The Ddevelopment and Designation Testing of a New USEPA-Approved Fine Particle Inlet: A Study of the USEPA Designation Process

Abstract

This article discusses the practical challenge of meeting USEPA requirements for equivalency between novel particulate matter monitoring instruments and the USEPA WINS PM_2.5 Impactor (i.e., the Federal Reference Method sampler for fine particulate matter). A project was undertaken to develop a new PM_2.5 instrument in which the WINS impactor was substituted by a cyclone, to give superior performance over long sampling periods under heavy loading. Empirical cyclone models were used to develop a new generation of very sharp cut cyclones (VSCC), together with a particular VSCC specimen suited to PM_2.5 sampling at 16.67 l min⁻¹. In laboratory tests, this VSCC demonstrated a precise 2.5 μm D₅₀ cutpoint and sharpness as good as the WINS. A formal application was then undertaken to achieve USEPA Class II Equivalency designation. The process included aerosol laboratory loading trials, with results showing no change in cutpoint after up to 90 days between cleaning cycles. Field trials to compare the VSCC to the WINS FRM were then performed in both western and eastern air sheds to demonstrate the precision and accuracy of the candidate VSCC FEM. The results showed that the VSCC instrument yielded precision and accuracy within USEPA requirements, although the USEPA data requirements for the field trials (in terms of aerosol size distribution and concentration) were not fully met. The outcome of the project was that the Class II equivalency designation was achieved, but not without major difficulties in gathering suitable and sufficient data to meet the stringent test requirements laid down by USEPA. Some changes in the designation procedure are recommended in light of this experience.

INTRODUCTION

Methods for measuring ambient concentrations of particulate matter (PM) may be designated as Federal Reference Methods (FRM) or Federal Equivalent Methods (FEM) in accordance with the Code of Federal Regulations (40 CFR Part 53). A list of designated reference and equivalent methods is updated periodically and is available from the USEPA (Federal Register 2002).

Reference methods for fine particulate matter (PM_2.5) were first designated in 1998. The PM_2.5 FRM samplers utilize an impactor, the Well Impactor Ninety Six (WINS), as a fractionator to select PM_2.5 downstream from a wind-tunnel-proven inlet (Tolocka et al. 2001) operating at a flow rate of 16.67 l min⁻¹. This inlet's performance was well characterized during the development of an earlier generation of sampling instruments for PM₁₀, and this eliminated the need for all but confirmatory wind tunnel testing of the new generation of PM_2.5 samplers. FRM samplers collect a single sample of PM_2.5 onto a filter for gravimetric analysis. Three classes of equivalent methods are envisaged; Class I equivalent samplers utilize the same inlet and size fractionator as the FRM method but are adapted to collect sequential filter samples of PM_2.5; Class II equivalent samplers can utilize an inlet and/or size fractionator of different design to the FRM but collect a single sample onto a filter for gravimetric analysis; and Class III equivalent samplers are any devices not falling into either Classes I or II. For example, instruments that collect and analyze PM continuously rather than producing filter samples for analysis would fall into Class III.

The stated purpose of allowing equivalent methods is to encourage the development of improvements to reference methods; however, all candidate equivalent method samplers are required to undergo rigorous testing and meet exacting performance requirements in order to be designated, as detailed in the Federal Register (1997). There are generic laboratory tests of performance that apply to all three Classes of FEM, as well as additional specific tests that are required for Class II and Class III equivalent samplers, depending on the nature of the design deviations from the FRM inlet and fractionator systems. In the case of inlet deviations, full wind tunnel testing of the alternative inlet would be required. In the case of fractionator deviations, the alternative fractionator is required to be tested in terms of its particle size separation and loading characteristics. Finally, in addition to these laboratory performance tests, field tests are required for all three classes of FEM samplers to demonstrate precision and accuracy in comparison with colocated FRM samplers. For Class II and Class III samplers, these field tests must take place in two different air sheds, possessing specified characteristics in terms of particle size distribution and concentration in order to highlight any biases attributable to the design deviations in the fractionator system.

