Environ. Sci. Technol. 2006, 40, 1172-1177
Residual Indoor Contamination from World Trade Center Rubble Fires as Indicated by Polycyclic Aromatic Hydrocarbon Profiles J O A C H I M D . P L E I L , †,‡ WILLIAM E. FUNK,‡ AND STEPHEN M. RAPPAPORT‡ Methods Development and Applications Branch, HEASD/ NERL/ORD, United States Environmental Protection Agency, Research Triangle Park, North Carolina 27711, and Department of Environmental Sciences and Engineering, School of Public Health, University of North Carolina, Chapel Hill, North Carolina 27599
The catastrophic destruction of the World Trade Center (WTC) on Sept. 11, 2001 (9/11) created an immense dust cloud followed by fires that emitted smoke and soot into the air of New York City (NYC) well into December. Outdoor pollutant levels in lower Manhattan returned to urban background levels after about 200 days as the fires were put out and the debris cleanup was completed. However, particulate matter (PM) from the original collapse and fires also penetrated into commercial and residential buildings. This has created public concern because WTC dust is thought to cause adverse pulmonary symptoms including “WTC cough” and reduced lung capacity. Additionally, some recent studies have suggested a possible link between exposure to WTC contamination and other adverse health effects. Distinguishing between normal urban pollutant infiltration and residual WTC dust remaining in interior spaces is difficult; efforts are underway to develop such discriminator methods. Some progress has been made in identifying WTC dust by the content of fibers believed to be associated with the initial building collapse. There are also contaminants created by the fires that burned for 100 days in the debris piles of the building rubble. Using WTC ambient air samples, we have developed indicators for fire related PM based on the relative amounts of specific particle bound polycyclic aromatic hydrocarbons (PAHs) and the mass fraction of PAHs per mass of PM. These two parameters are combined, and we show a graphical method for discriminating between fire sources and urban particulate sources as applied to samples of settled dusts. We found that our PAHs based discriminator method can distinguish fire source contributions to WTC related particulate matter and dusts. Other major building fires or large open burn events could have similar PAHs characteristics. We found that random samples collected ∼3.5 years after the WTC event from occupied indoor spaces (primarily residential) in the New York area are not
* Corresponding author phone: (919) 966-7319; fax: (919) 9664711; e-mail:
[email protected]. † United States Environmental Protection Agency. ‡ University of North Carolina. 1172
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 4, 2006
statistically distinguishable from contemporary city background.
Introduction The collapse and subsequent prolonged fires of the World Trade Center (WTC) disaster released an enormous amount of dust, smoke, and fine particulate matter into the New York City geographic area. This has resulted in a variety of acute and chronic pulmonary health effects; the most pervasive is “WTC cough”, observed mostly in rescue workers (1, 2), but a host of other respiratory ailments with immediate and delayed onset have also been reported (2-11). WTC contamination has been associated with adverse reproductive effects (7, 12, 13), asthma, and other pediatric morbidities (5, 14). Psychological stress could also be an important factor contributing to adverse reproductive outcomes (13) and may have led to the observed increased frequency of ventricular arrythmias both locally in NYC (15) and remotely in a Florida study (16). The disaster also impacted the mental health status of the population (17, 18), and transgenerational effects of post-traumatic stress disorder have been observed (19). It has been shown that numerous carcinogens and mutagens were present in the dust and smoke that may impact the appearance of long-term latency diseases (20-23). Although the ambient contamination levels dissipated over a few months due to dilution by wind and rain, WTC dusts and smoke penetrated and settled in the local residential and commercial buildings and remain an ongoing concern of the public (24-27). The EPA has established an expert panel including Government and University experts that meet in public every 2 months to address these issues specifically (http://www.epa.gov/wtc/panel/index.html). That panel met first in March of 2004 and 10 times since then. A primary topic of discussion has been the design of a new EPA study to sample the indoor environment in buildings near the WTC site to determine the extent of residual WTC contamination. This study entails the collection of dust samples to ascertain whether such dust can be associated with the collapse of the WTC Towers and ensuing fires. The bulk of the initial dust is from the mechanical pulverization of the two WTC towers and the WTC 