Characterization of Short-Term Particulate Matter Events by Real-Time Single Particle Mass Spectrometry

Single particle measurements were made in Baltimore, Maryland from March to December 2002 using a real-time single particle mass spectrometer, RSMS-3. Particle composition classes were identified that indicated how the aerosol composition changed with time. The results were compared with collocated instruments giving particle number concentrations and size distributions, sulfate, nitrate, organic, and elemental carbon mass concentrations and total mass. Examination of these measurements revealed several particulate matter (PM) events in which the 24 h averaged PM _2.5 mass exceeded 30 μ g/m ³ . Three of these events were studied in further detail by comparing number and mass concentrations obtained by RSMS-3 with standard methods. For all three events, the number concentrations obtained with RSMS-3 and a scanning mobility particle sizer were highly correlated (R ² ∼ 0.7). For the event characterized by a high sulfate mass concentration, the RSMS-3 provided an accurate measure of time-dependent nitrate and carbon mass concentrations, but not for sulfate and total mass. For the two events characterized by high carbon mass concentrations (one from a transcontinental wildfire and the other from stagnation during a period of high traffic), RSMS-3 provided an accurate measure of time-dependent nitrate mass, carbon mass and total mass when the aerosol was not dominated by particles outside the size limit of RSMS-3. While the time dependencies were strongly correlated, the absolute mass or number concentrations determined by RSMS-3 were sometimes off by a constant value, which permitted the relative detection efficiencies of some particle classes to be estimated. Other factors that inhibit reconciliation of mass- and number- based concentration measurements are discussed including the difficulty of detecting ammonium sulfate by laser ablation/ionization and the varying size ranges of different particle measurement methods.

INTRODUCTION

Ambient aerosol physiochemical properties vary by source and atmospheric transformation. Continuous measurements of particulate matter and its chemical components permit the observation of transient events that filter based methods cannot capture. Real-time information about the size distribution and composition of particles in urban air allows for the investigation of source strengths, atmospheric chemical processes, and the health effects these particles induce. Toward that end, continuous monitors capable of measuring mass concentrations including total mass (PM_2.5), sulfate, nitrate, and carbon (both organic and elemental) were deployed at the Baltimore Supersite. These monitors were complemented by instruments capable of measuring number concentrations including a scanning mobility particle sizer (SMPS) and a real-time single particle mass spectrometer (RSMS-3), the latter providing size-dependent chemical composition. The RSMS-3 is a single-particle, dual-polarity, time-of-flight mass spectrometer that has been deployed in several urban sites. (Rhoads et al. 2003; Lake et al. 2003; Phares et al. 2003; Tolocka et al. 2005). In Baltimore, over 380,000 particles were analyzed by RSMS-3 between March and December 2002.

Number concentration measurements from an SMPS are not typically used for reconstructing mass concentrations (Brundish et al. 2005) because of the variation of density and shape of the ambient particles (Morawska et al. 1998). Only a few correlations have been made between particle counting systems and mass based measurements. Sioutas et al. (1999) found that SMPS measurements for monodisperse aerosols were in good agreement with gravimetric analysis for particle sizes greater than 300 nm. However, the SMPS was found to overestimate the mass concentration by a factor of ca. 1.5 for both monodisperse and polydisperse size distributions less than 300 nm. Morawska et al. (1999) examined the correlation between particle mass and number for both ambient and laboratory aerosols. For high concentrations of laboratory aerosols (50–1,000 μg/m³) the correlations were quite high (R² > 0.95, accounting for 95% of the variability). For mass concentrations less than 10 μg/m³, correlations rarely accounted for more than 60% of the variability due to changes in particle density. The SMPS has also been used in conjunction with a mass based instrument, the Aerodyne Aerosol Mass Spectometer (AMS), to determine particle morphology and density by combining the mobility and vacuum aerodynamic diameter for the same particle populations (DeCarlo et al. 2004) and to characterize an aerodynamic particle inlet (Zhang et al. 2004).

Another single particle mass spectrometer, the ATOFMS, has been used to reconcile chemically resolved number and mass concentrations at two locations (Allen et al. 2000; Bhave et al. 2002; Wenzel et al. 2003). Initially, ATOFMS detection efficiencies were calculated from size segregated mass concentration data obtained with a micro-orifice impactor (MOI), and the time-dependent ATOFMS data for particles between 0.3 and 1.8 μ m dia. were found to correlate well with total mass and individual chemical species measurements by other methods (Allen et al. 2000). Subsequently, scaling factors were used to determine an underreported particle class by ATOFMS, presumably ammonium sulfate (Wenzel et al. 2003).

