Review of the Department of Defense Enhanced Particulate Matter Surveillance Program Report (2010)

The Department of Defense (DOD) Enhanced Particulate Matter Surveillance Program (EPMSP) analyzed samples of particulate matter (PM) that were collected with the methods described in Chapter 2. This chapter will review the analytic methods used and the evaluation of the resulting data on chemical composition. The objective of the EPMSP was to identify the general chemical characteristics of PM at 15 sites in the Middle East with methods that were similar to those used for ambient-air monitoring networks in the United States, including the Environmental Protection Agency’s Chemical Speciation Network and the Interagency Monitoring of Protected Visual Environments network. The analytic approaches used in the EPMSP are standard methods that have been widely used for the characterization of ambient aerosol samples.

The Teflon filter samples were analyzed for mass with gravimetry and for elemental composition with energy dispersive x-ray fluorescence (EDXRF) and inductively coupled plasma-mass spectrometry (ICP-MS). Anions on quartz filters were determined by using ion chromatography (IC) and cations with inductively coupled plasma-optical emission spectroscopy (ICP-OES). Carbon on quartz filters was measured as organic carbon (OC) and elemental carbon (EC) with a thermo-optical method. X-ray diffraction was used to determine the mineralogy of the samples by determining the spacing of layers in crystalline materials. In addition, samples were collected for examination with scanning electron microscopy (SEM) to characterize individual particles and provide information on particle structure and composition. Those are standard analytic methods that are applied to samples in the United States. In this chapter, the committee examines the validity of those approaches as applied to samples collected in the DOD EPMSP and reviews the comparability of the resulting data with those collected in the networks in the United States.

Engelbrecht et al. (2008) selected the appropriate particle-collection media and analytic methods. The selected protocols have been extensively tested and

used by a wide array of organizations, including universities and research centers, government (for example, the Chemical Speciation Network), and industry. All methods used are state-of-the-art methods and have adequate sensitivity for the types of samples collected.

The mass on the filters was measured with gravimetry. Using standard methods (40 CFR 50 [2010]), the blank filters were equilibrated at 23ºC and 35% relative humidity for 24 hours and then weighed before they were sent to the field. After sample collection, the filters were sent back to the laboratory for a second equilibration and reweighing. Airborne-mass concentration was calculated as the difference between filter weights divided by the volume of air sampled.

Although there were limitations in the sampling methods, as described in Chapter 2, there are no questions about these laboratory measurements because the weighing can be highly precise and accurate. However, there is concern about the potential loss of PM from the filters during shipping. The measurements showed PM concentrations high enough to suggest sampler overload, as discussed in Chapter 2.

EDXRF was used to measure 40 chemical elements nominally: sodium, magnesium, aluminum, silicon, phosphorus, sulfur, chlorine, potassium, calcium, titanium, vanadium, chromium, manganese, iron, cobalt, nickel, copper, zinc, gallium, arsenic, selenium, bromine, rubidium, strontium, yttrium, zirconium, molybdenum, palladium, silver, cadmium, indium, tin, antimony, barium, gold, mercury, thallium, lead, lanthanum, and uranium. For samples of airborne particles, x-ray fluorescence (XRF) was operated by using the thin-sample approach in which the sample is assumed to be thin relative to the range of the excitation or emission x rays in matter. For low-atomic-number elements or heavy elements in which outer-shell x rays are used for analysis, there can be attenuation of the emitted x rays, and particle-size-correction factors are often applied to compensate for absorption of the x rays in the sample. In the analysis of the EPMSP samples, however, no particle-size corrections were applied, and low-atomic-number elements are probably substantially underestimated.

The mass concentrations observed in the EPMSP are much higher than those currently measured in the United States. However, they are comparable with values observed 30-40 years ago when studies of airborne-particle composition began to use XRF as an analytic tool (Jaklevic et al. 1981). The MiniVol sampler uses a 47-mm filter that permits a reasonable mass loading while providing a uniform, thin film sample that is suitable for analysis. However, when

the total mass loading exceeds around 1,000 μg, there are increased problems of self-absorption that require the application of appropriate corrections.

