Aerosol Light Extinction Measurements by Cavity Attenuated Phase Shift (CAPS) Spectroscopy: Laboratory Validation and Field Deployment of a Compact Aerosol Particle Extinction Monitor

Abstract

We present laboratory and field measurements of aerosol light extinction ( σ_ep ) using an instrument that employs Cavity Attenuated Phase Shift (CAPS) spectroscopy. The CAPS extinction monitor comprises a light emitting diode (LED), an optical cavity that acts as the sample cell, and a vacuum photodiode for light detection. The particle σ_ep is determined from changes in the phase shift of the distorted waveform of the square-wave modulated LED light that is transmitted through the optical cell. The 3-σ detection limit of the CAPS monitor under dry particle-free air is 3 Mm^–1 for 1s integration time. Laboratory measurements of absolute particle extinction cross section ( σ_ext ) using non-absorbing, monodisperse polystyrene latex (PSL) spheres are made with an average precision of ± 3% (2-σ) at both 445 and 632 nm. A comparison with Mie theory scattering calculations indicates that these results are accurate within the 10% uncertainty stated for the particle number density measurements. The CAPS extinction monitor was deployed twice in 2009 to test its robustness and performance outside of the laboratory environment. During these field campaigns, a co-located Multi Angle Absorption Photometer (MAAP) provided particle light absorption coefficient ( σ_ap) at 635 nm: the single scattering albedo ( ω) of the ambient aerosol particles was estimated by combining the CAPS σ_ep measured at 632 nm with the MAAP σ_ap data. Our initial results show the high potential of the CAPS as lightweight, compact instrument to perform precise and accurate σ_ep measurements of atmospheric aerosol particles in both laboratory and field conditions.

1. INTRODUCTION

Aerosol particles affect the Earth's radiation balance directly by scattering and absorbing solar and terrestrial radiation, and indirectly by acting as cloud condensation nuclei (CCN). It is now recognized that the atmospheric loading of aerosols generated through human activities can have a radiative impact comparable in magnitude with that one of greenhouse gases (IPCC 2007). Accuracy and precision in estimating the optical properties of the atmospheric aerosols is therefore critical to effectively reduce the uncertainty in the aerosol effects on climate. In this context, knowing the partitioning of the atmospheric aerosol optical extinction (σ_ep) between light scattering (σ_sp) and light absorption (σ_ap) is required to understand the direct climate effects of aerosol particles. Accurate measurements of σ_ep, σ_sp, and σ_ap are often problematic due to the nonuniform spatial-temporal distribution and compositional variability of atmospheric particles. An additional challenge is represented by the need of defining the wavelength dependence of such aerosol properties for the correct representation of aerosols in climate models and for the validation of satellite retrievals (Anderson et al. 2005; Bates et al. 2006; Yu et al. 2006).

The development of instruments that can adequately measure aerosol optical properties is a key to precisely quantify the direct radiative effect of atmospheric aerosols by reducing the uncertainty in estimates of particle single scattering albedo ω (Schwartz 2004), where ω = σ_sp/σ_ep. The σ_sp coefficients are usually measured by nephelometry (Heintzenberg and Charlson 1996). Values of σ_ap can be directly obtained via commercial filter-based techniques such as particle soot absorption photometer (PSAP), multi-angle absorption photometry (MAAP), and aethalometers (Bond et al. 1999; Petzold et al. 2005; Weingartner et al. 2003, respectively) as well as by more recent photo-acoustic systems (Arnott et al. 1999; Lack et al. 2006). Cavity ring-down (CRD) is an accurate and sensitive technique (Brown 2003) and in the last few years several custom-built instruments employing CRD have been developed and deployed for direct measurements of aerosol particle σ_epcoefficients (e.g., Strawa et al. 2003; Moosmuller et al. 2005; Baynard et al. 2007; Massoli et al. 2009a). Research grade, custom-built CRD-based instruments are rather complex and, to the best of our knowledge, there are no other commercially available techniques that are capable of providing a comparable level of sensitivity for direct σ_ep measurements. At the same time, the applicability of indirect methods for estimating σ_ep (e.g., via sum of σ_sp and σ_ap) is limited by the significant sources of bias that characterize most of the commonly used optical techniques for the measurement of both the σ_sp and σ_ap coefficients (Lack et al. 2008; Cappa et al. 2008; Bond et al. 2009; Massoli et al. 2009b).

