Continuous and Semicontinuous Monitoring Techniques for Particulate Matter Mass and Chemical Components

By Solomon, Paul A Sioutas, Constantinos

ABSTRACT The U.S. Environmental Protection Agency (EPA) established the Particulate Matter (PM) Supersites Program to provide key stakeholders (government and private sector) with significantly improved information needed to develop effective and efficient strategies for reducing PM on urban and regional scales. All Supersites projects developed and evaluated methods and instruments, and significant advances have been made and applied within these programs to yield new insights to our understanding of PM accumulation in air as well as improved source-receptor relationships. The tested methods include a variety of continuous and semicontinuous instruments typically with a time resolution of an hour or less. These methods often overcome many of the limitations associated with measuring atmospheric PM mass concentrations by daily filter-based methods (e.g., potential positive or negative sampling artifacts). Semicontinuous coarse and ultrafine mass measurement methods also were developed and evaluated. Other semicontinuous monitors tested measured the major components of PM such as nitrate, sulfate, ammonium, organic and elemental carbon, trace elements, and water content of the aerosol as well as methods for other physical properties of PM, such as number concentration, size distribution, and particle density. Particle mass spectrometers, although unlikely to be used in national routine monitoring networks in the foreseeable future because of their complex technical requirements and cost, are mentioned here because of the wealth of new information they provide on the size-resolved chemical composition of atmospheric particles on a near continuous basis. Particle mass spectrometers likely represent the greatest advancement in PM measurement technology during the last decade. The improvements in time resolution achieved by the reported semicontinuous methods have proven to be especially useful in characterizing ambient PM, and are becoming essential in allowing scientists to investigate sources of particulate pollution and to probe into the dynamics and mechanisms of aerosol formation in the atmosphere.

INTRODUCTION

Numerous epidemiological studies have found persistent associations between ambient concentrations of particulate matter (PM) and significant adverse health effects.1 An improved understanding of factors influencing human exposure to PM is important for the development of effective control strategies designed to reduce the health impacts of airborne particulate matter.2 This requires a better understanding of the composition and physical properties of PM^sub 2.5^ (particles less than 2.5 [mu]m aerodynamic diameter [AD]), which depends on high-quality measurements of size, concentration, composition, other parameters of PM^sub 2.5^, and related species.

The U.S. Environmental Protection Agency (EPA) responded to this need with an ambient monitoring research program, commonly referred to as the Particulate Matter Supersites Program.3 The primary goals of the program are to provide unprecedented physicochemical characterization of ambient PM that contributes to a better understanding of PM sources and formation mechanisms, support health effects and exposure research, and conduct development and testing of novel state-of-the-art sampling methods of PM mass and chemical speciation. Atlanta, GA, and Fresno, CA, were chosen as initial Supersites Projects4,5 for Phase I on the basis of their ongoing and planned research activities and their distinctively different airsheds. In January 2000, Phase II started with Supersites Projects in Baltimore, MD; Fresno, CA; Houston, TX; Los Angeles, CA; New York, NY; Pittsburgh, PA; and St. Louis, MO.

This paper focuses on the third goal of testing methods, specifically in this case, for near continuous measurement of physical and chemical PM properties. Many of the methods developed and/or tested through the Supersites Program are capable of resolving PM size and composition at high time resolution (hourly or shorter), but had not been thoroughly evaluated or previously deployed in regulatory monitoring networks. The need to develop monitors that measure particle properties in relatively short time intervals is important for improving our understanding of adverse health impacts, atmospheric chemistry, and sources of PM. This is because sources, meteorology, and atmospheric processes of PM pollutants often vary on time scales substantially shorter than 24 hr. The daily averaging times used in many current networks tend to smooth out much of this variability, which limits our understanding of the factors influencing PM accumulation in air and exposures that result in the adverse impact of PM. Moreover, attempting to routinely obtain a better time resolution for ambient particle concentrations for large monitoring networks is presently impractical with the traditional time-integrated measurements that are based on field sampling and subsequent laboratory analysis of particle mass or chemical composition.

The purpose of this paper is to review results from the evaluation of continuous and semicontinuous methods for the measurement of mass, chemical composition, and physical properties of PM with an emphasis on the question, "Are there advanced monitoring techniques that should be instituted into routine networks so that we will have better information in a few years?" Many of the studies reviewed in this paper have been published in one of the special journal issues on research performed within the Supersites Program,6-9 and although this review is not limited to those papers, it is not an exhaustive review of the work conducted over the last 5-7 yr.

The first part of this paper covers methods measuring physical properties of ambient PM. One section describes development and testing of continuous methods for the direct measurement of mass concentration. The aim of these methods was to minimize potential positive or negative sampling artifacts (e.g., absorption or volatilization of semi-volatile compounds) associated with collecting atmospheric PM on a filter. Also reviewed in this section are mass monitors developed for PM size ranges different from PM^sub 2.5^, including ultrafine and coarse PM mass. In the following section, the indirect measurement of mass concentration is described from the combination of size distribution data obtained by different methods. Lastly, the first part reports on performance studies of a new water-based condensation particle counter (WCPC) for total number concentration, an approach to measure the density of particles in a given size range, and a system to measure the water content of ambient aerosol by comparison of ambient (wet) and dried size distributions.

The second part describes near continuous methods related to the determination of the chemical composition of PM in ambient air. Reviewed in this section are semicontinuous methods for sulfate, nitrate, elemental carbon (EC), organic carbon (OC), and trace elements. In addition, particle mass spectrometry (MS) methods are briefly mentioned.

PHYSICAL PROPERTIES OF PM

Continuous Gravimetric Methods for Mass Concentration

Continuous and integrated methods for mass have been developed with two different objectives. Some methods were developed, at least initially, to be comparable to the Federal Reference Method (FRM) for PM^sub 2.5^ mass. These are methods that might be considered as Federal Equivalent Methods (FEMs) and provide regulatory related data on a daily basis or even an hourly basis. However, the FRM is a regulatory standard and not an analytical standard (Fehsenfeld et al.10) and has been found to be influenced by sampling artifacts (e.g., loss of semi-volatile material) under certain conditions as described later in this paper. Other methods, and often the goal of the continuous methods developed and evaluated within the Supersites Program, include measuring the actual PM mass as found in ambient air, by accounting for loss of semi-volatile species, water retention, or other sampling and analysis artifacts. Methods based on both types of approaches are described below.

Tapered Element Oscillating Microbalance (TEOM). The TEOM monitor provides a continuous measure of collected mass.11 It uses a filter for particle collection and is therefore subject to the same artifact and interference problems that other filter-based methods suffer, including negative artifacts from the loss of semi-volatile material collected during sampling, and positive artifacts from absorption or adsorption of gaseous components on deposited particles and/or the filter media. The original version of the TEOM was operated with the filter heated to 50 [degrees]C to eliminate or limit the interference from particle-bound water.11,12 Heating to a controlled temperature is also necessary to make the response of the sensor, which is sensitive to temperature, more stable, resulting in reproducible measurements of ambient coarse PM (PM^sub 10^) or PM^sub 2.5^. However, the heating results in enhanced losses of semi- volatile material from the collected particles.12 Real-Time Ambient Mass Sampler (RAMS). To address the issue of semi-volatile losses during sampling with the TEOM, a real-time monitor was developed using a TEOM with a "sandwich" filter, which was designed to retain semi-volatile material evaporating from collected particles. 13-15 The RAMS13 represents the first attempt to measure PM^sub 2.5^ mass free of both positive and negative sampling artifacts. The complexity of the system described below begins to suggest how difficult it is to measure artifact-free PM^sub 2.5^ mass.

