Aerosol characterization studies at Great Smoky Mountains National Park, summer 2006

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114,, doi: /2008jd011274, 2009 Aerosol characterization studies at Great Smoky Mountains National Park, summer 2006 Douglas Lowenthal, 1 Barbara Zielinska, 1 Brooks Mason, 1 Shar Samy, 1 Vera Samburova, 1 Don Collins, 2 Chance Spencer, 2 Nathan Taylor, 2 Jonathan Allen, 3 and Naresh Kumar 4 Received 9 October 2008; revised 12 January 2009; accepted 18 February 2009; published 30 April [1] A study was conducted during summer 2006 at Great Smoky Mountains (GRSM) National Park (NP), TN, to address issues related to estimating aerosol light extinction in the Interagency Monitoring of Protected Visual Environments (IMPROVE) network. The revised IMPROVE equation calculates PM 2.5 light scattering (Bsp) from ammonium sulfate, ammonium nitrate, organic carbon mass, and fine soil concentrations; dry scattering efficiencies; and factors that account for hygroscopic growth. Organics are assumed to be nonhygroscopic. The organic compound mass (OCM)/organic carbon (OC) ratio is assumed to be 1.8. Experiments involving in situ and laboratory measurements were conducted to address issues related to (1) concentration-varying scattering efficiencies; (2) aerosol hydration state; (3) the OCM/OC ratio; and (4) the organic hygroscopicity. Filter-based measurements indicated that sulfate was acidic, with an average NH 4 + /SO 4 = molar ratio of Ambient State Hygroscopic Tandem Differential Mobility Analyzer measurements of the ambient hydration state rarely indicated deliquescence. The frequency of hysteresis ranged from 29% to 46% for 0.05 and 0.2 mm particles, respectively. There was a clear relationship between dry particle mean diameter and volume at GRSM, supporting the assumption that an increase in particle size during transport increases both the scattering efficiency and concentration. Water-soluble organic carbon (WSOC) was isolated from water extracts of high-volume filter samples using XAD solid-phase absorbents. The average ratios of OCM measured gravimetrically to OC measured by thermal optical reflectance in residues of isolated WSOC and dichloromethane (DCM) extracts were 2.4 ± 0.3 and 1.9 ± 0.2, respectively. Hygroscopic growth factors (GF) of aerosols generated from WSOC extracts averaged 1.10 ± 0.02, 1.13 ± 0.03, and 1.19 ± 0.04 at 80%, 85%, and 90% RH, respectively. These results indicate that, at GRSM during summer, at least some of the organic aerosol was hygroscopic. Citation: Lowenthal, D., B. Zielinska, B. Mason, S. Samy, V. Samburova, D. Collins, C. Spencer, N. Taylor, J. Allen, and N. Kumar (2009), Aerosol characterization studies at Great Smoky Mountains National Park, summer 2006, J. Geophys. Res., 114,, doi: /2008jd Introduction [2] Light extinction by atmospheric aerosols is an important component of the earth s radiation balance although direct and indirect aerosol effects remain uncertain in climate models [Charlson et al., 1992; Intergovernmental Panel on Climate Change (IPCC), 2007; Haywood and 1 Division of Atmospheric Sciences, Desert Research Institute, Reno, Nevada, USA. 2 Department of Atmospheric Sciences, Texas A&M University, College Station, Texas, USA. 3 Department of Environmental Health Exposure, Epidemiology and Risk Program, Harvard School of Public Health, Boston, Massachusetts, USA. 4 Environment, Electric Power Research Institute, Palo Alto, California, USA. Copyright 2009 by the American Geophysical Union /09/2008JD Schulz, 2007]. Measurements of particle size distributions and size-resolved chemistry have been used to estimate aerosol light extinction (scattering and absorption) at urban, rural, and marine locations [Hayasaka et al., 1992; Zhang et al., 1994; Lowenthal et al., 1995; McInnes et al., 1998; Malm et al., 2003, 2005]. Particle scattering (Bsp) is generally the dominant component of aerosol light extinction. Bsp, in units of Mm 1, is the product of the particle concentration and its scattering cross section: Z Bsp ¼ s nd ð Þ dd where s is the scattering cross section, D is particle diameter, and n(d)dd is the number concentration over the diameter interval dd. s depends on the incident light ð1þ 1of11

2 wavelength (l), D, and the complex index of refraction (m). Equation (1) may be reformulated as follows: Z Bsp ¼ ½ð3=2D rþqšcd ð ÞdD where r is the particle density, Q is the optical scattering efficiency, equal to s/pd 2 /4, and C(D) dd is the mass concentration over the size interval dd. The expression [(3/ 2Dr) Q] defines a mass scattering efficiency. If aerosol chemical components are assumed to be externally mixed, Bsp takes the functional form Bsp ¼ S j E j C j ; where E j is the mass scattering efficiency of the jth chemical component and C j is its mass concentration [White, 1986]. The Interagency Monitoring of Protected Visual Environments (IMPROVE) light extinction reconstruction equation has been used to estimate aerosol light extinction from PM 2.5 chemical concentrations measured in the Interagency Monitoring of Protected Visual Environments (IMPROVE) network [Malm et al., 1994]. Assuming that chemical components are externally mixed, the original IMPROVE equation for PM 2.5 light scattering is Bsp ¼ 3f ðrhþ½amsul þ AMNITŠþ41:4OC ½ Šþ1½SoilŠ ð4þ Dry mass scattering efficiencies of 3, 3, 4, and 1 m 2 /g are applied to ammonium sulfate (AMSUL), ammonium nitrate (AMNIT), organic compound mass (OCM), and soil mass (SOIL) concentrations (mg/m 3 ), respectively. The simplifying assumption that sulfate and nitrate are present as ammonium sulfate and ammonium nitrate is made because ammonium is not routinely measured in the IMPROVE network. While Malm et al. [1996] held that sulfate was generally completely neutralized in the western U.S., Day et al. [1997] demonstrated that sulfate was acidic at Great Smoky Mountains (GRSM) and Shenandoah national parks during summer. PM 2.5 nitrate was primarily associated with sea salt particles during the Big Bend Regional Aerosol and Visibility Observational Study [Lee et al., 2004]. Lee et al. [2008] concluded that only 2% of PM 2.5 nitrate was associated with ammonium at GRSM during summer, A factor of 1.4 