Atmospheric Measurement Techniques Evaluation and Application of a Semi-continuous Chemical Characterization System for Water Soluble Inorganic Pm 2.5 and Associated Precursor Gases

Water soluble inorganic particles components (Cl − , SO 2− 4 , NO − 3 , and NH + 4) and concentrations of their associated precursor gases (HCl, SO 2 , HNO 3 , NH 3) were semi-continuously measured using the Dionex Gas Particle Ion Chromatography (GPIC) system. Sampling was conducted adjacent to a high traffic street in downtown Toronto, Canada from June 2006 to March 2007. This study evaluated the precision and accuracy of field sampling measurements with the GPIC both relative to filter based measurements and other co-located semi-continuous instruments (R&P 8400N Nitrate Monitor, API Fluorescent SO 2 Gas An-alyzer, and Aerodyne C-ToF-AMS). High temporal resolution PM 2.5 mass reconstruction is presented by combining GPIC measured inorganic species concentrations and Sunset Laboratory OCEC Analyzer determined organics concentrations. Field sampling results were also examined for seasonal and diurnal variations. HNO 3 and particulate nitrate exhibited diurnal variation and strong partitioning to the gas phase was observed during the summer. Ammonia and particulate ammonium also demonstrated seasonal differences in their diurnal profiles. However, particulate sulphate and SO 2 showed no diurnal variation regardless of season suggesting dominant transport from regional sources throughout the year.


Introduction
Exposure to atmospheric aerosols has been implicated in mortality from cardio-pulmonary causes (Dockery et al., 1993;Pope et al., 1995;Hoek et al., 2002;Pope et al., 2002).Moreover, degradation in visibility (Watson, 2002) and environmental damages (Schindler, 1988;Jacobson, 2001;Driscoll et al., 2003) also result from atmospheric aerosols.These negative consequences may only be addressed once variations in the chemical composition, size distribution and mixing state of aerosols are better understood.Specifically, the mass, composition and size of fine particulate matter can be influenced by the partitioning of species between the gas and particle phase.
Two limiting factors in understanding dynamic atmospheric processes have been the low time resolution associated with many existing aerosol and gaseous chemical characterization instruments and the inability to measure chemical components in both the gaseous and particulate phases simultaneously.Instrument measurement inadequacies prompted the first objective of this study: commissioning the semi-continuous Dionex Gas Particle Ion Chromatography (GPIC) system.The GPIC simultaneously measured bulk concentrations of ambient water soluble inorganic species in particulate matter with an aerodynamic diameter smaller than 2.5 µm (PM 2.5 ) and the associated precursor gases with a fifteen minute time resolution.A number of additional semi-continuous instruments, including an R&P Tapered Element Oscillating Microbalance (TEOM), R&P 8400N Nitrate Monitor, API SO 2 Analyzer and Aerodyne C-ToF-AMS, were deployed along side of the GPIC to evaluate this recently developed instrument.
Published by Copernicus Publications on behalf of the European Geosciences Union.The second objective of this paper was to use GPIC measurements to better understand temporal fluctuations in the chemical composition observed for urban aerosols and their precursor gases.Focus was given to studying the diurnal and episodic variation of inorganic species within and between seasons.

Sampling site
This study sampled air collected from downtown Toronto, Ontario, Canada in the University of Toronto's Wallberg Building located at latitude 43.66 N and longitude 79.40 W (Fig. 1).The Wallberg Building is situated on the north side of College Street, which experiences weekday traffic of about 33 000 vehicles per day, a volume representative of traffic on many major streets in downtown Toronto.The sampling inlet was approximately 15 m from College Street and 5 m above the ground.
Filter, denuder and TEOM measurements used for comparison were collected on the roof top of the three story Gage building located approximately 170 m west of the Wallberg Building sampling site.A second TEOM operated by the Ontario Ministry of the Environment (MoE) was situated at the high traffic intersection of Bay and Wellesley Streets with an inlet 10 m above ground, approximately 875 m from the Wallberg Building sampling site.

Gas Particle Ion Chromatography (GPIC) system
The GPIC system was used to obtain semi-continuous measurement of soluble gaseous and particulate phase species simultaneously.Monitored species included particulate chloride (pCl − ), particulate sulphate (pSO 2− 4 ), particulate nitrate (pNO − 3 ), particulate ammonium (pNH + 4 ), HCl (g) , SO 2(g) , HNO 3(g) , and NH 3(g) .Each of the gas phase and particulate phase constituents were independently analyzed on two separate channels each sampling at 5 L min −1 operating with 15 min time resolution.The gas channel sampled ambient air via a 0.6 cm diameter, 3 m long piece of Teflon tubing while the particle channel employed 0.9 cm diameter stainless steel tubing of the same length for sampling.The use of Teflon lines for gas species sampling was to minimize the loss rate of HNO 3 which is known to readily adsorb onto many sampling inlet materials.All sampling lines were located indoors where a temperature of 23 • C was maintained; increased HNO 3 adsorption to Teflon inlet tubing has been noted to occur at temperatures below 10 • C (Neuman et al., 1999).
On the particle channel, the air stream passed through a PM 2.5 sharp cut cyclone prior to introduction into the etched plexiglass parallel plate wet denuder (PPWD).A 0.5 mM H 2 O 2 solution wetted the denuder and served to collect soluble gaseous constituents from the ambient air.All liquid flow was controlled using an eight channel peristaltic pump.Gas species dissolved in the dilute H 2 O 2 solution were collected from the denuder and sent to waste.Gas denuded air was then routed to a plexiglass air-tight particle collector vessel (PC), which was outfitted with a polytetrafluoroethylene (PTFE) filter located at the exit of the chamber.This hydrophobic filter prevented particles from leaving the PC and aided in the process of capturing soluble particulate matter.An 18.2 M dionized water stream was injected into the high velocity sample air flow path in the PC.Upon entrance into the collection vessel the water formed a fine mist.Any water soluble particulate matter was dissolved into solution.Negative artifacts, due, for example, to evaporation after sampling, were minimized since the volatile aerosol species were continuously being dissolved.Mist quickly accumulated in the PC causing the water droplets to coalesce due to the chamber's high humidity.The base of the PC was conical allowing for the liquid to be easily aspirated from the bottom of this reflux-condenser-like vessel.Once extracted from the PC, components within the dissolved particulates were routed into a series of cation and anion concentrator columns for conductivity and IC analysis, respectively.
Cation concentrators were eluted with a 15 mM NaOH solution.Liquid subsequently passed through a porous membrane device (PMD) (Dionex,.A concentration gradient was the acting driving force behind this shell and tube like exchanger; any NH 3 in the concentrated alkaline sample passed through the PMD shell diffusing into the inner stream.The high purity water, which was flowed through the inner PMD tubing, served to hydrolyze the collected NH 3 back into NH + 4 .Flow was then routed through a conductivity detector (Dionex, CD25A) providing a measure of the total ammonium content.The anion concentrators were eluted with 20.5 mM KOH and then pumped (GS50) into the ion chromatography system (Dionex, ICS-2000) for analysis of the remaining species.Once in the dissolved anion form, particles were routed into the anion guard column (AG11-HC, 2×50 mm) and then the separator column (AS11-HC, 2×250 mm), both of which were heated in an oven maintained at 30 • C. Following these columns, the sample solution passed through an ASRS-Ultra 52 mA suppressor and 2 mm carbonate removal device (Dionex, CRD) prior to entering the thermally stabilized conductivity cell.The system used electrodialytically generated (EG40) 20.5 mM KOH mobile phase.It should be noted that the GPIC system was modified from its original form to function with the additional post suppressor CRDs to minimize the magnitude of the carbonate in the sample stream, consequentially reducing interference with sulphate and nitrate peaks.This enhanced peak resolution also improved detection limits for the nitrate and sulphate related atmospheric gases and particle components measured on the IC system.
The air stream in the gas channel was directly introduced into another PPWD which collected soluble gases in a weak H 2 O 2 solution.The peroxide solution, now containing dissolved gas species, was drained from the denuder and then sent to a series of cation and anion concentrator columns for subsequent analysis identical to that on the particle side.
A computer using Dionex Chromeleon Software was connected to the GPIC system and allowed continuously automated valve control in 15 min intervals.This software package was also used to collect both conductivity detector signals and the ion chromatography output.The limits of detections were established as 100 ng m −3 for all particle anion species, 25 ng m −3 for all gas anion species and 100 ng m −3 for both gas and particle cation constituents.Further details regarding this instrument's design have been previously described by Al Horr et al. (2003) and Ullah et al. (2006).The measurement uncertainty for individual GPIC species was determined over a period of stable operating conditions and concentrations.As the magnitude of sampled component concentrations governs the associated uncertainty, it is useful to note that these uncertainties were determined for blank samples of high purity water with much lower ionic content than typical ambient samples.Hence these blanks samples provided an upper estimate of the uncertainty.For GPIC particulate species the measurement reproducibility was found to be pCl − =8%, pSO 2− 4 =9%, pNO − 3 =11%, pNH + 4 =13% and the following for gas phase compounds: HCl=6%, SO 2 =6%, HNO 3 =10%, and NH 3 =5%.The number of samples included for each uncertainty calculation ranged between 75 to 192 replicates.

