Two collocated, eight-stage rotating drum impactors were deployed at Trinidad Head (California) during the spring of 2002 as part of the Intercontinental Transport and Chemical Transformation 2002 (ITCT 2K2) experiment. One of the samplers operated at ambient relative humidity while the other was operated at a relative humidity of 55%. The impaction substrates from these samplers were analyzed using synchrotron X-ray fluorescence (SXRF) to provide continuous measurements of the size-resolved aerosol elemental composition with 3-hour time resolution. The aerosol elemental composition data identified three significant mineral dust episodes near the beginning of the time series. The backward air mass trajectory calculations from the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model and the PM10 to PM2.5 elemental mass ratios are consistent with the long-range transport of mineral dust from Asian sources. The data from the paired ambient relative humidity and low-relative-humidity samplers show that the aluminum, silicon, and iron elemental mass distributions are a function of relative humidity. In each case, the elemental mass distributions shifted toward smaller sizes as the relative humidity was reduced. This behavior indicates that the mineral dust transported from Asia to the west coast of the United States is somewhat hygroscopic upon its arrival. The hygroscopic nature of the aged mineral dust should increase its ability to nucleate cloud droplets (i.e., act as cloud condensation nuclei). Measurements of transported Asian mineral dust made at a high-elevation mountain site in Oregon (i.e., Crater Lake National Park) during the spring of 2002 show a strong correlation between the silicon and sulfur elemental mass concentrations. The ratio of calcium to sulfur makes it unlikely that this coarse sulfur is derived from gypsum (i.e., CaSO4). Instead, it indicates that the coarse mineral dust most likely accumulates sulfate coatings either near the source region or during transport across the Pacific Ocean.
 Mineral dust is one of the most abundant aerosol species in the atmosphere with an estimated global emission rate on the order of 1000 Mt yr−1 [e.g., Tegen and Miller, 1998; Lunt and Valdes, 2002]. Although much of the mineral dust mass is removed by dry deposition near the source regions, a significant fraction of the particles with aerodynamic diameters < 2.5 μm (i.e., PM2.5) are transported intercontinental distances [e.g., Prospero et al., 1970, 1989; Duce et al., 1980; Perry et al., 1997; Husar et al., 2001; VanCuren and Cahill, 2002]. Aerosol optical thickness data from the National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) clearly show the global nature of this transport [Husar et al., 1997]. What the global AVHRR aerosol optical thickness measurements do not reveal, however, are the extensive physical and chemical transformations that occur as the mineral dust is transported downwind from the source regions. For example, the mass mean diameter of the transported dust decreases over time because of the size-dependent nature of the removal by dry deposition. This evolution of the mineral dust mass distribution results in a mass scattering efficiency that increases over time [Perry et al., 1997]. The net result is that the direct effect of mineral dust particles on the tropospheric radiative balance persists for a much longer distance downwind than a simple estimate based upon the rate of dry deposition would indicate.
 Chemical transformations of the mineral dust can also occur during transport [e.g., Dentener et al., 1996; Zhang and Carmichael, 1999; Song and Carmichael, 2001]. These transformations include both gas-phase surface chemistry and aqueous-phase chemical reactions within cloud droplets [e.g., Xiao et al., 1997; Miller and Grassian, 1998; Goodman et al., 2000; Underwood et al., 2001; Usher et al., 2002; Krueger et al., 2003]. Both of these processes typically result in an internal mixture of many different chemical species. Some of these species, such as black carbon, retain much of their original shape as they adhere to the surface of the mineral dust particles. By contrast, soluble species, such as sulfates and nitrates, tend to form a much more uniform coating on the mineral dust particles. The inhomogeneous nature of these internally mixed, nonspherical mineral dust particles greatly complicates the calculations of their direct radiative effects [Sokolik et al., 2001]. The presence of soluble coatings enhances the ability of the mineral dust particles to act as cloud condensation nuclei (i.e., CCN) by permitting hygroscopic growth in subsaturated conditions. There is even some evidence that coated mineral dust particles can initiate cirrus cloud formation by heterogeneous nucleation [Hung et al., 2003].