Here we present aerosol σ_ep coefficient measurements obtained using a recently developed instrument that uses Cavity Attenuated Phase Shift (CAPS) technology (Herbelin et al. 1980; Herberlin and McKay 1981). This method has already been already successfully employed for measurements of ambient NO₂ concentrations (Kebabian and Freedman 2007; Kebabian et al. 2008). The CAPS instrument uses a square wave modulated light emitting diode (LED) as a light source and measures the change in the phase shift of the distorted waveform of the modulated light leaving a highly reflective optical cell. The aerosol extinction σ_ep is derived from such change in the phase shift.

This work aims to provide an initial validation of the CAPS extinction monitor in laboratory and field environments. The laboratory measurements are performed at wavelengths of both 445 nm and 632 nm using monodisperse polystyrene latex (PSL) spheres. The field data are from the 2009 Study of Houston Area Radical Precursors (SHARP) campaign in Houston, TX, and the 2009 Queens College summer study in New York City, NY. In both field studies the CAPS instrument was deployed in the Aerodyne Research, Inc. (ARI) mobile laboratory (Herndon et al. 2005), where a co-located MAAP instrument measured the σ_ap coefficient at 635 nm. Such setup allowed coupling the CAPS-based σ_ep collected at 632 nm with the MAAP-based σ_ap data to provide the ambient particle ω.

Our initial results presented in this paper show that the CAPS extinction monitor is capable of providing state-of-the-art performance while dramatically reducing the complexity of optical instrumentation for directly measuring the σ_ep coefficient of atmospheric aerosols.

2. INSTRUMENTS AND EXPERIMENTAL METHODS

2.1. CAPS Extinction Monitor

The principal components of the CAPS extinction monitor and a simplified schematic of such technique are depicted in Figure 1.

FIG. 1 Schematic representation of the main components of the CAPS extinction monitor (LED, optical cell, vacuum photodiode). Both the square and distorted waveform (before and after modulation, respectively) are shown at the left and right end of the optical cavity.

The CAPS extinction monitor consists of a 26 cm long near-confocal optical cavity, with high reflectivity (R > 99.98%), or “low loss,” mirrors located at both ends. At this level of mirror reflectivity, the effective path length within the cell is approximately 2 km. Each mirror is maintained under a constant particle-free air flow of ∼ 12 cm³ min^–1 (purge flow) to keep the aerosol sample away from the mirror's reflective surface. The total aerosol sample flow through the cell is determined and kept at the constant value of 0.85 lpm by a critical orifice.

Square wave modulated light from a broadband LED (Lumileds) enters the optical cavity through one mirror, and exits at the other mirror as a “distorted” wave with respect to the original waveform; the phase shift is 35–40° at the modulation frequency. The light exiting the cavity is collected by a vacuum photodiode detector. A 10 nm wide band-pass filter, located in front of the detector, is used to define the spectral range of the measurement. The relationship between the measured phase shift and light extinction σ_ep is described in Equation (1):

where θ is the measured phase shift, θ_∘ is the phase shift under particle-free air conditions, c is the speed of light, and f is the modulation frequency. A more detailed description of the CAPS technology is given elsewhere (Kebabian et al. 2007, 2008).