The RAMS removes gas-phase components from the stream of sampled air to avoid positive artifacts on the TEOM filter and then uses a reactive filter to capture any semi-volatile material that may evaporate from the particles during sampling. This is accomplished in multiple steps. First, a 2.5-[mu]m cut point cyclone is followed by a particle concentrator that removes 70% of the gas-phase, thereby concentrating fine particles.15-17 An annular denuder then removes nitrogen dioxide (NO^sub 2^) and part of the ozone (O3) from the remaining gas-phase. The denuder is followed by a Nafion diffusion dryer to remove water. The sample flow is then split into two equal parts and delivered to two nearly identical sampling lines, one of which is preceded with a quartz-fiber (QF or quartz) filter and represents a blank for online correction. Both lines then include a Brigham Young University Organic Sampling System (BOSS) diffusion denuder17 to remove gas-phase compounds that can be absorbed by charcoal, followed by an additional annular denuder to remove NO^sub 2^ formed in the BOSS denuder and another Nafion diffusion dryer. The sample stream finally passes the TEOM sandwich filter, in which a Teflon filter is backed up by a charcoal- impregnated filter (CIF) to adsorb ammonium nitrate (NH^sub 4^NO^sub 3^) and semi-volatile organic compounds lost from particles during sampling.

The RAMS has been evaluated in a number of locations where semi- volatile material was a significant fraction of the ambient PM. Two of these sets of studies are highlighted below, and additional discussions are given in following sections in comparison to other samplers where the RAMS is considered the historical reference sampler. The RAMS was evaluated in Salt Lake City, UT,18 during three intensive sampling periods (winter 1999- 2000, summer 2000, and winter 2000-2001). In these studies, a RAMS was collocated with a TEOM monitor (50 [degrees]C in summer and 30 [degrees]C in winter) and a Particle Concentrator-Brigham Young University Organic Sampling System (PC-BOSS19) diffusion denuder-integrated sampler. The PC-BOSS is the filter-based, time-integrated counter part to the RAMS and uses filters rather than a TEOM to collect particles. Mass is determined gravimetrically in the laboratory. Excellent agreement, based on linear regression of 24-hr averaged data across all three study periods (R^sup 2^ = 0.90, slope of 1.00 +- 0.02, and zero intercept), was observed between the RAMS and sum of the species from the PC-BOSS. The sum of the species as measured by the PC-BOSS provides a good estimate of nonvolatile and semi-volatile PM mass excluding water (see Long et al.18 and references within). Similar regression statistics were observed for shorter average sampling periods as well. One-hour average PM^sub 2.5^ RAMS versus TEOM data showed statistically significant differences between the two methods with lower values reported by the TEOM by as much as 31% for samples collected during the winter study and 42% for samples collected during summer with R^sup 2^ in the range of 0.4-0.6. These differences were due to a loss of semi-volatile material from the TEOM sampler, which as described by Long et al.18 and discussed later in this paper, appeared to be a reasonable estimate of nonvolatile PM^sub 2.5^ mass. Filter-based FRM PM^sub 2.5^ concentrations were found to under-report PM^sub 2.5^ only during warm dry periods. Loss of semi-volatile material from the TEOM when operated at 50 [degrees]C is well established in the literature and has motivated the research community and the manufacturer to implement additional measures to ensure improved mass measurements by the TEOM by minimizing losses of semi-volatile material and the influence of water. These methods are discussed in the following sections.

The RAMS also was tested in a series of field studies by comparing its concentrations to those obtained with a TEOM and a PC- BOSS.13 These studies occurred in the dry winter climate of Provo, UT; in the humid summer climate of Philadelphia, PA; and in Atlanta, GA, during the 1999 Atlanta Supersites project.5 In Atlanta, the TEOM was heated at 30 [degrees]C and preceded by Nafion dryer to reduce relative humidity (RH) to below 40%. At the other two sites, the TEOM was heated to 50 [degrees]C without a dryer. On average, the TEOM reported 82% of RAMS PM^sub 2.5^ concentrations in Provo and Atlanta, whereas in Philadelphia, the TEOM varied more, measuring 50-85% of RAMS concentrations (see Table 1). The higher content of semi-volatile material found in Philadelphia indicated that the higher PM^sub 2.5^ mass determined with the RAMS was due to both semi-volatile nitrate and organic material not measured by the TEOM.

Sample Equilibration System (SES) TEOM. To minimize volatilization associated with the 50 [degrees]C TEOM, it was operated at 30 [degrees]C; however, water removal was limited, and in a variety of environments 30 [degrees]C was not sufficient for stable operation. To help overcome these problems, a SES was developed in which the 30 [degrees]C TEOM was preceded by a diffusion dryer.20 When sampling generated particles of ammonium sulfate, copper sulfate, and potassium nitrate in the laboratory, Schwab et al.21 observed that the 30 [degrees]C TEOM with a dryer clearly responded to step changes in humidity; however, the mass gain was less than that without the dryer and just slightly higher than the bare tapered element sensor with no filter. Results also indicated that most of the mass gain appeared to be due to water collected by the filter because only a slight increase in mass was observed with a particle-laden filter as compared with a clean filter. The 30 [degrees]C SES-equipped TEOM was even less sensitive to changes in humidity at a lowered flow rate of 1 min^sup -1^ as compared with the usual flow rate of 3 min^sup -1^, likely because of the Nafion dryer being more efficient at removing water at the lower flow rate.

The SES TEOM, RAMS, and Continuous Ambient Mass Monitor (CAMM22) were operated in comparison studies in Atlanta, Baltimore, and Pittsburgh by Lee et al.23 The CAMM is described later in this paper. Among the continuous samplers, the SES TEOM was in best agreement with the FRM mass concentrations (with R^sup 2^ >/= 0.9 and slopes close to 1 at all three sites), suggesting that both had similar losses of semi-volatile materials. This is consistent with results from other studies that reveal the TEOM loses semi-volatile aerosol components during sampling. The RAMS appeared to measure total ambient PM, including semi-volatile materials. The mass differences observed between the RAMS and either the CAMM or SES TEOM could be accounted for by the loss of NH^sub 4^NO^sub 3^ and semi-volatile organic material.

Lee et al.24 also evaluated the SES TEOM (at 30 [degrees]C) during the summer in Houston, TX, and winter in Seattle, WA, and compared it to collocated RAMS, CAMM, and filter-based samplers. Measurements were obtained after 30 or 60 min of sampling. Reasonably good agreement was observed between pairs of the continuous mass samplers (see Table 2). RAMS measurements had slightly higher, but comparable PM^sub 2.5^ mass to the integrated filter measurements in warm and dry conditions in Houston (slope = 1.10 at R^sup 2^ = 0.89), whereas RAMS PM^sub 2.5^ mass concentrations appeared to be slightly lower, but comparable to filter-based mass during cold and humid periods in Seattle (slope = 0.92 at R^sup 2^ = 0.78). Similar dependence on RH and temperature was found for differences between RAMS and SES TEOM in Houston, which suggested that the loss of semi-volatile NH^sub 4^NO^sub 3^ contributed to the mass difference. An explanation could be that volatilization of semi-volatile materials from the filter-based measurements of PM^sub 2.5^ is expected to decrease during the colder winter period.18 The results also corroborate the earlier finding that setting the TEOM at a lower temperature than the standard configuration would still lose semi-volatile materials.

