Revisiting the size selective performance of EPA's high-volume total suspended particulate matter (Hi-Vol TSP) sampler

ABSTRACT

Under the National Ambient Air Quality Standard (NAAQS) for airborne lead, measurements are conducted by means of a high-volume total suspended particulate matter (Hi-Vol TSP) sampler. In the decade between 1973 and 1983, there were 12 publications that explored the sampling characteristics and effectiveness of the Hi-Vol TSP, yet there persists uncertainty regarding its performance. This article presents an overview of the existing literature on the performance of the Hi-Vol TSP, and identifies the reported sampler effectiveness with respect to four factors: particle size (reported effectiveness of 7%–100%), wind speed (−36% to 100%), sampler orientation (7%–100%), and operational state (107%–140%). Effectiveness of the Hi-Vol TSP was evaluated with a solid, polydisperse aerosol in a controlled wind tunnel setting. Isokinetic samplers were deployed alongside the Hi-Vol TSP to investigate three wind speeds (2, 8, and 24 km h⁻¹), three sampler orientations (0°, 45°, 90°), and two operational states (on, off) for aerosols with aerodynamic diameters from 5 to 35 µm. Results indicate that particle diameter was the largest determining factor of effectiveness followed by wind speed. Orientation of the sampler did not have a significant effect at 2 and 8 km h⁻¹ but did at 24 km h⁻¹. In a passive state, the Hi-Vol TSP was collected between 1% and 7% of available aerosol depending on particle size and wind speed. Results of this research do not invalidate results of previous studies but rather contribute to our overall understanding of the Hi-Vol TSP's size-selective performance. While results generally agreed with previous studies, the Hi-Vol TSP was found to exhibit less dependence on these four factors than previously reported.

© 2017 American Association for Aerosol Research

EDITOR:

• Kihong Park

Introduction

In 1978, the first National Ambient Air Quality Standards (NAAQS) for Pb was set by the U.S. Environmental Protection Agency (U.S. EPA) at 1.5 µg m⁻³ averaged on a quarterly basis. These airborne Pb measurements were conducted by means of a high-volume total suspended particulate matter (Hi-Vol TSP) sampler (Reference Method for the Determination of Suspended Particulate Matter in the Atmosphere [High-Volume Method] 1983). The Hi-Vol TSP sampler mandated for Pb sampling was first adopted during the 1950s for ambient particle concentration measurements but its performance was largely unknown. When the U.S. EPA recognized the importance of Pb associated with particulate matter in 1973, evaluation of the sampler's performance was initiated. This increase in characterization of the Hi-Vol TSP is reflected in the number of articles that were published in the decade between 1973 and 1983. In the decade prior to 1973, only two peer-reviewed literature articles or technical reports were published; the following decade saw that number rises to 12 publications.

Although the ‘‘cut point,” or particle aerodynamic diameter at which 50% of particles are effectively collected, of the Hi-Vol TSP was poorly characterized, it was generally assumed that the majority of particles were captured on the filter media. Because the nominal 25 cm s⁻¹ upward capture air inlet velocity of the Hi-Vol TSP sampler equals the settling velocity of 100 µm aerodynamic particles, it was generally assumed that particles up to approximately 100 µm would be captured efficiently by the sampler independent of ambient wind speed. However, initial evaluation of the Hi-Vol TSP performance revealed errors in this assumption almost immediately regarding the sampler's size-selective performance. Issues affecting sampler performance were identified, including ambient mean particle aerodynamic diameter, sampler orientation relative to wind direction, wind speed, and the potential for deposition or loss of material on the sampler filter during passive periods (McKee et al. 1972; Bruckman and Rubino 1976; Chahal and Romano 1976; Thanukos et al. 1977; Wedding et al. 1977; McFarland and Rodes 1979; Swinford 1980; Sweitzer 1980; Hofschreuder et al. 1983; Watson et al. 1983; van der Meulen et al. 1984; Hollander et al. 1990).

One reason for the large variation in reported Hi-Vol TSP efficiency is the loose design specifications described in 40 CFR Part 50, Appendix B: “The sampler cover or roof shall overhang the sampler housing somewhat….and shall be mounted so as to form an air inlet gap between the cover and the sampler housing walls. This sample air inlet should be approximately uniform on all sides of the sampler.” The use of terms such as “somewhat” and “approximately” has led to a variety of differing designs of Hi-Vol TSP samplers. The lack of a performance standard with which to evaluate, these designs has resulted in the reporting of significant variations in performance such as the particle collection target diameter range of 25–50 µm specified in 40 CFR Part 50, Appendix B. This variation in performance is also observed in the acceptable sample flow rate range of 1.1 m³ min⁻¹ to 1.7 m³ min⁻¹ (39–60 ft³ min⁻¹) and particle capture air velocity through the inlet gap of 20–35 cm s⁻¹. The lack of strictly defined performance standard and broad acceptable performance ranges provided in the 40 CFR Part 50, Appendix B, has led to differences in reported values and difficulties with intercomparisons between studies.

With the recognition that the NAAQS should reflect a health-based standard, the U.S. EPA, in 1987, changed the measurement of particulate matter from a TSP to a PM₁₀ standard. This shift reflected a move of the NAAQS toward a “performance” standard and away from a “design” standard. Despite the change of the NAAQS to the PM₁₀ (and later PM_2.5) standard, measurement of airborne Pb continues to be performed by Hi-Vol TSP samplers in accordance with 40 CFR Part 50, Appendix B.