Regular switching between aerosol sample flow and particle-free air is performed to ensure good instrument performance; in fact, variations in the instrument response (i.e., cotθ_∘) can occur via baseline drift and because of the contribution of gas phase absorption to the total extinction. Usually, baseline drift is minimal and it is related to changes in the temperature of the sampling cell, whereas gas phase absorption (mainly represented by NO₂ in the visible range) can be significant in polluted environments (Baynard et al. 2007). The CAPS-based σ_ep coefficient is also corrected for the fact that the aerosol sample is restricted to a limited fraction of the optical cavity due to the dilution of the sample flow by the mirrors' purge flow; we correct for such dilution with a multiplicative factor obtained as the ratio of the slopes of the extinction values measured without and with purge flow at varying amounts of nitrogen dioxide (NO₂). The approximation used for this correction, i.e., that gas and particle flow behave identically in the cell, is the limiting factor affecting the accuracy of the σ_ep measurements. The CAPS monitor operates at ambient pressure; pressure and temperature in the cavity are continuously monitored and used to correct for short term fluctuations in the gas phase Rayleigh scattering.

The simple CAPS design allows for easy wavelength conversion by changing the wavelength-specific components, i.e., the LED, high-reflectivity mirrors and band-pass filter. In this study we report data separately obtained with a single cavity CAPS instrument at both 445 nm and 632 nm for the laboratory validation tests, and at 632 nm for the field deployments.

2.2. Laboratory Measurements

The performance of the CAPS extinction monitor was tested in the laboratory by using non-absorbing polystyrene latex (PSL) spheres (Duke Scientific Corp., Palo Alto, CA). The PSL particles were generated using a Collison-type atomizer (Collison 1935) and dried to ≤ 25% relative humidity (RH) in a tube filled with silica gel. The PSL diameters covered the submicron range from 200 to 1000 nm (nominal particle diameter). A custom Differential Mobility Analyzer (DMA) was used to eliminate any particle agglomerates that were contained in the generated particle flow stream. A condensation particle counter (CPC, TSI, 3022A) was used to determine the particle number density. The CPC measures the total number of particles larger than 7 nm and it has a stated uncertainty of ± 10% in the reported particle number concentrations (CN) at the flow of 300 cm³ min^–1 used in this study. In order to minimize statistical errors, both the CN and σ_ep measurements were integrated for 120 s.

2.3. Field Measurements

The CAPS extinction monitor was deployed outside of the laboratory environment for the first time in 2009 during the Study of Houston Area Radical Precursors (SHARP) in Houston, TX and the Queens College study in New York City. In both cases, the instrument was located on board the ARI mobile laboratory, a facility equipped with a suite of gas phase monitors as well as aerosol counting and sizing instruments (e.g., CPC, DMA) and a MAAP (Thermo Carusso, Model 5012) for measurements of the σ_ap coefficient. The spectral band-pass of our particular MAAP model had a center wavelength of 635 nm (as measured using an Ocean Optics, Inc. diode array spectrometer) hence virtually overlapping the CAPS wavelength of 632 nm.

The MAAP collected data at 3 s time resolution; the data were processed using the algorithms based on Petzold and Schönlinner (2004) and Petzold et al. (2002, 2005). Such algorithms account for the radiation transmitted and scattered from the MAAP filter at multiple detection angles, and aim to correct for possible artifacts due to multiple reflection between the filter matrix and the aerosol layer deposited on the filter. Petzold et al. (2005) report a detection limit for the MAAP-based σ_ap coefficient of ∼ 1.5 Mm^–1for an integration time of 120 s, with an overall error of ± 12% for both laboratory and ambient aerosols. Based on Petzold et al. (2005) we assume that the uncertainty in our MAAP instrument is approximately 12%, although other studies (Slowik et al. 2007) have reported discrepancies up to 20% between MAAP and PAS techniques for aerosol absorption measurements of soot aerosols. Unfortunately, during our field deployments the lack of other co-located σ_ap measurements did not allow for further testing of the MAAP-based σ_ap via intercomparison exercises. Despite its uncertainty, the MAAP is considered the most reliable filter-based method for aerosol σ_ap coefficient measurements (Andreae and Gelècser 2006).