NAPS sample collection and analysis
Integrated 24 h ambient aerosol samples, acid gases (HNO 3 , HNO 2 , SO 2 ) and ammonia were collected on the rooftop of the Gage Building in Toronto, one of the National Air Pollution Surveillance (NAPS) sites.All PM samples were collected in parallel with R&P Partisol-Plus 2025-D sequential dichotomous particulate samplers (Dichot) and with the R&P Partisol Model 2300 Speciation samplers (Partisol) once every three days (Rupprecht & Patashnick Co., Inc.,).The dichotomous particulate sampler was operated at a total flow rate of 16.7 L min −1 which was split between two channels: PM 2.5 and coarse PM (PM 2.5−10 ).Two separate mass flow controllers maintained the flow rates of the fine and coarse particle stream at 15 L min −1 and 1.7 L min −1 , respectively.The Partisol speciation sampler was comprised of three distinct Harvard designed Chemcomb ® cartridges (A, B, and C) connected to a low volume air sampler operating with a total flow rate of 30 L min −1 .Each channel maintained a flow rate of 10 L min −1 .Aluminum inlets were used on cartridges A and B while a Teflon inlet was installed on cartridge C. The Chemcomb cartridge A contained pre-fired quartz fiber filter for collecting PM 2.5 for the analysis of elemental carbon (EC) and organic carbon (OC).Cartridge B contained a Teflon filter and a backup pre-fired quartz filter in series.The Teflon filter was analyzed for PM 2.5 and archived for other analyses.The backup quartz fiber filter samples particlefree ambient air and provided an estimate the amount of gaseous organic species adsorbed on the quartz fiber filter (Cartridge A).The cartridge C had two honeycomb glass denuders placed upstream of two filter packs with Teflon and nylon filters.A Na 2 CO 3 -coated denuder was used for collection of acid gases (HNO 3 , HNO 2 , SO 2 ) and a citric acidcoated denuder was used for collection of ammonia.The Teflon filter was used for the collection of PM for the analysis of water-soluble anions and cations, and water-soluble trace metals.The volatile nitrate that was released from the Teflon filter was captured by the nylon backup filter.
PM mass was determined by weighing all Teflon filters, before and after sampling, at controlled temperature (23±3 • C) and relative humidity (45±5%) with a Mettler 5 Microbalance (1-µg readability; MT-5, Mettler-Toledo Inc., Highstown, NJ), after the filters (samples and blanks) had acclimatized to these conditions for at least 24 h.The detection limit for gravimetric measurements was 0.3 µg m −3 which represented 1 to 2% of mean aerosol mass.Exposed quartz fiber filters were analyzed for OC and EC content using a DRI Model 2001 Thermal/Optical Carbon Analyzer (Atmoslytic Inc Calabasas, CA).In this study, the Interagency Monitoring of Protected Visual Environments thermal/optical reflectance protocol (Chow, 2001) was used.A total carbon loading of less than 0.2 µgC cm −2 in the analyzer oven was achieved prior to field sample analysis.In addition, before analyzing a sample batch, the instrument was calibrated for CH 4 or CO 2 .EC and OC concentrations were determined to have a detection limit of 140 and 610 ng m −3 , respectively.The uncertainty for OC was determined to be 20% and 30% for EC.Ion chromatography analysis was used to determine the soluble ion contents of the various extracts applying a dual-channel Dionex DX-500 and/or DX-600 and/or ICS-3000 (Dionex, Sunnyvale, CA, USA) ion chromatographs.Each channel operated a self-regenerating SRS-ULTRA suppressor with internal DDW regeneration mode, a CD20 conductivity detector, and a GP50 gradient pump.For the water-soluble cations in extracts of Teflon filters (Cartridge C) and ammonium in the extracts of citric acid-coated denuders, IonPac CS-12 analytical columns with methanesulfonic acid as eluent were used.Water-soluble anions in the extracts of Teflon filters (Cartridge C) were analyzed using IonPacAS15 column.For the anions in extracts of nylon filters and carbonate-coated denuders, a Dionex-AS4A column was used with carbonate/bicarbonate eluent.NIST-traceable calibration standards were run prior to field sample analysis and every 10 to 15 samples.Anions and cations were found to have detection limits in the range of 2 to 30 ng m −3 and 1 to 6 ng m −3 , respectively.Uncertainty for sulphate, nitrate and ammonium were in the order of 10%.