 Soluble coatings on mineral dust particles are commonly observed in the atmosphere. Several studies in the Mediterranean found significant sulfate coatings on transported African dust and ample evidence for aerosol-cloud interactions [e.g., Levin et al., 1996, 2001; Wurzler et al., 2000; Falkovich et al., 2001]. A study by Li-Jones et al. , however, found only minor amounts of hygroscopic materials associated with African dust transported to the tropical Atlantic Ocean. The disparity between these studies could simply be the result of different air parcel trajectories with different precursor gas concentrations. It could also result from different cloud histories along the air parcel trajectories. Several studies of Asian dust have shown strong correlations between non-sea-salt (nss) calcium, nitrate, and nss sulfate [e.g., Parungo et al., 1995; Dentener et al., 1996; Chen et al., 1997]. What are lacking from these previous studies, however, are direct measurements of the hygroscopic behavior of coated mineral dust particles.
 The goal of this study is to use measurements of the size-resolved aerosol elemental composition to determine if mineral dust transported from Asia to the west coast of the United States is hygroscopic upon its arrival. This study will make use of data from a network of ground-based aerosol samplers that were deployed in Oregon and California during the spring of 2002 as part of the Intercontinental Transport and Chemical Transformation 2002 (ITCT 2K2) experiment. Although measurements were made at five sites during the experiment, this study will focus on data collected at Trinidad Head (41.05°N, 124.15°W, 119 m mean sea level (msl)) and Crater Lake National Park (42.89°N, 122.14°W, 1963 m msl). Data from the Trinidad Head site will be used to diagnose the hygroscopic behavior of the mineral dust transported from Asia, while data from the Crater Lake site will be used to describe the size-resolved aerosol elemental composition of the transported mineral dust. For both of these locations, continuous measurements of the size-resolved aerosol elemental composition were made with 3-hour time resolution from late April through the end of May 2002. Detailed descriptions of both the aerosol samplers and the analytical technique used in this study are provided in section 2.
 The rotating drum impactor, which was developed at the University of California, Davis, operates on the same principles as the Micro-orifice Uniform Deposit Impactor (MOUDI) and Cascade impactors, but is configured for continuous, autonomous operation [Raabe et al., 1988; Cahill and Wakabayashi, 1993]. Continuous operation is achieved by slowly rotating the impaction substrates under the slotted orifices. This process results in a linear sample that is uniform in one dimension, yet preserves the time variation of aerosols in the other. The rotating drum impactor exists in both eight-stage and three-stage configurations. The eight-stage drum impactor physically separate aerosols into the following aerodynamic diameter ranges: 10.0–5.0, 5.0–2.5, 2.5–1.15, 1.15–0.75, 0.75–0.56, 0.56–0.34, 0.34–0.24, and 0.24–0.09 μm. An upper size cut of 10.0 μm is achieved by using a URG Corp. PM10 inlet head operating at a flow rate of 16.7 L min−1. Both the eight-stage and three-stage impactor configurations used Mylar impaction substrates that were greased with a 1% solution of Apiezon Type-L grease dissolved in toluene. The purpose of the grease is to minimize particle-sizing errors caused by particle bounce. The three-stage version of the drum impactor operates at a flow rate of 23.1 L min−1 and separates aerosols into the following aerodynamic diameter ranges: 2.5–1.15, 1.15–0.34, and 0.34–0.09 μm. The upper size cut of the three-stage rotating drum impactor is normally achieved by using a PM2.5 cyclone identical in design to those of the Interagency Monitoring for Protected Visual Environments (IMPROVE) aerosol network [Malm et al., 1994]. In this study, however, the PM2.5 cyclone was removed from the system and replaced with an inlet that provided an upper size cut of ∼12.5 μm. Thus the first stage of the three-stage rotating drum impactor collected particles with aerodynamic diameters between 12.5 and 1.15 μm. This change was made because this study required measurements of the coarse fraction of the transported mineral dust.
 The flow rates for both the eight-stage and three-stage impactors were controlled by a critical orifice. The volumetric flow rate through a critical orifice is controlled by the speed of sound (equation (1)) [Weast, 1986].
speed of sound, m s−1;
specific heat of air at constant pressure, equal to 1004 J K−1 kg−1 for dry air;
specific heat of air at constant volume, equal to 717 J K−1 kg−1 for dry air;
universal gas constant, equal to 8.3143 J K−1 mol−1;
average molecular weight of the air at sea level, equal to 0.02897 kg mol−1 for dry air;
molecular-scale temperature, K.