Because of the documented variability in Hi-Vol TSP performance, research was undertaken by the U.S. EPA to quantify the effects of several environmental and operational parameters on the sampling efficiency of the Hi-Vol TSP. This article details Hi-Vol TSP testing recently performed in the U.S. EPA Aerosol Test Facility (wind tunnel) in Research Triangle Park (RTP), NC, USA, and provides analysis of several variables on Hi-Vol TSP sampling efficiency.

Previous Hi-Vol TSP research

Two seminal papers characterizing Hi-Vol TSP sampler performance were published in the 1970s. The first, by Wedding et al. in 1977, evaluated the Hi-Vol TSP as a function of both particle size and sampler orientation. In 1979, a conference proceeding paper by McFarland and Rodes described efforts to evaluate the sampling effectiveness of two different Hi-Vol TSPs as a function of wind speed, orientation, and aerosol diameter.

The study by Wedding et al. was conducted in a 3.65 m × 1.83 m (W × H) wind tunnel operated at 16.46 km h⁻¹ (15 ft s⁻¹). Electrically neutralized, fluorescently tagged monodisperse oleic acid particles ranging from 5 to 50 µm in aerodynamic diameter were generated by a vibrating orifice aerosol generator (VOAG). The Hi-Vol TSP was tested under two different orientations, 0° (defined as the sampler ridge parallel to wind direction) and 45°. Under these conditions, test results obtained by Wedding et al. (1977) demonstrated that a significant bias exists in the sampling effectiveness of the Hi-Vol TSP with respect to both particle diameter and sampler orientation. For example, a Hi-Vol TSP oriented at 0° and presented with particles of 5 µm at a wind speed of 16.46 km h⁻¹ demonstrated a sampling effectiveness of 97% as compared to six isokinetic samplers. However, when subsequently presented with particles of 15, 30, and 50 µm aerodynamic diameters, the Hi-Vol TSP sampling effectiveness decreased to 35%, 18%, and 7%, respectively. With the Hi-Vol TSP orientation changed to 45° and presented with identical wind speed and particle diameters, the sampling effectiveness of the Hi-Vol TSP was found to be 100%, 55%, 41%, and 34% for particles of 5, 15, 30, and 50 µm, respectively. Wedding et al. (1977) postulated that the dramatic decrease in sampling effectiveness observed with the Hi-Vol TSP at 0° was because particles approaching the blunt face of the sampler had to negotiate a drastic turning angle to enter the sampler. This hypothesis was supported by visible evidence of particle loss on the blunt face of the Hi-Vol TSP. Additional testing of the Hi-Vol TSP as a function of wind tunnel turbulence intensity ranging from <1% to 8% (wind speed of 16.46 km h⁻¹ and particle diameter of 15 µm) revealed no influence on sampling effectiveness.

McFarland and Rodes (1979) evaluated two different Hi-Vol TSPs, differentiated by their outer dimensions (Hi-Vol TSP#1, 0.38 m × 0.38 m; Hi-Vol TSP#2, 0.29 m × 0.36 m), for sampling effectiveness as a function of wind speed, orientation, and aerosol diameter. A 1.5 m × 1.5 m wind tunnel was used to evaluate the samplers. The Hi-Vol TSPs were placed inside the wind tunnel on a platform that could be rotated at one revolution per minute. Similar to the study by Wedding et al. (1977), McFarland and Rodes (1979) generated a neutralized, fluorescently tagged monodisperse oleic acid particle cloud with a VOAG. Filters from nine isokinetic samplers and the Hi-Vol TSP were extracted with ethanol. Particle size was evaluated by comparison of fluorescence from these liquid extracts and was used as a measure of Hi-Vol TSP sampling effectiveness and aerosol uniformity within the wind tunnel.

McFarland and Rodes (1979) evaluated the Hi-Vol TSPs for sampling effectiveness as a function of particle size and wind speed with the two samplers operating at a flow rate of 1.41 m³ min⁻¹ (50 ft³ min⁻¹). Particle sizes ranged from 1 to 30 µm and wind speeds tested were 2, 8, and 24 km h⁻¹. During these tests, the samplers were rotated at one revolution per minute to minimize the orientation effects observed by Wedding et al. (1977). Both Hi-Vol TSP#1 and Hi-Vol TSP#2 displayed similar sampling effectiveness curves as a function of particle size and wind speed. For all three wind speeds, particle sizes less than 10 µm were sampled with approximately 90% efficiency. Differences were observed when sampling larger particles (28 µm), with sampling efficiencies for the Hi-Vol TSP#1 approaching 72% at 2 km h⁻¹, 50% at 8 km h⁻¹, and 43% at 24 km h⁻¹. Using identical particles sizes and wind speeds, the Hi-Vol TSP#2 sampling efficiencies were 81% at 2km h⁻¹, 58% at 8 km h⁻¹, and 43% at 24 km h⁻¹.