There is little description of the procedures used in the XRF analyses, particularly for PM₁₀ and total suspended particulate samples, which have higher mass loadings and larger individual particles. Thus, there could be additional errors in the estimation of the composition of the particle samples if the mass-loading and particle-size corrections were inappropriate. In addition, it is unlikely that all the elements in the XRF protocol can be measured adequately. For example, the primary uranium M x rays have energies of 3.171 and 3.337 eV. The potassium x-ray energy for its Kα line is 3.312 eV and for its Kμ line is 3.589 eV (Lide 1991). Given the resolution of the detectors in commercial XRF systems and the concentrations of uranium in samples, the uranium lines will be lost in statistical fluctuations in the potassium x-ray emissions. Thus, there needs to be a more careful review of the elements that can be reported using XRF analyses.

ICP-MS is a destructive method that requires dissolution of a sample, so an entire sample is used. Although problems with sample collection outlined in the previous chapter may be applicable to the XRF analyses, there would not be a problem in analyzing heavily loaded samples with methods in which samples are leached or solubilized, such as ICP-MS. In those cases, a combination of nitric acid and hydrochloric acid was used to solubilize selected elements, including antimony, arsenic, beryllium, cadmium, chromium, lead, manganese, nickel, vanadium, zinc, mercury, and strontium. If uranium was an element of interest, it would have been possible to measure it with this procedure. It would also have been easy to identify other elements that could be used to compare with the XRF results and to provide support for additional quality-assurance (QA) comparisons.

Aliquots of the water-soluble species that were obtained by leaching the quartz-fiber filters were analyzed with ICP-OES. In ICP-OES, the heat of the plasma causes the elements of interest to emit light of specific wavelengths that can be separated and used for quantititative analysis. There are unlikely to be substantial problems with this assay because the approach used was standard and straightforward.

Another aliquot of quartz-fiber leachate was analyzed with IC for the water-soluble anions sulfate (SO₄^2-), nitrate (NO₃^-), chloride (Cl^-), and phosphate

(PO₄^3-) and the ammonium (NH₄⁺) cation. There are unlikely to be substantial problems with this assay because the approach used was standard and straightforward.

In the OC-EC analysis, the National Institute for Occupational Safety and Health method was used (NIOSH 2003). A small punch of quartz filter was heated in helium gas to evolve the organic compounds associated with the PM. The evolved compounds were oxidized to carbon dioxide, the carbon dioxide was converted to methane, and the methane was quantitatively measured with a flame ionization detector. After the organic compounds were evolved at the highest temperature, the gas was changed to a mixture of a few percent oxygen in helium, and the refractory carbon, EC, was oxidized. The resulting carbon dioxide was again converted to methane and measured.

There are several protocols for the sequence of temperatures and times of exposure at each temperature. As the organic compounds are evolved, the filter darkens because of pyrolysis, and corrections are made for the carbon that was pyrolyzed and not evolved. These protocols produce somewhat different results (Chow et al. 2004), particularly in the amount of EC.

A major problem in OC measurements was noted by Engelbrecht et al. (2008). They were high, and the authors hypothesized that organic material evolved from the plastic containers in which the quartz filters were stored because of the very high temperatures at which they were exposed. Thus, no OC data were reported. However, the increased organic material would also contribute to additional pyrolytic carbon, which is difficult to separate from EC. Thus, there is considerable uncertainty as to whether the EC data in the EPMSP study were adequately characterized.

Nuclepore filters were used for computer-controlled scanning electron microscopy (CCSEM) to characterize individual particles. Nuclepore filters were also used for SEM analysis to analyze individual particles for shape, surface coatings, and chemical composition. CCSEM is a combination of backscattered electron imagery and energy-dispersive spectroscopy that automatically analyzes a large number (1,000-1,500) of individual particles for size and chemical composition. The particles are grouped in “bins” by chemical composition and particle size. The x-ray data are qualitative but can provide a basis for classifying particles into classes of similar composition (Kim et al. 1987; Xie et al. 1994). In the EPMSP, the classes were determined by expert judgment rather than through the application of specific data-analysis methods. These methods

are adequate for the purposes of the EPMSP, although more information could be gleaned from their results if more thorough data analysis was applied.