The ability of RSMS-3 to determine mass concentrations has not been fully explored. Unlike the version of the ATOFMS used in the above referenced experiments, RSMS-3 permits single particle measurements down to ca. 0.05 μ m dia., but with limited capability above 1 μ m dia. While this size range provides more complete characterization of fine and ultrafine particles, reconstructed mass concentrations with RSMS-3 may be in error if a majority of the mass resides in particles 1 μ m dia. or larger.

Six particulate matter (PM) events, defined here as time periods when the 24 hour averaged PM2.5 exceeded 30 μ g/m³, were observed in 2002 at the Baltimore Supersite (Park et al. 2005). During these events, a single chemical component, either sulfate or carbon was found to dominate the particulate mass. In addition, the ambient number concentration and RSMS-3 particle hit rate were correspondingly high. The high mass and number concentrations, coupled with the singular nature of the aerosol, provide a unique opportunity to investigate correlations between the RSMS-3, a particle number-based instrument, and other mass- and number-based measurements. Factors that inhibit the reconciliation of number- and mass based measurements are discussed.

EXPERIMENTAL

RSMS-3 measurements at the Baltimore Supersite have been previously discussed (Lake et al. 2003, 2004; Tolocka et al. 2004a, 2004b, 2005). In brief, particles were aerodynamically selected and passed through a series of differentially pumped stages into the mass spectrometer. Individual particles were ablated and ionized in-flight using an excimer laser at 193 nm. Positive and negative ion spectra were collected from 0 to +/−250 m/z for each particle. The positive ion spectra were classified using Art2-a (Carpenter et al. 1991; Phares et al. 2001). Over 99% of the 380,000 particles analyzed could be described by 10 major chemical composition classes (Tolocka et al. 2005). The most prevalent composition class, representing almost 40% of the particles in Baltimore aerosol over the eight-month period, corresponded to internally mixed particles containing organic and possibly elemental carbon, ammonium nitrate and ammonium sulfate (OCANS). A carbon class accounted for approximately 30% of the particles in Baltimore aerosol, and an ammonium nitrate class accounted for approximately 10% of the particles. A potassium/potassium-sodium class (including particles from a wildfire event that will be discussed below) accounted for another 10% or the particles. The remaining 10% of particles were split among several transition and/or heavy metal classes. Because of the difficulty in detecting sulfate in positive ion spectra, no distinct sulfate class was identified by Art 2-a analysis other than OCANS. Instead, particulate sulfate was monitored through the negative ion signature of particles in the various classes described above. Details concerning sulfate detection are given in Lake et al. (2004).

Continuous mass measurements were made with an R&P Model 1400a Tapered Element Oscillating Microbalance (TEOM). Continuous monitors of elemental and organic carbon (R&P Model 5400 East Greenbush, NY), sulfate (R&P Model 8400S, East Greenbush, NY), and nitrate (R&P Model 8400N, East Greenbush, NY) were also employed (Harrison et al. 2004; Park et al. 2005). A Scanning Mobility Particle Sizer (SMPS) (Electrostatic Classifier Model 3080 and CPC Model 3025A; TSI Inc., St. Paul, MN) provided the number concentration and particle size distribution for mobility diameters between 10 and 440 nm as a function of time.

For detection efficiency measurements of laboratory aerosols, polydisperse aerosols of desired composition were generated with a TSI Constant Output Atomizer (Model 3076, TSI Inc., St. Paul, MN). The aerosol output line was split to the RSMS-3 and SMPS. Conductive tubing (1/4″ o.d.) was used to transport the generated aerosol to RSMS-3 and to the SMPS. The SMPS was used to measure the particle size distribution leaving the atomizer. During the analysis of each solution, the distribution was measured at least three times to ensure output consistency. Ethanol or 50/50 ethanol-water solutions containing ammonium nitrate, ammonium sulfate and oleic acid were atomized. A 1:1 oleic acid and ammonium sulfate solution and a separate 1:1:1 oleic acid, ammonium sulfate, and ammonium nitrate solution were analyzed.