R&P TEOM monitor
Total PM 2.5 mass concentrations were measured using a TEOM (Model 1400A, Rupprecht & Patashnick Co. Inc.) with five minute resolution.As the mass collected on the TEOM filter was very sensitive to ambient water vapour concentrations, the ambient air inlet stream was heated to 30 or 50 • C to eliminate the effect of water associated with hygroscopic particles (Patashnick and Rupprecht, 1991;Mignacca and Stubbs, 1999).The heated inlet however, causes volatilization losses of pNO − 3 and OC (Wilson et al., 2006).Additional details are discussed in Sect.3.1.3.This study employed the use of two TEOMs: one run by Environment Canada on the rooftop of the Gage Building (30 • C heater temperature) and the second operated by the Ontario Ministry of the Environment with a 50 • C heated inlet at another downtown location.

Sunset Laboratory Semi-Continuous OCEC Analyzer
The Sunset Laboratory OCEC Analyzer collected PM 2.5 on a quartz fibre filter in an oven at a flow rate of 8 L min −1 for a period of 108 min per sample.Following the completion of the aerosol sampling period, deposited particles were heated in the quartz oven where elemental and organic carbon concentrations were individually quantified.EC and OC concentrations were measured using thermal methods: particles were heated in the quartz oven to oxidize the carbon aerosol into CO 2 which was measured using a Non-Dispersive Infrared detection system.Heating was conducted in various staged temperatures, each volatizing different carbonaceous components.The first stage raised ambient temperatures to a maximum of 840 • C and introduced pure helium gas into the oven in an effort to minimize the amount of black carbon (BC), also known as EC, oxidation while enhancing OC volatilization (Chow et al., 2005).Measured carbon during this stage was classified as OC.The oven temperature was re-duced to ambient conditions and purged of helium before the commencement of the second stage.Oxygen was then introduced into the system and temperatures were increased to a maximum of 850 • C so that remaining carbonaceous content (BC) of the collected bulk aerosol could be quantified (Chow et al., 2005).
In addition to thermal measurements of EC and OC concentrations, this Sunset instrument also used optical methods to correct thermal measurements.Although OC charring should be avoided in non-oxidizing environments, losses are still possible given the operating oven temperatures.By transmitting the beam of a laser through the collected aerosol sample, the intensity of light transmitted or reflected at each stage of thermal analysis was used to quantify the amount of OC charred: As OC became increasingly charred light absorption was enhanced, attenuating transmission.BC volatilization, occurring mostly in the second stage heating phase, increased laser transmission.(It is reasonable to expect some BC volatilization will occur prior to the second stage heating phase as the temperature at which this species evolves is dependent on the aerosol's chemical compositions and a number of other factors (Reid et al., 1998), though first stage volatilization is minor as compared to second stage volatilization).Using this optical detection method, the Sunset Laboratory OCEC instrument is able to correct for charring and first stage volatilization of BC, providing a better estimate the true concentration of each type of carbonaceous aerosol present in the sample and minimize errors from sampling processing.
The uncertainty for EC and OC measurements by the Sunset OCEC was determined to be 4% and 8%, respectively.Chow et al. (2005) provides an in depth discussion of this instrument's operating philosophy.The Sunset Labs OCEC was located in the Wallberg Building laboratory.

API fluorescent SO 2 gas analyzer
Five minute averaged ambient SO 2 gas samples were collected at the Wallberg Building sampling site using this API SO 2 instrument (Model 100A).

R&P 8400N Nitrate Analyzer
The Ambient Particulate Nitrate Monitor (Model 8400N, Rupprecht and Patashnick Co. Inc.) sampled air through a PM 2.5 impactor which was then heated to 350 • C.This temperature was adequately high to volatilize all collected nitrate, causing decomposition into NO x which was analyzed using the API chemiluminescent NO x analyzer (Model 200A) in 15 min averaged intervals.

Aerodyne C-ToF-AMS
An Aerodyne Compact Time of Flight Mass Spectrometer (C-ToF-AMS) (Aerodyne Research Inc., Billerica, MA) deployed at the Wallberg Building sampling site in January and February 2007.This instrument determined the size-resolved non-refractory chemical composition of particles with submicron vacuum aerodynamic diameter (PM 1.0 , particulates with an aerodynamic diameter of less than 1 µm) with 1 min time resolution.This instrument has been previously described by Drewnick et al. (2005).

Results and discussion
3.1 GPIC data evaluation

GPIC comparison with NAPS filter measurements
Inorganic particulate and gas concentrations generated by the GPIC were compared with these 24 h PM 2.5 Partisol and Dichot filter and denuder samples collected by the NAPS program on the roof of the Gage Building.The proximity of the Gage Building suggested similar ambient air concentrations and previous study have indicated that the two sampling locations yielded similar inorganic particulate measurements upon comparison (Tsai, 2002).
GPIC fifteen minute particulate measurements were averaged to a 24 h basis comparable to the NAPS filters' resolution.At least 75% of the fifteen minute GPIC measurements were required for a daily average to be considered valid.For the entire campaign duration, raw GPIC particulate measurements were less than inorganic concentrations given by the corresponding filter data: overall linear regressions between raw GPIC particulate averaged measurements and NAPS Partisol filter measurements indicated a slope of 0.45, 0.41, and 0.49 (µg m −3 ) GPIC /(µg m −3 ) Gage Filter for sulphate, nitrate and ammonium concentrations, respectively.However, good GPIC-NAPS Partisol filter temporal agreement was indicated by high Pearson r values of 0.93 (sulphate, N =76), 0.86 (nitrate, N =77) and 0.85 (ammonium, N=68) for each regression.
Consistency between the three main particle phase species and low month-to-month variation between the two measurement methods suggested a systematic bias.Non-unit GPIC/NAPS Partisol filter ratios for all inorganic components of interest suggested incomplete particle collection, air flow leakage, low flow rate, or line losses in the semicontinuous instrument.The possibility of low flow in the GPIC itself as the cause was eliminated as the mass flow rates were verified intermittently throughout the sampling with an independent flow meter.The reason for the nonunit GPIC/NAPS Partisol filter ratios could not fully be determined during the study as the issue was only identified at the end of the sampling period, once the filter results became available.However, in-leakage of air due to incomplete seals in the sampling lines upstream of the particle trap in the GPIC was one issue that was identified.This problem caused a 20-30% negative artifact that was aggravated by frequent maintenance-related disconnections.In a follow-up study, a discrepancy inherent in the GPIC design, relating to the fixed volume loop used for calibration was identified (Yao et al., 2008).This error caused a systematic 30% negative artifact in the GPIC concentrations results.Hence the difference between the GPIC and filter samples was eventually identified and accounted for.
To address the difference between the GPIC and NAPS results, NAPS Partisol filter measurements were selected as the basis for correcting the particulate mass concentrations.Given that the filter sampling methodology had been designed to minimize losses by volatilization species, specifically nitrate, it was reasonable to use the filters as a basis for semi-continuous instrument evaluation and calibration.Raw GPIC particulate concentrations were calibrated against the Gage Partisol filter particulate measurements on a month by month basis in an effort to obtain more accurate temporally resolved GPIC results.The final recoveries for pSO 2− 4 , pNO − 3 , pNH + 4 were 98% (r=0.93),100% (r=0.93),97% (r=0.94),respectively.The total propagated uncertainty associated with these particulate GPIC results calibrated with NAPS filter measurements are 13, 15, and 16% for pSO 2− 4 , pNO − 3 , and pNH + 4 , respectively.