Passive deposition, the tendency for windblown particles to deposit on Hi-Vol TSP filters during periods of non-sampling, was also evaluated by McFarland and Rodes (1979). Only the Hi-Vol TSP#2 was evaluated for passive deposition, which was calculated as the ratio of mass deposition rate (µg day⁻¹) to mean aerosol concentration (µg m⁻³) as measured by an isokinetic sampler, at two different wind speeds, 2 and 8 km h⁻¹, and particle sizes ranging from 4 to 30 µm. At 2 km h⁻¹, the deposition parameter increased linearly from approximately 2 m³ day⁻¹ for particles of 4 µm to 100 m³ day⁻¹ for particles of 30 µm. For a wind speed of 8 km h⁻¹, passive deposition followed a linear increasing trend from 10 m³ day⁻¹ at particle sizes of 5 µm to approximately 200 m³ day⁻¹ for particles of 30 µm.

In addition to wind speed, particle size, and passive deposition, McFarland and Rodes (1979) evaluated the effect of wind velocity and sampler orientation on sampler effectiveness. With the sampler oriented 45° to the wind direction, the sampler was less sensitive to wind speeds, with sampler effectiveness differing by <15% between wind speeds of 2 and 24 km h⁻¹. At 0° and 90° (roof ridge parallel and perpendicular to wind direction, respectively), sampler effectiveness displayed differences >50% between 2 and 24 km h⁻¹.

Several other researchers have examined Hi-Vol TSP performance characteristics such as passive deposition, effect of wind speed, and the potential for particle blow-off from filters. Bruckman and Rubino (1976) identified an average of 14.9% of the total weight collected by a Hi-Vol TSP operating on a typical 1-in-6 day schedule was due to passive deposition. This value compared favorably with the 12% determined by Chahal and Romano (1976). The increased filter weight due to passive deposition was found to lead to increases in reported TSP ranging from 10% to 17% for 1-in-6 days schedules, and up to 25% for Hi-Vol TSPs operated on a 24 h schedule.

Similar to these studies, Thanukos et al. (1977) evaluated the sampling effectiveness of Hi-Vol TSP samplers in high wind conditions. In this evaluation, the researchers outfitted one Hi-Vol TSP with an Andersen impactor (Thermo Scientific, Waltham, MA, USA), while a second Hi-Vol TSP was outfitted normally. All samplers were operated for 24 h on a 3-in-4 day schedule, thereby ensuring that two of the three filter samples would be collected immediately after completion of sampling while the third filter was allowed to remain in the sampler for 24 h. Thanukos and coworkers reported similar TSP concentrations for Hi-Vol TSP and Andersen impactor filters collected immediately after sampling, whereas TSP concentrations from Hi-Vol TSP filters allowed to remain in the samplers for 24 h after sampling were found to be 5.7% less than TSP concentrations determined from the Andersen impactor. The impact of high ambient TSP concentrations on Hi-Vol TSP performance was also evident in the work of Thanukos et al. (1977). During days with high TSP concentrations (TSP > 150 µg m⁻³), geometric average TSP values recorded by the standard Hi-Vol TSP were 35.6% lower than those reported by the Andersen impactor. These high TSP days were typically characterized by high winds with hourly averages in excess of 14 mi h⁻¹. Because both samplers had similar external geometries, Thanukos et al. (1977) hypothesized that lower sampling effectiveness observed with the standard Hi-Vol TSP was due to particle blow-off from the filter.

McKee et al. (1972) performed a field comparison of 12 identical, collocated Hi-Vol TSPs. While excellent agreement was found among 11 of the Hi-Vol TSPs, with relative standard deviations of 3.0% for a single analyst and 3.7% between different laboratories, loss of material from the filter was observed during a 4-day equilibration period after sampling. McKee et al. (1972) reported weight loss up to 5% of the initial filter weighing. This loss of material was postulated to be due to loss of volatile organic material.

Limited field evaluations have been conducted of the Hi-Vol TSP sampler against other samplers. Hofschreuder et al. (1983) tested the Hi-Vol TSP against several samplers including isokinetic samplers. During these tests, monodisperse dioctylphthalate (DOP) particles were generated by spinning-top generators at a location 30 m upwind of the Hi-Vol TSP. Although limited experimental information was presented, Hofschreuder et al. (1983) reported a Hi-Vol TSP efficiency of 33% as compared to the isokinetic samplers for 17 µm particles.

Hollander et al. (1990) conducted a series of comparisons between the Hi-Vol TSP and the wide range aerosol classifier (WRAC) throughout Germany in both industrial and rural areas. For these tests, the Hi-Vol TSP was operated at a flow rate of 68.4 m³ h⁻¹, significantly lower than the 85.0 m³ h⁻¹ used for McFarland and Rodes (1979). Hollander et al. (1990) reported linear correlation coefficients exceeding 0.95 for comparisons between the Hi-Vol TSP and the WRAC in urban and suburban areas. In the coastal area, however, this correlation dropped to 0.83, with a significant increase in the average ordinate intercept of the linear trend line (2.92 for urban and suburban areas to 35.92 for coastal areas). The decrease in Hi-Vol TSP-WRAC correlation was accompanied by an increase in the sample standard deviation from 2.99 for urban and suburban areas to 37.68 for the coastal site. This increase in variability of the Hi-Vol TSP-WRAC data set, together with average wind speeds at the coastal site (6 m s⁻¹) being 50% greater than at all other sites, led Hollander et al. (1990) to conclude that sampling efficiency is determined by particle inertia.