RESULTS AND DISCUSSION

Events Description

Six events, designated as “A” through “F” by (Park et al. 2005), were observed in which the 24 h averaged PM_2.5 mass exceeded 30 μ g/m³. Three of these events were studied in more detail to investigate the specific challenges of reconciling RSMS-3 measurements with standard methods. Event “D” began on August 11 and continued to August 14. About 80% of the PM mass was attributable to ammonium sulfate, while 15% was due to organic carbon. The aerosol is believed to have been transported from the Ohio River Valley (Park et al. 2005). High sulfate mass concentrations were observed in Events A, B, and E. Because of their similarity with Event D, they are not discussed here. Event C, characterized by a high carbon mass concentration, was associated with long range transport of aerosol from Canadian wildfires (Sapkota et al. 2005) between the 5th and 8th of July. Event F occurred between the 20th and the 21st of November and was characterized by a high carbon mass concentration owing to stagnant air during high traffic conditions (Park et al. 2005).

RSMS-3 Number and Mass Concentrations

Particle number concentrations were calculated from the rate of single particle data collection with RSMS-3:

where N is the number of particles (either total number of particles or the number of particles in a particular composition class) detected per time interval, Δ t; V is the volumetric flow rate of aerosol into the mass spectrometer; and log (σ_g)² represents the width of the transmission function between 16% and 85% of the size distribution. The detection efficiency, η, is defined as the fraction of particles entering the inlet that are ablated and detected. Several studies have shown that the detection efficiency is size and composition dependent (Allen et al. 2000; Gross et al. 2000; Kane and Johnston 2000; Kane et al. 2002; Wenzel et al. 2003). There are two main contributions to the detection efficiency:

the sampling rate, S_r, which is the rate at which particles are sampled through the inlet and into the laser beam when the laser fires, and the ablation efficiency, E_a, which is the fraction of particles the fraction of particles in the laser beam when the laser fires that are actually ablated and detected (Kane and Johnston 2001). Both particle size and morphology (the latter ultimately given by the chemical composition) influence S_r. In general, the largest sampling rates are achieved with large, spherical particles. Spherical particles are more tightly focused into a collimated beam than irregularly shaped particles, and large particles are less subject to Brownian motion that counteracts aerodynamic particle focusing that is achieved. Similarly, E_a is both size and composition dependent. Small particles are less efficiently ablated since fewer ions are produced per particle, making the likelihood of detection lower. Particles composed of ammonium nitrate or elemental carbon give the highest values of E_a. Particles composed of organic carbon or inorganic salts give values of E_a that range from 2–5 times smaller than ammonium nitrate. Particles composed of ammonium sulfate have such a low value of E_a (orders of magnitude smaller than other compositions) that detection is normally determined by the level of impurities, for example, alkali metals, present in the particle (Kane and Johnston 2000; Kane et al. 2001, 2002). If the impurity level is high, then particles containing ammonium sulfate are efficiently detected. If the impurity level is low, that is, the particle is nominally “pure” ammonium sulfate, it is unlikely the particle will be detected. For this reason, the ability of laser ablation in general and RSMS-3 in particular to detect ammonium sulfate in ambient particles is determined by the internal/external mixing state of the aerosol.

Size-dependent detection efficiencies for RSMS-3 averaged over all ambient particle composition classes in Baltimore aerosol are shown in Figure 1 (Lake et al. 2003). Not surprisingly, the detection efficiency decreases with decreasing particle size. Also shown in Figure 1 are detection efficiencies for laboratory generated particles having related compositions—ammonium nitrate, oleic acid (representative of an organic particle), and mixtures of oleic acid with ammonium sulfate and/or ammonium nitrate. As expected from previous work with an earlier version of RSMS (Kane and Johnston 2000), ammonium nitrate particles give higher detection efficiencies than the ambient particle average because the ablation efficiency is inherently large and the hygroscopic character of these particles promotes a nearly spherical morphology. The difference between the ammonium nitrate and ambient particle detection efficiencies is relatively low at small particle sizes because ambient particles in this size range are composed primarily of chemical species that are easily detected (ammonium nitrate, elemental carbon, potassium). As expected, particles composed of oleic acid with or without ammonium sulfate/nitrate are much smaller than ambient particles at small particle sizes as these chemical compositions give lower ablation efficiencies. As the particle size increases, internally mixed OCANS (organic carbon, ammonium nitrate and sulfate) particles become the dominant composition type in ambient aerosol. Therefore, it is not surprising that as the particle size increases, the ammonium nitrate detection efficiency increases more rapidly than that for ambient particles. What is surprising is that the detection efficiencies of oleic acid particles with or without ammonium sulfate and nitrate eventually become greater than those for ambient particles as the particle size increases. The large detection efficiencies for the laboratory particle compositions most likely are the result of the liquid oleic acid component that maintains a spherical particle morphology. In contrast, ambient OCANS particles are more likely to contain solid organic matter that combined with solid ammonium sulfate and/or nitrate do not produce a spherical morphology. The high detection efficiencies of laboratory-generated oleic acid—ammonium sulfate and/or nitrate particles suggests that ambient OCANS particles are also efficiently detected.