is used to convert organic carbon (OC) to organic carbon mass (OCM). The f(rh) are factors which account for increases in scattering caused by hygroscopic growth of particles as a function of relative humidity (RH) and were based on a modified hygroscopic growth curve for ammonium sulfate [Tang and Munkelwitz, 1994]. [3] Recent advances have led to revisions of the IMPROVE equation [Pitchford et al., 2007]. Studies by Malm et al. [2003], Lowenthal and Kumar [2004], Ryan et al. [2005], McMeeking et al. [2005], and Malm and Hand [2007] suggested that sulfate, nitrate and OCM scattering efficiencies increase with increasing concentration at U.S. national parks. The revised IMPROVE equation accounts for concentration-varying dry scattering efficiencies. The rationale for this is that gas-to-particle conversion, particularly in clouds, increases both particle ð2þ ð3þ size, which increases the optical scattering efficiency, and the bulk aerosol concentration. [4] The OCM/OC ratio of 1.4 used in the original IMPROVE equation was derived for urban aerosols [White and Roberts, 1977]. Our current understanding is that OCM in aged aerosols such as those transported to remote locales is characterized by an OCM/OC ratio that could be higher than 2 [Turpin and Lim, 2001; El-Zanan et al., 2005; Malm et al., 2005; Yu et al., 2005; Chen and Yu, 2007; Malm and Hand, 2007]. A value of 1.8 was adopted for the revised IMPROVE equation. The f(rh) in the revised equation are based on the hygroscopic growth curve for ammonium sulfate estimated with the AIM thermodynamic model (Aerosol Inorganics Model, Clegg et al. [1998]). Growth is assumed to follow the upper branch of the hysteresis loop where ammonium sulfate particles retain water in a metastable state below the deliquescence RH (80%) to the crystallization RH (37%). This assumption could overestimate the amount of water associated with AMSUL if it was on the lower leg, i.e., crystalline, at RH < 80%. On the other hand, acidic sulfates such as ammonium bisulfate and sulfuric acid retain water at RH < 37%. [5] There are several aspects of the revised IMPROVE equation that merit further investigation: (1) The assumption that hygroscopic growth is always on the upper branch of the hysteresis loop of AMSUL may overestimate the amount of water and thus light scattering by inorganic aerosols; (2) the assumption that organics are not hygroscopic may not be accurate. Studies by Dick et al. [2000] and Dinar et al. [2006, 2007] demonstrated that ambient water-soluble organic carbon (WSOC) are both hygroscopic and can act as cloud condensation nuclei; (3) the assumption that the OCM/OC ratio is 1.8 derives from limited experimental evidence; and (4) concentration-varying chemical scattering efficiencies depend on a direct relationship between dry particle size and concentration. [6] A multiyear study is under way to address these issues at several U.S. national parks, including Great Smoky Mountains National Park, TN (GRSM) during summer and winter. Studies are planned for Grand Canyon National Park, AZ (GRCA), Puget Sound, WA (PUSO), and Acadia, ME (ACAD) during summer. GRSM represents a high sulfate concentration, high RH environment. ACAD is another northeastern U.S., sulfate-dominated site but with different sources than those impacting GRSM. GRCA is a dry environment in the southwestern U.S. with similar sulfate and organic aerosol concentrations during summer. PUSO represents a humid environment with relatively high organic content (OC represents nearly a third of reconstructed mass). This paper presents results from a study conducted at GRSM during the summer of 2006 which involved extensive field and laboratory measurements. 2. Methods 2.1. Field Measurements [7] Measurements were conducted in GRSM at the IM- PROVE monitoring site at Look Rock, TN (longitude , latitude , elevation 815 m ASL) during July and August The campaign had three main elements: (1) collection of PM 2.5 (particles with aerodynamic diameters <2.5 mm) filter samples for chemical and 2of11

3 Figure 1. Illustration of a hysteresis loop: hygroscopic growth for ammonium sulfate. The solid line is the deliquescence branch. The dashed line is the efflorescence branch. The dotted line is the continuous growth curve for ammonium bisulfate. physical characterization in the laboratory; (2) continuous measurement of the ambient hydration state; and (3) continuous measurement of the dry particle size distribution and hygroscopic growth as a function of RH PM 2.5 Aerosol Sampling [8] Daily (24-hour duration) aerosol samples were collected from 19 July to 17 August Four Tisch Environmental, Inc. high-volume (hivol) samplers equipped with size-selective PM 2.5 inlets collected particles on Zefluor filters (#P5PJ001, Pall Corporation) from 19 July 2006 to 31 July 2006 and on Teflon Impregnated Glass Fiber Filter (#7215, TIGF) filters (Pall Corporation) from 1 August 2006 to 17 August 2006 at a flow rate of 40 CFM. The filters were precleaned by sonicating them in methanol and dichloromethane (DCM) followed by drying. The hivol samples were used to collect sufficient mass of organic material for separation of WSOC from inorganic ions for direct determination of hygroscopic growth and the OCM/ OC ratio in the laboratory. A single-channel medium volume (medvol) sampler preceded by a PM 2.5 size-selective inlet (Bendix/Sensidyne Model 240 cyclone, Tisch Environmental) collected particles on 47 mm quartz-fiber filters (2500 Pallflex QAT-UP, Pall Corporation) at a flow rate of 80 SCFH. The quartz filters were prefired at 900 C for four hours. These were analyzed for soluble inorganic ions, elemental carbon (EC), and OC. Three Teflon, TIGF and quartz filters were taken to and from the field and used for a dynamic blank correction. All filters were stored under refrigeration before and after sampling. [9] Quartz-fiber filters can adsorb volatile organic compounds (VOC) during sampling, resulting in a positive artifact [Turpin et al., 1994]. This is not the case for Teflon filters. No attempt was made to account for this bias, for example, by using a VOC denuder or backup quartz filter [Subramanian et al., 2004]. [10] Particulate ammonium nitrate (AMNIT) exists