Evaluation of airborne particulate matter and metals data in personal, indoor and outdoor environments using ED-XRF and ICP-MS and co-located duplicate samples

Abstract

Factors and sources affecting measurement uncertainty associated with monitoring metals in airborne particulate matter (PM) were investigated as part of the Windsor, Ontario Exposure Assessment Study (WOEAS). The assessment was made using co-located duplicate samples and a comparison of two analytical approaches: ED-XRF and ICP-MS. Sampling variability was estimated using relative percent difference (RPD) of co-located duplicate samples. The comparison of ICP-MS and ED-XRF results yields very good correlations (R² ≥ 0.7) for elements present at concentrations that pass both ICP-MS and ED-XRF detection limits (e.g. Fe, Mn, Zn, Pb and Cu). PM concentration ranges (median, sample number) of 24-h indoor PM₁₀ and personal PM₁₀ filters, and outdoor PM_2.5 filters were determined to be 2.2–40.7 (11.0, n = 48) μg m⁻³, 8.0–48.3 (11.9, n = 48) μg m⁻³, and 17.1–42.3 (21.6, n = 18) μg m⁻³, respectively. The gravimetric analytical results reveal that the variations in PM mass measurements for same-day sampling are insignificant compared to temporal or spatial variations: 92%, 100% and 96% of indoor, outdoor and personal duplicate samples, respectively, pass the quality criteria (RPD ≤ 20%). Uncertainties associated with ED-XRF elemental measurements of S, Ca, Mn, Fe and Zn for 24-h filter samples are low: 78%–100% of the duplicate samples passed the quality criteria. In the case of 24-h filter samples using ICP-MS, more elements passed the quality criteria due to the lower detection limits. These were: Li, Na, K, Ca, Si, Al, V, Fe, Mn, Co, Cu, Mo, Ag, Zn, Pb, As, Mg, Sb, Sn, Sr, Th, Ti, Tl, and U. Low air concentrations of metals (near or below instrumental detection limits) and/or inadvertent introduction of metal contamination are the main causes for excluding elements based on the pass/fail criteria. Uncertainty associated with elemental measurements must be assessed on an element-by-element basis.

Introduction

Air pollutants, especially airborne particulate matter (PM) and metals in PM, have been associated with both short-term and long-term adverse health effects including chronic respiratory disease, heart disease, lung cancer, and damage to other organs (Costa and Dreher, 1997, Ghio et al., 1999, Allen et al., 2001, Vincent et al., 2001, Prieditis et al., 2002, Osonio-Vargas et al., 2003, Rasmussen, 2004, Lingard et al., 2005, Williams and Wheeler, 2007, Niu et al., 2008). Characterization of PM components, including inorganic elements, is of central importance in proposing mechanisms for health effects and in source apportionment studies (Butler et al., 2008, Butler et al., 2009). One of the most effective strategies to address the increasing public concern from criteria air pollutants and their impact on human health is through undertaking personal exposure monitoring. The Windsor, Ontario Exposure Assessment Study (WOEAS; Williams and Wheeler, 2007), launched in 2005, aims to provide accurate and representative human exposure data collected across the city for indoor and outdoor residential locations, and personal environments. To meet the study goals of WOEAS, development of accurate and reliable sampling approaches is a critical focus, since sampling variability is one of the most important contributors to the overall uncertainty of exposure measurements of PM and PM-bound transition metals. Efforts in this area are essential to obtain accurate and representative information on daily exposures of urban populations to air pollution.