Given the technical challenges and costs involved in undergoing the required testing programs for equivalency, there has been almost no progress in developing and designating FEM samplers since 1998. Several instrument manufacturers have developed prototype sequential samplers (Class I FEM) but none have yet been formally submitted for designation. The first and only designated Class II FEM for PM_2.5 was approved in 2002 and is described in this article. Many of the state authorities responsible for regular monitoring of PM_2.5 against national ambient air quality standards would prefer, for practical reasons, to use Class III equivalent automatic PM_2.5 samplers as alternatives to the FRM. Unfortunately, although several commercially available instruments exist, the requirements for PM_2.5 Class III equivalent method designation (for automated methods) are considered technically impossible and cost prohibitive, and therefore no commercial instrument company has pursued this endeavor to date.

It is interesting to note that since 1982, when the first ambient particulate reference method for TSP was designated, all reference and equivalent USEPA methods since approved remain designated regardless of any subsequently reported data and operational faults. Rules for performance and data acceptance are included in 40 CFR Part 53; however, the USEPA has never challenged or removed a method for poor performance, once it is designated. Field and performance data for several of the previously designated reference methods have demonstrated poor precision and accuracy, and some have a failure rate not acceptable for state and local agency use. Additionally, there is a growing list of designated methods that are obsolete and no longer commercially available but that remain as published USEPA methods.

Given the extremely demanding requirements for both PM_2.5 Class II and III equivalency, the instrument we describe in this article is the first and only approved equivalent method (FEM) to be designated since promulgation of the USEPA fine particle method in 1997. This sampler uses a very sharp cut cyclone (VSCC) as the PM_2.5 fractionator in place of the WINS impactor but is otherwise identical to an FRM sampler. An earlier project to develop an alternative fractionator to the WINS impactor was undertaken as a collaboration between BGI Instruments (BGI), Rupprecht & Patashnick Company, and the UK Health and Safety Laboratory (HSL) in 1996 through 1999 (Kenny et al. 2000). The resulting sharp-cut cyclone (SCC) was submitted for verification by the USEPA and its contractor, who found it to exhibit small sampling biases with respect to the USEPA WINS impactor (these data remain unpublished). For this reason, BGI and HSL undertook a new project in 2001 to further improve the SCC. The VSCC is the result of that work, and it is now a designated Class II equivalent method. As we show in this article, achieving Class II equivalent method designation is a protracted and technically challenging process. However, it is feasible, at least if the design modifications to the FRM are limited to the fractionator system.

DEVELOPMENT AND TESTING OF THE VSCC

Rationale

Based upon the published opinions of prominent epidemiologists (Schwartz et al. 1996) regarding excess deaths from sub-10 μm particles, in 1995 The American Lung Association brought a suit against EPA and forced a review of the methods existing at that time for ambient particulate sampling, which were based on the sub-10 μm fraction of ambient aerosol, PM₁₀ (Federal Register 1997). While there was complete agreement that a size-selective median cutpoint lower than 10 μm was required for health-related particulate monitoring, there was no consensus as to what it should be. Opinion ranged from 4 μm down to 1 μm. In the absence of available hard scientific information, PM_2.5 was selected for historical reasons. To meet the requirement for a secondary PM_2.5 fractionator, USEPA staff considered both cyclone and impactor devices (Peters et al. 1996) and, based upon laboratory testing, selected a well-type wetted surface impactor (Peters et al. 2001a). This device was given the appellation Well Impactor Ninety Six (WINS). It is important to note that because of the court-imposed time schedule, no definitive field trials were conducted prior to its introduction.

Part of the published FRM requirement was that the secondary PM_2.5 fractionator should meet the following performance requirements:

1.	D50 cut to be 2.5 ± 0.2 μm.
2.	Sampling bias for three predefined ambient particle size distributions to be less than 5%.

Initial laboratory testing of the WINS indicated that it could meet these criteria. After a few years of experience in the field, two areas of concern with the WINS surfaced, in the form of “frozen” WINS oil at low ambient temperatures and the effect of loading on the cutpoint and bias. The USEPA studied both of these problems and published their findings (Vanderpool et al. 2001a,b).