7 building just north of the main complex across Vesey St. The major constituents have been characterized in settled dusts and in airborne dusts (20, 23, 28-32). Candidate dust fingerprints are under study based on trace metals and fiber content (33, 34). Offenberg et al. have suggested that patterns of organic compounds are identical in indoor and outdoor settled WTC dust but very different in dust samples collected in a rehabilitated home (26). Tang et al. have characterized New York City background levels of WTC contaminants (including PAHs and dioxins) using wipe samples from upper Manhattan apartments; they found no evidence of PAHs contamination and only moderate levels of dioxins (35). In contrast to this background work, Lioy et al. found large amounts of PAHs and other products of combustion in the WTC dusts, suggesting that the organic compounds could indeed serve as a WTC fires/smoke marker (20). Dioxin-like compounds are another major contaminant class that results from fires. Lorber has identified a profile of dioxin congeners associated with the WTC fires that has been observed in both ambient air and settled dust near the collapse site (36) but, due to a similarity to profiles from a building fire in Philadelphia, suggests that this may be more a general profile rather than one specific to WTC fires. Rayne et al. have analyzed wipe 10.1021/es0517015 CCC: $33.50
2006 American Chemical Society Published on Web 01/14/2006
samples of exterior window films from lower Manhattan and have suggested unique patterns of 2,3,7,8-substituted dioxin congeners as compared to background sites (37). Despite the aforementioned work, the residual contributions of the initial combustion of jet fuel and the subsequent lingering fires in the WTC rubble piles represent a second source of particulate matter for which an unambiguous identifying signature is not yet established. In prior work, we have developed methods for retrospective assay of particlebound polycyclic aromatic hydrocarbons (PAHs) from archived Teflon fiber filters used to collect fine fraction airborne particulate matter (PM2.5) defined as having up to 2.5 µm aerodynamic diameter (38). The specific PAHs analytes are representative of the general class of all PAHs and were a subset of the standard EPA-16 PAHs list (38). Subsequently, we analyzed a series of samples collected in the vicinity of WTC ground zero collected for 200 days after the disaster (22). We found a shift in PAHs character with respect to time after September 11, 2001 that has been attributed empirically to a reduction of WTC fires and relative stability of diesel activity. The ensuing discussion proposes to use certain observed relationships among individual particle bound PAHs to establish a metric for determining the likelihood that residual indoor particulate matter is related to the WTC fires. Additionally, we investigate the overall mass contribution of PAHs to particulate matter (or settled dusts). Because all samples are mixtures of many particle sources, we expect appreciable variability in both parameters. However, based on a two-dimensional discrimination method using outdoor airborne data, we demonstrate a general pattern and then further apply the technique to contemporary dust samples collected in late 2004. We further compare to dust samples collected locally in North Carolina, unknown samples from NY and NJ Port Authority buildings, and from Governor’s Island.
Materials and Methods PAHs AnalysissPM25 Filters. The detailed analytical methods have been described elsewhere (38). Briefly, Teflon filters are extracted with 10 mL of dichloromethane (DCM) containing deuterated PAHs as internal standards. After concentration of the extracts to 50 µL under N2, 2 µL aliquots are analyzed by gas chromatography-mass spectrometry (GCMS) in selective ion monitoring mode. On the basis of external calibration, quantitative results are reported for nine carcinogenic and/or mutagenic PAHs, namely, benz(a)anthracene (BaA), chrysene (Chr, benzo(b)fluoranthene (BbF), benzo(k)fluoranthene (BkF), benzo(e)pyrene (BeP), benzo(a)pyrene (BaP), indeno(1,2,3-c,d)pyrene (Ind), dibenz(a,h)anthracene (DaA), and benzo(g,h,i)perylene (BgP). These compounds are defined as representative for the larger class of particle bound PAHs and (except BeP) are included in the standard EPA-16 PAHs list (39). PAHs AnalysissWhole and Vacuumed Dust Samples. Since the work with Teflon filters was completed (22), we have modified the procedures slightly to accommodate whole dust samples (those scooped directly into containers from thick layers) and dust samples collected over larger areas with vacuum cleaners. Settled dusts or vacuumed dusts were size fractionated either via resuspension or sieving (23). Sample aliquots were subsequently weighed (typically 1580 mg) and placed directly into 4 mL of DCM extraction solvent with internal standards; this range of dust mass and the associated PAHs content were well within the linear range of the analytical methodology (22, 38). After settling (or centrifugation, if necessary), the supernatant was transferred to new vials and blown down under nitrogen to about 500 µL. A quartz-wool packed injection liner was used to avoid injecting excessive particles onto the GC column. The
