (cf. [151,152]), we can only achieve a coarse apportionment of sources detected at a particular monitoring station.
Still, UNMIX routines provide dependable information by imposing strong criteria of reliability and acceptability on the source profiling results. At the same time, negative values are allowed. These aspects enable creating "sources," whereby a group of vanishing VOCs is aligned with an appearance of another group. For instance, a negative value may represent a deficit in concentration of a species undergoing partial removal, the removal having dynamics uncorrelated with its emission dynamics. However, these dynamics may be correlated with arrivals of some other species, if the removal and the arrivals have the same cause, such as moving air masses. It appears that such effects are quite natural and frequent and they rationalize the plethora of profiles generated by the UNMIX routines.
Assuming such mechanisms are at work in the UNMIX routines, we can speculate that positive percentage values of an apportionment (occasionally higher than 100%) represent the initial emissions unaffected by the journey to the sensor, while negative values would represent removal of vapors from the air. Further studies are needed to fully explain the UNMIX results and to exploit them for better understanding the information obtained by remote sensing. The most problematic are trace concentrations originating from point sources: environmental influences destroy the relationship between the source emission pattern and the measurements performed by a distant station.
The large number of species under scrutiny, which may be limiting when directly building many-VOC models (cf. [45,46]), may be turned into an advantage: many reliable partial chemical fingerprints may be spliced into exhaustive, however approximated, source profiles. Also, arguably, there is a potential to consider "negative sources", i.e., the processes of removal and depletion. Thus, a better understanding of the environmental fates of VOCs-combined with robust modeling (cf. [153]), including modeling of the VOC depletion and removal-may result in useful depictions of environmental health hazards.