Differential TEOM. The Differential TEOM (D-TEOM) was developed by Patashnick et al.25 to further minimize positive and negative sampling artifacts, including effects observed with humidity. Initially, the system consisted of an inlet, dryer, two electrostatic precipitators (ESPs), and two TEOM monitors in parallel. In its commercial form, ambient aerosol is sampled through a standard inlet followed by a diffusion dryer, which is followed by an ESP and then a single TEOM. When the ESP is turned on, all particles are collected by the ESP26 and the mass change detected by TEOM is due to potential negative or positive sampling artifacts as described above. The ESP is switched on and off with a switching time that is long enough for measurable volatilization to occur but sufficiently short enough not to allow influence by atmospheric changes. These assumptions were confirmed for NH^sub 4^NO^sub 3^.27 In these laboratory studies, the dynamics of volatilization from the D-TEOM was examined at two operating filter temperatures: 30 [degrees]C and 35 [degrees]C. Results showed that losses of NH^sub 4^NO^sub 3^ from the TEOM filter occurred on a time scale that was longer than the 5-min cycle time used by the D-TEOM. This was important because it established that vaporization measured during the alternate 5-min periods could be used effectively as a reference baseline value for particle mass measurements. Thus, the D-TEOM is self referencing,25 where the measurement with the ESP turned on was used to correct the prior measurement of PM when the EPS was off. This configuration has been tested in several studies, as described below. Schwab et al.21 also tested a D-TEOM during their laboratory studies of the SES TEOM. Their analysis indicated that the D-TEOM could be a robust technique for the continuous real-time measurement of ambient aerosol mass, even in the presence of semi-volatile components and condensable gases, such as sulfur dioxide (SO^sub 2^) and NO^sub 2^. Figure 1 shows the response of the D-TEOM to laboratory-generated NH^sub 4^NO^sub 3^. Note that the calculated D- TEOM mass concentration measurements report the presence of NH^sub 4^NO^sub 3^ during generation and a mass concentration of zero afterward. Note also that the negative mass measured when the ESP is turned on is an estimate of the loss of semi-volatile material.

Hering et al.,27 Jaques et al.,28 and Lee et al.29 performed field evaluations of the D-TEOM in Southern California in locations with high nitrate concentrations, including Claremont and Rubidoux, the latter having some of the highest nitrate concentrations in the country. Jaques et al.28 field-tested two collocated PM^sub 2.5^ D- TEOM monitors in Claremont, CA, from September 2001 to August 2002. The D-TEOM monitors used 5-min switching periods for the ESP at a TEOM sensor temperature of 30 [degrees]C; 24-hr time-integrated mass concentrations of the D-TEOM were compared with collocated Micro- Orifice Uniform Deposit Impactor (MOUDI, MSP Corp.) and dichotomous Partisol (Dichotomous Partisol-Plus, Rupprecht and Patashnick (R&P) Model 2025; R&P is now part of Thermal Electron, Inc.) gravimetrically determined mass measurements. A high correlation (R^sup 2^ = 0.94, slope = 0.98) between the 24-hr integrated concentrations of the two D-TEOMs was obtained, indicating the high precision of the instruments. Comparison of the 5-min data showed that the response of the two instruments tracked each other well, that significant changes in ambient PM levels occurred on a time scale as short as 15 min, and that these changes are successfully tracked by the D-TEOM.28 PM^sub 2.5^ mass concentrations from collocated MOUDI and Partisol samplers correlated well with the D- TEOM (R^sup 2^ = 0.86 and 0.83, respectively). The Partisol measured lower PM^sub 2.5^ concentrations than the MOUDI, whereas the D-TEOM measurements were higher than the MOUDI, suggesting that the MOUDI still encounters loss of semi-volatile species, including semivolatile organics, albeit less than typically observed by integrated filter measurements.30 Also in Claremont, Hering et al.27 showed that under ambient conditions (February- June 2002), the vaporization reference signals (negative artifact measured by the reference) from the D-TEOM tracked well the ambient particulate nitrate concentration measured with a collocated cascaded Integrated Collection and Vaporization System (ICVS)31 (Figure 2). The ICVS is described later in this paper.

In nearby Rubidoux, CA, Lee et al.29 compared the PM^sub 2.5^ mass obtained by the D-TEOM with that of the RAMS, CAMM, and FRM Partisol, the latter an integrated filter-based method for PM^sub 2.5^ mass. The D-TEOM performed better in measurements of semi- volatile species than the CAMM or RAMS, and the RAMS typically measured more PM^sub 2.5^ mass than the CAMM. The FRM showed significant loss of semi-volatile material. Near artifact-free NH^sub 4^NO^sub 3^ was measured by a collocated Harvard-EPA annular denuder (HEADS)32 and the measured nitrate was approximately equivalent to the difference of D-TEOM and FRM mass concentrations.

Schwab et al.,21 Hering et al.,27 Jaques et al.,28 and Lee et al.29 concluded that the D-TEOM monitor provides a good estimate, under the high nitrate conditions tested, of the actual ambient PM present on a near-continuous basis.

Filter Dynamics Measurement System (FDMS). The FDMS33 was developed to account for loss of semi-volatile species relative to the previously mentioned TEOM methods. However, the FDMS uses a cold filter at 4 [degrees]C to remove particles and obtain particle-free air, rather than an electrostatic precipitator. The air is first conditioned with an SES diffusion dryer to remove water, after which the FDMS uses a switching cycle of 6 min, alternately sampling ambient and particle-free air. The loss or gain of semi-volatile material by the TEOM sensor is then accounted for by the period of particle-free air, similar to that of the D-TEOM. The FDMS was tested by Grover et al.34 in two field studies to determine how well it measured total fine PM, including the semi-volatile ammonium (NH^sub 4^^sup +^), nitrate, and organic material. Results from the FDMS were compared with results from a TEOM heated to 30 [degrees]C, D-TEOM, RAMS, and integrated PC-BOSS sampler in Lindon, UT, in February 2003. In July of the same year, the FDMS was compared with a TEOM (50 [degrees]C), RAMS, and PC-BOSS in Rubidoux, CA. A continuous R&P nitrate monitor (R&P Model 8400N) and a continuous Sunset carbon analyzer (Sunset Laboratories Inc.) were collocated in Rubidoux with the continuous mass monitors. The FDMS and D-TEOM monitors needed little attention from site operators during their studies and appeared to be rugged units. Results from these studies indicated that on average the FDMS and D-TEOM monitors both measured total PM^sub 2.5^, including the semi-volatile PM. However, for short-term high concentration episodes (38 of 474 data points), the FDMS was on average 21 [mu]g/m^sup 3^ higher than the D-TEOM (noting that general concentrations ranged from 20-60 [mu]g/m^sup 3^). Excluding these peak periods, the FDMS was on average only 1.2 [mu]g/ m^sup 3^ higher than the D-TEOM at an average PM^sub 2.5^ concentration of 35 [mu]g/m^sup 3^, and the two methods were well correlated (R^sup 2^ = 0.85). Results from Grover et al.34 also indicated that the FDMS agreed better with the RAMS and PC-BOSS measurements than did the D-TEOM, even during peak periods, although, in general, the comparisons were within expected uncertainties. Therefore, the authors recommended additional studies to determine whether the D-TEOM or FDMS is more accurate.