in equilibrium with gaseous ammonia and nitric acid (HNO 3 ) in the atmosphere [Stelson and Seinfeld, 1982]. AMNIT may be lost through evaporation from filters during sampling. This negative artifact may be accounted for by removing HNO 3 from the sample stream with a denuder and collecting the volatilized nitrate on an absorptive backup filter. These measures were not taken in this study. However, during summer under hot and dry conditions, most nitrate is expected to exist in the gas phase [Chow et al., 2005]. As noted above, most of the particle nitrate at GRSM during summer is expected to be associated with mineral or sea-salt aerosols [Lee et al., 2008] Ambient Hydration State [11] Figure 1 illustrates hygroscopic growth for pure ammonium sulfate [Tang and Munkelwitz, 1994]. Growth along the hysteresis loop is described by the ratio of the particle diameter at a given RH (D RH ) to the dry diameter (D Dry ). On the lower leg (deliquescence branch), the particle remains crystalline from dryness until the RH increases to the deliquescence RH (DRH = 80%), at which point the particle absorbs water. As the RH increases above the DRH, the particle continues to absorb water and grow in equilibrium with the atmosphere. On the upper leg (efflorescence branch), the particle evaporates in equilibrium with the atmosphere as the RH decreases to the DRH. As the RH decreases further, the particle remains hydrated in a metastable (supersaturated) state until it crystallizes at the crystallization RH (CRH = 37%). Figure 1 also shows the hygroscopic growth curve for ammonium bisulfate [Tang and Munkelwitz, 1994]. In this case, hygroscopic growth is in equilibrium with atmosphere under all RH conditions, where the variation of D RH /D Dry with RH is continuous as a function of increasing or decreasing RH. [12] The ambient hydration state was inferred from measurements with an Ambient State Hygroscopic Tandem Differential Mobility Analyzer (AS-HTDMA) [Santarpia et al., 2004]. Aerosols were classified at ambient RH with an ambient diameter of D a and at a final diameter (D f ) following three RH-conditioning paths: (1) Path 0: dried to <20%, typically below the CRH of most aerosols; (2) Path 1: dried to <20%, then increased to ambient RH; and (3) Path 2: dried to <20%, humidified to 85 95% RH, above the DRH of most salts, and returned to ambient RH, which averaged 74 ± 13% on an hourly basis. [13] Table 1 illustrates the conceptual model by which the ambient hydration state is inferred from AS-HTDMA measurements. Nonhygroscopic particles display D f /D a =1on Paths 0 2. Particles initially on the lower leg (deliquescence branch) of a hysteresis loop display D f /D a =1on Paths 0 and 1 and > 1 on Path 2. Particles initially on the upper leg (efflorescence branch) of a hysteresis loop will display D f /D a < 1 on Paths 0 and 1 and = 1 on Path 2. Hygroscopic aerosols whose growth is continuous with increasing and decreasing humidity will display D f /D a <1 on Path 0 and equal to 1 on Paths 1 and 2. Each of the four ambient states is defined by a unique combination of three D f /D a ratios for Paths 0, 1, and 2. The conceptual model underlying the AS-HTDMA is strictly valid for pure compounds. Chemically complex aerosols do not behave like pure compounds. For example, Day et al. [2000] did not observe crystallization of ambient aerosols at GRSM, even at RH as low as 10%. AS-HTDMA measurements generally reveal some hygroscopic growth for particles on the lower leg of a hysteresis loop. Examples of real-world growth 3of11

4 Table 1. Determination of Ambient Hydration State From AS- HTDMA Measurements Path 0 a Path 1 b Path 2 c D f /D a 1 NH d NH NH HL-LL e HL-LL EQ EQ f HL-UL <1 HL-UL g HL-UL EQ >1 HL-LL a Path 0: ambient RH! dry (<20%). b Path 1: ambient RH! dry! ambient RH. c Path 2: ambient RH! dry! high RH (85% 90%)! ambient RH. d NH: Nonhygroscopic. e HL-LL: Hygroscopic, hysteresis loop, lower leg. f EQ: Hygroscopic, smooth equilibrium growth. g HL-UL: Hygroscopic, hysteresis loop, upper leg. curves under conditions of increasing and decreasing RH are shown in the work of Santarpia et al. [2005, Figure 4]. [14] Continuous measurement of the hydration state over the diurnal cycle was used to assess the validity of the assumption that hygroscopic growth of sulfate, assumed to be present as ammonium sulfate in a 24-hour average IMPROVE sample, can be accurately described by the efflorescence branch of the ammonium sulfate growth curve Dry Particle Size Distribution [15] The dry particle size distribution was measured between 12 and 750 nm in 150 size bins (83 per decade) with a scanning differential mobility analyzer (DMA). To maximize count rate when sampling in clean environments, a high-flow DMA (HF-DMA) [Stolzenburg et al., 1998] was used which operates at flow rates about 7 times that of the commonly used TSI 3081 long-column DMA for classification of a given particle size when applying equivalent voltages. Further reduction in sampling time was achieved using software that maximized flow rates for each particle size analyzed. To balance the need to count enough particles to produce a statistically robust distribution with the desire to minimize the time spent analyzing the properties of particles of one size, rapid voltage scans were repeated for a given particle size until a prescribed number of particles were counted. The scanning inversion is described by Collins et al. [2002]. The ambient aerosols were dried by passing the sample air stream through a custom Nafion tube bundle Laboratory Measurements Soluble Ion and Carbon Analysis [16] Daily medvol quartz-fiber filters were analyzed for water-soluble sulfate, nitrate, and chloride by ion chromatography (Dionex DX 3000) and for ammonium ion by automated colorimetry (Astoria 301A analyzer). Ion analysis of water extracts of the hivol filters was done similarly. Elemental and organic carbon (EC and OC) on the medvol quartz filter were analyzed with the DRI thermal-optical reflectance (TOR) carbon analyzer [Chow et al., 1993] Isolation of WSOC [17] Extraction of nonpolar organic material and WSOC was done using methods similar to those described by Decesari et al. [2001], Varga et