GPIC comparison with NAPS denuder measurements
Similarly to the GPIC particle channel measurements, all the fifteen minute data was averaged to a 24 h basis while implementing the same criteria for validating averaged results as for the particulate channel.Again the NAPS denuder measurements for HNO 3 , NH 3 and SO 2 gases were assumed to be accurate and the GPIC concentrations were calibrated to these standards.Comparison of the denuder and GPIC concentrations indicated an overprediction by the GPIC for NH 3 and SO 2 measurements and underprediction for HNO 3 : overall linear regression slopes and corresponding Pearson r values of 0.25 (r=0.76,N=39), 1.43 (r=0.83,N=56), and 1.42 (r=0.97,N=70) ((µg m −3 ) GPIC /(µg m −3 ) Gage Denuder ) were found for HNO 3 , NH 3 and SO 2 , respectively.The high Pearson r values showed the strong temporal consistency between the two sets of measurements.Significant in-leakage of lab air (54%) was found at the end of the study on the gas channel.However, there was no reason to suspect that the indoor lab air had higher concentrations of NH 3 and SO 2 than the outside air.Consequently it was not clear why this in-leakage could cause the NH 3 and SO 2 species to be overpredicted.HNO 3 line losses likely contributed to low GPIC measurements for this species.Losses may have occurred along the walls of the inlet sample air tubing and any water droplets accumulated in the inlet line may have caused further losses.Furthermore, comparison of nitric acid measurements made at other locations to this study indicated that the GPIC concentrations were lower than those typically observed (Hering et al., 1988;Boring et al., 2002;Huang et al., 2002;Fischer et al., 2006;Lin et al., 2006;Morino et al., 2006;Ullah et al., 2006).For the gaseous species, the ratio of the GPIC to NAPS concentrations did not show as much month to month variability as the particle channel.Hence the GPIC-NAPS denuder calibrations were made based on two seasonal linear correlations: Summer-Fall (June 2006to December 2006) and Winter (January 2007to March 2007).GPIC calibration with seasonal denuder measurements yielded final relationships of: y=1.0x (r=0.9,N =70), y=1.0x (r=0.61,N =35), and y=1.1x (r=0.81,N=56) for SO 2 , HNO 3 , and NH 3 respectively.These calibrated gas phase GPIC results have propagated measurement uncertainties of 12, 14 and 11% for SO 2 , HNO 3 , and NH 3 , respectively.
3.1.3Mass closure study PM 2.5 mass in Toronto is composed of the major secondary inorganic species, including sulphate, nitrate, and ammonium, and organic matter and elemental carbon.Crustal elements and other inorganic species have been reported to account for the remaining PM 2.5 mass (approximately 80%) on an annual basis in Toronto (Brook et al., 1999a).As a further verification of the accuracy of the GPIC measurements, the particulate mass was reconstructed using the inorganic GPIC particulate measurements and EC and OC mass concentrations made by the Sunset Laboratories OCEC instrument, both co-located in the Wallberg Building for the months of July 2006 and October 2006 to January 2007.
Measurements collected by the Sunset Laboratory OCEC Analyzer included OC and EC which were both validated against 24 h NAPS Partisol filter results in a separate study.A recovery of 0.71 for Sunset OCEC's OC measurements to filter results was found with an associated Pearson r=0.73.Loss of semi-volatile organic species during sampling may have caused a negative artifact in the Sunset OCEC results.Conversely, condensation of semi-volatile organic carbon overnight on the particulate matter already collected on the filters may have created a positive artifact for the NAPS Partisol results.Comparison of the EC measured by the Sunset OCEC to the Partisol filter EC resulted in a slope of 1.08, Pearson r=0.58.
On average for this period, the major inorganic species represented 62% of the total reconstructed mass and carbonaceous compounds accounted 26%.It should be recalled that all the inorganic GPIC concentrations used in the mass closure study were corrected against the monthly NAPS filter measurements as previously described.Total reconstructed PM 2.5 mass from the Wallberg Building GPIC and OCEC instrument measurements were compared to both the Gage and Ministry of the Environment Downtown TEOMs.Successful comparison between the semi-continuous and filterbased measurement techniques required the Sunset OCEC's organic carbon results to be converted to an organic mass (OM) concentration.A factor of 1.4 was used to account for the contribution of elements other than carbon to this mass (Watson, 2002;Harrison et al., 2003).This conversion fac-  tor was expected to have a seasonal dependence; however, the degree to which the relative amounts of primary and secondary organic carbon varied between seasons was unknown and thus a constant conversion ratio was employed.
To evaluate both the Gage and MoE TEOM semicontinuous mass measurements, a comparison was conducted with 24 h NAPS Dichot filter mass data.Lower PM 2.5 mass measurements were made by both TEOM instruments as compared to the NAPS Dichot filter mass.Furthermore, high particulate mass concentrations were more seriously underpredicted by both TEOMs.It is possible that episodic values influenced this divergence given that the mean difference between the NAPS Dichot and each of the TEOMs was greater than the median concentration difference.Underprediction of PM mass concentrations by TEOMs has been well documented (Allen et al., 1997;Ayers et al., 1999;Green et al., 2001;Charron et al., 2004); the heated inlets of this instrument act as an effective denuder, removing all semivolatile compounds (in particular ammonium nitrate and OC) from the PM sample.Wilson et al. (2006) have further noted that semi-volatile compounds losses by the TEOM is a function of meteorological conditions; cold and humid conditions better stabilize ammonium nitrate and OC, minimizing TEOM PM mass concentration underestimation.Although TEOMs suffer from semi-volatile material losses, it is also pertinent to recall that gravimetric filter measurements may also experience losses of ammonium nitrate and OC ranging from less than 10% to greater than 50% (Wilson et al., 2006).
Individual monthly comparisons between each TEOM and the NAPS Dichot filter further elucidated the discrepancy noted between the two measurement techniques (Fig. 2).TEOM mass concentrations illustrated maximum underpredictions in the winter months and showed approximately equivalent measurements to the NAPS Dichot mass filters during the summer.As the Partisol mass filter measurements were taken as the standard for evaluating the semi-continuous measurement techniques, all semi-continuous TEOM mass concentrations were calibrated to the NAPS Dichot measurements on a month to month basis.
The cause for this winter TEOM underprediction was likely increased loss of semi-volatile species induced by the instrument's heated air stream: as the contribution of semivolatile species to the total particulate mass increased at the colder ambient temperatures, more particulate semi-volatile species removal would have occurred (Table 1).Hence, the magnitude of the TEOM mass underprediction was a function of particle composition, if all other factors are considered to be equal: as nitrate production was thermodynamically favoured at lower temperatures, the highest particulate nitrate concentrations were measured during the winter months.Elevated ambient relative humidity may have also contributed to increased differences between the two mass measurement methods as ambient relative humidity controls particle liquid water content and therefore the partitioning of semi-volatile species such as nitrate.Measured particulate nitrate was assumed to all be in the form of ammonium nitrate.
The TEOM experienced maximum particulate loss during the winter which corresponds to the standard NAPS Dichot mass filter trends and peak concentrations of volatile nitrate species.Least square linear regressions were conducted between the PM 2.5 mass reconstructed from the GPIC and Sunset OCEC measurements and (1) calibrated and (2) raw TEOM results for the months of November 2006 to March 2007.These months were associated with the greatest contribution of ammonium nitrate to the total Wallberg PM 2.5 mass measurement.Figure 3 was used to evaluate if the deviations of the TEOM measurements from the NAPS Dichot filter were approximately equivalent to the mass of ammonium nitrate present within the particulates.The linear relationship established between the reconstructed PM 2.5 mass and raw TEOM PM 2.5 measurements was y=1.55x and y=1.57x for the MoE and Gage, respectively (Fig. 3a and b).With a slope greater than 1, raw TEOM values underpredicted the reconstructed mass from the Wallberg measurements which only included major secondary inorganic and carbonaceous species, excluding the inorganic soil species which was estimated to account for approximately a further 8% of the total Toronto PM 2.5 mass (Brook et al., 1999a).Subtraction of the ammonium nitrate mass from the reconstructed Wallberg PM 2.5 mass yielded a significant reduction in the previously described slopes yielding values of about 0.8.The difference between the TEOM and NAPS Dichot filter mass measurements were attributed to partial volatilization of nitrate in the heated inlet air stream.It is worth noting that subtracting the ammonium nitrate slightly improved the correlation between the Wallberg PM 2.5 results and the TEOM measurements, suggesting that the day to day variation in the NH 4 NO 3 mass contributed to the variation in accuracy of the TEOM measurements.Furthermore, the different temperatures used in the two TEOMs did not cause significant dissimilarities in  the extent of semi-volatile mass loss; a temperature of 30 • C should, in theory, have been sufficient to volatilize all the ammonium nitrate.
Comparison of the calibrated TEOM MoE and Gage data with reconstructed GPIC and OCEC Wallberg Building PM 2.5 mass concentrations for the entire study period yielded a relationship y=0.93x and y=0.95x, respectively (Fig. 4).Excluded inorganic soil compounds possibly accounted for the non-unit slope in addition to instrument measurement errors.However, correlation of the filter mass measurements for the NAPS 24 h Partisol (y) and Dichot (x) filters also gave a slope of 0.95 (Pearson r=0.93).The propagated total measurement uncertainties on average for the reconstructed Walberg Building and NAPS 24 h Partisol PM 2.5 mass concentrations were 12% and 18%, respectively.Hence the 5% difference was also within the range of the experimental error.