In the ARI mobile laboratory, MAAP and CAPS sampled aerosols using a common inlet line. Aerosols were drawn through a cyclone impactor (URG Inc., 2000-30EN) with a nominal aerodynamic cutoff diameter of 2.5 μ m at a flow rate of 10 lpm. At the actual MAAP flow rate of ∼ 17 lpm (of which 0.85 lpm was used for the CAPS monitor) the nominal cutoff diameter was reduced to 2 μ m. All measurements were performed at ambient temperature and RH. We are not able to make any particular evaluation of MAAP performance and response at ambient RH conditions; therefore we assume that possible artifacts due to, e.g., water deposition on the MAAP filter are accounted for by the multiple scattering correction algorithms mentioned above.

3. RESULTS

3.1. Instrument Precision and Stability

The stability of the CAPS monitor was tested by means of an Allan analysis (Allan 1966). The σ_ep coefficient was recorded continuously in the laboratory for about 25 h at 445 nm and 7 h at 632 nm under a flow of dry particle-free air to provide an accurate characterization of measurement noise and drift over relatively long integration times. Measurements were done at 445 nm first and then at 632 nm by changing the wavelength specific components, i.e., the LED, high-reflectivity mirrors and bandpass filter. Figure 2 shows only the results of the Allan analysis obtained with particle-free air at 632 nm, i.e., the field version of the extinction monitor. The results for 445 nm are reported in the text.

FIG. 2 Allan analysis of the CAPS extinction monitor at 632 nm. The root mean square (rms) noise calculated as a function of the sampling integration time is 0.8 Mm^–1 at 1 s, 0.08 Mm^–1 at 100 s, and 0.16 Mm^–1 at 1000 s.

The raw time series data (σ_ep coefficient vs. acquisition time in hours) are shown at the top of Figure 2, whereas the root mean square (rms) noise calculated as a function of the sampling integration time in seconds is shown at the bottom. The dashed black line displays the ideal rms noise behavior (or “white” noise), which decreases linearly with the square root of the integration time. For a sampling period of 1 s, the measured rms noise represented by the y-axis intercept is 0.8 Mm^–1 at 632 nm (1.1 Mm^–1 at 445 nm). The rms noise initially decreases as the square root of the integration time, reaching a floor level of 0.08 Mm^–1 for 632 nm at ∼ 200 s (0.12 Mm^–1 for 445 nm). At this point, the noise begins to increase because of baseline drift which is mainly caused by small changes in the temperature of the instrument and only incidentally by minor changes in the mirror reflectivity: such changes have no effect on the measurement sensitivity or accuracy. By the end of the period shown in the figure (1000 s), the rms noise has increased only slightly (up to 0.16 Mm^–1). In fact, under typical operating conditions, the level of baseline drift in 1 hour is far less than 1 Mm^–1. Higher values of precision can be achieved using longer integration times and more frequent baseline measurements. At 60 s integration time and for a 50% zeroing-50% sampling duty cycle, a limiting precision of ∼ 0.2 Mm^–1 in the measured σ_ep can be reached.

3.2. Laboratory Results

The CAPS performance was tested in the laboratory with non-absorbing PSL spheres. Monodisperse particles were generated in the size range 200–1000 nm, selected by the DMA, and sampled through the CPC and CAPS as described in section 2.2. By dividing the measured CAPS σ_ep by the CPC-based CN, we first obtained the particle extinction cross section (σ_ext) for each monodisperse size. Figure 3 shows the σ_ep measured at 632 nm plotted versus CN for PSL particles of diameter 240 nm, 500 nm, and 800 nm, respectively (the other tested PSL particle sizes show similarly linear behavior and are not depicted). The σ_ext values are derived from the slopes of the linear fits to these data. The 2σ precision of σ_ext obtained as the average values of the linear fits for all PSL sizes is 3.2%.