At the outset of the Windsor sampling program, a pilot study of 24-h personal, indoor and outdoor levels of PM_2.5 and associated metals was conducted to develop standard operating procedures for field sampling and laboratory analysis (Rasmussen et al., 2007a). The present study is a continuation of this effort to improve monitoring quality, by using co-located duplicate samples to identify uncertainties associated with monitoring PM-bound metals. Recently Lippmann (2009) identified the need for such information, pointing out that problems in interpretation may arise from readings of elements in airborne PM which are near or below the lower detection limits. The aim of using co-located duplicate PM samples is to unravel all the sources of uncertainty in the multi-element data and to provide a framework for assessing such uncertainties as part of the larger monitoring study in Windsor (ON, Canada). Effective quality control criteria were developed to assess data derived from integrated 24-h personal and indoor PM₁₀ samples, and 2-week outdoor PM_2.5 samples.

There are many challenges in the WOEAS approach. One is the small particle mass collected on the filter samples due to the short sampling time (24 h) and the low-flow rates (4 LPM) (Niu et al., 2007a, Rasmussen et al., 2007b). A 24-h sampling time was employed in the present study as 24-h measurements are the standard in the majority of air pollution guidelines (US EPA, 2006, CESI, 2008). Light-weight battery-operated low-flow samplers are required for personal monitoring, and to minimize noise and disruption during indoor monitoring, while the participants are carrying out their normal daily routines (Rasmussen et al., 2006a, Rasmussen et al., 2007a). Although there are many advantages to recently developed real-time PM monitoring techniques (Butterfield and Quincey, 2007), gravimetric analysis techniques remain important for studies involving characterization of PM-bound metals. In addition, there is a continued requirement for accurate and reliable 24-h PM sampling as a comparison method to calibrate continuous monitoring techniques (Lanki et al., 2002; Lippmann, 2009).

The low sample mass poses a major obstacle in obtaining reliable elemental data, as many metals (particularly those present in trace concentrations in the airborne particles) are below the detection limits of the Energy Dispersive X-ray Fluorescence (ED-XRF) spectrometer, as observed in a preliminary residential air study in Ottawa (Rasmussen et al., 2006a). Problems associated with using the more sensitive and increasingly popular Inductively Coupled Plasma Mass Spectroscopy (ICP-MS) approach arise not from the instrumental detection limits but from issues related to sampling and analysis: sample contamination during collection, processing, transporting, weighing, extraction/digestion, and analysing procedures.

Challenges also arise from data correlation and equivalence (Butterfield and Quincey, 2007) between different science-based techniques which sometimes show bias or even unrelated results for PM monitoring (Lanki et al., 2002, Braniš and Hovorka, 2005). Both ED-XRF and ICP-MS are commonly employed but fundamentally different techniques for analysing PM-bound elements. Thus, data comparability between these methods is an important issue. There is very limited work on comparisons of ED-XRF and ICP-MS (Herner et al., 2006). A direct and detailed data correlation study to determine the degree to which the air quality data obtained using ICP-MS can be compared with ED-XRF in the present study will be helpful to address the information gap in this field.

The present study establishes the relative advantages and limitations of each instrumental approach, but goes further to demonstrate that the capacity of the field and lab personnel to collect and maintain uncontaminated samples during the whole measurement process (from sampling to analysis) is critical. Most epidemiological studies investigating health effects of airborne metals rely on occupational data (e.g. Wild et al., 2009), due to the current lack of reliable population-based airborne metals data. The few epidemiological studies of metals in urban air which have been published (e.g. Hibbs et al., 2002) would be more likely to be incorporated into risk assessments if the analytical and sampling uncertainties were better quantified and constrained. As future urban air studies are undertaken to address the airborne metals data gap, researchers will be increasingly obligated to monitor and report quality assurance data if the results are to be accepted as valid by regulators and risk managers. Uncertainty arising from sampling variability, the first stage of a measurement (Ramsey, 1997), is an issue of great concern due to its major contribution to the overall measurement uncertainty (Ramsey et al., 1995, Ramsey, 1997, Horowitz, 1997, Jorhem et al., 2006, Rasmussen et al., 2006a, Niu et al., 2007a). There is scarce information for identification of metal contamination and related sources. The present study focuses on this field to provide helpful information for evaluating date reliability to obtain accurate and reliable results for PM-bound elements analysis.