al. [2001], Alves et al. [2002], Duarte and Duarte [2005], and Rinehart et al. [2006]. Daily hivol samples were combined and cut into one-inch squares, transferred to a clean 200 ml beaker, covered with 150 ml of ultrapure water, and sonicated without heating for 10 min. The extract was transferred to a flask, and the filter material was extracted again in 100 ml of ultrapure water with sonication for 5 min. Extraction in water was done at room temperature to better simulate hydration in the atmosphere. Thus the definition of WSOC is operational. A harsher water extraction procedure, e.g., boiling under pressure, could have extracted more organic material. However, we don t believe this would have been realistic in terms of atmospheric processes. The water extracts were concentrated on a rotavap to about ml volume. After the water extraction, the filters were dried overnight in a vacuum oven at room temperature. The following day, the filters were extracted on an accelerated solvent extractor (ASE) with dichloromethane (DCM). Unlike the water extraction, which was done to simulate ambient conditions, the ASE extraction was done under high pressure and temperature. The DCM extract was concentrated to about 20 ml. The DCM extraction procedure is probably not 100% efficient. Some fraction of ambient organic material may not be soluble in any organic solvent or combination of solvents. [18] Water-soluble organic carbon (WSOC) was measured by total dissolved organic carbon (TOC) analysis with an OI Analytical Aurora 1030W (wet chemical oxidation) TOC Analyzer. A subset of samples was also analyzed for WSOC using a modification of the (TOR) method [Yang et al., 2003]. [19] The water extracts were evaporated to near dryness with a rotavap under gentle vacuum to 15 to 30 ml. WSOC was isolated from inorganic ions using the procedure of Duarte and Duarte [2005]. Two nonionic macroporous resins, XAD-8 and XAD-4, were applied in tandem. These resins have been widely used to isolate organic material from natural waters [e.g., Santos et al., 2001]. Dinar et al. [2006] used XAD-8 resin to isolate humic-acid-like substances (HULIS) from ambient aerosols. Duarte and Duarte [2005] demonstrated that XAD-8 collects higher molecular weight, less hydrophilic material while XAD-4 collects lower molecular weight, more hydrophilic organics. The aqueous extracts were acidified to ph = 2.2 with HCl before being applied onto the XAD-8 column. Effluent from the XAD-8 column was collected and applied to the XAD-4 column. After these two concentration stages, both columns were washed with one column volume of ultrapure water to remove inorganic species. The organic matter retained on both resins was then back-eluted with a mixture of methanol/water (40% methanol). The XAD-8 column was eluted again with 50% methanol in water to improve recovery of the retained organic matter. WSOC recovery and inorganic ion removal efficiency was determined after each pass through the XAD columns. The methanol/water eluates containing isolated WSOC were combined and evaporated to near dryness to remove methanol and then redissolved in ultrapure water for further analysis OCM/OC Ratio [20] Quartz filter punches were prefired at 900 C for four hours and preweighed on a Mettler MT5 microbalance to 1 mg. An aliquot of the isolated WSOC was spotted and dried on the punch. The punch was reweighed and the mass 4of11

5 growth curves from dryness to 95% RH and from 95% RH to dryness. The particles were dried and introduced into the HTDMA where the first DMA selected particles with a dry diameter of 0.07 mm. The approach used to sequentially characterize the upper and lower legs of any hygroscopic growth hysteresis loops is described by Santarpia et al. [2005]. Figure 2. Time series of chemical composition of PM 2.5 aerosol at GRSM during the summer of of the WSOC residue was determined by difference. The OC content on the filter punch was then determined by TOR. The DCM extract was treated similarly. The uncertainties of mass (OCM) and OC concentrations were determined from replicate analyses and averaged 3 and 5%, respectively. OCM/OC ratios were calculated separately for the water and DCM extracts. This approach is similar to that used by El-Zanan et al. [2005] Hygroscopic Growth of WSOC Aerosols [21] Aerosols derived from WSOC water extracts were subjected to HTDMA analysis in the laboratory for measurement of hygroscopic growth factors (GF = D RH /D Dry )as a function of RH. High-flow DMAs, as described above, were used in the HTDMA. The inter-dma RH was computer-controlled by varying the pressure of an initially water-saturated air stream that served as the purge flow in the custom Nafion tube bundle. The HTDMA was completely enclosed in a temperature-controlled acrylic-walled frame to minimize variability of RH caused by temperature gradients or fluctuations. The laboratory HTDMA was calibrated by injecting dry polydisperse ammonium sulfate (AMSUL) aerosol generated with a TSI 3076 constantoutput atomizer and comparing size-selected particles with Duke Scientific polystyrene latex (PSL) standards. Calibration of the RH sensors was achieved through comparison of measured GFs at multiple RHs > CRH AMSUL with those determined using empirical relationships for solution density and water activity reported by Tang and Munkelwitz [1994]. Calibration of the zero GF assignment was achieved through comparison of measured GFs at RH < CRH AMSUL with 1.0. On the basis of the results of repeated calibrations on multiple days, the uncertainty of measured growth factors is estimated to be ±0.02. [22] The TSI model 3076 constant-output atomizer was used to generate an aerosol from the WSOC extracts. In order to obtain sufficient particle concentrations for precise growth curves, the composite samples contained WSOC concentrations ranging from 130 to 500 mg/l depending on the sample WSOC concentration [Gysel et al., 2004]. Approximately 30 ml of extract was required to measure 3. Results and Discussion 3.1. Chemical Composition [23] A daily time series of PM 2.5 chemical composition is shown in Figure 2. Organic mass (OCM) is plotted as OC * 1.8. Average sulfate, ammonium, OC, EC, and nitrate concentrations