FIG. 3 σ_ep coefficient (Mm^{− 1}) measured at 632 nm plotted versus the particle number concentration CN (cm^{− 3}) for 240 nm, 500 nm, and 800 nm non-absorbing PSL particles. The slope for each linear fits represents the optical cross section (σ_ext) of each PSL size. The average σ_ext precision (2σ) is 3.2%.

The extinction efficiency (Q_ext) is then obtained as the ratio of the particle σ_ext to the known geometric cross section (σ_geom). These experimental Q_ext values are compared in Figure 4 to those calculated with Mie scattering theory for the specific wavelength, using a real part of the refractive index (n) of 1.597 and an imaginary part (k) of 0.00 (Mätzler 2002). Figure 4a shows the Q_ext values derived at 632 nm: the results at 445 nm are shown in Figure 4b. In both cases, we report the average of 2 or 3 values of Q_ext obtained for each particle size. The size of the error boxes represents the uncertainties of 3.2% in the Q_ext based on the σ_ext precision (vertical dimension) and 3% in the mean PSL size as stated by the PSL supplier (horizontal dimension). The solid lines are the Q_ext calculated from Mie theory. The uncertainty region around the Mie-based Q_ext value (dashed lines) is defined based on the level of accuracy in the particle number density measured by the CPC (± 10%). Our data show excellent closure with Mie scattering theory, with better agreement at 632 nm and for particle sizes > 400 nm. Larger discrepancies between theory and experimental Q_ext values for small particles and at lower wavelengths have been also observed in similar laboratory tests that employed CRD-based instruments (Pettersson et al. 2004; Abo-Riziq et al. 2007; Baynard et al. 2007). A possible explanation for such discrepancies could be the high sensitivity of the Mie scattering calculation to the index of refraction assumed for the PSL particles (Abo-Riziq et al. 2007). Overall, these results show that the CAPS monitor performance in the laboratory is both qualitatively and quantitatively comparable to CRD instruments for the same materials and methods.

FIG. 4 Extinction efficiency (Q_ext) of non-absorbing monodisperse PSL particles in the size range 200–1000 nm at 632 nm (panel a) and at 445 nm (panel b). Each data point is 1 to 3 min average. Errors boxes are 3.2% in the y-axis and 3% in the x-axis.

3.3. Field Measurements

During SHARP (Houston, 20 April to 30 May 2009) the ARI Mobile Laboratory performed day-round monitoring of industrial pollutants in the Houston area and characterization of vehicular emissions along major highways. The CAPS was deployed in SHARP during May 20–25, 2009. During this 5-day period, ambient RH ranged between 75 and 85% depending on the time of the day, and changed on a day-to-day basis with the meteorological conditions. Figure 5a shows the 1 min average time series of ambient CAPS-based σ_ep(solid thin line), MAAP-based σ_ap (solid thick line), and particle ω (dashed line) obtained combining the σ_ep and σ_ap coefficients (ω (σ_ep−σ_ap)/σ_ep) for a period of 6 hrs (0400 to 1000, CDT) on May 22. Prior to 0530 and after 0830, the ARI laboratory was stationed at a rural site located in Baytown, TX: during this time, overall aerosol loading is typical of a moderately polluted urban background with σ_ep∼ 40 Mm^–1 and σ_ap∼ 5 Mm ^–1, producing an average ω of 0.88. Periods of σ_ep and σ_ap values above background (0530–0830) pertain to times when the ARI mobile laboratory sampled on-road fresh vehicular emissions, for which aerosol particles have usually a strong light-absorbing character. Each single “plume hit” is characterized by decrease in ω to values as low as 0.6. Figure 5b is a snapshot of 3 s time resolution measurements of σ_ep reported on the same day between 0525 and 0529, during which at least two major traffic plumes were encountered. Along with the real time σ_ep data we show real time measurements of carbon dioxide, CO₂ (ppm), measured by a commercial instrument (LI-COR Biosciences, Model 6262) that has a time response of ∼ 3 s. The CO₂ concentration and the σ_ep time series track each other with excellent qualitative agreement, showing that the same “plume” features are captured by both instruments. These data demonstrate that the CAPS performance is satisfactory for the acquisition of fast time response data at any σ_ep level, and therefore can potentially be used for detailed characterization of point sources as well as the determination of extinction emission factors (a subject outside the scope of this article).