Researchers have also performed statistical modeling of Hi-Vol TSPs to estimate their performance under a variety of ambient particle size distributions. Watson et al. (1983) used sampling effectiveness curves as a function of wind speed generated by McFarland and Rodes (1979) along with hypothetical particle size distributions reconstructed from reported urban, background and aged urban plume, and marine particle size distributions. Each particle size distribution was modeled with 33% and 50% of the particle mass being classified as fine (0.2 to 0.7 µm) and the remainder being coarse particles (3 to 30 µm). Watson et al. (1983) determined Hi-Vol TSP sampling effectiveness for average urban particles ranged from a high of 92% for particles with 50% fine mass and a wind speed of 2 km h⁻¹ down to 70% for particle size distributions of 33% fine mass and wind speeds of 24 km h⁻¹. Background and aged urban plumes were estimated to be sampled with an effectiveness greater than 90% for all particle size distributions and wind speeds, while marine particles had the lowest estimated sampling effectiveness, with values ranging from 92% at particle size distributions of 50% fine mass and wind speeds of 2 km h⁻¹ to 72% at particle size distributions of 33% fine mass and wind speeds of 24 km h⁻¹. Similar to Watson et al. (1983), van der Meulen et al. (1984) used Hi-Vol TSP performance curves generated by McFarland and Rodes (1979) to estimate Hi-Vol TSP sampling effectiveness with synthetic particle size distributions. They determined that the Hi-Vol TSP sampler underestimated TSP by up to 30% when coarse particles dominated the particle size distribution, falling to 5% underestimation when fine particles were dominate. van der Meulen et al. (1984) also estimated Hi-Vol TSP performance for particle size distributions related to traffic emissions. They determined an estimated minimum of 70% of the direct automotive emissions would be collected by a Hi-Vol TSP. When resuspended road dust was included in the calculation, estimated sampling effectiveness ranged from 70% to 120% near the roadway, with estimated Hi-Vol TSP sampling effectiveness dropping to 70%–110% when the sampler was located 50 m from the roadway.

Historical data gaps

The historical data presented here point out several limitations in our existing knowledge of Hi-Vol TSP sampler performance. The previously discussed sampling effectiveness of the Hi-Vol TSP with wind speed, orientation, ambient particle size, and passive deposition during non-sampling periods was obtained under differing conditions, such as wind speed, sampler flow rates, and particle types. The lack of uniform testing procedures makes developing appropriate sampling effectiveness curves difficult, if not impossible. Table 1 summarizes the uncertainty ranges reported as a function of sampler operational state, orientation, particle diameter, and wind speed.

Table 1. Reported ranges of Hi-Vol effectiveness based on investigated experimental factors.

Factor	Reported Hi-Vol TSP efficiency
Operational state^1	107%^2, 110%–125%^3, 117%^4, 140%^5
Orientation	7%–100%^6, 30%–88%^7
Particle diameter	7%–100%^6, 43%–100%^7, 70%–95%^8, 72%–98%^9
Wind speed	7%–100%^6, 43%–100%^7, −36%^10

¹ Reported Hi-Vol TSP efficiency for operational state reflects increase in TSP due to passive deposition on filter.

² Chahal and Romano (1976);

³ Swinford (1980);

⁴ Sweitzer (1980);

⁵ Bruckman and Rubino (1976);

⁶ Wedding et al. (1977); depends on wind speed and particle size;

⁷ McFarland and Rodes (1979); depends on wind speed and particle size;

⁸ van der Meulen et al. (1984); depends on wind speed and particle size;

⁹ Watson et al. (1983); depends on wind speed and particle size;

¹⁰ Thanukos et al. (1977); negative value reflects particle loss due to filter blow-off.

Experimental

Aerosol generation

Tests of the Tisch Hi-Vol TSP Model TE-5170 VFC+ (Tisch Environmental, Cleves, OH, USA) were performed to evaluate effectiveness as a function of particle diameter, wind speed, orientation, and operational state (on/off) in EPA's Aerosol Test Facility located in RTP, NC. The Tisch Hi-Vol TSP has overall dimensions of 0.48 m wide × 0.48 m long × 1.37 m tall, with a 0.38 m square body. Additional details of the Tisch Hi-Vol TSP can be found on the Tisch Environmental website.

The wind tunnel in EPA's Aerosol Test Facility is divided into two sections, the human exposure test section (HETS) and the sampler test section (STS) as illustrated in Figure 1. The large cross-sectional area of the wind tunnel (HETS: W = 3.7 m, H = 3.1 m; STS: W = 1.8 m, H = 1.5 m) allows for evaluation of large samplers such as the Hi-Vol TSP in the STS while blocking less than 14% of the wind tunnel.

Figure 1. EPA's aerosol wind tunnel with air flow direction, distances in meters, and important sections labeled: (a) aerosol generator, (b) oscillating fans to provide secondary flow and enhance horizontal mixing, and (c) test sampler location. Not drawn to scale.