Semi-continuous inter-instrument comparison
To further verify the GPIC system was successfully mea-

R&P 8400N Nitrate Analyzer comparison
Comparison of the co-located R&P 8400N Nitrate Analyzer and the GPIC in the Wallberg Building showed an overall good correlation between the two measurement techniques: 8400N versus GPIC regression (N=4099) yielded a slope and associated 95% confidence interval of 0.83±0.03(µg m −3 ) R&P 8400N /(µg m −3 ) GPIC with an intercept of 0.32±0.11µg m −3 (Pearson r=0.86) and a recovery slope of 0.89.However, upon examination of the individual   months, the ratio relating the two instruments was found to approach unity in the wintertime (December 2006 to March 2007) which suggested better agreement was possible when higher ambient nitrate concentrations were present.This good agreement provided support for the validity of correcting the GPIC particle channel results based on the NAPS Partisol filter.Both nitrate measurements tracked each other well (Fig. 6a and b); however, discrepancies between the GPIC and 8400N, for both high and low nitrate concentrations, were noted for periods of elevated relative humidity.Toronto experienced a period of high temperature (T >30  (Long and McClenny, 2006).
Although the 8400N and GPIC exhibited excellent temporal correlation, the apparent 8400N underprediction warrants discussion.Comparable underprediction of the 8400N as compared to other measurement techniques (noted as the slope in Table 2) has been cited previously at various US urban and rural centers (Harrison et al., 2004;Hogrefe et al., 2004;Wittig et al., 2004;Grover et al., 2006;Long and McClenny, 2006;Rattigan et al., 2006).A summary of in-   strument intercomparison relationships is presented in Table 2 where all centres were urban environments with the exception of Whiteface Mountain, NY.Consistent underpredication by the 8400N was evident for each intercomparison and this divergence was not limited to comparisons with the GPIC.Moreover, each summarized study reported a similar mass concentration range as that observed in the current study.Nitrate measurement loss by the 8400N may have been in part due to decreased Molydenum converter efficiency which was located in the system's pulse analyzer (Long and McClenny, 2006).Low efficiency may have resulted in incomplete NO 2 to NO conversion causing nitrate measurement underestimations.It is also worth noting that Long and McClenny (2006) observed only a negative bias with the 8400N as compared to the GPIC during ambient sampling; the 8400N accurately measured high nitrate concentrations when operating under controlled laboratory conditions.These results suggested that interfering species may have contributed to the divergence between measurement techniques at high nitrate concentrations.Additional differences between these techniques may be caused to the 95% maximum 8400N sampling efficiency for particle diameters between 100 and 800 nm (Stolzenburg and Hering, 2000).