Grover et al.34 also compared the 30 [degrees]C and 50 [degrees]C TEOM systems and FRM to the FDMS. Both of the heated TEOM samplers reported consistently lower PM^sub 2.5^ concentrations than the FDMS. In the case of the 50 [degrees]C TEOM, the difference was similar or slightly lower than the summed concentration of semi- volatile organics and NH^sub 4^NO^sub 3^, indicating almost complete loss of semi-volatile matter for this heated TEOM, which suggests a potential continuous method for measuring either the nonvolatile or semi-volatile components of PM mass, as illustrated in Figure 3.

PM^sub 2.5^ and PM^sub 10^ Beta Attenuation Monitor (BAM). The BAM provides a direct continuous measurement of PM (PM^sub 2.5^ or PM^sub 10^) mass in air on the basis of the absorption of beta particles by PM deposited on a filter. In general, RH does not affect the measurement so the filter is not intentionally heated, as with the TEOM. BAM monitors are commercially available from several vendors. The BAM employs a radioactive beta source (typically ^sup 14^C or ^sup 85^Kr) that emits beta particles that are absorbed by the PM collected on a filter that is located between the source and detector. The filter is a continuous tape that advances at specified times or particle loadings. The absorption is proportional to the amount of PM collected on the filter. The newly advanced filter is used as a reference. A more detailed description of the BAM is given by Wen et al.35 The standard BAM has not changed much during the last 5 yr so further discussion of the standard BAM will not be given here. However, a BAM designed to measure ultra- fine PM is discussed below and represents a significant new development employing the BAM monitor.

Ultrafine Mass Monitor (UF BAM). Ultrafine PM mass concentrations vary drastically over short time scales in the atmosphere and are potentially toxic,36,37 thus, having important human health implications. Chakrabarti et al.38 modified a BAM to measure quasi- ultrafine (i.e., <0.15 [mu]m or PM^sub 0.15^) mass concentrations nearly continuously. The ultrafine BAM (UF BAM) is preceded by a 0.15-[mu]m cut point impactor, which is designed to have a very low- pressure drop. The UF BAM was tested at a downwind receptor site in the Los Angeles Basin in Claremont, CA. It was operated using a 2- hr cycle and collocated with a Scanning Mobility Particle Sizer (SMPS), Aerodynamic Particle Sizer (APS), and MOUDI. The UF BAM and MOUDI were in excellent agreement (R^sup 2^ = 0.92, slope = 0.97), as shown in Figure 4. The SMPS, however, underestimated UF BAM PM^sub 0.15^ levels, with the underestimation being highest in early morning and nighttime and least toward the middle of the day. The authors suggested that this could be caused by differences in the classification of fractal-structured UF particles. Several fractal- agglomerate particles, with effective densities substantially lower than 1 g/cm^sup 3^,39 may be detected by the UF BAM on the basis of their aerodynamic diameter (AD), but could be classified by the SMPS in higher size ranges on the basis of their physical or mobility diameter.40 No correlation (R^sup 2^ = 0.001) was found between the PM^sub 2.5^ mass concentrations, as determined by the SMPS, APS, and PM^sub 0.15^ mass concentrations, thereby indicating that ultrafine mass concentrations, at least in Los Angeles, are independent of PM^sub 2.5^ concentrations. Little or no correlation was observed between concurrently measured particle number concentrations measured by the sum of the SMPS size channels, and the PM^sub 0.15^ (UF BAM) mass concentrations (R^sup 2^ = 0.06). These findings demonstrate the need for a continuous UF mass monitor to ensure effective assessment of the short-term variations in UF aerosols. CAMM. The CAMM for PM^sub 2.5^ is described by Babich et al22. The measurement principle is based on an increased drop in pressure across a membrane filter (Fluoropore) during particle sampling. The pressure drop is proportional to the mass of the PM deposited. After passing through a PM^sub 2.5^ inlet, the airstream is dried below 40% RH using a Nafion diffusion dryer to remove particlebound water (PBW) followed by collection of the particles on a filter tape at a flow rate of 0.3 min^sup -1^. The tape advances about every 30-60 min so that particles remain close to equilibrium with the sampled air during collection. This also results in minimal volatilization and adsorption artifacts during sampling. The CAMM has a detection limit above 5 [mu]g/m^sup 3^ for PM^sub 2.5^ concentrations averaged over 1 hr.22 The performance of this monitor was investigated in a series of field studies conducted in seven cities with presumably different PM^sub 2.5^ chemical composition.22 Twenty-four 1-hr CAMM measurements were averaged and compared with Harvard Impactor (HI) 24-hr PM^sub 2.5^ integrated measurements. On the basis of 211 valid sampling days, the measurements obtained from the HI and the CAMM were highly correlated (R^sup 2^ = 0.90). The average CAMM-to-HI concentration ratio was 1.07 +- 0.18. Other comparisons of the CAMM with the D-TEOM, SES TEOM, and RAMS were discussed in previous sections. Coarse PM TEOM and BAM. Two continuous PM^sub c^ samplers (PM^sub c^ = PM^sub 10^-PM^sub 2.5^, i.e., particles in the size range between PM^sub 2.5^ and PM^sub 10^) are commercially available, one using the TEOM, and the other using the BAM as the detector. Both methods measure PM^sub c^ directly on filters. An indirect method using the APS is also discussed below. The PM^sub c^ TEOM41 was developed and evaluated as part of the Southern California Supersites Project, and therefore is described in more detail here. The operating principle of the PM^sub c^ TEOM is based on enriching particles in the 2.5-10 [mu]m range by a factor of approximately 25, while leaving fine mass (PM^sub 2.5^) at its ambient level. The resulting aerosol mixture is then sampled with a standard TEOM at 2 min^sup -1^, the response of which is dominated by the contributions of the coarse PM, due to concentration enrichment. The enrichment of coarse particles is achieved with a PM^sub 10^ sampling inlet operating at 50 min^sup -1^ followed by a 2.5-[mu]m cut point round nozzle virtual impactor whose ratio of minor to total flow is 25 (Figure 5). This represents the ideal ratio, and wall losses and other factors may slightly influence it up or down. This continuous coarse particle monitor (CCPM) was tested in Los Angeles in a field comparison with integrated sampling methods using a MOUDI and a Partisol (Dichotomous Partisol-Plus, R&P Model 2025) employing sampling times from 90 to 210 min. Coarse mass concentration measured by the CCPM and either MOUDI (Figure 5) or Partisol were well correlated (with R^sup 2^ = 0.88 and 0.85, respectively) and had ratios between 26 and 27.41 There was no difference observed in the response of the CCPM when the TEOM was heated to either 30 or 50 [degrees]C under the atmospheric conditions tested. Misra et al.41 showed that the enriched concentrations allow for reliable measurements at time intervals as short as 5 min and suggested that its simplicity and reliability make it ideal for use in large scale monitoring networks. The PM^sub c^ BAM monitor also separates fine from coarse particles using a virtual impactor, but it operates at an inlet flow rate of 16.7 min^sup -1^ with a concentration factor of 10 rather than 25. In the commercial BAM, coarse and fine particles are measured with a time resolution of 1 hr.