FIG. 1 Detection efficiency versus vacuum aerodynamic diameter (D_va) for ambient Baltimore aerosol (Lake et al. 2003) and four laboratory generated aerosols (ammonium nitrate; oleic acid; oleic acid and ammonium sulfate; and oleic acid, ammonium nitrate, and ammonium sulfate).

Within the constraint of particle-to-particle variations of the detection efficiency, the direct comparison of RSMS-3 number concentration measurements with those taken by the SMPS requires conversion to mobility diameter:

where the ratio of the vacuum aerodynamic (D_va) and electrical mobility (D_m) diameters is given in terms of the particle density ρ_p and shape factor χ, and C_c (D) is the Cunningham slip correction factor calculated using the particle diameter within the parentheses. The dependence of χ on the flow regime, or Knudsen number, has been neglected (Jimenez et al. 2003; DeCarlo et al. 2004). Since the observed size distribution with RSMS-3 is the composite of several particle composition classes (Tolocka et al. 2005), the correlation between RSMS-3 and SMPS particle size distribution functions depends on the particle composition as indicated in Equations (1) and (3). For the PM events considered in this work, the aerosol is dominated by one or a few particle classes. As will be seen, the main effect of the chemical composition dependencies of Equations (1) and (3) under these conditions is to induce a constant offset between the number and/or mass concentrations of RSMS-3 and the other monitors.

The mass concentration, (dM/dlog D _p ), of the entire aerosol or an individual particle class can be determined from the appropriate RSMS-3 number concentration:

Because the particle sizing is based on aerodynamic selection (Phares et al. 2002), conversion to volume equivalent diameter is required to directly compare these measurements with those taken by the TEOM. Here, once the vacuum aerodynamic diameter is known, as calculated in Equation (3), it is converted to it volume equivalent diameter through the following

where ρ_o is the standard density (1 g/cm³). Once D_ve has been calculated, integrating Equation (4) over the range of particle sizes analyzed by RSMS-3 gives the total mass of the entire aerosol or an individual particle class (DeCarlo et al. 2004; Slowik et al. 2004). The sum of these particle classes gives the overall mass to directly compare to the TEOM. Note that the densities of ammonium sulfate, ammonium nitrate and carbon were taken to be 1.77, 1.73, and 2.00 g/cm³, respectively. The shape factors for ammonium nitrate and sulfate were assumed to be 1.08, while the shape factor for elemental carbon will be discussed below.

Sulfate PM Event

As discussed above, the ability of RSMS-3 to detect ammonium sulfate is determined by the mixing state of the aerosol. This ability is explored for Event “D,” a three day period in August 2002 in which both sulfate mass and PM_2.5 were quite high. Figure 2 shows how number and mass concentrations determined by RSMS-3 compare to standard methodology. Figure 2a compares the total particle number concentrations obtained with RSMS-3 and SMPS. The two measurements are highly correlated (R² = 0.77) suggesting that the great majority of ambient particles during this event were efficiently detected by RSMS-3. Figure 2b compares the total mass concentrations obtained with RSMS-3 and TEOM during the same time period. While the mass concentrations are of the same order of magnitude, the correlation is poor (R² = 0.26). A similarly poor correlation is obtained when the TEOM measurements are compared with mass concentrations determined from the SMPS size distribution (data not shown).

FIG. 2 Number and mass concentration measurements during Sulfate PM Event “D.” (A) RSMS-3 vs. SMPS. (B) RSMS-3 vs. TEOM.

