Comparison of Integrated Filter and Automated Carbon Aerosol Measurements at Research Triangle Park, North Carolina

Abstract

Carbon aerosol measurements from the Rupprecht & Patashnick Series 5400 carbon analyzer, the Magee Scientific AE-21 Aethalometer and a filter-based Andersen RAAS2.5-410 Chemical Speciation sampler with quartz filters analyzed by Thermal Optical Transmittance (TOT) were compared. The Series 5400 total carbon (TC) and organic carbon (OC) were moderately correlated (r = 0.64 and 0.67) with the RAAS TOT method and the elemental carbon (EC) was poorly correlated (r = 0.37). The 5400 underestimated the RAAS TC and OC by 64% and 78%, respectively. The underestimate is attributed in part to the positive OC artifact of the filter-based sampling method. Another difference between the 5400 measurements and the RAAS TOT is a correction for char. The lack of correction for any char that occurs in the 5400 could cause and underestimate of OC and an overestimate of EC. The 5400 overestimated RAAS EC by 89%. The Aethalometer black carbon (BC) was compared to the 5400 EC and the RAAS TOT EC measured. The Aethalometer BC correlated well (r = 0.86) with the RAAS EC, but the Aethalometer overestimated the RAAS EC by 30%. The 5400 EC was compared to Aethalometer BC both with and without a PM_2.5 size selective inlet. The correlations were 0.92 (inlet) and 0.55 (no inlet). The 5400 overestimated the mean Aethalometer BC by 17% with the inlet and 39% without. The improvement in results may not be due to the addition of the PM_2.5 inlet, but instead may be due to a difference in the amount of char formed by the 5400. Factors contributing to the differences in these results are discussed.

INTRODUCTION

A national monitoring program was implemented by the United States Environmental Protection Agency (USEPA) to support the PM_2.5 (fine particulate matter) National Ambient Air Quality Standard (NAAQS) for gravimetric mass in 1997 (62 FR 38763). PM_2.5 is the mass of particles that have an aerodynamic size of ≤ 2.5 μ m. The NAAQS was put in place to protect public health and the environment from excessive levels of PM_2.5 mass. To support PM_2.5 NAAQS monitoring efforts, a chemical speciation trends network (STN) was implemented to characterize the primary chemical components of mass in urban areas and provide data for trends across the U.S. Results from the STN and other PM speciation studies have shown carbon to be a major constituent of PM_2.5 mass in many parts of the U.S. and may account for as much as 50% of the total fine particle mass in the atmosphere (Gray and Cass 1998; Tolocka et al. 2001; USEPA 2002). The carbon fraction consists of elemental carbon (EC), organic carbon (OC), and carbonate carbon. The inorganic carbon fraction or carbonate carbon found in minerals is assumed to be a very small component of the PM_2.5 carbon mass and is not routinely measured (Chow and Watson 2002; Lim et al. 2003). EC is a primary emission from incomplete combustion of fossil fuels and biomass burning and serves as a tracer for combustion-derived particles. Particulate OC has both primary and secondary emission sources. Primary emissions originate from combustion, chemical, fossil fuel, and biogenic sources, whereas OC is secondarily formed from gas-to-particle transformation of anthropogenic and biogenic precursor gases to particulate matter (Seinfeld and Pandis 1998). Both EC and OC are significant in relation to atmospheric visibility (EC as a light-absorbing material and OC as a light-scattering material), source attribution, influence of climate by radiative forcing, and the development of control strategies for attaining the NAAQS. Since carbon represents a significant portion of the total PM_2.5 mass, control strategies for carbon will be considered for reductions in PM_2.5.

Current routine methods of collection and analysis of carbon aerosol involve filter-based sample collection with thermal optical analysis (TOA). These methods provide low time resolution (24 h) data and are subject to both sampling artifacts and analysis method uncertainties (Turpin et al. 2000; Chow et al. 1993). The nature and complexity of carbon aerosol requires time-resolved data for improved temporal resolution to help develop control strategies, determine and understand the sources, develop air quality models, and understand the relevance of carbon to the effect of PM_2.5 on health. Continuous and semicontinuous monitoring techniques are commercially available for the collection of carbon aerosol at very high time resolution (as short as 1 min) and can play a key role in providing these data. However, before these technologies can be considered for use in routine monitoring networks, method demonstration and comparison studies are needed to determine how they compare with standard, filter-based sampling methods.

There are several published methods for OC and EC measurements in ambient air. Most of these techniques involve the use of time-integrated, filter-based collection on a quartz filter substrate and analysis by thermal optical analysis (Huntzicker et al. 1982; Cachier et al. 1989; NIOSH 1999; Birch and Cary 1996; Chow et al. 1993; Peterson and Richards 2002). Methods of this type are the predominant technique in use today for OC and EC determinations. Newly developed continuous and semicontinuous, automated measurement techniques are commercially available. There are two automated systems currently available in the U.S. for OC and EC determinations: the Series 5400 ambient particulate monitor from the Rupprecht & Patashnick Company, Inc. (Albany, NY, USA; Rupprecht et al. 1995) and the carbon aerosol analysis field instrument from Sunset Laboratories (Tigard, OR, USA).

In addition to methods for OC and EC, automated methods for the determination of the light-absorbing carbon aerosol or black carbon (BC) are also available. These include the Aethalometer from Magee Scientific (Berkeley, CA, USA) the particle soot absorption photometer (PSAP) from Radiance Research Inc. (Seattle, WA, USA), and the multiangle absorption photometer (MAAP) from Thermo Electron Corporation (Woburn, MA, USA). There are “research-grade” instruments available for BC measurements, including the resonant photoacoustic spectrometer (Moosmuller et al. 1998; Arnott et al. 1999, 2003), the integrating plate (Bond et al. 1999; Lin et al. 1973), and the integrating sphere (Hitzenberger et al. 1996). Of the commercially available instruments, the Aethalometer, which measures BC collected on a quartz filter tape by optical absorption, is the most commonly studied BC measurement technique (Hansen and McMurry 1990; Kulbusch et al. 1998; Moosmuller et al. 1998; Allen et al. 1999; Hitzenberger et al. 1999; Lavanchy et al. 1999; Babich et al. 2000; Watson and Chow 2002; Lim et al. 2003). The MAAP is newly available and measures BC using a radiative transfer scheme on particle-loaded glass fiber filters (Petzold et al. 2002). The PSAP operates in a manner similar to the Aethalometer and measures light absorption of particles collected on a glass fiber filter (Bond et al. 1999; Mertes et al. 2004).

For this study, the Thermo Andersen RAAS2.5-410, PM_2.5 integrated, filter-based sampling system (Smyrna, GA, USA) was used. Quartz filters from the RAAS were analyzed by a thermal optical transmittance (TOT) analysis technique for carbon aerosol measurements. The RAAS has been evaluated in previous studies (Tolocka et al. 2001; USEPA 2001; Watson and Chow 2002) and will be used with TOT as the reference for the automated methods. The Series 5400 Ambient Particulate Monitor (5400) from Rupprecht & Patashnick Company was selected for study. At the time of purchase, it was the only commercially available, continuous OC and EC monitor available. The Series AE-21 Aethalometer was selected to measure BC due to its wide use in previous studies.