Recent studies42 were conducted by EPA at four locations (Research Triangle Park, NC [January 2003], Gary, IN [March 2003], Phoenix, AZ [May 2003, January 2004, May 2005], and Riverside, CA [July 2003]), to evaluate the PMc TEOM and PM^sub c^ BAM (Kimoto Electric Co. Ltd., Japan) as potential FRM or FEM samplers. An APS was initially included in the study to provide size distribution data to understand better potential differences between the two direct reading PM^sub c^ samplers, but results suggested it might also prove suitable as a PM^sub c^ sampler as described below. The initial field comparisons in 2003 and 2004 (before modifications as mentioned below) indicated that the prototype commercial PM^sub c^ TEOM was low by 20-30% (TEOM/reference) under some conditions and relative to the reference method (PM^sub 10^ FRM minus PM^sub 2.5^ FRM); however, the results were highly correlated (>0.95). The low values were believed to result from three factors. First, the inlet on the prototype PM^sub c^ TEOM had a cut point of 9 [mu]m rather than 10 [mu]m. Secondly, the TEOM was heated to 50 [degrees]C, as is the case for the standard TEOM. This latter effect may have resulted in the loss of semi-volatile coarse particle components, which in this size range could be organic species such as lower molecular weight polycyclic aromatic hydrocarbons (PAHs).43 Finally, the concentration conversion ratio may have been too high, using the ideal value rather than one adjusted for wall losses.41 R&P subsequently increased the cut point of the inlet from 9 to 10 [mu]m, lowered the operating temperature from 50 [degrees]C to 40 [degrees]C, and modified the concentration conversion algorithm to more accurately reflect influences on the ratio, such as wall losses, from 25:1 to 23:1. Preliminary results from the 2005 study in Phoenix indicate better agreement for the PM^sub c^ TEOM with the reference method, on average, agreement within 5% and R^sup 2^ > 0.98. Recently, R&P also has developed a fine and coarse particle sampler employing the FDMS (PM^sub c^ FDMS) approach, which should minimize the loss of volatile species from the coarse and fine particles.

The PM^sub c^ BAM (enrichment factor 10) had good agreement to the difference method for PM^sub c^ in all the studies reported to date42 (regression slopes between 0.88 and 1.17, with R^sup 2^ values exceeding 0.95); however, PM^sub 2.5^ was significantly overestimated at all locations and times of year by at least 40%. PM^sub 10^ values were also higher than those obtained by the PM^sub 10^ FRM. Laboratory studies with Arizona road dust and sodium chloride confirmed the overestimation for PM^sub 2.5^ as well as the good agreement for PMc. The PM^sub c^ BAM was modified before the 2005 studies. The modifications included reduction for the PM^sub 2.5^ deposition area and a change from mass flow control to volumetric flow control. In the 2005 Phoenix study, the PM^sub c^ continued to agree well, but PM^sub 2.5^ was still significantly overestimated.42

As noted above, an APS was initially included in the study design to provide size distribution data to better understand potential differences between the two direct reading PM^sub c^ samplers. However, the APS had reasonably good agreement with the difference method at most locations; slopes with the FRM difference method within 10% and R^sup 2^>0.85, with the exception of Gary, IN, where the APS was low by 30%. Subsequently, a heater was installed to stabilize the RH to below 45%, as transport losses of large hygroscopic particles in humid environments were determined to be part of the reason for the low values. However, the RH in Phoenix in 2005 was below 45% most of the time so the smart heater modifi- cation was not enabled. Operational problems also plagued the APS during the 2005 Phoenix study.

The 2005 field studies in Phoenix, AZ, and Birmingham, AL, also included other samplers. Besides the modified PM^sub c^ TEOM, PM^sub c^ BAM, and APS, the study included a GRIMM Technologies, Inc. instrument based on detecting single particles by light scattering, as well as the PMc FDMS as mentioned above. Preliminary results for these samplers as observed in the 2005 Phoenix study are presented in Vanderpool et al.42 Results for the 2005 Birmingham, AL, study are not yet available.

Continuous Indirect Methods for Mass Concentrations

SMPS-APS. The size distributions of a SMPS and APS can be combined to obtain a unified size distribution covering particle sizes from a few nanometers up to 10 [mu]m. Stanier et al.44 provide an excellent example of the use of these combined systems for Pittsburgh, where strong diurnal patterns are observed showing the direct effect of particle sources such as traffic and other combustion sources as well as atmospheric nucleation that appears to be regional in nature and occurs on 30-50% of the days. These number- based size distributions (number of particles/cm^sup 3^ in specific size bins) can be used to calculate volume and mass concentrations as indicated below. The SMPS (Model 3934 and 3936) typically measures aerosol particles in the size range from 10 to 800 nm. These size ranges can be adjusted by adjusting the aerosol and sheath air flows of the instruments. A newer version of this instrument, the TSI Nano-SMPS, which uses a shorter-length differential mobility analyzer (nano-DMA, Model 3085), allows classification of particles in the size range of 3-150 nm. The lower detectable size range of this instrument is important in atmospheric studies investigating aerosol formation mechanisms. Examples include nucleation events as well as exposure assessment studies conducted inside or near freeways and busy roadways, where a larger fraction of the aerosol is associated with sub-10 nm particles. 45,46 The Model 3321 APS spectrometer provides high-resolution, real-time aerodynamic measurements of particles from 0.5 to 20 [mu]m.

Aerosol volume concentrations are determined from the number- based size distributions measured by the SMPS-APS system, assuming that all particles are perfect spheres. Conversion of volume concentration to mass concentration also requires an estimate of particle density. For PM^sub 2.5^ mass, the shape factor is often assumed to be 1 and the density from approximately 1.5 to 1.6 g/ cm^sup 3^.40,47 Estimates of particle density for PM^sub c^ seem to range between 2 and 2.2 g/cm^sup 3^.40 The ability of this method to accurately measure PM^sub 10^, PM^sub 2.5^, and/or PM^sub c^ concentrations was evaluated by Shen et al.40 and Khlystov et al.47 In these studies, the SMPS-APS was tested in comparison with other continuous PM measurement devices and with time-integrated mass samplers at various sites in the Los Angeles Basin and in Pittsburgh, respectively. Size distributions between the SMPS and APS were merged based on algorithms by Sioutas et al.48 in Los Angeles and for Pittsburgh on the basis of Shen et al.40 as well as a second approach fitting the APS to SMPS size distributions to yield a size correction factor that was related to the ratio of particle density and the shape factor.