This equation shows that the speed of sound is independent of atmospheric pressure and weakly dependent upon atmospheric temperature. The temperature fluctuations encountered by the samplers during the ITCT 2K2 experiment (i.e., −10° to 25°C) resulted in a sample flow rate uncertainty of ±3% for this study.
 With the exception of one of the eight-stage rotating drum impactors at Trinidad Head, all of the aerosol samplers were operated at ambient relative humidity (RH). Humidity control on one of the samplers at Trinidad Head was achieved using an Omega HX92AV-RP1 humidity probe wired to an Omega CNi3222 process controller. The process controller maintained the relative humidity at 55% by supplying power (on an as needed basis) to a heat tape that was wrapped around the inlet stack. When turned on, the rectangular (2.54 × 60.96 cm.) silicone rubber heat tape (Omega SRFG-124/5) supplied 0.775 W cm−2 (120 W total). Pipe insulation was placed over the heating tape to shield it from the environment and increase the efficiency of the energy transfer to the sample airstream. Because the Trinidad Head site is located along the coast and is well within the marine boundary layer, the ambient relative humidity was almost continually >55%.
 Measurements of the size-resolved aerosol elemental composition were made from the greased Mylar impaction substrates using synchrotron X-ray fluorescence (SXRF). The SXRF analysis was performed using beamline 10.3.1 at the Advanced Light Source (Lawrence Berkeley National Laboratory). The Advanced Light Source (ALS) is a Department of Energy (DOE) national user facility that generates intense ultraviolet and soft X-ray beams for scientific and technological research. The ALS is the first third-generation synchrotron light source in its energy range and is capable of producing light in the X-ray region of the electromagnetic spectrum that is one billion times brighter than the Sun.
 The X-ray microprobe at beamline 10.3.1 is the X-ray analog to an electron microprobe. It is designed to operate at photon energies over the range from 1.5 to 18.5 keV using a white (i.e., nonmonochromatic) beam configuration. This energy range permits quantitative analysis of the elements sodium through uranium when the samples are analyzed under vacuum. Light from the beamline is collimated prior to entry into the sample chamber. The resulting beam spot is typically 500 μm wide in the nontime axis of the impaction substrate and 10–50 μm wide along the time axis. Previous tests have shown that the sample deposit from the drum impactors is extremely uniform along the nontime axis [Bench et al., 2002]. A Si(Li) detector is used to record the resultant fluorescence X-rays, with the photon energy identifying the element and the intensity its concentration. The beam is 100% polarized which greatly reduces the background signal and dramatically improves the signal-to-noise ratio. Deconvolutions of the raw X-ray spectra are performed using the Quantitative X-ray Analysis System (QXAS) X-ray peak-fitting software package that was developed by the International Atomic Energy Agency (IAEA) Laboratories in Seibersdorf, Austria. Quantitative analysis is performed by calibrating the response of the system to a comprehensive set of 40 single-element and multielement NIST-traceable standards (Micromatter, Inc.).
3. Results and Discussion
3.1. Long-Range Transport of Mineral Dust During the ITCT 2K2 Experiment
 The long-range transport of mineral dust from Asia to the west coast of the United States is a well-known phenomenon that occurs each year during the spring [e.g., Jaffe et al., 1999, 2001; Husar et al., 2001; McKendry et al., 2001; Tratt et al., 2001; Thulasiraman et al., 2002; VanCuren and Cahill, 2002; Peterson and Tyler, 2003; VanCuren, 2003; Liu et al., 2003]. Transport in other seasons is possible as well. For example, VanCuren and Cahill  demonstrated that much of the fine (i.e., Dp < 2.5 μm) mineral dust measured throughout the year at high-elevation sites within the western United States originates from Asia. Ground-based instrumentation at Trinidad Head observed three significant mineral dust episodes in the early part of the ITCT 2K2 experiment (Figure 1). These three dust pulses were observed on 22, 24, and 26 April 2002 (i.e., Julian days 112, 114, and 116). The PM10 soil mass concentrations for these episodes can be estimated from the elemental concentration data by assuming that the mineral dust elements are in the form of their most common oxides [Seinfeld, 1986; Malm et al., 1994]. The estimated maximum PM10 soil mass concentrations for these three transport episodes were 5.9, 5.2, and 4.0 μg m−3. The corresponding coarse (i.e., 2.5 < Dp < 10 μm) to fine (i.e., Dp < 2.5 μm) ratios of the mineral dust for these episodes were 0.46, 1.59, and 0.93. These ratios are significantly smaller than the average coarse to fine soil mass ratios typically observed throughout the western United States. For example, Eldred et al.  calculated the average coarse to fine ratios for all of the typical soil elements using data from 36 IMPROVE aerosol monitoring network sites located in the western United States. The average coarse to fine ratios for aluminum, silicon, and iron from the Eldred et al.  study were 5.5, 5.5, and 4.3, respectively. These ratios exceed those observed at the Trinidad Head site during the ITCT 2K2 experiment by more than a factor of 10. Thus the fine nature of the mineral dust observed at Trinidad Head is consistent with a nonlocal source that has lost much its coarse soil fraction during transport.