It has been previously noted that the use of liquid test aerosols may not accurately represent the collection characteristics of large atmospheric particulate matter (Wang and John 1987; Koehler et al. 2011). Evaluations of the Hi-Vol TSP were conducted using Arizona Test Dust (ATD, Powder Technology Inc., Arden Hill, MN, USA), a dry, solid polydisperse test aerosol. Custom mixtures of ATD, Table 2, were dispensed from a volumetric feeder (Model 102, Schenck Accurate, Whitewater, WI, USA) equipped with a variable-speed screw drive. ATD from the feeder is drawn into the aerosol generation system where it is aerosolized by a stainless-steel sonic nozzle (Model JS-90M-316SS, Vaccon Company, Medway, MA, USA) operating at a supply pressure of 5400 mmHg (90 psig). Sheer forces within the sonic nozzle are sufficient to aerosolize the ATD as discrete particles as verified by collection and microscopic examination in the laboratory. Supply air to the sonic nozzle is fed through a pulsed DC ion generator (Model 261, Meech International, Oxfordshire, England) and the aerosol output from the nozzle is carried through a steel tube (L = 30 cm, D = 0.15 m) by the output of four additional pulsed DC ion generators. The five-port array of ion generators is controlled by a Meech pulsed DC controller Model 977CM and effectively neutralizes the ATD aerosol.

Table 2. Custom dust mixtures.

	% (m/m)
Wind speed (km h^−1)	A4	10–20	20–40
2	20	80	—
8	—	20	80
24	—	20	80

Active and passive sampling tests used the same dust mixture for a given wind speed.

Wind speeds of 2, 8, and 24 km h⁻¹ were generated in the STS by volumetric control of a direct-drive, vane-axial fan (Twin City Fan and Blower, Minneapolis, MN, USA). Temperature and relative humidity for all tests was maintained at 25°C and 25%, respectively. The aerosol generation system functioned as a point source and was centered vertically and horizontally in the HETS located approximately 11 m upstream of the STS. The aerosol generation system was oriented in the upwind direction such that the flow from the sonic nozzle being introduced to the major flow within the HETS section formed a particle plume to assist in dispersion of the aerosol. Additional mixing in the HETS was accomplished by the use of five oscillating 0.75 m diameter mixing fans providing secondary flow perpendicular to the major tunnel flow. Velocity profiles in the STS were determined by performing multi-point measurements with a thermal anemometer (Model AQTI-01, Dwyer Instruments, Michigan City, IN, USA). The tunnel configuration as shown in Figure 1 was capable of achieving sufficient velocity and aerosol uniformity in the STS for all three wind speeds as determined by EPA's wind tunnel testing requirements and acceptance criteria in Subpart D - Procedures for Testing Performance Characteristics of Methods for PM10 (1987).

Aerosol collection and setup

Determination of the Hi-Vol TSP's size-selective performance required measurement of the size distribution of the ‘‘challenge” aerosol in the STS during the period of testing. Reference aerosol samples were collected using two identical, custom-designed 90 mm diameter filter holders designed to operate at an inlet flow rate of 100 actual liters per minute (aLpm). For the 90 mm filter holders, a series of custom-design sharp-edged, stainless-steel isokinetic nozzles were machined with internal nozzle diameters of 6.18 cm, 3.09 cm, and 1.78 cm, which correspond to isokinetic sampling at simulated wind speeds of 2, 8, and 24 km h⁻¹, respectively. The correct volumetric flow rate was maintained by a custom-designed flow control system equipped with automatic flow valves, flows were continuously monitored using calibrated flow sensors (Model 4045, TSI Inc., Shoreview, MN, USA). Prior to each test, the sampling inlet and vacuum system were leak checked and the volumetric flow rate verified using a TetraCal flow calibration system (Mesa Laboratories, Butler, NJ, USA).

Polycarbonate membrane filters were used to ensure efficient removal of collected particles for subsequent size distribution determination, the isokinetic reference samplers were equipped with 90 mm pre cleaned, 1 µm pore size polycarbonate Whatman filters (GE Healthcare, Piscataway, NJ, USA). The Hi-Vol TSP was equipped with 200 mm × 250 mm pre cleaned, 3 µm pore size polycarbonate Sterlitech filters (Sterlitech Corporation, Kent, WA, USA). Due to the nature of membrane filters and small pore size, pressure drop in the Hi-Vol at 85 m³ h⁻¹ (50 actual cubic feet per minute [aCFM]) was approximately 500 mmHg. To maintain sufficient volumetric flow, a custom vacuum system was developed consisting of a Roots rotary lobe blower Model: 33 U-RAI (Dresser Inc., Houston, TX, USA), driven by a premium efficiency motor Model: AEHH8P (Teco Westinghouse, Round Rock, TX, USA), and controlled by an inline air flowmeter with integrated PID driver Model: 504FTB-24 (Kurz Instruments, Monterey, CA, USA). A custom 203 mm × 254 mm (8′ × 10′) filter cassette was machined from 6.35 mm billet aluminum stock to provide sufficient structural rigidity to support the filter under the relatively high vacuum required (Figure S3). Prior to each test, the Hi-Vol TSP inlet and vacuum system were leak checked and the volumetric flow rate verified using a hiVolCal high volume air flow calibrator (BGI Incorporated, Waltham, MA, USA). The set point of the integrated proportional—integral—derivative (PID) controller/driver was adjusted so the hiVolCal read 85 m³ h⁻¹. Volumetric flow rate during the testing period was measured by a Roots meter Model: 5M (Dresser Inc., Houston, TX, USA), temperature and pressure of air entering the roots meter were measured along with elapsed time and were used to calculate the m³ h⁻¹ flow rate and total sample volume for an experiment. This is detailed in the online supplementary information (SI) Section S-3.