Table 1 provides a summary of previously published results for comparisons made in the last 15 years between the instruments selected for this study. Earlier comparisons used a thermal optical technique that did not use a laser transmittance or reflectance measurement for correction of char. More recent comparisons used filter-based TOA with TOT and thermal optical reflectance (TOR) as a reference for the automated methods. Comparisons of the Aethalometer BC with EC collected by TOR methods showed very good correlation that ranged from 0.87–0.98 (Allen et al. 1999; Babich et al. 2000; Watson and Chow 2002). The Aethalometer BC/EC TOR in all cases showed some underestimation of EC TOR by the Aethalometer. Mean Aethalometer/EC TOR ratios ranged from 0.62 in Dallas, Texas (Babich et al. 2000) to 0.92 in Fresno, CA (Watson and Chow 2002). One study compared the Aethalometer to an in situ TOT method (Lim et al. 2003). Comparison of the Aethalometer to the in situ TOT method showed very good correlation (r = 0.95), but contrary to the results found in comparison to the EC TOR, the Aethalometer was found to overestimate the in situ TOT by 12% (mean Aethalometer BC/EC TOT = 1.12). There were three studies (Berghmans et al. 1996; Lim et al. 2003; Watson and Chow 2002) that reported results from comparison of the Aethalometer BC to the 5400 EC. The correlations were > 0.71 and the mean ratio of 5400 EC/BC showed the 5400 EC to overestimate the Aethalometer BC by as much as 63%. The 5400 incorporates a PM_2.5 size-cut cyclone inlet as a standard device. Previous studies did not use a size-selective inlet on the Aethalometer, most likely because BC is thought to reside primarily in the submicron size fraction (Heitzenberg 1982; Wolff et al. 1982; McMurry and Zhang 1989). Because all of the other measurement methods used in this study incorporate a PM_2.5 inlet, it is of interest to evaluate the Aethalometer both with and without a size-selective inlet to determine any differences.

One study compared the 5400 to the TOR method (Watson and Chow 2002), and one study compared the 5400 to an in situ TOT method (Lim et al. 2003). Both studies found good correlation (r = 0.74 and r = 0.86) of the 5400 EC with the TOA methods, but there was an overestimation of EC by the 5400 (mean 5400 EC/TOA ratios of 1.53 and 1.20). Watson and Chow (2002) found very poor correlation (r = 0.42 and r = 0.31) of the 5400 with TOR TC and OC. The regression slopes were much less than 1.00 and the intercepts were large (> 5 μg C m⁻³). The mean ratio of 5400/TOR TC and OC showed the 5400 be 48% higher than the TOR method. Lim et al. (2003) found very good agreement of the 5400 TC and OC with the in situ TOT method. The correlation regression slopes and the mean ratios of 5400/TOT TC and OC were all near 1.00.

TABLE 1 Summary of comparisons between automated and filter-based measurements

Study location	Y	X	Slope (se ^a )	Intercept (se ^a )	r	Mean y/x ratio	N	Reference
Glendora, CA August 1986	EC TOA	BC Aeth.	1.05 ± 0.14 μg m^−3	−0.09 ± 0.45 μg m^−3	0.89	NR ^b	17	Hansen and McMurry (1990)
Mainz, Germany 28 Mar.–2 Oct. 1994	BC Aeth.	EC TOA	0.39 ± 0.02 μg C m^−3	0 ± 0.29 μg C m^−3	0.85	NR	36	Kulbusch et al. (1998)
Uniontown, PA Summer 1992	BC Aeth.	EC TOR	0.95 ± 0.04 μg m^−3	−0.2 ± 0.4 μg m^−3	0.96	0.86	63	Allen et al. (1999)
Antwerp, Belgium Oct.–Nov. 1995	BC Aeth.	5400 EC	1.2 μg m^−3	0.17 μg m^−3	0.89	NR	NR	Berghmans et al. (1996)
Bakersfield, CA 8–29 Jan. 1999	BC Aeth.	EC TOR	0.63 ± 0.06 μg m^−3	0.17 ± 0.10 μg m^−3	0.87	0.75	36	Babich et al. (2000)
Chicago, IL 10 Oct.–15 Nov. 1997	BC Aeth.	EC TOR	0.76 ± 0.03 μg m^−3	0.05 ± 0.05 μg m^−3	0.98	0.79	35	Babich et al. (2000)
Dallas, TX 5 Dec.–20 Jan. 1998	BC Aeth.	EC TOR	0.62 ± 0.04 μg m^−3	−0.01 ± 0.06 μg m^−3	0.95	0.62	34	Babich et al. (2000)
Philadelphia, PA 1 Aug.–1 Sept. 1998	BC Aeth.	EC TOR	0.62 ± 0.05 μg m^−3	0.12± 0.09 μg m^−3	0.95	0.70	18	Babich et al. (2000)
Phoenix, AZ 10 Dec.–25 Jan. 1998	BC Aeth.	EC TOR	0.71 ± 0.04 μg m^−3	0.36 ± 0.15 μg m^−3	0.96	0.80	33	Babich et al. (2000)
Riverside, CA 15 Aug.–20 Sept. 1997	BC Aeth.	EC TOR	0.83 ± 0.07 μg m^−3	−0.03 ± 0.13 μg m^−3	0.92	0.81	31	Babich et al. (2000)
Atlanta, GA Aug.–1 Sept. 1999	BC Aeth.	In situ TOT EC	1.23 μg C m^−3	−0.30 μg C m^−3	0.95	1.12	269	Lim et al. (2003)
Aug.–1 Sept. 1999	BC Aeth.	5400 EC	1.24 μg C m^−3	−0.87 μg C m^−3	0.90	0.93	542	Lim et al. (2003)
Aug.–1 Sept. 1999	In situ TOT TC	5400 TC	0.96 μg C m^−3	1.19 μg C m^−3	0.91	1.02	197	Lim et al. (2003)
Aug.–1 Sept. 1999	5400 OC	In situ TOT OC	0.99 μg C m^−3	1.15 μg C m^−3	0.85	0.92	197	Lim et al. (2003)
Aug.–1 Sept. 1999	5400 EC	In situ TOT EC	0.96 μg C m^−3	0.57 μg C m^−3	0.86	1.20	197	Lim et al. (2003)
Fresno, CA 1 Jan.–26 Dec. 2000	5400 TC	RAAS TC TOR	0.31 ± 0.02 μg m^−3	6.22 ± 0.17 μg m^−3	0.42	1.48	36	Watson and Chow (2002)
1 Jan.–26 Dec. 2000	5400 OC	RAAS OC TOR	0.20 ± 0.02 μg m^−3	5.42 ± 0.11 μg m^−3	0.31	1.48	36	Watson and Chow (2002)
1 Jan.–26 Dec. 2000	5400 EC	RAAS EC TOR	1.10 ± 0.09 μg m^−3	1.75 ± 0.28μg m^−3	0.74	1.53	34	Watson and Chow (2002)
1 Jan.–26 Dec. 2000	5400 EC	BC Aeth.	1.05± 0.05 μg m^−3	0.28 ± 0.05 μg m^−3	0.71	1.63	37	Watson and Chow (2002)
1 Jan.–26 Dec. 2000	BC Aeth.	RAAS EC TOR	0.81 ± 0.03 μg m^−3	0.13 ± 0.04 μg m^−3	0.98	0.92	51	Watson and Chow (2002)