resulting sample concentrates were subsequently analyzed identically to the Teflon filter sample extracts using GC-MS. Airborne Samples Set. To date, a collaborative effort between the U.S. EPA and the University of North Carolina has generated 243 measurements of nine individual particle bound PAHs. These were extracted from PM2.5 filter samples collected at sites at or near ground zero of the WTC disaster site from September 23, 2001 to March 30, 2002; the results of this study have been published (22). These samples represent all sources of PAHs; the earlier ones are more impacted by the WTC fires, whereas the later ones reflect New York City background. This empirical evidence is the basis for the discrimination methodology developed in this work. Settled Dusts Sample Sets. We received sets of contemporary dust samples from various New York City (NYC) locations during late 2004 and early 2005 collected to support the EPA expert panel’s efforts. Because 24 h particulate samplers were deemed intrusive and impractical for residential or office settings, EPA personnel chose bulk dust collection by sweeping/scooping or with a vacuum cleaner. Indoor sampling sites included the Deutsche Bank Building (DBB) complex in Manhattan; apartments and homes in the greater NYC area; Columbia Medical Center; Governor’s Island; and two Port Authority Buildings (PABs). The DBB was severely damaged by the collapse of the WTC South Tower that opened a 15 story gash in the 42 story building. We received two groups of samples from areas with apparent residual WTC contamination; nine from the main tower at 130 Liberty St. (DB1) and six from a portion of the complex behind the main tower at 4 Albany St. (DB2). These samples were collected in August and September 2004 by a sampling crew from the U.S. EPA’s Region 2 office and their prime contractor (Lockheed Martin, Edison NJ, EPA Contract 68-C99-223) and are considered to represent WTC dusts that have remained essentially undisturbed for about 3 years. From September 2004 to April 2005, we received a series of 17 dust samples from apartments, houses, and the Columbia Medical Center representing samples unlikely impacted by WTC fires from locations scattered about Manhattan, Long Island, Brooklyn, and the Bronx and four samples from apartments within one or two blocks of the WTC site. We also received eight samples from the PAB in Port Newark, NJ, nine from the PAB on Columbia St. in Brooklyn, and eight from various indoor locations of the Governor’s Island Coast Guard facilities just south of the tip of Manhattan that may or may not have been contaminated with WTC dusts. All samples were collected either in bulk or with a vacuum cleaner and subsequently sieved to achieve less than 150 µm diameter particle size. For illustration and comparison to WTC and NYC samples, we collected and processed 14 similar samples from local residences and eight from University and city parking garages in Chapel Hill, NC. All sieved dust samples were >100 mg; each was aliquoted into two independent samples ranging from 15 to 80 mg and analyzed to assess combined measurement precision and uniformity of PAHs in dust. Other PAHs Data. In addition to samples analyzed by our laboratory, we also analyzed data gleaned from other researchers and publications. Specifically, Offenberg et al. measured PAHs in fine settled dusts collected November 19, 2001 inside buildings close to the WTC site and impacted by the rubble fires (26). Swartz et al. analyzed airborne PM2.5 samples collected at the EPA sampling site at 290 Broadway, Manhattan, NY from September 26 to October 21, 2001 that were identified as impacted by the WTC plume by concurrent nephelometer measurements (40). Butt et al. collected semivolatile films coating external windows with solvent wiping from October 27-29, 2001 in lower Manhattan and Brooklyn (41). VOL. 40, NO. 4, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1173
Mathematical Parameters. Ambient data from the previous work demonstrated that the absolute and relative amounts of individual PAHs, as well as overall levels of total PAHs, changed with time elapsed after 9/11. This was attributed to the changes in the dominant sources and gradual diminution of fires and reduction of diesel engine operations associated with the cleanup; a mathematical model for this behavior has been developed (22). For the present work, we exploit the relationships among individual PAHs as compared to the sum of all measured PAHs and their time dependent profiles. We also calculate the total PAHs contribution to ambient particulate mass as ng of PAHs/mg of PM2.5 to determine any changes in source based PAHs enrichment.