Aerodyne C-ToF-AMS comparison
Co-located GPIC PM 2.5 and C-ToF-AMS PM 1.0 measurements were collected for the period of 22 January 2007 to 4 February 2007 and compared using 15 min averaged sulphate, nitrate, and ammonium particulate concentrations.Excellent temporal agreement between the two measurement techniques was indicated from Pearson r values for each species: 0.92, 0.92, and 0.86 for sulphate, nitrate and ammonium, respectively (Table 3).However, the C-ToF-AMS concentrations were typically 40% lower than those from the GPIC.This difference was mostly due to the different size cutpoints of the two instruments; PM 2.5 for the GPIC and PM 1.0 for the C-ToF-AMS.Mass concentrations calculated from TSI 3321 Aerosol Particle Sizer (APS) measurements (assuming spherical particles with unit density) indicated that the ratio of PM 1.0 to PM 2.5 was on average 0.63±0.21during the sampling period.The collection efficiency of the AMS may also have been lower than the value of one used in the calculations, and this also potentially contributed to the difference.
The time series for each particulate species indicated that the relationship between the GPIC and C-ToF-AMS was not constant throughout the measurement period (Fig. 7).This variability was partially attributable to fluctuations in the aerosol size distribution.However, scaling the C-ToF-AMS data based on the PM 1.0 and PM 2.5 ratio from the APS only improved the agreement some of the time, with some notable exceptions.For example, on 23 January 2007 the ratio of PM 1.0 to PM 2.5 rose to 0.9 while the C-ToF-AMS measurements remained 40% lower on average than those of the GPIC.A similar pattern occurred overnight on 1 February 2007 to 2 February 2007: the C-ToF-AMS to GPIC ratio was 0.55 whereas that for PM 1.0 to PM 2.5 was 0.83.Hence, particle size could explain the difference between the GPIC and C-ToF-AMS on average, but other as of yet unidentified fac-  tors also contributed to the divergence at times.It is worth noting that the GPIC agreed with the R&P 8400N on 23 January 2007 and was higher overnight on 1 February 2007 to 2 February 2007 (Fig. 6).

Sampling conditions
Measured gas and particulate concentrations were to a large extent governed by meteorological parameters (wind speed, wind direction, temperature, relative humidity and solar radiation).The measurement time of day and contributions from local and long range transported aerosol sources were also influential factors.NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) back trajectory analysis and plume dispersion modeling were used to elucidate the geographic origin of measured air masses.
Regarding local sources, the Wallberg Building sampling site was strongly influenced by vehicle traffic.Peak traffic densities of gasoline vehicles passing by the site occurred on weekdays between: 06:30 to 09:00 and 16:30 to 18:30.Nighttime vehicular traffic was significantly lower than during the day.
Long range transported air masses were frequently sampled at the Wallberg Building sampling site in downtown Toronto.The majority of industrial sources influencing sampled air masses were situated south and southwest of the sampling site.Thus, air masses which originated from the north and northwest were generally free from high aerosol concentrations.However, winds from the southwest had the potential to carry gases and particles from Hamilton, Ontario, Ontario Power Generation's Nanticoke coal-fired power generation station, and the Ohio River Valley, into Toronto.Westerly winds were influenced by industries and power generating facilities in, or near, Detroit, Michigan and Sarnia, Ontario.Furthermore, agricultural activities located to the southwest and west of the city also contributed to measured pollutant gas and aerosol concentrations.

High time resolution mass reconstruction
A high time resolution reconstruction of the PM 2.5 mass was created using inorganic and organic mass concentrations measured by the GPIC and Sunset Laboratory OCEC semi-continuous instruments.Each season was associated with a different signature of relative particulate concentrations dictated partially by temperature dependent chemistry.Figure 8 compares variations in percent composition of inorganic and organic components of PM 2.5 mass for periods of elevated and reduced concentrations during different sampling periods: the entire sampling period (July 2006-March 2007, A), in the fall (October and November 2006, B) and winter (December 2006-March 2007, C).Organic species dominated the PM 2.5 mass fraction when low PM 2.5 concentrations were measured, regardless of season.Furthermore, the percent mass contributed by organic species during the low PM 2.5 periods was the greatest during warmer months.Conversely, the percent contribution of pNO − 3 increased as temperature declined and during high PM 2.5 concentration periods.
Air mass geographic origin and meteorological parameters influenced resultant particulate concentrations.Large seasonal differences were observed, in particular for sulphate and nitrate, even though the seasonal variation in the emission rates from mobile and point sources was likely small.Southwestern winds typically brought elevated regional sulphate concentrations for all seasons.However, measured pSO 2− 4 concentrations were higher during the summer as compared to the winter as the availability of OH and H 2 O 2 for reaction with SO 2 increased during this season.Higher Concentration regimes are defined as low PM2.5 ( ) and high PM2.5 ( ). Organic matter was calculated as 1.4 times organic carbon concentrations.Organic matter was calculated as 1.4 times organic carbon concentrations.
organic matter was measured during the summer and fall as compared to the winter.Summertime increases in biogenic emissions and photochemical activity allowed for enhanced volatile organic compound reaction with OH, O 3 and NO 3 and consequently favoured the formation of secondary organic aerosols.
Stacked time series of inorganic particulates, elemental carbon and organic matter during October 2006 and January 2007 are shown in Figs. 9 and 10, respectively.These figures illustrate the rapid and extensive temporal variability that exists in the mass and speciation of Toronto's fine particulate matter.During period of 18 October 2006 to 22 October 2006, three distinct air masses passed through Toronto resulting in large changes in PM 2.5 mass.Southwesterly winds first introduced a sulphate driven episode of elevated PM 2.5 which commenced on the morning of 18 October 2007 and lasted for approximately 24 h.A shift in winds midday on 19 October 2007 brought air from the north into Toronto producing a dramatic drop in PM 2.5 sulphate, nitrate and ammonium.This allowed local urban sourced pollutants to dominate measurements and resulted in reduced PM 2.5 concentrations as compared to air masses originating from southwest of the city.The third air mass, originating from east of the city, caused a nitrate dominated episode that started on the morning of 21 October 2007.It should be noted both sulphate and nitrate events contained approximately similar concentrations of ammonium and organics constituents.Owega et al. (2006) also observed the most significant difference in PM 2.5 aerosol mass occurred between air masses originating from the north/ northwest and south/southwest of Toronto.The magnitude and frequency of pSO 2− 4 episodes declined in the winter and pNO − 3 concentrations increased.This was attributed to the semi-volatile nature of ammonium nitrate, which should partition more extensively to the particle phase at low winter temperatures.Consistently elevated nitrate concentrations were observed during winter sampling months while the sulphate mass contribution varied depending on the geographic origin of the air mass.Similar to summer and fall measurements, winds from the southwest (17 January 2007 to 19 January 2007) were associated with higher sulphate concentrations than those from the other directions (13 January 2007 to 15 January 2007).
Episodes of high particulate concentrations resulting from long range transported air masses, independent of geographic origin or season, typically lasted for approximately two to three days.Particulate concentrations during these pollution periods were not continuously elevated, but instead exhibited a number of interwoven local minima and maxima values due to a series of separate plumes or fronts passing across the sampling site.Temperature driven diurnal variations, as described next, were also superimposed on top of the synoptic variability.A similar concentration variation was noted during the 2001 PMTACS-NY summer and winter sampling campaigns conducted in New York City (Weimer et al., 2006).