The combination of a high correlation with number concentration and a low correlation with mass concentration suggests that the discrepancy is associated with a small number of large diameter particles that are not detected with RSMS-3 or SMPS. Figure 3 shows the size distributions during episode D at 0700 EST on 13 August. Here, the size bins from the SMPS were co-added to account for the differences in the size distributions of each size bin (Phares et al. 2002). As shown in the figure, the difference may arise from the upper size ranges for particle detection −2.5 μ m for TEOM versus 0.4 μ m for SMPS and ca. 0.6 μ m for RSMS-3. Alternatively, it is clear that the RSMS-3 did not detect some particles between 0.4 and 0.6 μ m dia. that were nominally “pure” ammonium sulfate.

FIG. 3 Number concentrations compared from the SMPS to the RSMS-3 plotted versus mobility diameter (D_m).

Correlations among mass concentrations from RSMS-3 particle classes and several continuous monitors are shown in Figure 4. The OCANS class encompasses internally mixed particles containing secondary aerosol components including organic carbon, ammonium sulfate and ammonium nitrate. One of these components, ammonium nitrate is also found in a separate particle class, although the mass concentration of the ammonium nitrate class is almost an order of magnitude smaller than that of the OCANS class during this PM event. Figure 4a compares the mass concentration of the OCANS and nitrate classes to the sum of the corresponding continuous monitors—sulfate, nitrate and organic carbon. Note that the sum of these monitors is obtained by multiplying the organic carbon value by 1.4. While there exists some debate about the absolute magnitude of this multiplier (Turpin et al. 2000), it appears to reconcile the speciation monitors with the TEOM data. The correlation between the RSMS-3 classes and the sum of the three continuous monitors is adequate (R² = 0.55) and the combined mass of the OCANS and nitrate classes is consistently smaller than the sum of the continuous monitors by a factor of ca. 2.5. The nature of this discrepancy revealed in Figure 4b, Figure 4c, Figure 4d. In Figure 4b, the sulfate mass determined by RSMS-3 and a continuous monitor are compared. The RSMS-3 values were obtained from the total mass of all particles showing indicator ions for sulfate in either the positive or negative ion spectra. If anything, these values would be expected to give an overestimate of the sulfate mass since the particles must contain other chemical species for efficient detection (and indeed other species are observed in the mass spectra). Nonetheless, the sulfate mass determined in this manner grossly underestimates the values obtained from the continuous monitor by a factor of 8 and the correlation is poor (R² < 0.1). As with the TEOM correlation in Figure 2b, RSMS-3 is missing a significant portion of the mass, because it exists in particles larger than 0.8 μm dia. and/or it exists in nominally “pure” ammonium sulfate particles that are not detected.

FIG. 4 Mass concentrations from RSMS-3 and continuous monitors. (A) RSMS-3: sum of OCANS and nitrate classes; continuous monitors: sum of sulfate, nitrate and organic carbon (×1.4). (B) RSMS-3: sum of particles exhibiting sulfate in negative ion spectra; sulfate monitor. (C) RSMS-3: ammonium nitrate class; nitrate monitor. (D) RSMS-3: carbon class; elemental carbon monitor.

While sulfate detection is clearly a problem, other chemical species are not. Figure 4c compares the nitrate mass determined from the ammonium nitrate particle class detected by RSMS-3 with the continuous monitor. Although nitrate is a minor component of ambient aerosol during this time period, the correlation is good (R² = 0.65). It should be noted, however, that the nitrate mass determined from the ammonium nitrate class is consistently higher than the nitrate monitor by about a factor of 2. The good correlation indicates that most of the ammonium nitrate in the aerosol is encompassed by the ammonium nitrate class. Although OCANS particles do show signature ions for ammonium nitrate along with other chemical species, the absolute amount of ammonium nitrate in these particles is relatively low. Other components in these particles, primarily organic carbon and ammonium sulfate, are less sensitively detected and therefore are likely to constitute a much larger portion of the particle mass. The overestimate of nitrate mass based on the ammonium nitrate particle class most likely arises from the high detection efficiency of these particles (Figure 1) and the presence of other chemical components in the particles. These other components are less sensitively detected by laser ablation and give much weaker ion signals in the mass spectra. The factor of 2 difference between RSMS-3 and continuous nitrate mass measurements suggests that the detection efficiencies (η in Equation [1]) for the ambient ammonium nitrate particle class are 2 times greater than the composition-averaged ambient particle detection efficiencies in Figure 1. The magnitude of this difference is consistent with the laboratory particle measurements in Figure 1.