Results and Discussion General Observations of PAHs Data. The most striking observation of the results is the high variability among samples despite spatial or temporal proximity to the WTC event. However, there are also definite trends that support the conjecture that samples can be characterized by their source contributions. We infer that all samples, regardless of size fraction, are likely mixtures derived from all potential sources and that the general character is reflective of the source ratios within the mixture. That is, even an ambient particle sample collected at ground zero during the highest levels of rubble fire will contain city background and resuspended WTC collapse dusts, and a nominal background city sample will contain contributions from various smaller fire sources, long-range transport, or traces of re-entrained remnant WTC dusts. This is expected for settled dusts as well; undisturbed contemporary dust collected within the DBB contains various levels of the collapse source, the fire source, and 3.5 years of background city contributions, whereas all background dusts have the possibility of contribution from random indoor fire sources such as smoking, cooking, etc. or infiltration from other outdoor sources. The variability is a function of the mixture: the greater the fire contribution, the greater the shift away from the nominal background levels of the two indicator parameters. Ambient Samples and Relative WTC PAH Composition. Nine individual PAHs were assayed for each of the 243 ambient EPA PM2.5 samples. As expected, the relative amounts among individual compounds, as well as the total amount of PAHs relative to particulate mass, varied between samples presumably depending on the origin of the particles. This variation has an overall temporal trend with respect to time after 9/11 and has been ascribed to primary sources of PAHs changing from WTC fires to diesel equipment and finally to city background (22). To investigate this behavior, we constructed a contribution parameter for each analyte PAH with respect to the remaining measured PAHs. Given concentrations Ci,j (ng/m3) for the ith PAH in the jth sample, we calculate n
Ni,j ) Ci,j/[(
∑C
k,j)
- (Ci,j)]
(1)
k)1
for each sample where the sum over k represents all n measured PAHs (in this case, n ) 9). On the basis of the behaviors of Ni,j (t), where t ) number of days after 9/11, we can gauge the differences (if any) among individual PAHs contributions to the sum. We found that the time (after 9/11) dependent behavior of the Ni,j parameters either rises or falls to a plateau depending on the compound. On the basis of scatter plots, we observed that the total PAHs are relatively enriched by DbA, BbF, BaA, and BkF soon after 9/11, that the total becomes more enriched by Ind, BgP, and BaP as time progresses, and that BeP and Chr contributions remain fairly 1174
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 4, 2006
FIGURE 1. Scatterplot of ln(S) vs MF for group A (squares representing WTC fires source), group C (circles representing NYC background) samples, and samples from Swartz et al. taken when the WTC plume impacted the EPA Broadway site (triangles representing WTC fires source). Upward and right directions on graph represent more distinct fire signatures. Dotted lines are estimated dividers to guide the eye. constant. BbF and BgP demonstrate the extremes of this time dependent behavior. Source Parameter. As described previously (22), the primary source of airborne PAHs measured soon after 9/11 was the WTC rubble fire, whereas later, the dominant sources are diesel activity from the WTC site and general city traffic background. The closer to the active fires from WTC (either temporally or spatially), BbF becomes more dominant relative to total PAHs with respect to BgP. To exploit these trends mathematically, we define a source parameter S as
S)
∑(9 PAHs - BbF) BgP/∑(9 PAHs - BgP)
BbF/
(2)
where the units for the compounds are in ng/m3 and the denominators refer to the total measured PAHs (nine analytes) minus either BbF or BgP. Although other mathematical constructs could be used (such as a simple BbF/ BgP ratios), there is a small statistical advantage to referencing each compound to their coanalytes as in eq 2. The behavior of S versus time after 9/11 exhibits clear decay with a large amount of variability at a given time due to meteorology and cleanup activities. (See scatterplot 1 and text in Supporting Information.) Ambient WTC Samples and PAHs Particle Enrichment. A second dimensionless parameter is the PAHs mass fraction (MF) defined as the sum of the PAH analytes (in ng) divided by the mass of PM2.5 (in mg). This was also found to have an overall decreasing trend in time after Septemeber 11, 2001 and thus is similarly attributed to the change in dominant particle sources. Specifically, during the initial 30 days, the median value was 288 ng of PAHs/mg of PM2.5 (n ) 27); from days 31-150, it was 103 ng of PAHs/mg of PM2.5 (n ) 152); and from day 151-200, it was 76.6 ng of PAHs/mg of PM2.5 (n ) 64). (See scatterplot 2 in Supporting Information.) Combining S and MF Parameters. The parameters S and MF can be independently associated with the likelihood that a sample reflects WTC fires or NYC background. Each has a decay profile parametrized by elapsed time, and so we combine them into a log-log graph where the S value is plotted as the independent variable versus the MF value as the dependent variable for each sample. Figure 1 shows a scatterplot of the group A samples (representing WTC fires sources, initial 30 days) and the group C samples (represent-