Both studies indicated good agreement between PM^sub 2.5^ gravimetric mass and that estimated from the size distribution data. Results from Pittsburgh47 indicated an average standard error of approximately 20% using the alternative algorithm and an average aerosol effective density of 1.5 g/cm^sup 3^. However, results from both studies also indicated a significant difference (little correlation) between the measured mass for lowest stages of the MOUDI (<0.15 [mu]m) and the SMPS mass estimates. This discrepancy may be due to particle bounce in the MOUDI, causing larger particles to be collected in lower stages, and/or the effect of shape and density of local vehicular emissions in Los Angeles on the measurement of particle mobility by the SMPS. The latter would result in its underestimation of the mass below 0.1 [mu]m relative to the MOUDI. These results are also consistent with other findings in Los Angeles near Claremont, CA,38 as noted earlier, in which the SMPS also underestimated mass below 0.15 [mu]m in comparison to the UF BAM monitor. Results from Los Angeles40 are shown in Figure 6.

Shen et al.40 also compared the APS results to PM^sub 10^ measurements obtained by a Partisol sampler and found a weaker association between the SMPS-APS and Partisol for PM^sub 10^ than for PM^sub 2.5^, likely driven by the poor efficiency of the APS for measuring coarse particle mass. Their results and those cited within ref 40 indicated that, relative to gravimetric methods (impactors), the APS both overestimates and underestimates PM^sub c^ mass because of overcounting across all size ranges (2.5-10 [mu]m) and loss of particles because of wall losses in the APS. The latter has a significant influence for particles above 5 [mu]m. These findings corroborated the results of previous studies.49,50 However, more recent results by Vanderpool et al.42 in comparing the APS to the FRM difference method for PMc (described above) indicate better agreement for PM^sub c^ between these two methods.

Watson et al.51 compared size distribution data collected in Fresno, CA, by an SMPS (TSI Model 3936L10) and an optical particle counter (OPC; PMS Lasair 1003) to hourly mass concentrations obtained by a PM^sub 2.5^ BAM. The SMPS measured the size distribution between 9 and 392 nm, the OPC between 0.1 and 2 [mu]m. The SMPS and OPC agreed well (R^sup 2^ = 0.95 and slope = 0.79) in the overlapping range of their size distributions. The OPC particle concentrations, however, were systematically higher by an average of approximately 20%. The authors speculated that the longer residence time for the SMPS in the heated trailer environment (relative to the colder ambient temperatures in winter) could result in an apparent loss of SMPS particles from the overlapping region. No attempt was made to combine the size distributions by the two instruments or to convert them to volume and mass concentrations. However, they found a high correlation between the number concentration for particles greater than 0.3 [mu]m, as measured by the OPC, with PM^sub 2.5^ measured by the BAM (R^sup 2^ = 0.91), as shown in Figure 7.

Results from the studies reviewed here suggest that size distribution data can be used to estimate PM^sub 2.5^ mass concentrations, but not ultrafine or coarse particle mass. Results for PM^sub 10^ also show a significant difference between PM^sub 10^ mass estimated by the SMPS-APS and gravimetrically determined PM^sub 10^ related to the measurement efficiency of the APS for coarse particles.

Size Distribution and Number Concentration

McMurry52 provides a thorough review of well-established methods for obtaining size distribution and total number concentration data as well as other physical properties of PM. Here we focus on newer methods that were developed and/or evaluated as part of the Supersites Program and related studies. A discussion of the SMSP and nano-SMPS is given above.

WCPC. A new water-based laminar-flow condensation particle counter WCPC has been developed (Aerosol Dynamics Inc. and Quant Technologies, LLC)53,54 and is commercially available (TSI, Inc.). The WCPC uses water rather than butanol to grow particles by condensation to a size where they can be counted by optical means. The use of water is particularly advantageous because it is a safer chemical than butanol, which is combustible, has a strong odor, and could contaminate other concurrent measurements. In this method, the wetted condenser walls are warmer than the air entering it, creating a condition of high supersaturation within the aerosol stream with rapid condensation of water vapor onto the particles. In contrast, because of the relatively low mass diffusivity of the butanol vapor, the butanol CPCs employ condenser walls that are cooler than the entering air to create the zone of supersaturation. The saturated butanol vapor then condenses onto the particles where they are grown and can be counted optically.

The WCPC (Model 3785) has an aerosol-sampling rate of 1 L/min and a lower detection limit near 5 nm. Calibration of this instrument with aerosol from a freeway tunnel and with ambient aerosol shows lower detection cut points of 4.3-4.8 nm, similar to the 5-nm cut point obtained for ammonium sulfate and NH^sub 4^NO^sub 3^ aerosol54 (Figure 8a). Somewhat higher cut points are found for laboratory- generated hydrophobic organic aerosols.53 The time response of the instrument is characterized by an exponential decay time constant of 0.35 sec, which is faster than for other unsheathed instruments.

Biswas et al.55 compared the performance of the WCPC (a prototype of the TSI WCPC Model 3875) to the widely used TSI CPC Model 3022A (TSI Inc.), a butanolbased CPC, to varying particle composition, concentrations, and size. Results of the study indicated good correlation between the WCPC and butanol-based CPC (R2: 0.74-0.99). Good agreement was found between the two instruments for particle concentrations up to 40,000/cm^sup 3^, with ratios of WCPC to CPC3022A concentrations varying between 0.8 and 1.2 (Figure 8b). Because of differences in the photometric mode calibration of these instruments, this ratio drops to 0.6-0.8 between 40,000 and 100,000/ cm^sup 3^. The photometric mode of these counters is used at higher number concentrations, during which multiple particles are present in the scattering volume, and the particle concentration is derived from the light scattered from the cloud of particles within the scattering volume.

In addition to the TSI Model 3785 WCPC, three other water-based CPCs are commercially available: Models 3781, 3782, and 3786. All WCPCs operate based on the same principle as the 3785. The 3786 UF WCPC detects particles as small as 2.5 nm with an extended concentration range to as high as 100,000/cm^sup 3^ with single- particle counting. The 3782 is similar to 3785 but operates at a lower flow rate, and thus, detects particles at a higher concentration range in the single count mode than the 3785. The 3781 was designed to detect particles at concentrations up to 5 x 10^sup 5^ particles/cm^sup 3^ in a single count mode, with a Dp^sub 50^ of 6 nm. These WCPC instruments are relatively new commercial methods and require additional field and laboratory testing before use in routine monitoring networks.