There have not been many comparisons between filter-based TOT methods and the automated instruments selected for this study. Additional comparisons are needed to demonstrate the use of automated systems for routine monitoring networks. The purpose of this article is to compare filter-based TOT and automated carbon measurements methods. The RAAS filter-based method (TOT analysis), the Series 5400 ambient particulate monitor, and the Series AE-21 Aethalometer were intercompared. A summary of the relationships between collocated carbon measurements collected during the summer and winter of 2002 in Research Triangle Park (RTP), NC are provided, along with a discussion of potential causes for the differences.

EXPERIMENTAL METHODS

The following section provides the monitoring site description, details on the filter-based and automated monitoring techniques, the definition of TC, OC, EC, and BC used in this study, and details regarding the data treatment prior to analysis.

Site Description

The site was located on a grass-covered field in a wooded area of the Research Triangle Park (RTP) of Durham, NC (35° 53′ N, 78° 52′ W). The population of Durham is approximately 203,000 and an estimated 50,000 people are employed in the RTP (RTP 2002). The RTP is a suburban business park with one major interstate passing through it and several major highways leading into and out of the area. It is centrally located between the mountains in the west and the coast on the east and south. The area terrain is rolling and the site elevation is about 150 m. The filter-based sampling system was located on the wooden platform with the PM_2.5 inlet located about 5 m above ground. The automated carbon sampling systems were housed about 10 m away in the temperature-controlled shelter, with the PM_2.5 inlets located about 2 m above the trailer roof and about 6 m above ground. The sampling inlets were about 2 m apart. The immediate surrounding area consists of mixed deciduous and pine forest, highways and corporate and government research facilities. The site is located about 1200 m south of Interstate 40, a six-lane highway that carries significant traffic loads during the peak morning and evening rush-hour periods. It is also located about 100 m south of a four-lane road that parallels the Interstate and about 100 m east of a heavily traveled secondary road. Due to the close proximity of the site to roadways and forested areas, the expected primary and secondary sources of carbon aerosol include mobile sources, biogenic aerosol, and secondarily formed organic aerosol in the summer, and primarily mobile sources in the winter.

Carbon Measurement Methods

All monitors were operated at the recommended default settings specified by each manufacturer. A Thermo Andersen RAAS2.5-410, PM_2.5 filter-based sampling system was used to collect 24 h integrated samples. The RAAS2.5-410 is a six-channel Speciation Sampler (Symrna, GA, USA). Only three channels were used to collect PM_2.5 on 47 mm Teflon^® -, nylon- and quartz-filter media at nominal flow rates of 16.7, 7.3, and 7.3 ± 10% liters per minute (lpm), respectively Quartz filters (Whatman, QM-A, Clifton, NJ, USA) were used for collection of carbon aerosol and were the only filters evaluated in this study. Gas-phase carbon denuders or backup filter and sorbent systems were not used for sample collection. Filters were prefired at 900°C for at least 3 h and placed in a freezer at ≤ −15°C until shipped to the field for use. Filters were returned to the analytical laboratory from the field in coolers at ≤4°C. Upon receipt in the lab, they were stored in the freezer until ready for analysis. The quartz filters were analyzed by Research Triangle Institute (RTI) International in Research Triangle Park, NC using the TOT method and a Sunset Laboratory Carbon analyzer (Tigard, OR, USA). The TOT method (Peterson and Richards 2002) is a modification of the NIOSH 5040 method used for occupational exposure to particulate diesel emissions (NIOSH 1999; Birch and Cary 1996). The TOT method involves the evolution of OC in pure helium up to a temperature of 900°C and the evolution of EC in a 98%:2% helium:oxygen (He:O₂) atmosphere up to 920°C. The temperature profile used for analysis was He for 60 s at 310°C, followed by He for 60 s at 480°C, He for 60 s at 615°C, He for 90 s at 900°C; then He:O₂ for 35 s at 600°C, He:O₂ for 45 s at 675°C, He:O₂ for 45 s at 750°C, He:O₂ for 45 s at 825°C, and He:O₂ for 120 s at 920°C. The estimated method detection limit (MDL) provided by RTI for OC, EC, and TC was 0.13 μg C m⁻³. The sampling system was leak checked before every sample collection, and the sampling flow rate was calibrated monthly using a triCal volumetric flow calibrator (BGI, Inc., Waltham, MA, USA).

Three-hour samples were collected with the Series 5400 ambient particulate monitor. The 5400 Monitor (Rupprecht et al. 1995) collects continuous samples with dual, alternating collection streams. The sample inlet incorporates a PM_2.5 sharp-cut cyclone operated at a flow rate of 16.7 ± 10% lpm. The air flows through the inlet to a multiple-holed, stainless-steel impaction plate. The PM_2.5 cyclone and cutpoint of the impactor provide a particle collection range between 2.5 and 0.14 μ m (Rupprecht & Patashnick Co., Inc. 2002). No filter substrate material was used, and the baseline impactor temperature is maintained at 50°C. During the analysis phase, the particle-laden impactor is incorporated into a closed gas circulation loop that includes an afterburner, circulating pump, and nondispersive infrared (NDIR) CO₂ detector. The NDIR CO₂ detector measures the amount of CO₂ released when the sample is oxidized at elevated temperatures. The CO₂ value is used together with the sample volume in order to determine the concentration of carbon in the air during the sampling period. Air was present at all times during collection and analysis. The instrument measures the OC (the carbon evolved at 340°C) and TC (the carbon evolved at both 340°C and 750°C) concentrations in μg C m⁻³. Measurements of EC were obtained by subtracting the OC from the TC measured. The instrument was operated using the following temperature profile: an initial temperature increase from 50°C to 340°C (dwell time of 480 s) and then to a final temperature of 750°C (dwell time of 360 s). No temperature ramp control was used. The manufacturer's estimate of the MDL for the 5400 on a 3 h time base is 0.10 μg C m⁻³. The sample collection path was leak checked weekly. The CO₂ detector was calibrated initially and audited weekly with 414 and 1417 ppm CO₂ gas standards (Scott-Marrin Inc., Riverside, CA, USA). The collection flow rate was calibrated monthly with a triCal volumetric airflow calibrator.