Independent experiences reported with the WINS stimulated the development of a replacement device (Kenny et al. 2000). The main drive for this work was the desire to develop continuous monitors in order to reduce the personnel costs of monitoring. The loading effects noted for the WINS indicated that it would not be suitable for continuous monitoring in many situations because the loading effects precluded extended operation between cleaning cycles. Cyclonic devices were demonstrated to have significant advantages over impactors for extended operation; however, problems remained in eliminating bias between the candidate cyclones and the WINS.

Design and Validation of the Penetration Curve

The very sharp cut cyclone is a tangential, round-entry cyclone based on the design of the SRI-II cyclone originally described by Smith et al. (1979). The SRI cyclones were developed for size-selective stack sampling but have since proved useful for a range of applications stemming from a re-examination of their characteristics by Kenny and Gussman (1997). This work showed that cyclones could be dimensionally scaled using an empirical “family model” peculiar to a given cyclone geometry. The family model relates cyclone cutpoint (i.e., penetration D₅₀) to cyclone body diameter and flow rate. For a given cyclone geometry the family model has only two empirical parameters, which means that the model can be fitted from minimal experimental data, for example, D₅₀ values measured at only two flow rates for a single specimen of the cyclone family. Once the family model is known, it can be used to calculate the dimensions for other family members that will yield a specified D₅₀ at a specified flow rate.

Our work on the SRI cyclones led to the development of three useful families of cyclones. The GK family, with similar geometry to the SRI IV cyclone, has a very gradual (i.e., unsharp) penetration curve ideally suited to occupational hygiene sampling. The SCC family, with similar geometry to the SRI III cyclone, has a much sharper penetration curve and offers comparable size selectivity to the WINS impactor for ambient PM_2.5 sampling, but with much better performance under loading (Kenny et al. 2000). Finally, the VSCC cyclone described in this article is closely modeled on the SRI-II cyclone described by Smith et al. (1979).

Early experience with the SRI cyclones demonstrated that small differences in cyclone geometry sometimes have a large impact on the sharpness of a cyclone's penetration curve. This was further explored in a systematic investigation of the effects of geometrical modifications on cyclone penetration (Kenny and Gussman 2000). The results from this work indicated how to modify cyclone geometry in order to improve the sharpness of cyclone selectors and suggested that the SRI-II cyclone should have an extremely sharp penetration curve. This prediction was confirmed by the results of tests carried out on the Andersen AN3.68 cyclone (Peters et al. 2001b), which on investigation proved to have almost identical geometry to the SRI-II. Utilizing all available data on the SRI-II and near clones, a “working” family model for a new VSCC cyclone family was estimated. The working model was used to calculate the dimensions for a prototype VSCC cyclone expected to give a D₅₀ cutpoint close to 2.5 μm at a flow rate of 16.67 l min⁻¹. A schematic diagram of the cyclone showing the dimensions is given in Figure 1.

FIG. 1 Schematic diagram of the VSCC cyclone, showing the principal dimensions.

Our previous experience with pooling data from various sources led us to expect that there would need to be minor adjustments to the prototype design, in order to compensate for uncertainty in our initial estimates of the “working” family model parameters. Fortunately, the work on cyclone geometry suggested a simple way to optimize the prototype. We had observed a linear relationship between D₅₀ and vortex finder length (S in Figure 1), whereas vortex finder length did not significantly change the sharpness of the penetration curve. This finding allows the D₅₀ of the penetration curve of a cyclone to be “tuned” by a few tenths of a micrometer by adjusting the vortex finder length (keeping all other parameters constant). The prototype VSCC was made with two alternative tops, one with a short vortex finder and one with a long one. The optimum vortex finder length was deduced experimentally by linear interpolation of the D₅₀ data obtained with the different vortex finders (which bracketed the desired value). The final VSCC design was then constructed and calibrated at a flow rate of 16.67 l min⁻¹ in the normal way. Additional tests at several other flow rates were carried out in order to confirm the final VSCC family model. The experimental methods used throughout to test the cyclones were similar to those described in detail by Maynard and Kenny (1995). For each aerodynamic diameter range, the average particle number counted with the cyclone present was divided by the average number counted without the cyclone present to determine the aerosol penetration for that diameter. The penetration values were analyzed using the software package Tablecurve (Jandel Scientific) in order to locate the D₅₀, D₁₆, and D₈₄ diameters by interpolation. The sharpness values were calculated as