Particle Density

Particle density is crucial in determining where particles of a given size deposit in the respiratory system. Particle density also is an important property of aerosols because it is required to convert measured number distributions to mass distributions and to relate AD to Stokes diameter. Bulk particle density is often estimated from the measured chemical composition of PM. However, this method has considerable uncertainty because the mass balance is not always complete, organic species have a wide range of densities, and particle shape is unknown but often inaccurately assumed to be spherical, which is often not an accurate assumption. Other methods of determining particles density take advantage of the fact that particle effective density can be found if one of the following combinations is known: mobility size-aerodynamic size, mobility size- particle mass, or aerodynamic size-particle mass. A brief summary of the methods is given in Mc- Murry et al.52 and Ristimaki et al.56

McMurry et al.39 describe a method for measuring particle density on the basis of first selecting particles of a known electrical mobility with a DMA and then measuring their mass using an aerosol particle mass analyze57 (APM, now available commercially, Model APM- 10, Kanomax USA). The APM classifies particles according to their mass to electrical charge ratios. Classification occurs between the narrow annular spaces, also termed the operating space, available between two rotating coaxial cylindrical electrodes. Particles thus experience both radial electrical and centrifugal forces, which act in opposite directions. When the forces balance each other, the particles will penetrate through the rotating cylinders to the downstream detector. For spherical particles, the electrical mobility equivalent diameter equals the geometric diameter. Density is determined by knowing the geometric diameter and the mass. For nonspherical particles, their measurements can provide insightful information on dynamic shape factors, fractal agglomerate content, and "effective" densities. McMurry et al.39 used this technique to measure the effective density of ambient aerosols in Atlanta, GA, during August 1999. Their study showed that particles of a given mobility (~0.1 [mu]m and 0.3 [mu]m AD) often have several distinct densities and shapes, the most abundant of which consists of spherical hygroscopic particles with effective densities in the range from 1.54 to 1.77 g/cm^sup 3^ at 3-6% RH. These values agreed to within approximately 5% of values calculated based on the measured size-resolved composition. Particles that were more and less massive than these were also observed. The less massive particles had effective densities of 0.25-0.64 g/cm^sup 3^ and the more massive particles had effective densities of 1.7-2.2 g/cm^sup 3^. McMurry and co-authors39 hypothesize that the less massive particles consist of chain agglomerate soot. Ristimaki et al.56 described the use of the electrical low pressure impactor (ELPI), which measures AD in near real time, in combination with an SMPS, which measures mobility diameters. These measurements are performed in parallel, as opposed to the McMurry et al. method above, and then the two size distributions are compared to estimate the density of particles measured. Using test aerosols of known density, Ristimaki et al. showed that the SMPS-ELPI could provide a reasonable estimate of submicron particle densities. Geller et al.58 compared density estimates using an SMPS-ELPI with those obtained with a DMA-APM.39 Agreement to within approximately 10% was observed for each method for aerosols of known density, with better agreement between the methods (~5%) for ambient aerosols. Results indicated that as particle sizes increase from 50 to approximately 400 nm, their effective densities decreased; this was especially true near freeways and in particular the freeway with the higher fraction of diesel vehicles.58 In addition, as particle sizes increased, two or more density peaks occurred at a given size. In general, particle densities for 50-nm particles were greater than 1 g/cm^sup 3^ and decreased to less than 0.5 g/cm^sup 3^ as particle sizes increased to near 300-400 nm. Variations regarding these findings occurred depending on the site location; near one of the two freeways (gasoline- dominated vs. a relatively high fraction [25%] of diesel vehicles), a source-influenced site at the University of Southern California (USC), and a receptor site in Riverside, CA. Results indicated that as the particle size increased, the particles consisted more of chain-like agglomerates as suggested by McMurry et al.39 Fractal shape factors also were estimated in both of these studies.

Particle-Bound Water (PBW)

A number of methods have been described in the literature for the uptake of water associated with ambient particles. Both laboratory and field studies have been conducted and are briefly summarized in Stanier et al.59 For example, one of the most common methods used to date is the hygroscopic tandem DMA (H-TDMA).60 The HTDMA consists of two DMA units in series, one operated under dry conditions following a drying column and the other at RH approaching 90%. Particles of a given size are usually classified into two types by this method: less hygroscopic and more hygroscopic.60 Switching of the first DMA size range allowed more than one particle size to be examined. Authors typically report a growth factor, which is the ratio of the humidified to dry particle diameters; however, the water content of the aerosol is typically not determined by this method because it is limited to particles of only a few sizes.

Dry and Ambient Aerosol Size Spectrometer (DAASS). An automated system to measure the water content of ambient aerosol (3 nm to 10 [mu]m) and its hygroscopic growth, the DAASS, has been designed and was tested for 1 yr in Pittsburgh.59 In this method, a nano-SMPS, SMPS, and APS sampled in parallel and measured overlapping size distributions from 3 nm to 10 [mu]m. The SMPS and APS size distributions were merged as described by Khlystov et al.47 The system alternately sampled ambient and dried air with a switching period of 7 min. RH was controlled to 30% using Nafion dryers. The difference in ambient and dry size distributions and the corresponding integrated volumes reveal the physical state of the aerosol and is expressed as a growth factor (ratio between ambient and dried volumes) or an absolute amount of aerosol water every 15 min, after having measured one ambient and one dry size distribution. In the laboratory, the DAASS was tested first with ammonium sulfate and good agreement with the known hygroscopic properties of these particles was achieved. Preliminary results from the field study showed that the water content of the particles was between zero and 20 [mu]g/m^sup 3^ (0-50% of the ambient, hydrated mass). Figure 9 shows an example of ambient and dried size distributions. Discussions of extended field results in Pittsburgh are given by Khlystov et al.61 who found that the seasonal behavior of aerosol water content followed that of the aerosol acidity, so that in summer at lower RH (even below 30% RH), particles could contain more water than in winter at higher RH because the summer aerosol was more acidic. The relationship between water content and aerosol acidity is illustrated in Figure 10 for hourly data obtained in Pittsburgh during July 2001, where the maximum in water content occurs during periods of greatest aerosol acidity.62 The acidity ratio is calculated as:

Acidity Ratio = 2[SO^sub 4^ ^sup -2^]/[NH^sub 4^ ^sup +^]

where [SO^sub 4^^sup -2^] is the measured molar concentration of aerosol sulfate, and [NH^sub 4^ ^sup +^] is the measured molar concentration of NH^sub 4^ ^sup +^. A ratio of approximately 1 indicates neutral conditions; ratios greater than 1 indicate acidic conditions when some bisulfate is likely present; and ratios of 2 or greater indicate that all of the sulfate present is likely in the form of bisulfate.

The seasonal variation in water content is illustrated in Figure 11.62 The average water content for the summer was approximately 16% (3.9 [mu]g/m^sup -3^) of the PM^sub 2.5^ mass, whereas in the winter it was approximately 8% (0.9 [mu]g/ m^sup -3^) of the PM^sub 2.5^ mass.62 Rees et al.62 were able to achieve a PM^sub 2.5^ FRM mass balance (daily average) within the uncertainties of the data using hourly data, including results from the DAASS along with improved estimates of the loss of semi-volatile chemical components (Figure 11). However, it should be noted that the organic matter (OM)/OC factor of 1.8 used by Rees et al.62 may be highly uncertain and variable, depending on season and time of day, as the OC composition varies over the course of these periods because of its differing sources and formation mechanisms. Use of 1.8 likely minimizes the uncertainty, because the likely range of the OM/OC ratio is from 1.4 to 2.163

PM CHEMICAL COMPOSITION

The major chemical components of PM typically include anions (sulfate and nitrate, sometimes chloride and organic acids), cations (NH^sub 4^ ^sup +^, sometimes water-soluble sodium, potassium, calcium, and magnesium); carbon, including OC (actually composed of hundreds of compounds) and elemental carbon (EC); and a series of trace elements. Methods to measure these species, typically on a 1- hr time resolution or better, are described below.