Ten-minute measurements were made using the Series AE-21 Aethalometer. The Aethalometer (Hansen et al. 1982) is a semicontinuous method that measures light absorbing BC using continuous filtration and optical transmission to determine the attenuation of light at a fixed wavelength through a quartz filter tape. The Series AE-21 is a dual-channel unit that measures BC by optical absorption at 880 nm in the near-infrared and optical absorption at 370 nm in the near ultraviolet (UV). The BC measurement was of interest in this study; therefore, the readings from the UV channel were not evaluated. The Aethalometer converts light attenuation to BC mass using a fixed specific attenuation cross section (σ) of 16.6 m² g⁻¹ of BC. This was the default value set by the manufacturer for a wavelength of 880 nm (Hansen 2002). In previous studies, the Aethalometer was operated with a σ value of 19 m² g⁻¹ (Babich et al. 2000; Lavanchy et al. 1999; Allen et al. 1999; Liousse et al. 1993; Petzold et al. 1997; Kuhlbush et al. 1998; Moosmuller et al. 1998; Sharma et al. 2002). The manufacturer's current default σ value of 16.6 m² g⁻¹ was used in this study and was also used by Lim et al. (2003) in Atlanta. The default is being used since the goal is to operate the instrument at the conditions recommended by the manufacturer for normal, routine operation. The flow rate was maintained at 5.0 ± 8% lpm for the study. The estimated MDL for this instrument collecting 10 min data at 5.0 lpm is 0.10 μg C m⁻³. This value was determined from the manufacturer's estimated instrument sensitivity (3 times the noise equivalent of 1 ng BC) and the sample volume for a 10 min measurement. The 10 min values are mathematically averaged to 3 h increments; therefore, the 3 h MDL is assumed to be the same since no additional sample volume is collected. The Aethalometer was operated both with and without a PM_2.5 cyclone inlet manufactured by BGI, Inc. The BGI sharp-cut cyclone model 1.829 is designed for a 2.5 μ m aerodynamic aerosol size-cut at 5.0 lpm. The sharp-cut cyclone was cleaned monthly. The instrument flow rate was checked weekly and calibrated monthly with a triCal volumetric flow calibrator.

OC, EC, TC, and BC Definitions for this Study

The differentiation of OC from EC and BC is highly dependent on the specific measurement or analysis technique and operational procedures used. Different instrument operating procedures may lead to different results, as demonstrated by intercomparison studies for the TOA methods (Birch 1998; Chow et al. 2001; Currie et al. 2002; Schmid et al. 2001). In addition, there are currently no reference standards for assessing the measurement of true OC, EC, and BC or a standardized method for distinguishing between these species. Method differences also depend on the nature and source of aerosol collected.

For clarification and the purposes of the work reported here, 5400 OC is defined as the carbon thermally evolved by the Series 5400 monitor during the 50–340°C temperature step and measured by the CO₂ detector. The carbon evolved during the second temperature step at 750°C plus the 5400 OC is defined as 5400 TC, and 5400 EC is the difference between the 5400 TC and 5400 OC measurements.

For the TOT method performed on quartz filters from the Andersen RAAS, the analysis is done in two phases. In the first phase, carbon is thermally evolved during the 310–900°C temperature steps in a pure helium atmosphere. The evolved carbon is oxidized to CO₂ and then reduced to methane and measured by a flame ionization detector (FID). During the second phase, both pyrolysis correction and EC measurement are performed. The oven temperature is reduced to 600°C, an O₂/He mix is added, and the temperature of the oven is raised to 920°C. The filter transmittance is monitored throughout the analysis. During the second phase, the point at which the transmittance reaches the initial value is considered the split between OC and EC. The carbon evolved in the first phase, and prior to the split, is OC and will be referred to as RAAS TOT OC. The carbon evolved after the split is EC and is referred to as RAAS TOT EC. For a detailed description and discussion of the TOT method, see Birch and Cary (1996). RAAS TOT TC is the sum of the RAAS TOT OC and RAAS TOT EC measurements.

BC is defined as the light-absorbing carbon aerosol measured by the Aethalometer with the default optical absorption cross section of 16.6 m² g⁻¹ of BC. The Aethalometer BC, 5400 EC and RAAS TOT EC are treated as indicators of light-absorbing primary carbon aerosol and are compared directly in this study. The 5400 TC and OC measurements are compared to the RAAS TOT TC and OC measurements.

About the Data Sets

Twenty-four-hour, time-integrated RAAS TOT, 3 h 5400 carbon measurements, and 10 min Aethalometer BC measurements were collected during the summer of 2002. The most intensive sampling period was from 1 July through 2 October 2002, when the filter-based sampling system and both automated methods were operated. The intensive sampling was done in the summer because the highest TC aerosol concentrations were expected. Only the 5400 and Aethalometer with a PM_2.5 inlet were operated during the winter (23 November through 31 December 2002). This comparison was done in the winter due to availability and delivery of the cyclone inlet for the Aethalometer.

Twenty-four-hour integrated filter samples were collected at a frequency of 1 every 3 days. For each 24 h RAAS TOT filter-sampling event, eight 3 h 5400 measurements were averaged to obtain 24 h values for comparison. The 24 h 5400 averages were considered valid where 75% completeness was obtained. Completeness is the percentage of valid data obtained from the amount of data expected to be collected. Data were excluded when measured below the MDL. Twenty-six valid matched data pairs were obtained for comparison of the 5400 and RAAS TOT OC and TC and 24 valid matched data pairs for EC.

Ten-minute Aethalometer BC measurements were averaged to obtain 24 h BC values to compare to the filter-based RAAS TOT EC measurements. Aethalometer BC averages were considered valid where 75% completeness was obtained and measurements were above the MDL. Twenty-one valid data pairs were obtained for the period.

The 5400 3 h EC measurements were compared to 10 min Aethalometer BC averaged over 3 h. The 75% completeness and MDL requirements were also applied. The Aethalometer BC data were collected both with and without a PM_2.5 size-selective inlet. Of the 752 possible 3 h measurements that could be made over the period (1 July through 2 October) for comparison of the 5400 to the Aethalometer without a size-cut inlet, 558 valid matched data pairs were obtained for comparison. Beyond completeness and data below the MDL, additional missed data were due to downtime required to perform weekly instrument calibrations and maintenance. A smaller data set was obtained for those Aethalometer data collected with the PM_2.5 inlet from 23 November through 31 December 2002. Of the 316 possible 3 h measurements 189 valid matched data pairs were obtained for comparison. The loss of data was due to the instruments being off-line for calibrations and maintenance, and 16 days (4–20 December) of lost data were due to an extended power outage resulting from a major ice storm. The missing data were primarily interspersed throughout the sampling period, except for the winter time comparison of 5400 EC to BC with PM_2.5 inlet. The 16 days of consecutively lost data for this comparison occurred on all days of the week and not just on a particular day of the week or only on the weekend. Since no diurnal comparisons are being made, no affects on the data set are anticipated.