Finally, the elemental carbon mass is compared in Figure 4d for the continuous monitor and the carbon particle class detected by RSMS-3. Although elemental carbon is a minor component of the ambient aerosol during this time period, the correlation is reasonably high (R² = 0.6) and the absolute values of the mass agree well. The agreement among the absolute mass values suggests that the detection efficiencies for the carbon particle class are approximately equal to the composition-averaged detection efficiencies in Figure 1. It should be noted that a constant shape factor, χ, was used to calculate these mass concentrations. When a variable shape factor (Slowik et al. 2004) was used for fresh combustion emissions, the amount of elemental carbon in these particles is was overestimated by a factor of 8 using RSMS-3, clearly a not realistic result.

Several key observations can be drawn from the sulfate PM event. This event represents a worst case situation for RSMS-3 analysis as ammonium sulfate is poorly detected and a considerable amount of the aerosol mass may exist beyond the upper size limit of the instrument. Nonetheless, RSMS-3 provides an accurate measure of time-dependent particle number concentrations. In addition, the time-dependent nitrate and elemental carbon mass concentrations can be determined even though these species represent minor components (ca. 10% each) of the total aerosol mass.

Carbon PM Event #1 (Stagnation During Heavy Traffic)

Event “F” occurred in November 2002 and was characterized by a high carbon mass loading as stagnant air and high traffic contributed to an elevated PM_2.5. Figure 5 shows how number and mass concentrations determined by RSMS-3 compare to standard methods during this time period. As with the sulfate event, number concentration measurements obtained with RSMS-3 and SMPS are highly correlated (R² = 0.68) as shown in Figure 5a. The smaller values obtained with RSMS-3 on the last day of the event are due to the underestimation of the number concentrations using the RSMS-3. During the latter part of the event the SMPS consistently measured number concentrations throughout the size range a factor of 1.5 higher than the RSMS-3. On this day, a significant number of particles were observed with the SMPS below the lower size limit of RSMS-3 (30 nm dia.). The RSMS-3 mass concentration correlates well with the TEOM data, R² = 0.74 as shown in Figure 5b. In particular, a large spike in mass concentration was recorded by both methods around mid-day on 20 November. Note that the SMPS number concentration data and the TEOM mass concentration data for the event were also associated (R² = 0.62), indicating that most of the particle mass was found in the submicron size range.

FIG. 5 Number and mass concentration measurements during Carbon PM Event “F” (stagnation). (A) RSMS-3 versus SMPS. (B) RSMS versus TEOM.

Correlations between RSMS-3 and continuous monitor speciation data are shown in Figure 6. Figure 6a compares the sum of the organic carbon, ammonium nitrate and ammonium sulfate continuous monitors (as described above for the sulfate event) with the combined mass concentration of the OCANS and ammonium nitrate classes observed with RSMS-3. Here the correlation is quite high (R² = 0.81) indicating that RSMS-3 efficiently detects most particles containing these components and that the detection efficiencies of the OCANS particles are approximately equal to the composition-averaged detection efficiencies in Figure 1. As with the sulfate event, Figure 6c and Figure 6d show that the nitrate particle class and carbon particle class mass concentrations correlate well with the nitrate (R² = 0.75) and elemental carbon (R² = 0.72) continuous monitors. Unlike for the previous event, no offset of the nitrate concentration is observed here as the majority of the nitrate mass is in the OCANS particle class rather than the ammonium nitrate particle class. Ammonium sulfate determined from RSMS-3 data (procedure described in the sulfate event section) shows a poor correlation with the sulfate monitor (R² < 0.1), again highlighting the difficulty in extracting sulfate mass concentration data from laser ablation mass spectra. While sulfate speciation is not possible, the high correlations in Figure 5b and Figure 6a show that the contribution of ammonium sulfate to total mass can be accounted for when the sulfate exists in internally mixed particles (i.e., OCANS) that do not exceed the upper size limit of RSMS-3.

FIG. 6 RSMS-3: sum of OCANS and nitrate classes; continuous monitors: sum of sulfate, nitrate, and organic carbon (×1.4). (B) RSMS-3: sum of particles exhibiting sulfate in negative ion spectra; sulfate monitor. (C) RSMS-3: ammonium nitrate class; nitrate monitor. (D) RSMS-3: carbon class; elemental carbon monitor.