and least-squares fitting to the data was used to obtain the final VSCC family model:

where D₅₀ is the penetration cutpoint in micrometers, D_C is the cyclone body inside diameter in centimeters, Q is the flow rate in liters per minute, and a and b are empirical constants determined using nonlinear least squares regression. Best fit values for a and b were a = 1.415 and b = 1.908 for the VSCC family.

To compare the VSCC penetration curve with the WINS impactor penetration curve, data from seven independent repeated tests of the VSCC penetration curve are compared to data from seven independent tests on the clean WINS impactor (data taken from Kenny et al. 2000) in Figure 2. Means and standard deviations of penetration values are shown. In the tests carried out at HSL the WINS was found to have an average D₅₀ value of 2.48 μm and a sharpness value of 1.22. The results of the VSCC final design tests give a cutpoint of D₅₀ = 2.50 μm at a flow rate of 16.67 l min⁻¹, and a sharpness value of 1.16. Note that the mean HSL penetration curve for the WINS impactor is not quite as sharp as the ideal WINS curve published by Peters et al. (2001a). However, it should be noted that the ideal USEPA curve was of necessity derived by fitting the fairly small dataset from initial USEPA tests on the WINS, whereas the USEPA have since obtained many additional data (see, for example, Vanderpool et al. 2001a,b). When all the USEPA WINS data are pooled the resulting fitted curve is indistinguishable from the HSL WINS data in Kenny et al. (2000). The VSCC penetration curve does not deviate significantly from the WINS curve, and the VSCC sharpness value is as good as the WINS.

FIG. 2 Comparison of WINS and VSCC showing curves fitted to data.

Testing the VSCC under Loading

Dust-loading tests were conducted on the VSCC in line with the criteria for Class II equivalency designation set out by USEPA 40 CFR part 53. This describes the dust-loading protocol to be followed. The candidate device is challenged with a concentration of ISO fine test dust, which is equivalent to an ambient concentration of 150 μg m⁻³ over a 24 h period. After “critical” loading periods of 14 days, 30 days, and 90 days, the candidate device should show no significant signs of change in cutpoint or sharpness of cut values.

As it is difficult to generate consistent dust concentration at levels as low as 150 μg m⁻³ for periods of up to 90 days, higher concentrations of test dust were generated for shorter intervals to give equivalent 24 h exposures of 150 μg m⁻³. This procedural change was agreed with the USEPA prior to the tests being carried out. The dust-loading tests were carried out inside the same chamber used to perform the VSCC penetration tests. This was to ensure that the cyclone was moved as little as possible during the tests, so that coarse particles deposited inside the cyclone would not become dislodged as a result of movement. This is important because the oversize dust inside the VSCC collects on dry deposition surfaces and could possibly be disturbed after sampling.

The test dust used to load the VSCC was ISO 12103-1 fine (commonly referred to as Arizona Road Dust or ATD by the USEPA). The dust was generated inside the chamber using a Wright dust feed (WDF), which was serviced before use and was fitted with the tungsten-carbide-tipped blade. The dust emitted from the WDF entered the chamber at the top, where it was mixed and neutralized using an ionizer fan. It then passed through aluminum honeycomb into the working section. The WDF was set to the minimum speed at which it would operate consistently (0.05 rpm) and the dispersion air was set to 0.5 bar pressure. At this speed the WDF should run for around 60 h before requiring refilling. The air flow through the chamber was adjusted using a butterfly valve situated at the base of the chamber to give a concentration of approximately 10 mg m⁻³ inside the working region. Exposure at this concentration for approximately 20 min would simulate a 24 h test at 150 μg m⁻³. The air velocity through the chamber was less than 0.04 ms⁻¹.