 Probable source regions for the three mineral dust episodes that occurred near the beginning of the ITCT 2K2 experiment were diagnosed using backward air mass trajectories (Figure 2). These trajectories were calculated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model created by the National Oceanic and Atmospheric Administration Air Resources Laboratory (NOAA/ARL) (R. R. Draxler and G. D. Rolph, HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) model, access via NOAA ARL READY Web site http://www.arl.noaa.gov/ready/hysplit4.html), 2003; G. D. Rolph, Real-Time Environmental Applications and Display System (READY) Web site http://www.arl.noaa.gov/ready/hysplit4.html), 2003). The HYSPLIT model was initialized with the FNL meteorological archive and used the model-derived vertical velocities to track the tagged air parcels. 10-day backward air mass trajectories originating at 500, 1000, and 1500 m above ground level were computed for each of the three dust episodes observed at Trinidad Head. These trajectories were initialized at 0600 UTC to match the timing of the maximum mineral dust concentrations observed at the ground. Although the accuracy of the calculated trajectories undoubtedly degrades over time, all of these trajectories pass over mineral dust source regions within Asia. The transit time from the mineral dust source regions in Asia to Trinidad Head ranged from 5 to 9 days.
 The timing of the local mineral dust maxima at Trinidad Head also supports a nonlocal source. In each of the three dust episodes the maximum mineral dust concentrations occurred at night (i.e., between 1800 and 0300 LST). Nighttime maxima are unlikely to be caused by local sources because anthropogenic activities that generate mineral dust typically occur during the day. In this case, it is hypothesized that the nighttime maxima result from a combination of synoptic-scale subsidence of an elevated aerosol layer and the relatively shallow atmospheric boundary layers that occur at night. Evidence for the synoptic-scale subsidence can be seen in the air parcel altitude time series shown in the bottom panels of Figures 2a, 2b, and 2c. No attempt was made to compare the chemical composition of the mineral dust at Trinidad Head to known Asian sources because the transported dust spent from 1 to 5 days within the marine boundary layer before arriving at the west coast of the United States. Transport within the marine boundary layer can significantly alter many of the elemental ratios typically used to diagnose the origin of the mineral dust. This alteration will be most severe for common mineral dust elements that also have a sea-salt source (i.e., Na, Mg, Ca, and K).
3.2. Hygroscopic Behavior of Mineral Dust
 In this section, simultaneous measurements of the size-resolved aerosol elemental composition made at ambient and low relative humidities will be used to show that the mineral dust observed at Trinidad Head during the three Asian dust transport episodes was somewhat hygroscopic. Since both of the eight-stage rotating drum impactors at Trinidad Head were equipped with a URG Corporation PM10 sample inlet and were operated at the same flow rate, the total particulate mass entering each sampler should be identical. Recall that the sample heating within the low-RH impactor occurs after the PM10 size cut has already been established. As a result, lowering the RH within the sample airstream will not affect the total particle mass collected by the eight-stage rotating drum impactor. It can, however, alter the particle mass distribution if the aerosol is hygroscopic. In a similar manner, it can be argued that the total aerosol elemental mass concentrations measured from the ambient RH and low-RH samplers should also be identical.