RESULTS AND DISCUSSION

Comparison of 5400 to RAAS TOT TC, OC, and EC

Three-hour 5400 measurements were averaged to obtain 24 h values for comparison to the RAAS filter measurements. A Deming, orthogonal regression and Pearson correlation (r) were used to compare the results. The Deming regression accounts for variability in both x and y variables. The regression slope, intercept, standard error, and correlation results are given in Table 2. The scatter plots are provided in Figures 1 through 3. The 5400 TC to RAAS TOT TC correlation was 0.64, the regression equation was 5400 TC = 0.38 (± 0.09) RAAS TOT TC–0.26 (± 0.64) μg C m⁻³. The OC results were very similar to the TC results: 5400 OC = 0.22 (± 0.05) RAAS TOT OC −0.03 (± 0.33) μg C m⁻³, and the correlation was 0.67. The OC measured comprises a large part of the TC, so the similarity of results is not surprising. The mean RAAS TOT OC was 6.14 and the mean RAAS TOT TC was 6.56 μg C m⁻³ (Table 3). Comparison of the 5400 TC and OC with TOR in Fresno, CA (Watson and Chow 2002) resulted in similar slopes (0.31 ± 0.02 for TC and 0.20 ± 0.02 for OC), but the correlation was not as good (r = 0.42 and r = 0.31) and the intercepts were much greater (6.22 and 5.42 μg C m⁻³) in Fresno. Comparison of the 5400 to an in situ TOT method in Atlanta, GA (Lim et al. 2003) resulted in much better slopes (0.96 for TC and 0.99 for OC), slightly larger intercepts (1.19 and 1.15 μg C m⁻³), and better correlations (r = 0.91 and 0.92).

TABLE 2 Deming ^a regression and Pearson correlation (r) results

	Sampler		Regression
Parameter	y	x	Slope (± se ^b	Intercept (± se ^b	r	Mean y/x ratio	Number of pairs
TC	5400	RAAS	0.38 (± 0.09)	−0.26 (± 0.64)	0.64	0.36	26
OC	5400	RAAS	0.22 (± 0.05)	−0.03 (± 0.33)	0.67	0.22	26
EC	5400	RAAS	0.95 (± 0.51)	0.43 (± 0.30)	0.37	1.89	24
EC	AETH	RAAS	1.17 (± 0.16)	0.06 (± 0.10)	0.86	1.30	21
EC	5400	AETH	0.94 (± 0.06)	0.29 (± 0.05)	0.55	1.39	558
EC	5400	AETH2.5	0.79 (± 0.02)	0.25 (± 0.02)	0.92	1.17	189

TABLE 3 Minimum (Min.), maximum (Max.), mean (μ), and median (Med.) concentrations by comparison and method

		OC, μg C m^−3				EC or BC, μg C m^−3				TC, μg C m^−3
y to x comparison	Method	Min.	Max.	μ	Med.	Min.	Max.	μ	Med.	Min.	Max.	μ	Med.
5400 to RAAS	5400	0.63	2.52	1.36	1.24	0.35	1.57	0.89	0.83	0.98	4.09	2.26	2.07
5400 to RAAS	RAAS	2.91	11.4	6.14	5.58	0.14	1.64	0.47	0.44	2.98	11.7	6.56	6.10
AETH to RAAS	AETH	—	—	—	—	0.29	1.95	0.65	0.56	—	—	—	—
AETH to RAAS	RAAS	—	—	—	—	0.14	1.64	0.50	0.46	—	—	—	—
5400 to AETH	5400	—	—	—	—	0.13	4.20	0.92	0.83	—	—	—	—
5400 to AETH	AETH	—	—	—	—	0.07	3.55	0.66	0.53	—	—	—	—
5400 to AETH2.5	5400	—	—	—	—	0.12	2.71	0.77	0.62	—	—	—	—
5400 to AETH2.5	AETH2.5	—	—	—	—	0.08	3.04	0.66	0.36	—	—	—	—

FIG. 1 Comparison of the R&P 5400 TC to the RAAS TOT TC concentrations (μg C m⁻³) measured 1 July to 2 October 2002. The solid line represents the Deming linear regression, and the dashed line represents the 1:1 line. The correlation (r), number of samples (N), and mean ratio are also shown.

FIG. 2 Comparison of the R&P 5400 OC and RAAS TOT OC concentrations (μg C m⁻³) measured 1 July to 2 October 2002. The solid line represents the Deming linear regression and the dashed line represents the 1:1 line. The correlation (r), number of samples (N), and mean ratio are also shown.

FIG. 3 Comparison of the R&P 5400 EC and RAAS TOT EC OC concentrations (μg C m⁻³) measured 1 July to 2 October 2002. The solid line represents the Deming linear regression and the dashed line represents the 1:1 line. The correlation (r), number of samples (N), and mean ratio are also shown.

The 5400 TC and OC measurements significantly underestimated the RAAS TC and OC results in this study. The mean 5400/RAAS ratios of TC and OC (0.36 and 0.22) were much smaller than those found Watson and Chow (2002) using the TOR method (1.48 and 1.48) and those found by Lim et al. (2003) using the in situ TOT method (1.02 and 0.92). The underestimation of TC and OC is attributed in part to the positive OC artifact expected from the filter-based sampling method. The filter sampling method did not incorporate gas-phase carbon denuders or backup filters/sorbents, and blank subtraction was not applied. The TOR method routinely incorporates a blank subtraction for OC measurements (Chow et al. 1993). The in situ TOT method used by Lim et al. (2003) incorporated a dynamic blank for the measurements collected. Another difference between the 5400 measurements and the RAAS TOT is a correction for char that occurs during pyrolysis. The RAAS TOT method use transmittance for correction of char; however, the 5400 does not incorporate a char correction. Lack of correction for char causes an underestimate of OC and overestimate of EC. These issues are discussed in more detail below.

The 5400 EC-to-RAAS TOT EC regression was 5400 EC = 0.95 (± 0.51) RAAS TOT EC + 0.43 (± 0.30) μg C m⁻³, and the correlation was 0.37. The ratio of mean 5400 EC is much higher than the mean RAAS TOT EC (ratio of 1.89). Both Watson and Chow (2002) and Lim et al. (2003) reported an overestimation of EC when comparing of the 5400 to the TOR and the in situ TOT method (ratio of 1.53 and 1.20, respectively). This overestimation of EC is indicative of the bias associated with the 5400 carbon measurements due to the lack of correction for char that may occur.

Berghmans et al. (1996) found an underestimate of carbon measured by the 5400 and performed lab experiments to identify a “saturation effect” for the 5400. A similar effect is seen in this study when assessing the difference in RAAS TOT OC and 5400 OC measurements as a function of RAAS TOT OC concentration (Figure 4). There is a strong relationship between the differences in measurements as a function of concentration. As the RAAS TOT OC value increases, the 5400 OC concentration does not increase accordingly. An inverse trend was apparent for the EC data (increasing difference as concentration decreased); however, the EC concentrations for both measurements were very low (less than 1.8 μg C m⁻³).