Carbon PM Event #2 (Wildfire)

Event “C” presents a special case because of the unique composition of the aerosol. While OCANS was normally the largest particle class in both number and mass concentration in Baltimore aerosol, during this time period the potassium and potassium-OCANS (K-OCANS) classes were larger owing to the impact of wildfire aerosol. Over the eight-month period, particles containing potassium were responsible for 10% of the total aerosol mass, which corresponded to ca. 1 μ g/m³. During the wildfire PM event, the potassium and K-OCANS classes together exceeded 50 μg/m³ and represented 90% of the total mass. It should be noted that the K-OCANS class was not observed any other time during the eight-month study. This particle class most likely corresponds to wildfire aerosol transformed by the addition of secondary components during long range transport.

Figure 7 shows how the total number and mass concentrations obtained from RSMS-3 compare to standard methods. Both are reasonably well correlated, R² = 0.71 and 0.65 respectively for number (Figure 7a) and mass (Figure 7b) concentrations. The RSMS-3 number concentration is about 50% larger than the SMPS number concentration, which most likely reflects higher detection efficiencies of potassium particles relative to the composition-averaged values from Figure 1 that were used in the calculation. It is interesting to note that during this time period, mass concentrations derived from the SMPS measurements did not correlate at all with the TEOM (R² < 0.1), despite the reasonably high correlation of the RSMS-3 data with both SMPS and TEOM measurements. The better correlation of RSMS-3 with the TEOM is a consequence of the greater size range of RSMS-3 relative to the SMPS coupled with the fact that most of the aerosol mass was associated with particles larger than the upper size limit of the SMPS. Nonetheless, Figure 6b shows that RSMS-3 under-reports the total mass during the peak of the event on 6 July, presumably because much of the aerosol mass exists in particles larger than the upper size limit of RSMS-3. As with the other events, the correlation between the sulfate mass determined by RSMS-3 and the continuous monitor was poor (R² = 0.36). Interestingly, the carbon class did not track well with the elemental carbon mass (R² = 0.35), perhaps because such a large amount of elemental carbon resides in the potassium and K-OCANS classes. We suspect that the carbon class, which contains mostly small particles that show only carbon cluster ions in the mass spectra (Tolocka et al. 2005), is indicative of local emissions only. No distinct nitrate class was observed during this time period, although nitrate was evident in mixed composition particles.

FIG. 7 Number and mass concentration measurements during Carbon PM Event “C” (wildfire). (A) RSMS-3 versus SMPS. (B) RSMS versus TEOM; also shown are mass concentrations of four particle classes identified by RSMS-3.

CONCLUSIONS

For all three PM events, particle number concentrations obtained with RSMS-3 and SMPS correlated well, R² ∼ 0.7. In contrast, mass concentrations obtained with RSMS-3 and TEOM did not always correlate well. When sulfate was a minor component of the aerosol (Events “C” and “F”), the discrepancy was small and appeared to arise from the smaller upper particle size limit of RSMS-3 (0.8 μm dia.) relative to the TEOM. When sulfate was the major component of the aerosol, RSMS-3 could not be used to obtain the total mass concentration owing to inefficient detection of particles composed primarily of ammonium sulfate in combination with a limited ability to characterize particles larger than 0.8 μ m dia. Nonetheless, even during periods of exceptionally high sulfate mass concentration, the nitrate and carbon particle classes identified by RSMS-3 gave mass concentrations that correlated well with the nitrate and elemental carbon continuous monitors. These correlations were obtained with a simple procedure based on Art2-a particle classification and particle detection efficiencies averaged over the entire ambient aerosol. In some situations, a constant offset between RSMS-3 and the continuous monitors was observed which permitted the relative detection efficiencies of different particle composition classes to be estimated. Based on these offsets, it appears that the detection efficiencies of ammonium nitrate particles are two times larger, potassium-organic particles are 1.5 times larger, OCANS and elemental carbon particles are approximately equal to, and ammonium sulfate particles are much less than the composition-averaged detection efficiencies reported here and previously.

Although the research described in this article has been funded by the United States Environmental Protection Agency through grant number R82806301, it has not been subjected to the Agency's required peer and policy review and therefore does not necessarily reflect the views of the Agency and no official endorsement should be inferred.