Investigation of passive or non-functional sampling artifacts of the Hi-Vol TSP was conducted in an identical manner to the active sampling, with the exception that the Hi-Vol vacuum system was not turned on. Any particles collected were deposited to the filter surface by gravitational settling and impaction.

Prior to each day's tests, the Hi-Vol TSP was cleaned, the filter cassette was installed, the two preloaded isokinetic samplers were positioned 30 cm on either side of, and at the same elevation as, the Hi-Vol TSP as shown in Figure 2. Run time for each effectiveness test was 90 min. Typically, 3 to 10 replicate tests were conducted for each experimental condition to obtain high-confidence test results. The measurement performance of all flow devices used during this study was periodically validated at EPA's Metrology Laboratory located in RTP, NC.

Figure 2. Hi-Vol TSP installed in aerosol wind tunnel: (a) isokinetic reference sampler on left side, (b) Hi-Vol TSP in 0° orientation, and (c) isokinetic reference sampler on right side. Embedded diagram of Hi-Vol TSP as viewed from above. Wind parallel to the roof ridge is 0°, wind perpendicular to the roof ridge is 90°, and wind approaching each corner is 45°.

Size distribution measurement

Extraction of collected particles from filters involved the use of pre-filtered electrolyte consisting of an 80:20 (% m/m) mixture of Isoton II and glycerol. The volume of electrolyte used for each filter was based on the filter size and the experiment's expected loading but was typically 100 mL for 90 mm diameter filters and 800 mL for the Hi-Vol TSP's 203 mm × 254 mm filters. To account for any inadvertent particle losses within the 100 Lpm isokinetic reference nozzles, the internal surfaces of the nozzles were carefully rinsed and the nozzle extracts were included in the determination of the total reference sampler's particle concentration. The face of the Hi-Vol TSP's filter cassette was carefully cleaned with a de-inchionized (DI) water/isopropanol alcohol wipe, and the filter surface and inside edges were then carefully rinsed, to determine the sampler's particle concentration. Each filter rinse was sonicated for 10 s immediately prior to analysis. The filter extraction procedure is detailed in S-3.

The size distribution (number concentration versus physical diameter) of particles in each solution was determined using a Coulter Counter (Model Multisizer IV, Beckman Coulter, Bree, CA, USA) equipped with a 100 µm aperture. The Coulter Counter's measurement principle provides sizing data in equivalent volume diameter, the aerodynamic diameter of each measured particle was calculated by accounting for the particle density of 2.5462 g cm⁻³ (measured by ultrapycnometry) and a dynamic shape factor for ATD of 1.4 (Endo et al. 1998; Mohler et al. 2008):(1) where:µm

µm

Effectiveness calculations

The sampling effectiveness (E) of each discrete particle size was calculated for the candidate inlet as follows:(2) where:

, #/mL

, #/mL

, #/mL

All evaluations of Hi-Vol TSP effectiveness were performed at fixed angles with respect to the approach of dust; the omni-directional effectiveness at each wind speed was calculated for each discrete size bin by summing the number of possible approach angles in Figure 2:(3) where:

To aid in the interpretation of results SAS (SAS 9.4, SAS Institute Inc., Cary, NC) was used to perform a pair-wise non-parametric analysis to evaluate if the distributions between the test scenarios were significantly different (Kruskal–Wallis, p < 0.001). A p-value less than 0.001 indicates that the impacts from either wind speed and/or wind direction significantly modify the Hi-Vol TSP's effectiveness for a given condition. The Kruskal–Wallis analysis was run on each effectiveness curve for 5 to 20 and 20 to 30 µm aerodynamic diameter thereby separating the effects of the lower and middle range of the effectiveness curve from the top end where very large particles and counting statistics may play a more significant role.

Results and discussion

Reference sampler performance

The actual aerosol concentration for each sampling period was established by two reference isokinetic samplers installed to the left (R₁) and right (R₂) of the Hi-Vol TSP. As illustrated in Figure 3, the ratio R₁/R₂ was consistently close to unity for aerodynamic particle sizes less than 25 µm, indicating that the measured concentration is equal on both sides of the Hi-Vol TSP. For aerodynamic particle sizes greater than 25 µm, the measurements remain close to unity; however, the confidence interval increases significantly. This increase in the confidence interval is primarily an issue of counting/statistics, since the dust blend used has a low percentage (<10%) of particles greater than 25 µm (details of the relative size distributions of ATD are illustrated in S-1). In addition, large particles (>25 µm) are not transported as effectively as small particles down the tunnel from the aerosol generation system to the STS sampling location. The net result is that there are few particles in each size bin greater than 25 µm. The compounding effects of a smaller number density of large particles and less efficient transport down the tunnel get progressively worse as the aerodynamic particle size increases.

Figure 3. Ratio of isokinetic sampler 1 (R₁) counts to isokinetic sampler 2 (R₂) counts for 2 km h⁻¹ and 90° orientation. Whiskers represent ± 1 standard deviation of replicate tests, n = 4.

Attempts were made to improve the number of particles collected in the larger size bins, however several factors limited the possible improvement. For optimum Coulter Counter performance, the electrolyte extract of the filters must be less than 10% solids by volume to limit coincidence counting, and the ATD dust is a naturally occurring sieved dust, meaning that even the largest size dust (20–40 µm) has significant numbers of particles below 10 µm in diameter (Figure S1). These two limiting factors were balanced and effectively optimized through iterative adjustment of ATD mixtures, sample collection period, and electrolyte extraction volumes.