Seasonal diurnal variation
Hourly averaged GPIC HNO 3 , pNO − 3 , NH 3 and pNH + 4 concentrations are plotted as diurnal trends for the months of July 2006, October 2006and January 2007 in Fig. 11.Measurements were subdivided into weekend and weekday cat-  egories and were conditioned to exclude outliers, defined as 1.5 times greater than the interquartile range of the monthly average of the measurements for the given time of day.Diurnal profiles were analyzed to better understand the influence of local and long range source contributions to the measured secondary aerosol species and their production pathways.Emissions, photochemical activity and the planetary boundary layer mixing height were parameters which affected diurnal patterns.Locally derived gas and particulate species would be expected to exhibit a distinct diurnal trend while long range transported pollutants would be more influenced by synoptic meteorology and therefore should not necessarily show a daily cycle.
No statistically significant diurnal variation was noted for pSO 2− 4 and gaseous SO 2 as both these species did not likely originate from local urban emission sources.Long range transport, generally from sources west or southwest of the sampling site, governed the magnitude of SO 2 and sulphate concentrations observed in this urban environment.Sulphate results are in line with previous studies: Fisseha et al. (2006) also noted that sulphate lacked any diurnal trend and Drewnick et al., (2004) only observed a weak diurnal pattern.The late afternoon peak reported by the latter study was postulated to be the result of regional pollution episodes coupled with in situ SO 2 photochemical processing.However, disagreement between studies exists in the literature regarding diurnal sulphate behaviour: a pronounced broad morning (07:00 to 12:00) sulphate peak identified by Weimer et al. (2006) was attributed with a rising boundary layer.pNH + 4 and pSO 2− 4 diurnal trends were both insignificant in the summer.This was expected as the pNH + 4 was most likely in the form of either (NH 4 ) 2 SO 4 or NH 4 HSO 4 (Zhang et al., 2006).Conversely, winter pNH + 4 favoured nitrate association (Harrison and MacKenzie, 1990;Wexler and Seinfeld, 1990); thus diurnal variation of pNH + 4 was observed in January.
As temperatures declined, gas phase ammonia concentrations also decreased.Interestingly, the diurnal profile for ammonia also shifted with season: a pronounced morning maximum rush hour peak (09:00 to 10:00) was noted in July which shifted to occur later in the day as the overall ambient temperature dropped: in January, the daily maximum was observed at 18:00.This profile likely implied that NH 3 originated from a local source, possibly vehicle traffic since this maximum did not occur on weekends.Furthermore, significant statistical correlation between NH 3 with CO was found (r=0.60).The morning rise in the NH 3 may not have been observed in January due to increased conversion of NH 3 to NH 4 NO 3 as a result of the colder temperatures.The late afternoon rise in the NH 3 diurnal profile for January appears to coincide with a drop in pNO − 3 and pNH + 4 concentrations.This observation potentially suggests that the same NH 3 from the vehicles that condensed in the morning subsequently evaporated once the temperature rose.Diurnal NH 3 trends have previous been reported in urban centres and were attributed to traffic emissions.Nowak et al. (2006) reported that during the summer in Atlanta, NH 3 concentration increased at 08:00 and remained consistently elevated throughout the day.However, during the winter in New York City, Li et al. (2006) noted a bimodal diurnal ammonia pattern; a rise in NH 3 commenced at 06:00 and attained peak levels at 09:00 which was followed by a gradual decline in concentration until 17:00 when a second peak was noted.
Although vehicle emissions appeared to influence the ammonia diurnal profile, agricultural fertilizer and livestock emissions also likely contributed to measured concentrations as monthly median NH 3 concentrations declined with temperature.This hypothesis was further investigated by comparing the total ammonium concentration (calculated as the sum of pNH + 4 and NH 3 ) variation for the period when livestock and agricultural ammonia emissions were likely contributing (June-October) with that for the months of when minimal livestock or agricultural emissions were assumed present (December 2006to March 2007).Given total ammonium concentrations were significantly lower (p<0.0001)during the colder months (N =4787) than the warmer months (N=5088), with mean total ammonium concentrations and associated standard errors of 3.07±0.3and 5.21±0.3µg m −3 respectively, agricultural and livestock source do potentially influence the total measured ammonium measurements in downtown Toronto.
On weekends, the traffic density passing the sampling site was reduced and this is possibly reflective of the weaker daytime rise in gaseous NH 3 .In addition, a gradual midday decline due to a rise in the planetary boundary layer mixing height, which effectively diluted concentrations, was more apparent, particularly in July.As the depth of the mixing height weakened in the winter as compared with summer months, the degree to which this primary gas emission was diluted during the day was expected to be accordingly reduced.
For the entire duration of the sampling period HNO 3 concentrations were consistently less than those of pNO − 3 .Gaseous nitric acid reached a peak concentration of 1.08 µg m −3 in the early afternoon (12:00 to 13:00) for the month of July coinciding with maximum daily solar radiance and temperatures.Particulate nitrate in July peaked at 2.1 µg m −3 between 10:00 and 11:00, prior to the HNO 3 peak.Diurnal HNO 3 formation and pNO − 3 removal in the summer was the most pronounced diurnal trend observed.HNO 3 was likely formed during the night via N 2 O 5 +H 2 O reaction yielding pNO − 3 in the early morning hours provided sufficient NH 3 was present (Ansari and Pandis, 1999;Weimer et al., 2006).As local vehicle sources further emitted primary NO x and NH 3 in the morning, more NH 4 NO 3 may have been produced through oxidation of the NO x to HNO 3 .As temperature and solar radiance increased throughout the day, gaseous HNO 3 concentrations continued to rise due to the evaporation of morning rush hour derived PM nitrate and the rapid oxidation of NO x .The concentration of both species declined in the afternoon, suggestive of HNO 3 dry deposition and removal by advection (Fisseha et al., 2006).
Particulate nitrate exhibited a strong seasonal dependence: the greatest contribution of particulate nitrate (in the assumed form of ammonium nitrate) to total nitrate (calculated as the sum of HNO 3 and pNO − 3 ) was observed in March (85%) and the lowest fraction was found in August (34%).Furthermore, the proportion of the total PM 2.5 mass, as measured by the TEOM, that was due to pNO − 3 was on average 11 times higher in March than in August.Similar seasonal and diurnal variation were also observed in Zurich where the pNO − 3 content of measured PM 10 was five times greater in March as compared to August (Fisseha et al., 2006).