FIG. 4 Difference in RAAS TOT and R&P 5400 measurements as a function of RAAS TOT concentration (μg C m⁻³) for (a) OC and (b) EC. The sampling period was 1 July through 2 October 2002. The number of samples for was (a) 26 and (b) 24.

The RAAS is a filter-based sampling system that collects aerosol on quartz filters that are subsequently analyzed by the TOT method. The 5400 collects aerosol on a heated metal impaction place with a 0.14–2.5 μ m cutpoint. Neither method uses a gas-phase carbon denuder or backup filter sorbents or filters. Factors to consider regarding the difference between the two types of measurements include the differences in thermal evolution dwell times, temperatures, and atmospheres; correction for char.; adsorption of gas-phase organics on the quartz-filter substrate; variation in the amount of volatilization of organics from the particles due to the differences in sampling flow rates (7.3 lpm for the RAAS and 16.7 for the 5400) and resulting pressure drops across the filter and impaction plate; the low-end cutpoint of the 5400 (0.14 μ m); and possible volatilization of organics from the particles due to the 5400 baseline collection temperature of 50°C. Losses due to particle bounce and resulting “blow-off” associated with classic impaction are reported not to be of consequence due to containment of the impaction plate (Rupprecht et al. 1995). The quantitative impact of each of these effects is unknown; however, additional discussion of these issues is provided below.

Different Operational Definitions

The differentiation of OC from EC is highly dependent on the specific measurement or analysis technique and operational procedures used the method. Different operational procedures can lead to highly variable results, as demonstrated by intercomparison studies for the TOA methods such as TOT (Birch 1998; Chow et al. 2001; Currie et al. 2002; Schmid et al. 2001). The RAAS TOT method determines OC in a pure helium atmosphere up to a temperature of 900°C, while the 5400 determines OC in air up to a temperature of 340°C. In addition, the TOT method determines EC in an O₂/He atmosphere up to 920°C, while the 5400 determines EC in air from 340 to 750°C. Different evolution temperatures, atmospheres, and dwell times for the 5400 would contribute to the differences in OC, EC, and TC when compared to the TOT method.

Correction for Char

An important difference between the 5400 measurements and the RAAS TOT is the correction for OC char. Some organic species in aerosol samples produce pyrolytically generated EC or char during thermal analyses (Huntzicker et al. 1982). The RAAS TOT method uses transmittance for correction of char; however, the 5400 does not incorporate a char correction. Although the formation of char during heating in air is minimized, it is not completely eliminated (Cadle et al. 1980) and could lead to an underestimation of OC and an overestimation of EC. With very low measured EC concentrations, the uncertainty due to char is magnified. An underestimation of OC and an overestimation of EC and poor correlation between the 5400 EC and RAAS TOT EC were found in this study. This is indicative of a bias associated with the 5400 carbon measurements due to the lack of a correction for char. The amount of char formed is not a constant and depends on a variety of factors, including the amount of OC, the temperature used in analysis, the residence time at each temperature, the presence of certain inorganic constituents, and the carrier gas used (Cadle et al. 1980; Yu et al. 2002). All of these factors, except the amount of OC and inorganic constituents, present in the sampled atmosphere, are different between the 5400 and RAAS TOT thermal methods.

Positive and Negative Artifacts

Sampling for OC is subject to positive and negative artifacts that are difficult to separate and quantify. Estimates of the positive (adsorption) and negative (volatilization) biases range from +50 to −80%, respectively (Eatough et al. 1990; Turpin et al. 1994). Adsorption of condensable organics on the quartz sampling filter is dependent on ambient temperature, surface area of the filter, partitioning coefficient (vapor pressure), and the presence of other competing adsorbates (Mader et al. 2001). However, metal impaction plates do not absorb gas-phase organics (Zhang and McMurry 1987). The volatilization of particle-phase semivolatile organic compound (SVOC) can be a function of the pressure drop (filter-face–velocity) across the filter or the impaction plate (McDow and Huntzicker 1990). Increasing pressure drop due to increasing flow rate can drive volatilization (Turpin et al. 2000). Positive artifacts can be reduced for filter sampling through the use of gas-phase denuders and backup filters and sorbents, but these may disrupt the gas–particle equilibrium and cause negative biases in the measurements. No denuders or backup filters and sorbents were used for filter sampling. The RAAS TOT method used in this study also does not incorporate an OC blank subtraction. The typical average annual OC blank value found for a RAAS sampler in the EPA STN is 1.3 μg C m⁻³ (Flanagan et al. 2002). This amounts to a positive artifact that is 23% of the mean OC and 21% of the mean TC measured in this study and does not fully account for the difference in results.

The 5400 uses an impaction plate for sample collection. Impactor OC measurements have been found to be 45–55% lower than quartz filter sampling methods (Hering et al. 1990). Zhang and McMurry (1991) predicted lower impactor carbon measurements due to losses of SVOCs as a result of the pressure drop across the impactor and the equilibrium ratio of gas-to-particle concentrations of the evaporative SVOC species. McMurry et al. (1996) shows very poor agreement between an impactor and filter carbon measurements. Impactor OC was higher than filter OC at two locations (Meadview, AZ and Hopi Point, AZ) and lower than filter OC for a site in Los Angeles, CA. Cabada et al. (2004), compared micro-orifice uniform deposit impactors (MOUDIs) to filter-based samples and found that about 50% of the organic mass was lost by the MOUDI in summer and good agreement (no loss) was found in winter. This was attributed to the greater loss of SOA (secondary organic aerosol, assumed to be SVOCs) present at higher levels in the summer. The 5400 TC and OC measurements in RTP were much lower than the quartz filter samples and are partly attributed to the lower OC expected to be measured by the impactor.

The 50°C baseline collection temperature of the 5400 monitor, used to prevent condensation of moisture, could cause additional loss of SVOCs. Zhang and McMurry (1989) tested the effect of 20°C versus 40°C temperature on the volatilization loss from impactors. They saw no loss for ambient aerosols but did see losses of SOA produced in smog chambers at the higher temperature. A loss of SVOC mass was demonstrated in the literature for the 50°C taper element oscillating microbalance (TEOM), which uses a filter and not an impaction plate. As much as 37% of the PM_2.5 mass was lost during data collection in summer (Ayers et al. 1999; Allen et al. 1997; Cyrys et al. 2001). Losses of SVOC mass at 50°C in the 5400 could account for a portion of the lower response in comparison to filter OC measurements, which are collected near ambient temperature. The loss of SVOC is expected to vary considerably as a function of ambient conditions and aerosol type, and therefore it is not expected to be constant.