Hi-Vol TSP sampling effectiveness

The Hi-Vol TSP was evaluated for size-selective sampling effectiveness for 5 to 35 µm aerodynamic diameter. Figure 4 represents the laboratory-evaluated effectiveness of the Hi-Vol TSP operated at 50 aCFM while sampling 2 km h⁻¹ wind at an orientation of 0° (roof ridge parallel to wind). Effectiveness from 5 to 10 µm is approximately 80%. Above 10 µm, effectiveness declines with increasing aerodynamic particle size. McFarland and Rodes (1979) have the only data point available from literature for comparable conditions of 2 km h⁻¹ wind speed and fixed orientation of 0°, they used an oleic acid monodisperse particle of 23.5 µm and measured an effectiveness of 71% (Hi-Vol TSP#2, 0.29 m × 0.36 m data are used for comparison as the geometry most similarly resembles the geometry of the Tisch Hi-Vol TSP used in this study), which falls within one standard deviation of the results of this study. The results of this study for 2 km h⁻¹ at orientations of 45° and 90° are consistent with those found at 0° as seen in Figure 5. Kruskal–Wallis analysis of the 2 km h⁻¹ results showed no significant effects of orientation (p > 0.001) contrary to the findings of McFarland and Rodes, which found orientation to have significant impact on sampler effectiveness. Complete results of the Kruskal–Wallis analysis can be found in the SI (S-5).

Figure 4. Effectiveness of the Tisch Hi-Vol TSP operated at 50 cfm while sampling 2 km h⁻¹ wind at an orientation of 0°. Whiskers represent ± 1 standard deviation of replicate tests, n = 4.

Figure 5. Directional effectiveness of the Tisch Hi-Vol TSP operated at 50 cfm. Dots represent this study, open diamonds represent McFarland and Rodes (1979) data. Replicates for each condition were typically between 4 and 7.

At 8 km h⁻¹, overall sampling effectiveness is very similar to that at 2 km h⁻¹. There is a slight increase in effectiveness with particles in the 10 to 15 µm range at 2 km h⁻¹, which is postulated to be a result of increased wind-driven penetration into the sampling region. This effect is most notable at the 90° orientation, where the roof of the Hi-Vol TSP forms a wing-like shape to the approaching wind, potentially increasing velocity under the roof and across the filter region where particles are most likely to be captured. For 23.5 µm particles, McFarland and Rodes reported effectiveness values of 59%, 66%, and 56% while this study found values of 62%, 70%, and 69% for 0°, 45°, and 90° orientations, respectively, as seen in Figure 5. The Kruskal–Wallis analysis shows significant differences for 0° to 45° and 0° to 90° for both 5–20 and 20–30 µm, but found 45° to 90° to be insignificant for both size ranges.

At 24 km h⁻¹, Hi-Vol TSP effectiveness decreases more rapidly as aerodynamic particle diameter increases. Sub-10 µm effectiveness is notably increased for the 0° and 90° orientations versus those at 2 and 8 km h⁻¹. This is likely due to higher static pressure on the leading side of the Hi-Vol TSP, driving more wind-driven penetration of small particles across the face of the filter. At the 45° orientation, there is significant effectiveness below 50% for particle sizes over 25 µm, possibly due to the more aerodynamic characteristic of the Hi-Vol TSP at 45° allowing large particles to pass around the sampler without entering the sampling region, an effect that is only apparent at the highest wind speed tested. Effectiveness for large particles was found to be slightly improved for 0° versus 45°, while showing significant increase at the 90° orientation. Again, it is postulated that this is the effect of the wing-like shape of the Hi-Vol TSP's roof at this orientation allowing greater velocity under the roof and across the filter-sampling region. McFarland and Rodes measured effectiveness in the 24 km h⁻¹ and 45° condition for 23.5 µm particles to be 57%, in strong agreement with this study's measured effectiveness of 51%. In a reversal of the trend found in this study, McFarland and Rodes found effectiveness for 23.5 µm particles at 0° and 90° to be 33% and 30%, respectively, while this study found 56% and 75% for the same conditions. The measured effectiveness in this study may be due to the effects of particle bounce allowing collection of particles following contact with the face of the Hi-Vol TSP. Kruskal–Wallis analysis showed significant effects of orientation on effectiveness (p < 0.001) at 24 km h⁻¹ in all conditions with the exception of 0° compared to 45° for 5–20 µm particles.

McFarland and Rodes reported omni-directional results for two aerodynamic diameters, 10 and 28 µm, which a comparison can directly be made with this study's findings. For aerodynamic particle diameters of 10 µm, McFarland and Rodes reported higher effectiveness values for 2, 8, and 24 km h⁻¹ during the 1 rpm rotational tests than calculated omni-directional values in this study as illustrated in Figure 6. At 28 µm, the trend is reversed and omni-directional effectiveness values for 8 and 24 km h⁻¹ are higher in this study with calculated values of 60% and 53% versus 50% and 43% for McFarland and Rodes. Interestingly, at 2 km h⁻¹ McFarland and Rodes and this study are in very close agreement with measured effectiveness values of 68% and 68%, respectively. The discrepancy between McFarland and Rodes and this study for larger aerodynamic particles for the two higher wind speeds appears to support the hypothesis that there is significant post-bounce collection for solid, atmospherically relevant particles versus the liquid aerosols used in McFarland and Rodes. It is worth noting that in McFarland and Rodes two different values are reported for effectiveness at 28 µm, one set from their figures reports 68%, 40%, and 43% while values in the text are reported as 81%, 58%, and 43% for 2, 8, and 24 km h⁻¹, respectively. While the reported values agree at 24 km h⁻¹, it is unclear why there is a discrepancy in reported values at 2 and 8 km h⁻¹. A summary of effectiveness values from McFarland and Rodes (1979) and Wedding et al. (1977) can be found in S-4.