were 9.0, 1.86, 2.9, 0.5, and 0.04 mg/m 3, respectively. Average uncertainties determined from replicate analyses for 24-hour sulfate, ammonium, OC, EC, and nitrate concentrations were 5, 5, 10, 17, and 20%, respectively. Concentrations measured during this study were typical for GRSM during summer. Average sulfate, ammonium, OC, EC, and nitrate concentrations measured with the IMPROVE sampler at Look Rock during July and August from 2000 to 2005 were 7.7, 1.7, 2.0, 0.41, and 0.14 mg/m 3, respectively. Average nitrate concentrations measured with the DRI and collocated IMPROVE samplers on 10 corresponding days during the study period were 0.05 and 0.12 mg/m 3 ; both are too low to be visible in Figure 2. The IMPROVE sampler collects particulate nitrate on a nylon filter preceded by an HNO 3 denuder. Higher nitrate concentrations from the IMPROVE sampler could be due to volatilization of ammonium nitrate from the DRI sampler. However, since most of the ambient nitrate was probably in the gas phase, and because most of the PM 2.5 nitrate was probably associated with mineral or sea-salt aerosols [Lee et al., 2008], higher IMPROVE nitrate may be due to lessthan-100% HNO 3 denuder efficiency and collection of breakthrough HNO 3 by the nylon filter. [24] Average OC concentrations measured with the DRI and collocated IMPROVE samplers on 10 corresponding days were 3.0 and 1.9 mg/m 3. IMPROVE front quartz filter OC concentrations are corrected for the positive VOC absorption artifact by subtracting the monthly median OC concentration from backup quartz filters at six IMPROVE sites around the country [Eldred, 2002]. The correction is not site- or sample-specific but Eldred [2002] suggested that sample and backup filters become saturated with absorbed VOCs in the automated IMPROVE sampler and during storage and transport. The DRI OC concentrations were blank-corrected using an average from three dynamic blanks. The higher DRI OC concentrations could reflect absorption of VOC by the quartz filters except for the fact that DRI EC concentrations, for which there is no absorption artifact, were also higher than IMPROVE EC concentrations (0.51 and 0.34 mg/m 3, respectively). The average EC/TC (TC = OC + EC) ratios for both samplers were 0.16 for the 10 corresponding samples. It is thus possible that the IMPROVE OC artifact correction was not accurate and that both DRI and IMPROVE OC measurements were highbiased to some degree by the VOC absorption artifact. To the extent that our measured OC may have been high-biased by a VOC sampling artifact, the estimated WSOC/OC ratio could be considered as a lower limit. 5of11

6 Figure 3. Molar ratio of ammonium (with ammonium associated with nitrate subtracted) to sulfate at GRSM, summer [25] Figure 3 shows daily-average molar ratios of ammonium to sulfate, When sulfate is completely neutralized as ammonium sulfate, the ratio is two. When sulfate is partially neutralized as ammonium bisulfate, the ratio is one. Sulfates were generally acidic, with an average molar ratio of 1.17 ± Sulfates were nominally more acidic in August than in July, when average molar ratios were 1.10 ± 0.21 and 1.26 ± 0.18, respectively. On average, the sum of ammonium and sulfate (10.8 mg/m 3 ) was roughly double that of OCM (OC * 1.8) (5.2 mg/m 3 ). These relative abundances were similar in July and August. [26] Figure 4 presents a time series of 24-hour average OC concentration and WSOC expressed as a percent of OC. The average WSOC percentage of OC was 22%. Percentages were higher in August (30%) than July (11%). The average WSOC mass concentration (WSOC * 1.8) was only 11% of the average sum of the ammonium plus sulfate concentrations. WSOC measured by TOC and TOR analysis gave equivalent results: TOR (mg/sample) = 0.99 TOC (mg/sample) (mg/sample), R 2 = 0.99, N = 20. Because the TOR method required less sample material (20 ml versus 2+ ml), this approach was used in subsequent WSOC analysis. [27] Several studies have found higher WSOC/OC ratios at remote and urban sites using methods similar to those applied in this work. Zappoli et al. [1999] reported WSOC/ TC ratios of 0.84 and 0.52 at rural sites in Sweden and Hungary, respectively. They sampled aerosols on quartz filters and did not correct for VOC absorption artifact. Krivácsy et al. [2001] collected aerosols on quartz filters and measured TC by thermal analysis. They did not correct for VOC sampling artifacts. BC (black carbon) was estimated from Aethalometer measurements and OC was estimated as TC-BC. They reported WSOC/OC ratios of 0.41, 0.65, and 0.60 at Mace Head, Ireland, K-puszta, Hungary, and Jungfraujoch, Switzerland. Alves et al. [2002] measured WSOC and OC at two urban sites in Portugal (Aveiro and Lisbon) and at Hyytiälä, a remote forested site in Finland. They apparently used quartz filters, since OC was determined by thermal-optical analysis. No correction was reported for the VOC sampling artifact. WSOC/OC ratios at Aveiro and Lisbon were 0.35 and 0.41, respectively, while at the remote Finnish site, the WSOC/OC ratio was 0.22, the same as that found in this study Ambient Hydration State [28] The ambient hydration state at GRSM was inferred from changes in particle diameter as ambient particles were exposed to three RH-conditioning paths in the AS-HTDMA, as described above. Ambient aerosols on the lower leg (deliquescence branch) were rarely observed; ambient aerosols were nearly always hydrated. [29] Figure 5 summarizes AS-HTDMA measurements over roughly 2-hour periods for contrasting cases on 3 and 14 August. Particles were selected with ambient diameters of 0.05, 0.1, 0.2, and 0.4 mm. This upper size limit was chosen because ambient particles larger than 0.4 mm could exceed the upper limit of the DMA after exposure to high RH. The average ambient RHs during the 3 and 14 August measurement periods were 60 and 59%, respectively. The y axis in Figure 5 is D f /D a, as described above. Figure 5 shows that ambient particles of all sizes on 3 and 14 August were hydrated because D f /D a was <1 on Path 0. The growth factor increased with increasing particle size. This may be explained by increasing fractions of less hygroscopic material with decreasing particle size. On 3 August (Figure 5a), ambient particles on Path 2 displayed D f /D a = 1, suggesting that the hydration state was either HL-UL (hysteresis loop, upper leg) or EQ (continuous equilibrium growth). However, values of D f /D a < 1 on Path 1 constrain the ambient hydration state to HL-UL, suggesting the presence of some ammonium sulfate. On the basis of the medvol filter sample on 3 August, the average molar ratio of ammonium to sulfate was 1.42, well above the study average ratio of [30] Figure 5b shows that by contrast, on 14 August, ambient particles displayed D f /D a = 1 on both Paths 1 and 2. This represents continuous equilibrium growth typical of acidic sulfate and is consistent with a lower average molar ratio of ammonium to sulfate (0.92) based on the 14 August Figure 4. Time series of organic carbon (OC) and corresponding fraction of water-soluble organic carbon (WSOC %) at GRSM, summer of11