The dust concentration inside the chamber was measured using two thin-walled samplers set up according to the Agarwal and Lui (1980) criteria. These were fitted with 25 mm GF/A glass fiber filters, and the dust-laden air was pulled through at 4 l min⁻¹ using Rotheroe and Mitchel sampling pumps. At the same time, the temporal variation in concentration was monitored using the Microdust 880 nm direct-reading dust monitor. The Microdust 880 is calibrated in the factory using ATD, and in these tests the calibration could be reverified using the gravimetric measurements. The calibrated Microdust 880 nm was used to calculate the exposure time required at any given concentration to give the equivalent 150 μg m³ 24 h exposure.

An important consideration in producing valid test results during loading is producing the correct size distribution. If a very fine aerosol is being produced, then the dust will pass through the VSCC and the tests will overestimate the cyclone's required cleaning interval. Similarly, if the aerosol is too coarse, a large fraction of the mass will be removed by the PM₁₀ inlet upstream of the size selector, and the cleaning interval may again be overestimated. For this reason the size distribution inside the chamber was measured using a Sierra 8-stage impactor.

Throughout the loading tests, the VSCC was fitted with the USEPA standard PM₁₀ lo-flow Dichotomous inlet and was operated at a calibrated volumetric flow rate of 16.67 l min⁻¹. The dust penetrating the cyclone was collected on a 47 mm GF/A filter mounted inside a FRM cassette at the outlet of the cyclone. The test protocol agreed with EPA was as follows:

1.	Verify the penetration performance curve of the clean cyclone using standard glass microspheres.
2.	Generate a concentration of approximately 10 mg m−3 ISO 12103-1 fine test dust into the chamber, measure the concentration using the thin wall samplers, and calibrate the Microdust 880 nm dust monitor. Determine the mass median diameter and standard geometric standard deviation of the test dust using the cascade impactor.
3.	During each loading test, monitor the dust concentration inside the chamber using the calibrated Microdust 880 nm to predict the required sampling period.
4.	Carry out penetration tests corresponding to loading intervals of 1, 2, 3, 4, 14, 30, and 90 days at an equivalent daily concentration of 150 μg m−3.

The intention of the test schedule was to verify the ability of the VSCC to perform within EPA criteria. The key tests were those verifying a two-week cleaning interval, a 30-day cleaning interval, and a 90-day interval. Penetration curves for the VSCC cyclone after the various loading intervals were analyzed, and the relationship of D₅₀ and sharpness to exposure interval (at an equivalent concentration of 150 μg m⁻³) is shown in Figure 3. The results indicate that the VSCC can operate at a dust concentration of 150 μg m⁻³ for at least 90 days without any significant change in the cyclone cutpoint. A small increase in the sharpness value (i.e., decrease in sharpness) was observed over the same period.

FIG. 3 Effect of dust loading on VSCC D50 and sharpness of cut.

In order to assess the impact of loading the VSCC cyclone on apparent PM_2.5 concentrations, numerical simulation was utilized to “sample” the three ambient aerosol distributions cited in the Federal Register. Numerical integration of the penetration curves with the aerosol size distributions was used to calculate the mass of particulate that would be collected, first with the VSCC cyclone and then with a fractionator following the “ideal” PM_2.5 curve (as specified in the FRM). The bias between VSCC and “ideal” FRM was calculated from the results. A detailed discussion of how these calculations are performed has been presented by Kenny et al. (2000). Calculations were made for each VSCC penetration curve corresponding to different loading times (1, 2, 3, 7, 15, 30, and 90 days). These calculations showed that biases between the VSCC and “ideal” samplers would be expected to be less than 1% for fine and typical aerosols, even with loadings up to 90 days. For the coarse aerosol, the VSCC sampling bias was between +2 and +3% for all loadings up to 90 days. Note that bias values in the range−5% to +5% are permissible for FRM or FEM samplers, and the VSCC easily passes this performance requirement even after 90 days of continuous operation at an equivalent concentration of 150 μg m⁻³. The small positive bias for coarse aerosol is due in part to coincidence-related measurement error in the tail of the penetration curve, which is very difficult to eliminate entirely for measurements made with our method (using the APS 3310). For comparison, similar tests and data treatment procedures on the WINS impactor data also produce a positive bias exceeding 1% for the clean WINS impactor (Kenny et al. 2000).