Figure 3 shows scatterplots of the PM10 aerosol elemental mass concentration measurements of the three most abundant mineral dust elements (i.e., Al, Si, and Fe) at ambient and low RH for the entire ITCT 2K2 experiment. For each of these elements, the Pearson correlation coefficients (i.e., R2) are >0.92 and the slopes of the least squares regression lines are between 0.95 and 1.00. The excellent agreement of these independent measurements of the PM10 elemental mass concentrations indicates that the sampling and analytical techniques are free from systematic biases that would preclude comparison to each other. Since the total mass of each element collected by the ambient RH and low-RH impactors is comparable, any differences observed in the particle mass distributions from these samplers must result from the hygroscopic behavior of the mineral dust particles.
 Although the total PM10 elemental mass concentrations of aluminum, silicon, and iron were not affected by reducing the RH (Figure 3), the mass distributions for these elements were significantly altered during the three Asian dust transport episodes. For example, the elemental mass measured on the first stage of the rotating drum impactors (i.e., coarse particles with aerodynamic diameters between 5.0 and 10 μm) decreased substantially when the RH of the airstream was reduced to 55% prior to the size-selective sampling (Figure 4). Since the PM10 elemental mass measurements from the paired samplers are in excellent agreement (Figure 3), the only explanation for a reduction in mass on the first stage of the low-RH impactor is that the aerodynamic diameter of the coarse mineral dust particles shifted to smaller sizes. As a result, a portion of the elemental mass that was collected on stage 1 of the ambient RH sampler was instead collected on other stages within the low-RH impactor. Evidence for this redistribution of mineral dust elemental mass to the smaller size ranges is also shown in Figure 5. This figure shows the average mass distributions of aluminum, silicon, and iron for both the ambient RH and low-RH conditions from 21 April through 26 April 2002. Since the total elemental masses of aluminum, silicon, and iron measured by the ambient RH and low-RH samplers during this time period agree to within 0.9%, 0.6%, and 1.1%, respectively, changes in the elemental mass within each size bin must result from the hygroscopic properties of the mineral dust.
3.3. Do Aged Asian Mineral Dust Particles Have a Sulfate Coating?
 Since the only evidence of hygroscopic mineral dust has been associated with trajectories that pass over polluted regions [e.g., Levin et al., 1996, 2001; Wurzler et al., 2000; Falkovich et al., 2001], it is reasonable to assume that the hygroscopic behavior of the aged Asian mineral dust is caused by coatings of secondary sulfates, nitrates, or organics. These coatings could have been formed either downwind of the source regions within the polluted Asian atmosphere or during the 5–9 day transit from Asia across the Pacific. Although it is beyond the scope of this study to identify the atmospheric processes responsible for depositing these coatings on the mineral dust particles, the SXRF data do provide some information that pertains to the types of coatings that might exist on the mineral dust particles. Specifically, the SXRF analysis provides size-resolved measurements of elemental sulfur mass concentrations. Unfortunately, the SXRF measurements do not provide data that can be used to identify the presence of either nitrate or organic coatings.
 Since the transport of Asian dust to the western United States is typically concentrated in an altitude zone ranging from 500 and 3000 m above mean sea level [VanCuren and Cahill, 2002], the frequency of mineral dust transport events to Crater Lake is much greater than at the low-elevation, Trinidad Head site. Thus a much more robust description of the size-resolved elemental composition of the Asian dust can be obtained using data from the Crater Lake site. Another advantage of using data from the high-elevation, Crater Lake site is the reduced probability that the observed coarse sulfur originated from within the marine boundary layer. In general, the Asian dust transport episodes are regional events that are large enough to simultaneously impact more than one state along the west coast of the United States. Thus the transport events that penetrated the marine boundary layer at the Trinidad Head site represent a subset of the more frequent, high-elevation transport events observed at the Crater Lake site.
Figure 6 shows a scatterplot of the 3-hour averaged coarse (i.e., Dp > 2.5 μm) silicon and sulfur mass concentrations measured at Crater Lake during the ITCT 2K2 experiment. The least squares linear regression of the data in Figure 6 shows a significant correlation between coarse silicon and coarse sulfur with a Pearson correlation coefficient (i.e., R2) of 0.71. The slope of the regression line indicates that the coarse sulfur mass concentrations are on the order of 10% of the coarse silicon mass concentrations. There are two possible explanations for this correlation: (1) Some of the mineral dust is composed of gypsum (i.e., CaSO4), or (2) the mineral dust has a sulfate coating.