7 Figure 5. Examples of ambient hygroscopic growth measured with the AS-HTDMA on (a) 3 August 2006, when the average ambient RH was 60% and (b) 14 August 2006, when the average ambient RH was 59%. Total AS-TDMA scan time was 93 min in both cases. Circles represent Path 0: sample taken from ambient RH to dryness. Triangles represent Path 1: sample taken from ambient RH to dryness then back to ambient RH. Diamonds represent Path 2: sample taken from ambient RH to high RH (>90%) then back to ambient RH. filter sample. There were about 300 cases such as those described in Figure 5. Hysteresis was indicated by D f /D a <1 on Path 1. Table 2 summarizes the percentage of cases where hysteresis was observed as a function of ambient particle size. The frequency of hysteresis ranged from 29 to 46% for 0.05 and 0.2 mm particles, respectively. [31] Day et al. [2000] measured the variation of f(rh), the ratio of Bsp as a function RH to dry Bsp (RH < 20%) during the Southeastern Aerosol and Visibility Study (SEAVS) at GRSM during summer, Bsp measurements were made with an RH-controlled nephelometer (humidograph). Day et al. [2000] concluded that hygroscopic growth was always continuous. While our AS- HTDMA results indicated that the ambient hydration state was continuous in the majority of cases, hysteresis (efflorescence branch) was observed in a significant number of cases. This should have been evident in the low-to-high RH scans with the humidograph. The humidograph responds to the bulk aerosol, which could have obscured variations as a function of particle size revealed by the AS-HTDMA. It is also possible that the ambient hydrations states were different in 1995 and Relationship Between the Dry Particle Size Distribution and Concentration [32] Malm and Hand [2007] examined the relationship between dry mass scattering efficiencies and concentration with a statistical approach. They found that a systematic bias between reconstructed and measured light scattering could be reduced with a correction factor applied to dry mass scattering efficiencies that increased with concentration. McMeeking et al. [2005] reported a direct relationship between dry particle geometric mean diameter and volume measured with a DMA and optical particle counter at Yosemite National Park during summer, The aerosol was dominated by organics from wildfire smoke. Mass scattering efficiencies derived from measured size distributions and estimated density and refractive index also varied directly with particle geometric mean diameter. [33] In the 2006 GRSM study, variation between dry particle size and volume was examined directly using measurements with a DMA. Shifts in the particle size distribution were evaluated with the particle mean diameter. Particle volume is related to mass concentration by density, which may vary as a function of time. However, the PM 2.5 filter measurements suggest that the chemical composition was relatively constant. If reconstructed mass (REC25) is taken as the sum of NH 4 +,SO 4 =,NO 3, 1.8* OC, and EC, then the average daily percentages of REC25 for NH 4 + plus SO 4 = and 1.8*OC were 64 ± 9 and 18 ± 5%, respectively. Hourly dry particle mean diameter and particle volume measured with the HTDMA are compared in Figure 6a. There was a clear relationship between dry particle size and Table 2. Frequency of Hysteresis as Revealed by D f /D a <1on Path 1 (Particle Taken From Ambient RH to Dryness Then Back to Ambient RH) Ambient Particle Diameter (D a )(mm) Total number of measurements Percentage of measurements with significant hysteresis of11

8 Figure 6. Relationship between hourly-average particle volume and particle mean diameter: (a) this study and (b) during the Southeastern Aerosol and Visibility Study (SEAVS) in volume (R 2 = 0.46). Lowenthal and Kumar [2004] examined relationships among particle size and scattering efficiency during SEAVS. Particle size distributions were measured with a humidity-controlled differential mobility analyzer and optical particle size spectrometer (RH-DMOPSS) by Aerosol Dynamics, Inc. [Kreisberg et al., 2001]. Figure 6b shows that the relationship between hourly-average particle mean diameter and particle volume over a narrow range of RH (60 70%) was similar (y = , R 2 = 0.86) to that determined from dry HTDMA measurements in this study (y = , R 2 = 0.46). The intercept for SEAVS was higher because the RH was higher (60 < RH < 70%) than in the current study where the particles were dried. The direct relationship between particle mean diameter and volume supports the use of concentrationvarying dry scattering efficiencies in the revised IMPROVE equation OCM/OC Ratios [34] The OCM/OC ratio was determined from the masses of residues of XAD-treated water and DCM extracts and their respective OC contents. Initial measurements indicated that WSOC mass from individual hivol samples was not sufficient for detection by gravimetric analysis. It was also found that two passes through the XAD system were necessary to remove >99% of the inorganic ions from the water extracts. This reduced the overall recovery of WSOC. To obtain sufficient WSOC mass for this analysis and for hygroscopic growth experiments on reaerosolized WSOC, it was necessary to combine samples. Table 3 describes the sample composites and provides