There was an interest in identifying particle types that would probably induce silicosis, with an emphasis on crystalline silica. High-resolution SEM was applied to the samples to provide the needed information. It was likely that a sufficient sample size was analyzed. However, no statistical analysis was provided to suggest that the observed subsample adequately reflected the larger population of samples.

The EPMSP report does not adequately address QA. A strong QA assessment includes details of the results of replicate analyses and other measures (for example, calibration procedures) to assess analytic precision, accuracy, and sensitivity. Furthermore, a strong QA analysis reports information on field or laboratory control samples (blanks) that could affect the reported values. Such data do not necessarily need to be included in the body of the report, but the details need at least to be included in an appendix. In addition, the lack of uncertainty estimates associated with the numbers reported prevents the reader from determining the degree of confidence in the data. For example, many of the reported elements (the committee estimated up to 40%)—especially uranium, gallium, rubidium, strontium, yttrium, zirconium, molybdenum, palladium, silver, cadmium, indium, tin, antimony, gold, and mercury—are reported at concentrations that may be below the concentrations at which they can be measured accurately. The reporting of those elements may lead to misinterpretation of the composition of the samples.

A commonly used QA approach for PM composition data is to perform a mass-closure analysis (Malm et al. 1994). In this approach, the measured chemical species are summed, and the concentration is compared with the gravimetric mass concentration as a benchmark. Some elements are not measured but can be estimated. For example, silicon would typically be found in airborne particles as SiO₂, and the mass of SiO₂ can be estimated from the silicon concentration and the appropriate gravimetric factor (molecular weight of SiO₂ divided by the molecular weight of silicon). Analogous factors are applied to aluminum, calcium, titanium, and iron. The organic mass is estimated from the measured OC by multiplying by a factor that estimates the amount of hydrogen, nitrogen, and oxygen associated with the carbonaceous material. The factors range from 1.4 (Malm et al. 1994) to possibly 2.1 (Lim and Turpin 2002). In the major U.S. monitoring networks, there is generally good mass closure, and this analysis serves as an important QA check of the data.

In the design of the EPMSP, however, only one type of filter was collected at each site during each sampling period. Therefore, Engelbrecht et al. (2008) could not conduct an analysis for all species at any site in the network, making it impossible to attempt a mass-closure assessment. Furthermore, the study design does not permit an understanding of the complete composition of the PM and its variance in time and space on the basis of the data presented.

As discussed in Chapter 2, the particle-sampling schedule influences the estimation of mean particle compositions measured in a network of sites and the chemical composition of the particulate mass. Because concentrations and compositions are not uniformly distributed in time, a given sampling frequency may not provide a truly representative set of samples for characterizing the PM in an area. Thus, care needs to be exercised in interpreting the results as accurate estimates of potential exposure.

The EPMSP attempted to compare data gathered from 15 Middle Eastern sites with data gathered in the United States. The committee surmised that the comparison was probably performed to demonstrate that there are no substantial differences in PM composition between the Middle East and the United States despite higher mass concentrations in the Middle East. As indicated above and in Chapter 2, the sampling protocol limits the ability to obtain a complete measurement of the PM composition of the samples. However, the data allow an initial assessment of the concentrations and a qualitative assessment of the major components found in PM in the region.

As expected, the major portion of PM in all size fractions was of geologic origin. The investigators were able to show linkages of PM composition to the regional composition of the soils in the area. The soils would have a different composition from soils in the United States and may have different types and amounts of flora and microorganisms although the species were not directly measured as part of this effort. The fine PM fraction of the Middle Eastern samples may have been affected by regional transport of PM from other areas, depending on wind patterns. The PM also included carbon, both OC (not measured but presumed) and EC. The OC was presumed because it could not be measured directly. However, without detailed understanding of OC composition, it is difficult to define its origin accurately.