FIELD COMPARISON OF THE VSCC AND WINS-BASED PM_2.5 SAMPLERS

Test Requirements

The Federal Register requires field tests to be carried out at two sites for Class II candidate equivalent methods. Requirements regarding the test conditions at the sites are stringent and difficult to meet given the variable nature of pollution episodes. The aerosol size distribution must be such that all acceptable sample sets from one site must have a PM_2.5/PM₁₀ ratio greater than 0.75, while at the other site they must have a ratio less than 0.40. The concentration must be such that at the first site (ratio > 0.75), a minimum of 3 acceptable sample sets must have an average PM_2.5 concentration > 40 μg m⁻³ (for 24 h samples), and a minimum of three acceptable sample sets must have an average PM_2.5 concentration < 40 μg m⁻³ (for 24 h samples), as measured by the reference method. At the second site (ratio < 0.40), a minimum of 3 acceptable sample sets must have an average PM_2.5 concentration > 30 μg m⁻³ (for 24 h samples), and at least three acceptable sample sets must have an average PM_2.5 concentration < 30 μg m⁻³ (for 24 h samples). For each site, at least 10 acceptable sample sets must be obtained. To be considered acceptable, sample sets must be in the range of 10–200 μg m⁻³. Also, the precision of the reference method measurements must be less than specified maximum values. These requirements mean that sampling sites and seasons must be carefully chosen and that several extra test days are built in to ensure that sufficient acceptable datasets are obtained.

The USEPA requires a minimum of 3 reference method and 3 candidate method samplers to be set up and operated at each test site, with each set of measurements consisting of 3 reference and 3 candidate measurements, all obtained simultaneously. All measurements must be either 24 or 48 h integrated measurements. A minimum sample period of 23 h is required for both the reference and candidate methods. At least two PM₁₀ samplers must be operated simultaneously with the candidate and reference methods to determine the PM_2.5/PM₁₀ ratio. The equivalency requirements specify a correlation coefficient of greater than 0.97 between candidate FEM and the FRM, and a precision, defined as the standard deviation of three collocated candidate FEM instruments, of less than 2 μg m⁻³

Test Methodology

For the VSCC equivalency field trials, sampling locations in Hartford, CT and Phoenix, AZ were used. The Hartford trials were conducted first by USEPA contractor Research Triangle Institute (RTI) as part of a larger series of tests. The Phoenix, AZ “super site” was then chosen for the second set of trials after discussion of historical field test data with EPA and the Technical Review Board. The Phoenix Super Site has been in operation for nearly 15 years, and is typically impacted by fine ambient PM from urban activity as well as coarse wind-blown particulate from the surrounding desert. This site and the one at Rubidoux, CA are two historical sampling sites that have been repeatedly used to test equipment due to their difficult aerosol and climate conditions.

At each field site the following Lo-Vol PM_2.5 and PM₁₀ samplers were operated daily:

• Three BGI PQ200 PM_2.5 FRM with WINS EPA

• Three Impactor, three (at AZ: 2 at CT) BGI PQ200 Candidate PM_2.5 FEM with BGI VSCC

• Three Cyclone, and three BGI PQ200 PM₁₀ Low-Vol FRM Samplers.

These samplers were set up in three rows, collocated on a single sampling platform meeting the site criteria for spacing, inlet height of 2 m, and operating parameters as stated in the CFR. A tenth FRM sampler was utilized for special-purpose sampling and as a backup sampler if one of the other nine low-flow samplers had a problem. This special-purpose sampler was utilized for PM₁ sampling and the data taken for the purpose of comparing PM₁₀, PM_2.5, and PM₁ concentrations.

A 22/23 h sampling time (the minimum acceptable) was used to allow testing on sequential daily operation with a view to completion of the entire test cycle within ∼ 20 days. For days with low concentrations of PM_2.5 the 23 h sample time could be extended to 48 h in an attempt to meet the required minimum or higher concentration averages. The flow rate, ambient temperature, and barometric pressure of samplers was checked throughout using a DeltaCal calibrator. Gelman^® Teflo™ PTFE (Teflon™) membrane filters were utilized in all samplers in the Phoenix field test. The presample and postsample mass of the filters was measured on a Sartorius Model C5 microbalance. USEPA guidelines regarding FRM filter preparation and analysis were rigorously followed throughout.