Conclusions
Comparison of the GPIC with other parallel measurements revealed good temporal agreement demonstrating the potential of this technique.However, numerous challenges were encountered in the continuous operation of the GPIC producing systematic errors in accuracy.Many of the underlying causes were identified and resolved but only after many months of sampling.Application of monthly and seasonal correction factors to particle and gaseous species concentrations, respectively, provided reasonably accurate data while maintaining the high time resolution of the GPIC measurements.Applying similar correction to the gas species data was less successful in resolving disagreement between different instruments.
A difference between TEOM and NAPS Dichot filter measured PM 2.5 mass concentrations was identified, in particular during the winter.The difference between the two measurement techniques was attributed to partial volatilization of nitrate in the TEOM's heated inlet air stream.Variation was normalized by calibrating the TEOM concentration to the NAPS Dichot filter data.Comparison of the calibrated TEOM data with reconstructed GPIC and Sunset Laboratory OCEC PM 2.5 mass concentrations exhibited excellent correlation.
Co-located inter-instrument comparison between the GPIC and API Fluorescent SO 2 Gas Analyzer showed an overall good correlation between the two measurement techniques (r=0.87);calibration challenges with the API SO 2 Analyzer likely contributed to an observed offset.Comparison of the GPIC with a co-located R&P 8400N Nitrate Analyzer (r=0.86) and Aerodyne C-ToF-AMS (r≥0.86)showed overall good correlation between the measurement techniques.
High time resolution PM 2.5 mass reconstruction was possible via the GPIC and Sunset Laboratory OCEC measurements.This enabled rapid PM 2.5 inorganic and organic species and precursor gases fluctuations to be characterized within and between seasons.
The strongest diurnal HNO 3 and pNO − 3 partitioning was observed during the summer: nighttime pNO − 3 formation and morning rush hour derived PM nitrate was converted to nitric acid gas midday as daily temperatures peaked.In the winter, low temperatures and high relative humidity induced gaseous HNO 3 and NH 3 condensation yielding NH 4 NO 3 aerosol.Consequently, the fraction of total nitrate (TNO 3 =HNO 3 +pNO − 3 ) in gaseous form was the lowest during the winter months.
Ammonia and particulate ammonium also demonstrated seasonal diurnal differences.During the summer, NH 3 exhibited a morning rush hour maxima on weekdays.The progression from summer to winter shifted the morning maxima to an afternoon/evening diurnal peak and the overall magnitude of the NH 3 concentrations also decreased.Ammonium only demonstrated diurnal variation in the winter.pNH + 4 and pSO 2− 4 both showed little diurnal variation in the summer; these trends were coupled as the measured pNH + 4 was mostly associated with pSO 2− 4 , in the form of either (NH 4 ) 2 SO 4 or NH 4 HSO 4 .Conversely, winter pNH + 4 favoured nitrate association; thus a diurnal variation in pNH + 4 was only observed during the winter.

Fig. 1 .
Fig. 1.Detailed map of downtown Toronto highlighting the Gage Building sampling location (A), the Wallberg Building site (B), and the Ontario Ministry of the Environment (MoE) downtown sampling location (C).

Fig. 1 .
Fig. 1.Detailed map of downtown Toronto highlighting the Gage Building sampling location (A), the Wallberg Building site (B), and the Ontario Ministry of the Environment (MoE) downtown sampling location (C).

Fig. 2 .
Fig. 2. Ratio of mass concentration as measured by the NAPS Dichot Filter to the TEOM for both the downtown Ministry of the Environment (MoE) and Gage Building sampling sites.Standard deviation for each TEOM ratio is indicated by the error bars.

Fig. 2 .
Fig. 2. Ratio of mass concentration measured by the NAPS Dichot Filter to the mass concentration measured by the TEOM for both the downtown Ministry of the Environment (MoE) and Gage Building sampling sites.Standard deviation for each TEOM ratio is indicated by the error bars.

Fig. 3 .BFig. 3 .
Fig. 3. November 2006 to March 2007, 2 hour resolution least square regressions between raw Ministry of the Environment (MoE) (A) and Gage Building (B) TEOM measurements and reconstructed PM 2.5 mass (grey squares) and reconstructed PM 2.5 mass minus the NH 4 NO 3 mass (black triangles) as measured by the GPIC and OCEC at the Wallberg Building.(N=566)

Fig. 4 .Fig. 4 .
Fig. 4. Comparative analysis of the reconstructed PM 2.5 mass as measured by the GPIC and OCEC instruments in the Wallberg Building to the Ministry of the Environment (MoE) (A) and Gage Building (B) TEOMs for the entire study period: July 2006 to March 2007.(N=722)

Fig. 5 .
Fig. 5. Semi-continuous, 1 hour resolution, SO2 gas concentrations measured by the GPIC and API Fluorescent SO2 Analyzer for the period of January 17 to February 17, 2007.

Fig. 5 .
Fig. 5. Semi-continuous, 1 h resolution, SO 2 gas concentrations measured by the GPIC and API Fluorescent SO 2 Analyzer for the period of 17 January to 17 February 2007.

Fig. 9 .Fig. 9 .
Fig. 9. Stacked fifteen minute resolution inorganic particulate concentrations and two hour resolution organic matter and elemental carbon concentrations as measured during October 2006 presented associated five minute averaged temporal wind direction.Organic matter was calculated as 1.4 times organic carbon concentrations.

Fig. 10 .
Fig. 10.Stacked fifteen minute resolution inorganic particulate concentrations and two hour resolution organic matter and elemental carbon concentrations as measured during January 2007 presented associated five minute averaged temporal wind direction.Organic matter was calculated as 1.4 times organic carbon concentrations.Periods of missing data for specific species (organic matter, elemental carbon, and particulate ammonium) are indicated with horizontal bars along the upper axis.

Fig. 10 .
Fig.10.Stacked fifteen minute resolution inorganic particulate concentrations and two hour resolution organic matter and elemental carbon concentrations as measured during January 2007 presented associated five minute averaged temporal wind direction.Organic matter was calculated as 1.4 times organic carbon concentrations.Periods of missing data for specific species (organic matter, elemental carbon, and particulate ammonium) are indicated with horizontal bars along the upper axis.

Table 1 .
Mean monthly temperatures and Gage TEOM concentrations presented in conjunction with the percentage of volatile species in the reconstructed PM 2.5 mass by the GPIC and Sunset OCEC measurements in the Wallberg Building (WB).

Table 2 .
Intercomparison of studies between GPIC, Particle Into Liquid Samples (PILS), and Partisol 24h Filter nitrate measurement against the R&P 8400N Nitrate Analyzer at rural and urban centres in the United States.All presented slopes have units of (µg m −3 ) R&P 8400N /(µg m −3 ) Other Instrument .