the data on which OCM/ OC estimates were made. Duarte and Duarte [2005] reported WSOC recoveries of 63 71% after one pass through the XAD system. WSOC recoveries from the composite extracts after two passes through the XAD ranged from 32 to 78% (Table 3). Corresponding sulfate removal efficiencies, also shown in Table 3, ranged from 99.1 to 99.9% and averaged 99.6%. [35] OCM/OC ratios for WSOC and DMC-extracted organic material ranged from 2.2 to 2.8 and from 1.7 to 2.0, and averaged 2.4 ± 0.3 and 1.9 ± 0.2 (average ± standard deviation), respectively. The uncertainties of ratios determined from replicate analysis were 3 and 6% for WSOC and DCM extracts, respectively. The higher ratio in WSOC relative to the DCM extract during this study is consistent with an expected higher proportion of polar, oxygenated compounds in the water extract, since nonpolar compounds are soluble in DCM. [36] Graber and Rudich [2006] reviewed numerous approaches for characterizing WSOC. There are clearly no standard approaches. Our results represent extractable organic material. Table 3 shows that the recovery of quartz filter-based OC in the DCM extracts ranged from 13 22%. As noted above, this may be considered a lower limit if the OC concentrations were high-biased by the VOC adsorption artifact. Also shown in Table 3 are WSOC/OC ratios (%) for the composites, which ranged from 10 40%, which also may be low-biased by the VOC adsorption artifact. While it is not possible to estimate the OCM/OC ratio for material that was not recovered either from the XAD columns or by the DCM extract, it is clear from Table 3 that there were no Table 3. OCM/OC Ratios Derived From Water and DCM Extracts a,b Start Date End Date Composite Sample Mass H 2 O c OC H 2 O d WSOC Recovery (%) e Sulfate Removal (%) WSOC/OC % OCM/OC H 2 O Mass DCM f OC DCM g DCM OC Recovery (%) OCM/OC DCM 19 July Jul 2006 S Jul Jul 2006 S Aug Aug 2006 S Aug Aug 2006 S Average = 2.4 Average = 1.9 SD = 0.3 SD = 0.2 a Concentrations are mg/filter punch. b Overall recovery of OC in the initial water and DCM extracts. c Mass residue from water extract. d OC residue from water extract. e Recovery of WSOC after two passes through the XAD columns. f Mass residue from DCM extract. g OC residue from DCM extract. 8of11

9 Figure 7. Hygroscopic growth factors (GF = DRH/DDry) for composite isolated WSOC samples from high-to-low RH scans with the HTDMA. Scans are shown for increasing RH (30% 90%) and decreasing RH (90% 30%). relationships between the OCM/OC ratio and WSOC/OC, XAD recovery, or DCM recovery. The problem of incomplete extraction is compounded in many studies which estimated the OCM/OC ratio based on selected [Turpin and Lim, 2001] or identified [Yu et al., 2005] compounds. For example, Yu et al. [2005] based their results on 46 identified compounds which represented only 1 6% of the bulk OC mass. [37] Our results are consistent with previous estimates of the OCM/OC ratio determined by various methods. For example, Turpin and Lim [2001] recommended an OCM/ OC ratio of 2.1 for nonurban aerosols based on aerosol organic compounds reported in the literature. Kondo et al. [2007] used Aerosol Mass Spectrometer measurements in Tokyo to estimate OCM/OC ratios of 1.7 and 2.2 in winter and summer, respectively. Chen and Yu [2007] used a combination of gravimetric and chemical measurements in Hong Kong and reported OCM/OC ratios of 1.9 and 2.2 associated with marine and continental air masses, respectively. Malm and Hand [2007] estimated the OCM/OC using regression of OC, sulfate, nitrate, and soil against PM 2.5 for all IMPROVE sites. Their estimate for GRSM was 1.7 ± 0.1. El-Zanan et al. [2005] measured OCM/OC directly on residues from DCM and acetone extracts from composited portions of IMPROVE samples collected at GRSM from and obtained an OCM/OC ratio of 2.0. Note that some of the water-soluble OC was probably extracted in the acetone. There was not enough mass in residues of water extracts to accurately weigh the OCM. El Zanan et al. also estimated the OCM/OC ratio using a mass balance. They estimated OCM as the difference between measured PM 2.5 and reconstructed mass excluding OC. The estimate for GRSM using this approach was 2.3 on an annual basis. As noted above, Yu et al. [2005] estimated a 24-hour average OCM/OC ratio of 1.9 ± 0.3 at GRSM during SEAVS based on 46 identified compounds in water and organic extracts (representing only 1 6% of organic mass) WSOC Hygroscopic Growth Factors [38] The four composite WSOC samples described in Table 3 were reaerosolized as described above and subjected to HTDMA analysis to measure their hygroscopic growth factors (GF) as a function of RH. The WSOC aerosols were subjected to RH scans from low-to-high (up-scan) and from high-to-low RH (down-scan). Figure 7 shows hygroscopic growth curves (GF = D RH /D Dry versus RH ranging from 30 to 95%) for the up-scans and downscans were nearly identical. In contrast with simple organic compounds such as glutaric and pinonic acids which deliquesce at high RH [Cruz and Pandis, 2000], the hygroscopic growth of ambient WSOC was continuous with increasing and decreasing RH. The composite samples exhibited similar growth curves, with average growth factors of 1.10 ± 0.02 (1.08 to 1.13), 1.13 ± 0.03 (1.11 to 1.17), and 1.19 ± 0.04 (1.16 to 1.24) at 80, 85, and 90% RH, respectively. Dinar et al. [2007] measured similar hygroscopic growth factors for WSOC isolated from aerosol samples representing fresh and aged wood smoke. For 9of11