Figure 6. Calculated omni-directional performance of the Hi-Vol TSP operated at 50 cfm. McFarland & Rodes 1 revolution per minute data points in solid markers.

Hi-Vol TSP passive sampling

Passive sampling tests were conducted for each wind-speed and orientation to determine the level of positive sampling that will occur when the Hi-Vol is in a non-operational state. The Hi-Vol TSP vacuum system was left in an unpowered state such that any particle deposition on the filter area was due to wind-driven effects. The passive sampling effectiveness was calculated in the same manner as the active sampling tests. Figure 7 shows the calculated passive omni-directional effectiveness for each wind speed. Passive deposition increases proportionally with wind speed, since more particles are driven into the Hi-Vol TSP at higher wind velocity. Directional values showed there was no significant effect of orientation on passive sampling so only omni-directional values are presented. Average omni-directional passive sampling over an aerodynamic diameter from 5 to 35 µm was measured to be 1.7 ± 0.4%, 3.1 ± 0.3%, and 5.5 ± 0.3% for 2, 8, and 24 km h⁻¹ wind speeds, respectively. The limited information on passive sampling available in literature is from ambient 1-in-6 day samples where Chahai and Romano (1976) found 12% of the mass on the filter was from passive deposition and Bruckman and Rubino (1976) who found 14.9% of the mass from passive deposition. Unfortunately, no wind speed data or time series ambient aerosol concentrations were presented in either study, making comparisons to data in this study problematic. Estimates for 1-in-6 day passive sampling can be made by multiplying the omni-directional passive sampling values by the number of non-functional days in the study, five, resulting in values of 8.5%, 15.5%, and 27.5% for 2, 8, and 24 km h⁻¹, respectively. If the typical wind speed during the five non-operational days of the field studies was between 2 and 8 km h⁻¹, the resulting mass measurement bias of 8.5% and 15.5% from our wind tunnel measurements are consistent with previous studies.

Figure 7. Omni-directional passive sampling artifacts when in a non-operational state.

Conclusions

A solid, polydisperse dust when properly dispersed and evaluated with a Coulter Counter can be used to generate more complete effectiveness curves in fewer experiments than traditional monodisperse techniques. The technique used in the article produces effectiveness data for approximately 45 discrete particle sizes for each test condition, rather than the relatively few sizes at which the Hi-Vol TSP had been previously tested. Results indicate that Hi-Vol TSP effectiveness is a near-monotonic function of aerodynamic particle size, as generally predicted by sampling theory.

Performance of the Hi-Vol TSP was evaluated with consideration of four factors: orientation, wind speed, particle diameter, and operational state. This study found that orientation does not have a significant effect (p > 0.001) on sampler effectiveness at 2 km h⁻¹, is significant in 4 out of 6 cases at 8 km h⁻¹ and is significant in 5 out of 6 cases at 24 km h⁻¹. In general, variability due to sampler orientation increases with wind velocity. We did not obtain results which support the level of variability previously reported in Wedding et al. (1977) of 7%–100% or McFarland and Rodes (1979) of 33%–88%. This study found that wind speed plays a significant role in sampler effectiveness, with that significance increasing with wind speed and particle size. Comparing 2 to 24 km h⁻¹, Kruskal–Wallis analysis found that in every condition the changes were significant for particles 20–30 um. At 24 km h⁻¹ wind speed, the slope of the effectiveness curve increased when compared to the lower wind speeds regardless of sampler orientation. This study reports effectiveness values that range from 42% to 92% based on particle diameter, across orientations and wind speeds. This range is comparable to some previously reported values and smaller than others as summarized in Table 1. This study employed the use of solid particles with dynamic shape factors, more representative of ambient aerosol that previous studies use of liquid phase, spherical aerosols.

Finally, operational state of the Hi-Vol TSP was evaluated for passive sampling effectiveness and was found to vary from less than 1% for small particles at 2 km h⁻¹, to upward of 6% for large particles at 8 and 24 km h⁻¹. These values are significant, but smaller than some previously reported values of up to 140%. It is worth noting that the positive measurements bias resulting from passive sampling is a function of wind speed, particle size distribution, and particle concentration, which the sampler experiences during non-sampling periods. Thus, it is virtually impossible to predict the magnitude of passive sampling errors or accurately adjust measured concentrations for a given sampling event.

Acknowledgments

The author would like to thank the RTI International, Alion Science and Technology, Jacobs Technology Inc., and Ali Kamal for their contributions to this work.

Funding

The United States Environmental Protection Agency through its Office of Research and Development funded and managed the research described here. It has been subjected to Agency review and approved for publication. Mention of products or trade names does not indicate endorsement or recommendation for use by the agency.