The region has a large particle contribution from combustion sources, including both stationary and mobile sources. For example, high concentrations of lead were observed relative to that in the United States, with the lead being attributed primarily to poorly controlled smelter operations and the use of leaded gasoline. Anecdotal evidence suggests that a major portion of combustion emissions from the military bases is due to poorly run vehicles with minimal emis-

sion controls. Another potential contribution is from open burn pits. Without more detailed composition data, it is difficult to identify the air toxicants that originate in the pits. It will be important to consider burn pits for future work because there is ample evidence that uncontrolled burning produces air toxicants.

The MiniVol sampler is not a Federal Reference Method sampler, nor has it been designated as a Federal Equivalent Method. Thus, the results of these measurements are not fully comparable with the results of samples collected in the regulatory monitoring network for determining attainment of the PM_2.5 or PM₁₀ National Ambient Air Quality Standards. There is a reasonable correspondence between the performance of the MiniVol and Federal Reference Method samples under United States air-quality conditions (Baldauf et al. 2001), but evidence based on sampling performed in Kuwait (Brown et al. 2008) suggests that the MiniVol may overestimate concentrations. (See discussion in Chapter 2.)

The data obtained in the EPMSP are not useful for source identification and apportionment, because the design does not permit complete characterization of the particle mass for any given set of samples. It is always possible to provide some apportionment on the basis of the composition data that are available, but the analysis would be incomplete.

The critical issue is that there needs to be more clearly defined objectives of the sampling and analysis scheme at the outset of the program. If source apportionment is a desired outcome, it is important that likely source types be identified and, from that information, that chemical constituents be determined so that an appropriate chemical-analysis scheme can be used. Planning for all the desired outcomes is essential in the design of an ambient-aerosol sampling and analysis program.

• The surveillance efforts undertaken by the EPMSP are commendable and, with some changes in approach, the data could be better suited for defining regional composition and could contribute to the design of health studies to address the potential role of air quality in health. The study design, in which dif-

ferent types of sampling media are not collected simultaneously, did not permit a quantitative understanding of PM composition. In addition, as indicated below, some analytic concerns may need to be better addressed to adequately characterize exposure for health studies. However, the ambitious survey of composition in the region provided a qualitative understanding of the composition of major components, and the observations will help to define questions and analytic issues for future surveillance efforts.

• Some of the samples that have already been collected may be analyzed further to improve understanding of additional chemical components in the region. There is an opportunity to identify selected particle-bound toxic species (for example, endotoxins and other biologic materials, polycyclic aromatic hydrocarbons, dioxins, and dibenzofurans) by analyzing composites of the quartz-filter samples. The additional measurements would lead to improved understanding of the chemicals in the region that may be appreciably toxic and could form a basis for improving the prioritization of objectives for future surveillance efforts.

• XRF, which was used to measure a major fraction of the reported data, may have some technical barriers that were not accounted for in the EPMSP report. The technical issues may affect the quality of the reported data. They include the need to correct for the size of particles and the potential for heavily loaded samples to interfere with the analysis of some elements. Only elements that are accurately determined by the method should be reported.

• The QA of the exposure-surveillance effort needs to be better defined, and the analytic certainty of the data should be reported to reflect the confidence in the measurements. A number of elements, particularly uranium, are reported at concentrations that are known to be poorly measured by XRF. Future efforts should analyze uranium with ICP-MS, and the reported data should include the analytic uncertainty.

• The suite of chemicals investigated in routine U.S. monitoring studies might not be appropriate for understanding exposures in the military environment of the Middle East. There may be a number of sources that contribute to PM in the ambient environment that were not considered (for example, burn pits and demolition debris).

• To the extent that it is possible to compare compositions measured by the EPMSP with those observed in the U.S., there appear to be no significant differences from U.S. regions that are substantially affected by geologic material. There were higher proportions of lead than are currently seen in the United States, presumably from smelters and the use of leaded gasoline. Further comparisons would require more detailed assessment of the carbonaceous component, including the volatile fraction generated from open-pit burning.

• As discussed in Chapter 2, the committee recommends the development

of a well-defined set of study objectives. It is important to define project objectives a priori so the exposure monitoring is designed to meet project objectives.