On completion of the 23 h sample event, each sampler stopped sampling and displayed the summary screen of data, including date, time, sample volume, ambient temperatures, barometric pressure, and system pressures. All these data were recorded manually onto “Single Filter PM_2.5 Sampler Run Data Sheets” for each of the nine samplers and each of the sample events. Local meteorology notes were made, and observations were noted on any aspect that may have been of later use in interpretation of the data. Each sampler was visited by the operator within the 1 h window of time between sample events. The sample filter with mass was collected and placed back into each metal transport container, and a new clean filter was placed in the sampler ready for the new sample event. All pertinent data were recorded and the requirements for flow rate and volume were met. The operator stayed at the site until the new sample event began and verified that all the samplers started and were operating correctly. In addition to each sample event, the operator periodically would remove the PM₁₀ inlet and place a BGI DeltaCal flow calibrator on each sampler to verify the actual sample rate was correct at the onset of sampling.

Results of the Field Comparisons

A total of 33 data sets were obtained at the Hartford, CT site; however, the PM_2.5/PM₁₀ ratio was < 0.40 for only 3 of the test days, and those 3 data sets had an average PM_2.5 concentration below the lower acceptable limit of 10 μgm⁻³. Thus, the data sets did not meet the USEPA requirements for a site having PM_2.5/PM₁₀ ratios < 0.40. This result determined the choice of Phoenix, AZ as the second field site.

The PM_2.5/PM₁₀ ratio was > 0.75 for 11 of the 33 data sets from CT, and of these 11 data sets, the average PM_2.5 concentration (as determined by the reference method) was < 40 μg m⁻³ for 10 data sets. The precision of the reference method measurements for all 10 of these data sets was well within the acceptable bounds, thus meeting the required minimum of 3 acceptable data sets with PM_2.5 less than 40 μg m⁻³. That left only 1 of the data sets having average PM_2.5 concentration greater than 40 μg m⁻³, whereas the rules require a minimum of three data sets at high concentration. Therefore, the data did not strictly meet the requirements for a site having PM_2.5/PM₁₀ ratios greater than 0.75 with at least 3 of the acceptable data sets having PM_2.5 concentrations greater than 40 μg m⁻³. However, one of the data sets had a PM_2.5/PM₁₀ ratio of 0.73 and average PM_2.5 concentration of 38 μg m⁻³, both of which are within 95% of acceptable. The next-closest data set had a PM_2.5/PM₁₀ ratio of 0.73 (97% of 0.75) and an average PM_2.5 concentration of 31 μg m⁻³ (77% of 40). The precision of the reference method measurements for both of these data sets was also acceptable. At least 11 data sets were fully acceptable, thereby meeting the requirement for a minimum of 10 acceptable data sets for the site.

The regression parameters (slope, intercept, and correlation coefficient) were calculated for both the 11 sets of data that were fully acceptable and for the 13 sets of data that included the two additional data sets that almost met the ratio and concentration requirements that are needed for the minimum of 3 in the “over 40 μg m⁻³” category. All three performance parameters for both data sets were well within the specified limits (see Table 1).

TABLE 1 Summary of results from the field comparisons of WINS and VSCC samplers

	CT data: 11 data sets	CT data: 13 data sets	Phoenix data: 15 data sets	USEPA performance requirement
Slope	1.013	1.005	1.04	1 ± 0.05
Intercept	0.17	0.23	0.859	0 ± 1
R^2	0.9993	0.9991	0.970	> 0.97

A total of 15 sets of data were obtained from the Phoenix site, and the PM_2.5/PM₁₀ ratio was < 0.40 throughout. However, the PM_2.5 concentration only exceeded the minimum acceptable level of 10 μg m⁻³on three of these days, and on no day did it exceed 30 μg m⁻³. The precision of the reference method measurements for all data sets was within acceptable bounds on all 15 days. The low concentrations at Phoenix mean that the data do not strictly meet the USEPA requirements for a site having PM_2.5/PM₁₀ ratios less than 0.40.