Low-End Impaction Cutpoint

The 5400 collects aerosol on a heated metal impaction place with a 0.14–2.5 μ m cut point. The low-end impaction cutpoint is an important issue in the comparison of the 5400 to filter-based methods. In theory, the filter collects all particles ≤ 2.5 μ m (with a PM_2.5 size-selective inlet) and has no loss of the ultrafine fraction (< 0.1 μ m) of particle mass. The 5400 loses all particle mass < 0.14 μ m. How much mass is unknown; however, very small particles do not contribute significantly to the mass and rarely account for more than a few percent of the total mass of airborne particles (Seinfeld and Pandis 1998; Keywood et al. 1999; Sturm et al. 2003). Mass distribution is a function of aerosol type and age. In the summertime (when these samples were collected), it is expected that SOA is present in the atmosphere due to photochemical activity. SOA is expected to be higher in both the ultrafine mode (< 0.18 μ m) and the accumulation mode (> 0.18 μ m and < 2.5 μ m) in the summer (Fine et al. 2004), so it is conceivable that losses of ultrafine particle mass by the 5400 due to the low-end cutpoint of 0.14 μ m would be greater in the summer.

Comparison of Aethalometer BC to RAAS TOT EC

Ten-minute Aethalometer BC measurements were averaged to obtain 24 h BC values to compare to the RAAS EC measurements. Refer to Table 2 for the correlation and regression results and Figure 5 for the scatter plot. The correlation of Aethalometer 24 h BC to RAAS filter EC was 0.89. The regression equation was BC = 1.17 (± 0.16) RAAS TOT EC + 0.06 (± 0.10) μg C m⁻³. Other studies have shown that filter-based EC measurements compare well with Aethalometer BC (Hansen and McMurry 1990; Allen et al. 1999). Hansen and McMurry (1990) reported filter EC = 1.05 (± 0.14) BC – 0.09 (± 0.45) μg m⁻³ and correlation of 0.89; Allen et al. (1999) reported BC = 0.95 (±0.04) TOR EC – 0.2 (± 0.4) μg m⁻³ and correlation of 0.96. The filter EC values are not subject to the same positive and negative sampling artifacts expected with OC (and subsequently TC). The RAAS TOT EC method used to analyze the EC is fundamentally different from the optical absorption method used by the Aethalometer. The Aethalometer overestimated the RAAS TOT EC on average by 30% (BC/RAAS TOT EC ratio of 1.30). Allen et al. (1999), Babich et al. (2000), and Chow and Watson (2002) reported mean BC/TOR EC ratios that were all < 1.00 (Table 1), which shows an underestimation of TOR EC by the Aethalometer. Measurements of EC by TOR have been shown to be greater (as much as a factor of 2) than EC measured by TOT (Chow et al. 2001). Lower EC by the TOT method would cause the ratio to be higher in this study and could explain the difference in results seen here versus those found previously for EC by TOR.

FIG. 5 Comparison of Aethalometer BC and Andersen RAAS filter EC concentrations (μg C m⁻³) measured 1 July to 2 October 2002. The solid line represents the Deming linear regression and the dashed line represents the 1:1 line. The correlation (r), number of samples (N), and mean ratio are also shown.

Liousse et al. (1993) Petzold et al. (1997), and Hitzenberger et al. (1999) suggest that the specific attenuation cross section (σ) used by the Aethalometer to determine BC mass is highly variable, changes with location, carbon source, mass fraction, and particle size, and that a site-specific value should be used. The σ of the Aethalometer was calibrated against the EC as determined by the TOT method. Multiplying the slope of the BC to TOT EC regression times the σ used to determine BC provides a site-specific estimate (Babich et al. 2000). Multiplying the slope (1.17) by the σ value of 16.6 m² g⁻¹ results in σ of 19.4 m² g⁻¹ for RTP, NC during the summer. When comparing the Aethalometer BC to mass of BC using a thermal method, Liousse et al. (1993) reported a range of σ from 5 m² g⁻¹ for a remote area to 20 m² g⁻¹ for an area dominated by biomass burning. Petzold et al. (1997) reported a range of values from 8 m² g⁻¹ for rural sites to 19 m² g⁻¹ for a street crossing in Germany, and Sharma et al. (2002) reported a range of values from 6.4 to 20 m² g⁻¹ for 6 sites in Canada. The σ of 19.4 m² g⁻¹ for RTP, NC is consistent with the σ of 19 m² g⁻¹of BC typically used for the Aethalometer (Babich et al. 2000; Lavanchy et al. 1999; Allen et al. 1999; Liousse et al. 1993; Petzold et al. 1997; Kuhlbush et al. 1998; Moosmuller et al. 1998; Sharma et al. 2002). It is slightly higher than the σ of 18 m² g⁻¹ found by Liousse et al. (1993) for an area dominated by anthropogenic sources with an aerosol high in sulfur and carbon and the value of 18 m² g⁻¹ identified by Allen et al. (1999) for a site in southwestern PA dominated by vehicular emissions and open burning. The aerosol at the RTP site in the summer is to be influenced by motor vehicle emissions due to its very close proximity to three major roadways and SOA due to photochemical activity.

Comparison of 5400 EC to Aethalometer BC

The 5400 3 h EC measurements were compared to 10 min Aethalometer BC measurements averaged over 3 h. The Aethalometer is typically operated without any size-selective inlet; however, the 5400 incorporates a PM_2.5 cyclone in its standard configuration. Since the 5400 incorporates a PM_2.5 cyclone inlet, it was of interest to determine the effects of a PM_2.5 inlet on the 5400 EC to Aethalometer BC comparison. The Aethalometer BC data were collected both with and without a PM_2.5 size-selective inlet. See Table 2 for the correlation and regression results. Data were collected without the PM_2.5 inlet (AETH) from 1 July through 2 October 2002 and with the PM_2.5 inlet (AETH2.5) from 23 November through 31 December 2002. The correlation of 5400 to AETH BC during summer was 0.55, and the regression equation was 5400 EC = 0.94 (± 0.06) AETH BC + 0.29 (± 0.05) μg C m⁻³. The correlation of 5400 EC to AETH2.5 BC during winter was 0.92, with a regression of 5400 EC = 0.79 (± 0.02) AETH2.5 BC + 0.25 (± 0.02) μg C m⁻³. The mean 5400 EC/AETH BC concentration ratio was 1.39, and the mean 5400 EC/AETH2.5 BC was 1.17. Without a PM_2.5 inlet on the Aethalometer, the mean 5400 EC concentration was 39% higher than the AETH BC. With the PM_2.5 inlet added to the Aethalometer, the mean 5400 EC concentration was only 17% higher. Figures 6 and 7 show the scatter plots both with and without the inlet. The addition of the PM_2.5 sharp-cut cyclone to the Aethalometer seemed to improve the correlation with the 5400 EC.

FIG. 6 Comparison of R&P 5400 EC and Aethalometer BC, without PM_2.5 inlet (AETH BC) concentrations (μg C m⁻³) measured 1 July to 2 October 2002. The solid line represents the Deming linear regression and the dashed line represents the 1:1 line. The correlation (r), number of samples (N), and mean ratio are also shown.