• If understanding PM chemical composition is a project objective, surveillance should be designed with an array of sampling media that are collected simultaneously. That will ensure that the sampling and analytic methods permit mass-closure testing as part of the overall QA process. Future efforts may omit total suspended particulates if the goal is to understand potential health effects from exposures.

• The comparisons of PM composition in the EPMSP report should be interpreted with caution because they may not fully represent the true underlying distribution of PM concentrations. Therefore, the variability in the PM composition may contribute substantial error and make it difficult to conduct accurate and complete apportionment studies. It is important not to overinterpret concentrations of potentially toxic elements. Uranium, for example, was found in measurable amounts, but the lack of confidence in the measurements renders the data unusable for assessing the risk of exposure to that material in theater.

Baldauf, R.W., D.D. Lane, G.A. Marotz, and R.W. Wiener. 2001. Performance evaluation of the portable MiniVOL particulate matter sampler. Atmos. Environ. 35(35):6087-6091.

Brown, K.W., W. Bouhamra, D.P. Lamoureux, J.S. Evans, and P. Koutrakis. 2008. Characterization of particulate matter for three sites in Kuwait. J. Air Waste Manag. Assoc. 58(8):994-1003.

Chow, J.C., J.G. Watson, L.W. Chen, W.P. Arnott, H. Moosmüuller, and K. Fung. 2004. Equivalence of elemental carbon by thermal/optical reflectance and transmittance with different temperature protocols. Environ. Sci. Technol. 38(16):4414-4422.

Engelbrecht, J.P., E.V. McDonald, J.A. Gillies, and A.W. Gertler. 2008. Department of Defense Enhanced Particulate Matter Surveillance Program (EPMSP). Final report. Desert Research Institute, Reno, NV. February 2008 [online]. Available: http://chppm-www.apgea.army.mil/foia/DOCS/Final%20EPMSP%20Report%20without%20appx%20Feb08.pdf [accessed Feb. 1, 2010].

Jaklevic, J.M., R.C. Gatti, F.S. Goulding, B.W. Loo, and A.C. Thompson. 1981. Aerosol Analysis for the Regional Air Pollution Study. Final report. EPA-600/S4-81-006. Environmental Science Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency, Research triangle Park, NC.

Kim , D.S., P.K. Hopke, D.L. Massart, L. Kaufman, G.S. Casuccio. 1987. Multivariate analysis of CCSEM auto emission data. Sci. Total Environ. 59:141-155.

Kim, D.S., and P.K. Hopke. 1988. The classification of individual particles based on computer-controlled scanning electron microscopy data. Aerosol Sci. Technol. 9(2):133-151.

Lide, D.R., ed. 1991. Handbook of Chemistry and Physics, 72^nd Ed. Boca Raton: CRC Press.

Lim, H.J., and B.J. Turpin. 2002. Origins of primary and secondary organic aerosol in Atlanta: Results of time-resolved measurements during the Atlanta supersite experiment. Environ. Sci. Technol. 36(21): 4489-4496.

Malm, W.C., J.F. Sisler, D. Huffman, R.A. Eldred, and T.A. Cahill. 1994. Spatial and seasonal trends in particle concentration and optical extinction in the United Sates. J. Geophys. Res. 99(D1):1347–1370.

NIOSH (National Institute for Occupational Safety and Health). 2003. Elemental Carbon (Diesel Particulate). Method 5040. Third Supplement to NIOSH Manual of Analytical Methods (NMAM), 4th Ed., P.C. Schlecht, and P.F. O’Connor, eds. NIOSH 2003-154. U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Institute of Occupational Safety and Health, Cincinnati, OH [online]. Available: http://wwwtest.cdc.gov/niosh/docs/2003-154/pdfs/5040f3.pdf [accessed Mar. 12, 2010].

Xie, Y., D. Wienke, and P.K. Hopke. 1994. Airborne particle classification with a combination of chemical composition and shape index utilizing an adaptive resonance artificial neural network. Environ. Sci. Technol. 28(11):1921-1928.