The regression parameters (slope, intercept, and correlation coefficient) were calculated for all 15 sets of data. The bias between the VSCC- and WINS-based samplers at the Phoenix site, i.e., slope of regression line, was slightly higher than the value of 2–3% predicted from numerical integration, but still just within the upper limit of 5% allowed by USEPA. The intercept and correlation coefficient were also within the specified limits (see Table 1). A graph of all the field test results from both Hartford, CT and Phoenix, AZ is plotted in Figure 4.

FIG. 4 Field comparison of VSCC and WINS PM2.5.

DISCUSSION

The laboratory measurements of the penetration curve of the VSCC show that it is at least as sharp as the WINS impactor (when clean) and, if anything, somewhat sharper than the WINS in the sub-2.5 μm range. The VSCC has been demonstrated to retain its D₅₀ and sharpness under conditions of heavy loading, whereas the WINS is known to suffer loading effects after relatively short periods of use at high concentration (Kenny et al. 2000). The effect of loading the WINS is to decrease the D₅₀ cutpoint and also to decrease the sharpness, thus leading to negative bias in sampled PM_2.5 concentrations. These known characteristics of the two size selectors might lead one to expect a small positive bias between the VSCC and WINS PM_2.5 data from field comparisons at high concentrations.

In the Hartford, CT field comparisons the aerosol size distribution, as indicated by the PM_2.5/PM₁₀ ratio, was typical to fine, and there was no significant bias observed between the VSCC and WINS. Under the conditions pertaining to this test the WINS would not be expected to manifest any measurable loading bias. In the Phoenix, AZ field comparisons the aerosol size distribution was coarse throughout but the concentration was very low, leading to higher experimental errors as a result of the low filter masses collected. The results show a small positive bias between the VSCC and WINS that could not reasonably be attributed to WINS loading effects at the concentrations experienced. The bias is too small to be considered significant in terms of the required performance characteristics and given the small number of data sets does not significantly differ from zero.

The field comparisons reported here highlight the difficulty of performing the equivalency tests strictly as required by EPA. In both field locations, one would need to plan a very lengthy sampling campaign in order to meet the USEPA's stringent data requirements for results with concentrations above 40 μg m⁻³ and 30 μg m⁻³. This adds greatly to what is already a very resource-intensive and hence expensive test protocol. On the basis of the data we have presented, one can argue that the data sets at high concentrations are not necessary if those obtained at low concentrations meet the performance requirements. In particular, there is not a strong case for excluding datasets with concentrations below 10 μgm⁻³, which are supposed to be subject to greater experimental error, where those data clearly meet the requirements for precision. Naturally, different considerations would apply to field tests of continuous monitors, for which loading effects would cause biases after extended use at high concentrations.

CONCLUSIONS

A VSCC cyclone selector for PM_2.5 has been developed and tested in both laboratory and field situations. The VSCC selector is at least as sharp as the WINS impactor when clean, and its size-selection characteristics are remarkably stable when it is subjected to loading. This contrasts with the known characteristics of the WINS impactor, which is known to suffer from serious loading effects on prolonged use. The VSCC was submitted for designation as a Class II Federal Equivalent Method and subjected to field comparisons with the WINS FRM at two different sites. The field tests covered a range of conditions as regards aerosol size distribution; however, concentrations were generally lower than required. Nevertheless, the data from both sites met the USEPA performance requirements for equivalency. Following extensive negotiation, the USEPA was persuaded to waive the three minor failures of the field data to meet all Class II FEM test requirements and, accordingly, consider the VSCC fully acceptable based on the excellent regression parameters determined for the candidate method.

Acknowledgments

We thank the State of Connecticut Department of Environmental Protection for cooperation and use of their Windsor, CT sampling station, the State of Arizona Department of Environmental Quality for cooperation and use of their laboratory and Phoenix, AZ Super Site, the USEPA and Contractor Research Triangle Institute, and the EPA Technical Review Committee.