FIGURE 2. Scatterplot of ln(S) vs MF for DB1, DB2, and Offenberg samples (representing WTC dust) and residential samples and Columbia Medical Center (CMC) samples (representing NYC dust). All aliquots and multiple samples are represented with 95% standard error bars. Upward and right directions on graph represent more distinct fire signatures. Dotted lines are estimated dividers to guide the eye. ing NYC background, days 151-200). We expect A points to fall into the upper right region and C points to fall into the lower left region of the graph. To confirm this approach, we show (as triangles) the only other published WTC airborne PAHs data from four measurements made on days September 26 and October 4, 12, and 20, 2001 when the WTC fires plume was observed to directly impact the PAHs sampler at the EPA 290 Broadway site (40, 42). On the basis of these results, there is a definite twodimensional PAHs pattern difference between airborne particle matter (PM2.5 fraction) with a WTC fire source and the NYC background. Presumably, this discrimination technique is valid for indoor dusts for discerning the impact from WTC fire particulate matter penetration. Indoor Particulate Matter Samples. Indoor whole dust samples were sieved to a 150 µm size fraction to avoid objects such as hair, carpet fibers, food particles, etc. that are not likely related to an external source. As such, these samples are not directly comparable to the airborne PM2.5 particle samples. Also, the original PM2.5 samples were collected at the time they were created by the local sources, whereas the contemporary dust samples (representing the putative WTC residuals) have aged for about 3.5 years. The analytical PAHs results for WTC dust samples from DB1, DB2, and presumed background samples from various residences are presented in Figure 2. We also include results from Offenberg et al. (26) for a set of indoor samples collected during the period of smoldering WTC fires. This graph is analogous to Figure 1 where we distinguish PAHs mass fractions and source parameter ln(S) on separate axes and the dividing lines are adjusted to visually distinguish between the WTC and the background samples. The WTC dusts from DBBs can be mostly confined to the upper right quadrant if we draw the vertical line (1) at ln(S) approximately -0.05. The data from Offenberg et al. (26) from samples collected November 2001 fall more obviously into this quadrant presumably due to smaller contributions from city background. We note that samples from the 22nd floor and some from the 35th floor of DB1 have been removed from this data set because these areas had been cleaned 6 months prior to collection (discussed later). The results also show that there are some discordant data; in this case, samples from one particular residence consistently appear in the upper right quadrant, one DB2 aliquot falls into the upper left quadrant, and some residential samples overlap with DB samples along the x-axis.
From these results, we speculate that there is some heterogeneity within dusts from the same area, that mixing with city background or dusts from the collapse of the WTC can reduce ln(S) without changing MF, and that mixing can also introduce a large inorganic fraction to the total dust mass that reduces MF without affecting ln(S). To illustrate these results, we have plotted all groups of analyses as means and x and y standard error bars (95%) in Figure 2. If we were to require WTC contamination to exhibit both traits simultaneously (upper right quadrant, using vertical line 1), then we estimate 89% sensitivity for identifying samples with WTC fires contamination and 98% specificity for identifying samples without WTC dust residuals. Under a more conservative approach, we can use the inverse definition and define non-WTC dusts to reside only in the lower left quadrant. In this case, we find 100% sensitivity for identifying samples with WTC fire contamination and 61% specificity for identifying samples without WTC dust residuals. These are approximations; other choices of the dividing lines defining the quadrants can also be employed to accommodate specific requirements. We could, for example, move the vertical dividing line to the right to ln(S) ) 0.7 (vertical line 2) and then only the Offenberg et al. (26) samples from November 2001 would be unambiguously identified as WTC fire related. Regardless of the choice of dividing lines, these samples can be considered to serve as a training data set for comparison to unknown samples. For most locations shown in the graphs, we analyzed multiple aliquots and/or sites; these are represented as a single mean with 95% standard error bars in both axes. We note that the PAHs results from window film wipes demonstrate similar internal fires signature estimated as ln(S) ∼0.96 for the two locations mentioned (41). These could not be plotted on the graphs because total particle mass values are not available. Contemporary Indoor Samples from Occupied NY Spaces. The contemporary samples in Figure 2 (denoted as stars) include both presumed background locations and four groups of samples collected within one or two blocks of the WTC site; visual inspection of data and graphs demonstrates no segregation based on proximity. Furthermore, upon grouping samples into four distance categories (