Anion and Cation Monitoring

Thermal Reduction with Detection by Gas Analyzers. In these systems, the chemical components of PM are heated to sufficient temperatures, usually in the presence of a catalyst, to form a gas- phase species of lower oxidation state (e.g., sulfate is measured as SO^sub 2^ and nitrate is measured as oxides of nitrogen [NO^sub x^]). The resulting gas species are then measured by an appropriate continuous gas monitor as described below. These methods do not measure the species of interest directly and conversion efficiencies are not always 100% so calibration factors are needed to convert the measured species to the desired species and to ambient concentrations.

ICVS. An automated system for the measurement of fine particulate nitrate in the atmosphere was developed by Stolzenburg and Hering,64 the ICVS. A similar system was developed for sulfate65 and prototypes of both systems were evaluated during the 1999 Atlanta Supersites Project.66 Results from Atlanta showed good agreement between the ICVS methods for sulfate and nitrate with other continuous methods and 24-hr averaged filterbased methods. However, generally more scatter was observed in these comparisons, than with the ion chromatography (IC)-based methods, with R^sup 2^ values usually near 0.7. The higher variability may have been due to the different analytical techniques employed where the ICVS method indirectly measures sulfate and nitrate as the thermally desorbed gases SO^sub 2^ and NO^sub x^, respectively.

R&P has commercialized both systems (Model 8400N for nitrate and 8400S for sulfate). As with the prototype, particles are grown by condensation of water to sizes large enough for efficient impaction on the collection strip. This process greatly reduces particle bounce from the collection strip and minimizes NH^sub 4^NO^sub 3^ volatilization before collection. Following humidification, particles are collected by impaction on either a stainless steel or a Nichrome strip for nitrate or a platinum strip for sulfate. After an 8-min sampling period, the material collected on the strip is flash vaporized into a nitrogen carrier gas and the generated gases of interest are then detected by either an ozone-chemiluminescence detector with molybdenum (Mo) catalyst for NO^sub x^ generated from the nitrate or SO^sub 2^ by ultraviolet (UV) fluorescence for sulfate. Both systems use a 2.5-[mu]m sharp-cut cyclone as the inlet followed by a honeycomb denuder (ChemComb Model 3500 Speciation Sampling Cartridge, Thermo Electron Corp.) that can be coated with different gas-phase adsorbing compounds to remove interfering gases. The humidifier follows the denuder. The system operates with a time resolution of 10 min. Stolzenburg and Hering64 have reported laboratory and field evaluations of the prototype nitrate system. Laboratory studies indicated collection efficiencies of 95% and higher for particles with diameters greater than 0.1 [mu]m; collection efficiencies for the sulfate system appear to be greater than 90%. High correlations between 24-hr average ICVS and denuder- filter measurements were observed for nitrate using the pre- commercial instrument (R^sup 2^ > 0.94) at three sites (Denver, CO; Rubidoux, CA; and Bakersfield, CA) with slopes from 0.96 to 1.06 and data recovery exceeding 97% during the field study.64 Wittig et al.67 evaluated the commercial sulfate and nitrate monitors (R&P 8400S and R&P 8400N, respectively) in Pittsburgh from June 2001 to March 2002. They found reasonable correlations (R^sup 2^ > 0.83) with EPA speciation sampler 24-hr average sulfate and nitrate results as reported directly by the commercial instrument (Figure 12a). However, after additional corrections were made to the data (e.g., blank, calibration, software correction, gas analyzer effi- ciency), the commercial units (8400N and 8400S) still reported values lower (slopes of 0.83 for nitrate and 0.71 for sulfate) than the filter-based methods, the latter of which were used as the reference measurement. Results were better below approximately 2 [mu]g/m^sup 3^ for nitrate and sulfate (Figure 12b). The short-term results (24-hr average) were calibrated to the 24-hr filter-based measurements to obtain improved estimates of temporally resolved nitrate and sulfate during the study period (Figure 12c). They used these data to investigate short-term phenomena in Pittsburgh that could not have been revealed with the 24-hr filter data. This information was used to determine daily and diurnal seasonally averaged variations of the gas-to-particle partitioning of nitrate, as illustrated in Figure 13.

The R&P 8400N also was evaluated in east Baltimore from February to November 2002.68 The continuous nitrate data were compared with 24-hr denuder/filter-based measurements employing 47-mm nylon filters preceded by a MgO denuder to remove nitric acid.69 On average, the R&P 8400N was 33% lower than the 24-hr filter-based measurements, suggesting a conversion efficiency considerably lower than one, assuming negligible positive artifact of the denuded filter measurement (i.e., 100% MgO denuder efficiency). Results indicated that the NO^sub x^ monitor was sensitive to the absolute pressure of gas in the analytical cell requiring corrections for cell pressure to the data. Measurements also were corrected for ambient temperature. Precision for both 10-min and 24-hr average data were less than 10%. Detection limits ranged from 0.17 and 0.24 [mu]g/m^sup 3^ for the 10-min and daily average results. Conclusions suggested that an independent filterbased measurement (denuded Nylon or Teflon with adsorbing backup) is needed to calibrate the data, as was suggested by Wittig et al.67 Grover et al.34 also used the R&P 8400N nitrate monitor in their field studies with the RAMS in Lindon, UT, and Rubidoux, CA. They found that the R&P 8400N nitrate concentrations were low with respect to the PC-BOSS fine PM nitrate (zero-intercept slope of 0.65 +- 0.07, intercept of 3.3 +- 2.4 [mu]g/ m^sup 3^, R^sup 2^ = 0.73) with the largest differences observed at higher concentrations as reported by the authors. The authors indicated that the lower values were likely due to incomplete volatilization of sampled NH^sub 4^NO^sub 3^.

EPA has been evaluating the R&P 8400N and 8400S at five sites since mid-2002 (Five-Cities Study). The sites included locations where the monitors were tested under different air quality and meteorological conditions. The sites were located at existing state air monitoring sites with existing filter-based chemical speciation samplers. The sites were located in Phoenix, AZ; Chicago, IL; Seattle, WA; Deer Park, TX; and Indianapolis, IN. These evaluations were focused on operational aspects of the samplers as used by state employees and comparisons to EPA’s STN filter-based results. Part of the evaluation included performance audits by an independent EPA laboratory located in Montgomery, AL.70 EPA conducted five audits on the R&P samplers. Five aqueous standards containing varying amounts of nitrate (KNO^sub 3^) and sulfate (NH^sub 4^)^sub 2^SO^sub 4^ were provided to the site operators to evaluate the 8400N and 8400S, respectively. In the fifth nitrate audit sample, one of the blind audit mixtures contained nitrate from four different salts as well as sulfate from ammonium sulfate. This last solution was included to test the instrument’s response to different forms of nitrate that may occur in ambient air. All the single salt solutions showed low recoveries. The mixed nitrate audit sample consistently had the lowest values at all sites for its challenge of the 8400N. The results obtained from the audits are consistent from year-to-year. Precision was excellent with triplicate samples, usually varying by less than a few percent. Results were linear with correlation coefficients usually greater than 0.95. However, significant intercepts were observed and recoveries of the standard (linear regression slopes) were low for nitrate, often by 20-35%, especially at higher concentrations similar to the results presented in Figure 12, a and b. The greatest difference for nitrate is for deposits greater than 50 [mu]g on the flash strip (~2-5 [mu]g/m^sup 3^ based on 24-hr sampling using the PM^sub 2.5^ FRM or Met O