 Although the direct single-particle measurements necessary to resolve the gypsum/sulfate coating question were not made at the Crater Lake site during the ITCT 2K2 experiment, some indirect evidence for sulfate coated mineral dust particles does exist. For example, Figure 7 shows the correlation between the coarse calcium and coarse sulfur elemental mass concentrations (Figure 7). If the coarse sulfur was associated with CaSO4, then there should be a linear relationship between sulfur and calcium. In this case, the slope of the linear relationship would depend upon the fraction of calcium that is composed of CaSO4. For reference purposes, Figure 7 shows the expected sulfur concentrations assuming that 25%, 50%, and 100% of the calcium was associated with CaSO4. The actual data for this site, however, show a distinctly nonlinear relationship between these variables. This behavior is contrary to what would be expected if the coarse sulfur was associated with CaSO4 and the transported Asian dust came from similar source regions throughout this time series. Although the data shown in Figure 7 do not prove that the coarse sulfur exists as a sulfate coating on the mineral dust particles, they do show that the alternative gypsum explanation is problematic. This contention is further strengthened by Nishikawa et al. , who presented sulfate weight fractions in soils of arid regions of China ranging from <0.01% to 0.46%. The equivalent sulfur content corresponds to 0.003–0.15%.
 The goal of this study was to determine if aged mineral dust particles that have undergone long-range transport from Asia to the west coast of the United States exhibit hygroscopic growth. Since fresh mineral dust near source regions typically does not exhibit this type of behavior, any observations of hygroscopic growth would demonstrate that important chemical transformations do occur to the mineral dust during transport across the Pacific. These chemical transformations can impact both the light scattering and cloud nucleating properties of the transported mineral dust.
 Direct measurements of the hygroscopic behavior of mineral dust were made at the Trinidad Head site in northern California from 21 April to 26 May 2002 as part of the ITCT 2K2 experiment. Collocated measurements of the size-resolved aerosol elemental composition were made at ambient and low relative humidities (i.e., RH = 55%) using eight-stage rotating drum impactors followed by SXRF analysis at the Advanced Light Source (Lawrence Berkeley National Laboratory). These sampling and analytical techniques provided continuous measurements of the aerosol elemental composition with 3-hour time resolution. The aerosol composition data were combined with backward air mass trajectories to identify three significant Asian dust transport episodes during the ITCT 2K2 experiment. These Asian dust episodes were characterized by coarse (i.e., 2.5 < Dp < 10 μm) to fine (i.e., Dp< 2.5 μm) soil elemental ratios ranging from 0.46 to 1.59. These ratios are more than an order of magnitude less than values typically observed in the western United States.
 Measurements from the paired ambient RH and low-RH samplers show that the elemental mass distributions of the most abundant mineral dust elements (i.e., Al, Si, and Fe) were dependent upon the RH during the Asian dust transport episodes. For each of the major soil elements, the mass distributions shifted toward smaller sizes as the RH was reduced. Since fresh mineral dust does not typically exhibit this behavior, the aged Asian mineral dust must have been altered either within the polluted Asian atmosphere or during transport across the Pacific. Measurements of the transported mineral dust made at a high-elevation mountain site show a strong correlation between the silicon and sulfur elemental mass concentrations. The ratio of calcium to sulfur makes it unlikely that this coarse sulfur is derived from gypsum (i.e., CaSO4). Instead, it indicates that the coarse mineral dust most likely accumulates sulfate coatings either near the source region or during transport across the Pacific Ocean.
 Future studies should be directed at quantifying the relative contributions of secondary sulfates, nitrates, and organics to the hygroscopic behavior of the aged mineral dust and to identifying the specific atmospheric processes that lead to the chemical coatings of the mineral dust. Both of these research topics need to be addressed before the hygroscopic behavior of mineral dust can be effectively incorporated into regional chemical transport models.
 Funding for both the field and laboratory work was provided by NOAA under award NA16GP2360. Support for the Advanced Light Source (Lawrence Berkeley National Laboratory) was provided by the Department of Energy, Office of Basic Energy Science. The authors also gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model.