FIG. 7 Comparison of R&P 5400 EC and Aethalometer BC with PM_2.5 inlet (AETH2.5) concentrations (μg C m⁻³) measured 23 November to 31 December 2002. The solid line represents the Deming linear regression and the dashed line represents the 1:1 line. The correlation (r), number of samples (N), and mean ratio are also shown.

As mentioned in the discussion earlier for the comparison of the 5400 to RAAS TOT, the 5400 does not incorporate any type of correction for char. This would lead to an underestimation of OC and overestimation of EC. The improvement in results for the comparison in winter may not be due to the addition of the PM_2.5 inlet but may be due to the lower amounts of OC in the winter and potentially less amount of char formed by the 5400. Less char would result in less overestimation of EC and less variability (scatter) in the data. When visually comparing Figure 6 to Figure 7, there is less scatter in Figure 7 and more data points with high 5400 EC relative to BC concentration in Figure 6. The affect of char is also supported by the difference in mean and maximum EC measurements by the 5400 and Aethalometer. The 5400 EC mean and maximum concentrations in the summer were 0.92 and 4.20 μg C m⁻³, versus 0.77 and 2.71 μg C m⁻³ in the winter. The Aethalometer BC concentrations were essentially unchanged (Table 3). The 5400 measured more EC in summer than in winter. The concentration of EC as a primary emission is not expected to change unless there are more primary EC emissions in the winter (e.g., woodsmoke). The site in RTP is not in a residential area and is mostly influenced by motor vehicle traffic due to its proximity to highways; therefore, primary EC emissions are not expected to change significantly.

Berghmans et al. (1996) found that for high concentrations measured by the Aethalometer, relatively low concentrations were measured by the 5400. Concentrations of BC in the RTP study did not reach very high levels (none > 5 μg C m⁻³); however, the slope of the 5400 to the AETH BC (Figure 6) dropped below the 1:1 line at values higher than 2 μg C m⁻³. This indicates a lower response by the 5400 at these concentrations, although there was a very large amount of scatter in the data set. The same trend was seen in the 5400 to AETH2.5 comparison (Figure 7), where much less scatter was observed and the trend below the 1:1 line is easier to discern. This trend can be seen through assessing the difference between AETH2.5 BC and 5400 EC measurements as a function of AETH2.5 BC concentration. The downward trend in AETH2.5 BC–5400 EC can be seen in Figure 8a as the BC concentration exceeds 2 μg m⁻³. This trend is also seen when no inlet was used (Figure 8b). The difference between measurements increases with increasing BC concentration, suggesting confirmation of a “saturation” effect as also seen in the comparison of the 5400 TC and OC to the RAAS TOT in this study.

FIG. 8 Difference in Aethalometer BC and 5400 EC as a function of Aethalometer BC concentration for (a) Aethalometer with the PM_2.5 inlet and (b) Aethalometer without the PM_2.5 inlet. The sampling period was 1 July through 2 October 2002, and the number of samples was (a) 558 and (b) 189. The sampling period was 23 November through 31 December 2002.

Test of Population Means

In order to evaluate the differences in the mean concentrations provided by the various methods statistically, a t test was used. The t test was selected to compare the mean concentration values obtained from each pair of measurement techniques compared. The null hypotheses (H₀) that the difference between the mean values μ 1 and μ 2 is not different from zero (H₀: μ 1-μ 2 = 0), was tested at the 5% significance level (α = 0.05). Equal variances were assumed. Based on these results, the null hypothesis is rejected at the 5% level of significance for all but two cases: the comparison of means between the 5400 EC and the AETH2.5 BC (p = 0.0720) and the AETH BC and the filter-based RAAS EC (p = 0.1149). The population means for the first case were marginally significantly different at the 0.05 level. In all but these two comparisons, the measurement methods provide significantly different mean concentrations for the aerosol tested. It is important to note that although the mean EC values measured in all cases are well above the MDL, the mean concentrations are all < 1 μg m⁻³. Where statistically significant differences in the mean EC concentrations were found, there were not necessarily large differences in the actual mean concentration.

CONCLUSIONS

All of the methods used in this study provide measurements of carbon either by thermal evolution or optical absorption. The dissimilarity in sampling and analysis methods contributes to the difference in results obtained. The 5400 did not compare well with the filter-based method. The 5400 TC and OC measurements significantly underestimated the RAAS TC and OC results by 64 and 78%, respectively. The underestimation of TC and OC in this study could be attributed in part to the positive OC artifact of the filter-based sampling method. No filter blank was used to correct for positive filter artifacts for the RAAS TOT method. Sampling to address positive filter artifacts and OC blank subtraction should be considered when using filter-based methods. Another important difference between the 5400 measurements and the RAAS TOT is a correction for char. Although oxidation in air is expected to minimize char, any charring that occurs would cause an underestimation of OC and overestimation of EC by the 5400. Other factors that contribute to the underestimation of TC and OC by the 5400 include the difference in operational procedures (evolution temperatures and atmosphere, and dwell times) used to define OC and EC by the two methods, the potential for loss of SVOCs from the impactor, particle mass lost as a result of the low-end cutpoint of the 5400 (0.14 μ m), and possible volatilization of organics from the particles due to the 5400 baseline collection temperature of 50°C. The volatilization of organics from the particles due to the differences in sampling flow rates and resulting pressure drop across the filter or impactor also play a role. The 5400 overestimated RAAS EC by 89%. The overestimation of EC by the 5400 is attributed to the lack of correction for char and the dissimilarity in sampling and analysis methods.

The Aethalometer overestimated mean RAAS TOT EC concentration by 30% using the default σ of 16.6 m² g⁻¹ set by the manufacturer. A t test showed that the mean concentration values from Aethalometer BC and the filter-based RAAS TOT EC are not significantly different. Since α is expected to change with location, carbon source, and mass fraction a site-specific value was determined. The Aethalometer was calibrated against RAAS TOT EC in this study. A value for σ of 19.4 m² g⁻¹ was determined for the aerosol measured at the RTP site.

For comparisons between the 5400 and the Aethalometer BC, data were collected with and without a PM_2.5 size selective inlet. This was done to be consistent with the 5400, which incorporates a PM_2.5 cyclone, and to explore any difference in measurements made with the Aethalometer configured with a PM_2.5 inlet. The correlation and mean ratio improved when the PM_2.5 size-selective inlet was installed on the Aethalometer during the winter. The improvement in results for the comparison may not be due to the addition of the PM_2.5 inlet, but may be due to lower amounts of OC in the winter and potentially lesser amounts of char formed by the 5400. It is unclear whether the PM_2.5 inlet on the Aethalometer had any affect on the comparison. Additional experiments are needed to evaluate the effect of a PM_2.5 cyclone inlet on the Aethalometer over different seasons.

Acknowledgments

This article has been subject to agency review and approved for publication. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

Notes

^a Standard error.

^b Not reported.

^a Orthogonal regression for method comparison when imprecision is present in both variables.

^b Standard error.