10 comparison, the growth factor for pure ammonium sulfate at 80% RH is 1.49 [Tang and Munkelwitz, 1994]. The contributions of WSOC to hygroscopic growth depend not only on its growth factors as a function of RH but its relative abundance with respect to other soluble components. On average, the ammoniated sulfate to WSOC mass ratio was about 10 and the corresponding ratio of their growth factors at 80% RH was about 1.5. This suggests that the relative contribution of WSOC hygroscopicity to particle light scattering during our study was relatively small. Nonetheless, the assumption in the revised IMPROVE equation that organics are not hygroscopic is not supported by our analysis. 4. Conclusions [39] A multiyear, multisite research study was begun at Great Smoky Mountains National Park (GRSM) during the summer of 2006 to elucidate the causes of aerosol light extinction in U.S. national parks. PM 2.5 aerosol concentrations at GRSM were dominated by sulfates and organic carbon from 19 July through 17 August. Water-soluble organic carbon (WSOC) comprised 22% of total organic carbon (OC), on average, and ranged from 11 to 30% of OC in July and August, respectively. The average molar ratio of ammonium to sulfate was 1.16 ± 0.21, indicating that sulfate was generally acidic, i.e., present as ammonium bisulfate. The ambient hydration state of the aerosol was measured continuously with the AS-HTDMA (Ambient State Hygroscopic Tandem Differential Mobility Analyzer). While deliquescence was not observed, a significant number of cases exhibited hygroscopic growth on the upper leg (efflorescence branch) of the hysteresis loop of particles containing some ammonium sulfate. In most cases, continuous hygroscopic growth was consistent with that expected for acidic sulfates, as indicated by the filter-based measurements of NH + 4 and SO = 4. Thus the measured ambient hygroscopic state at GRSM during summer, 2002 was consistent with the IMPROVE equation. [40] The relationship between the dry particle size distribution and concentration was examined in the field using scanning DMA measurements. There was a clear relationship between particle mean diameter and volume that was consistent with similar measurements during the Southeastern Aerosol and Visibility Study (SEAVS) at GRSM in Because dry mass scattering efficiency increases with particle size in the submicron size range and because particle volume is related to mass concentration by density, these results support the use of concentration-varying dry mass scattering efficiencies in the revised IMPROVE equation. The behavior of species-specific mass scattering efficiencies should be examined using size-resolved aerosol chemical measurements. [41] Water-soluble organic carbon (WSOC) was isolated from aerosol water extracts using XAD solid-phase absorbents. Organic material was also extracted in dichloromethane (DCM). The extracts from individual samples were combined to provide adequate organic mass for gravimetric analysis and hygroscopic growth experiments. The residues from water and DCM extracts were weighed and their OC content measured. Average ratios of OCM/OC in WSOC and DCM extracts were 2.4 ± 0.3 ( ) and 1.9 ± 0.4 ( ), respectively. These ratios were not related to the WSOC/OC ratio, recoveries from the XAD resins, or DCM extraction efficiency. These results are consistent with those of several recent studies. [42] WSOC in composite water extracts were reaerosolized and particles with a dry diameter of 0.07 mm were selected for hygroscopic growth measurements with the HTDMA. Growth factors (GF), i.e., the ratios of humidified to dry diameter were determined by scanning from low to high and then from high to low RH. The growth curves were smooth and continuous, did not exhibit deliquescence, and were nearly identical for the low-to-high and high-tolow RH scans. The composited samples exhibited similar growth curves, with growth factors ranging from 1.08 to 1.13, 1.11 to 1.17, and 1.16 to 1.24 at RHs of 80, 85, and 90%, respectively, for the high-to-low RH scans. These GFs are similar in magnitude to those obtained from isolated WSOC in previous studies. Finite WSOC growth factors, even in cases where WSOC concentrations are low compared to other soluble aerosol components, contradict the assumption in the revised IMPROVE equation that organics are not hygroscopic. [43] Acknowledgments. This work was supported by EPRI in Palo Alto, CA. This research would not have been possible without the cooperation of the National Park Service. We gratefully acknowledge the assistance provided by Jim Renfro, air quality specialist at Great Smoky Mountains National Park, and Roger Tanner and Bill Hicks, Tennessee Valley Authority. References Alves, C., A. Carvalho, and C. Pio (2002), Mass balance of organic carbon fractions in atmospheric aerosols, J. Geophys. Res., 107(D21), 8345, doi: /2001jd Charlson, R. J., S. E. Schwartz, J. M. Hales, R. D. Cess, J. A. Coakley Jr., J. E. Hansen, and D. J. Hofmann (1992), Climate forcing by anthropogenic aerosols, Science, 255, Chen, X., and J. Z. Yu (2007), Measurement of organic mass to organic carbon ratio in ambient aerosol samples using a gravimetric technique in combination with chemical analysis, Atmos. Environ., 41, Chow, J. C., J. G. Watson, L. C. Pritchett, W. R. Pierson, C. A. Frazier, and R. G. Purcell (1993), The DRI thermal/optical reflectance carbon analysis system: Description, evaluation and applications in U.S. air quality studies, Atmos. Environ., 27A, Chow, J. C., J. G. Watson, D. H. Lowenthal, and K. Magliano (2005), Loss of PM 2.5 nitrate from filter samples in central California, J. Air Waste Manage. Assoc., 55, Clegg, S. L., P. Brimblecombe, and A. S. Wexler (1998), A thermodynamic model of H-NH 4 -SO 4 -NO 3 -H 2 O at tropospheric temperatures, J. Phys. Chem., 102A, Collins, D. R., R. C. Flagan, and J. H. Seinfeld (2002), Improved inversion of scanning DMA data, Aerosol Sci. Technol., 36, 1 9. Cruz, C. N., and S. N. Pandis (2000), Deliquescence and hygroscopic growth of mixed inorganic-organic atmospheric aerosol, Environ. Sci. Technol., 34, Day, D. E., W. C. Malm, and S. M. Kreidenweis (1997), Seasonal variations in aerosol composition and acidity at Shenandoah and Great Smoky Mountains national parks, J. Air Waste Manage. Assoc., 47, Day, D. E., W. C. Malm, and S. M. Kreidenweis (2000), Aerosol light scattering measurements as a function of relative humidity, J. Air Waste Manage. Assoc., 50, Decesari, S., M. C. Facchini, E. Matta, F. Lettini, M. Mircea, S. Fuzzi, E. Tagliavini, and J.-P. Putaud (2001), Chemical features and seasonal variation of fine aerosol water-soluble compounds in the Po Valley, Italy, Atmos. Environ., 35, Dick, W. D., P. Saxena, and P. H. McMurry (2000), Estimation of water uptake by organic compounds in submicron aerosols measured during the Southeastern Aerosol and Visibility Study, J. Geophys. Res., 105, of 11