FIG. 5 (a) 60 s time average data of the CAPS σ_ep (thin solid), MAAP σ_ap (thick solid), and particle ω (dashed) between 0400 and 1000 on May 22, 2009 during SHARP. On road traffic emissions were measured approximately between 0530 and 0830. (b) is a 3 second data snapshot of σ_ep (thick solid) and CO₂ (thin solid) of vehicular exhaust plumes between 0525 and 0529.

The CAPS extinction monitor was again deployed on-board the ARI mobile laboratory during the Queens College Air Quality study that took place in Queens, New York, in July 2009. During 3 weeks of field deployment, the ARI mobile laboratory was mostly stationed at Queens College which is approximately half a kilometer south of the Long Island Expressway (LIE, I-495). The LIE is the major east-west motorway on Long Island and is characterized by heavy traffic of both gasoline and diesel powered vehicles (> 200,000 vehicles per day in Queens). One of the goals of this study was to determine how concentration of gaseous and aerosol species as well as particle optical properties vary with distance from highway and time of the day for both downwind and upwind locations. Characterizing pollution gradients in near-highway air is the object of increasing attention as recent epidemiological studies have shown that people who live within ∼ 100 m of highways are exposed to factor 2 higher amounts of pollutants than those living elsewhere (Zhu et al. 2002; Zhang et al. 2004a, 2004b, 2004c). We performed gradient studies in the time frame 0430–1000 (local time) to characterize the pollution build-up during the morning traffic rush hours under shallow and stable boundary layer conditions. The mapping of the spatial evolution of pollutants downwind the LIE was accomplished by driving the ARI mobile laboratory at a constant speed of 10 km h^–1; a complete downwind transit took approximately 45 minutes. Data were continuously recorded by all instruments; periods contaminated by self sampling were eliminated from the data set using wind direction data.

Figure 6 shows the spatial distribution and temporal evolution of particle ω values (obtained from σ_ep and σ_ap) for both neighborhoods upwind (south side) and downwind (north side) of the LIE, indicated by a solid black line. We show the data collected on July 28, 2009. On this day, the prevailing S-SW winds were light (∼ 2 m s^–1); ambient RH was 88% at 0500, and 75% at 0700. The lowest ω values were generally measured within less than 50 meters from the LIE at any given time during the experiment. Between 0500 and 0545 (Figure 6a) we observed that ω was ∼ 0.55 by the LIE, 0.7 between 50 and 250 meters away from LIE and 0.85 between 250 and 500 meters further downwind. Between 0615 and 0700 (Figure 6b), the ω values had markedly decreased everywhere: ω was < 0.5 directly downwind of the LIE, and averaged 0.72 everywhere else (with the exception of few single vehicle hits with ω < 0.7). This behavior is ascribed to the buildup of light absorbing particles associated with vehicle emissions during the early morning hours before vertical mixing becomes significant.

FIG. 6 Spatial distribution and temporal evolution of particle ω values for both upwind and downwind sides of the Long Island Expressway (LIE) on July 28, 2009 between 0500–0545 (a) and 0615–0700 (b). The lowest ω values were measured directly downwind of the LIE. The arrows in panel b indicate the direction of the ARI mobile laboratory as it transited through the downwind and upwind neighborhoods.

These data demonstrate that the combination of CAPS-based σ_ep and MAAP-based σ_ap provided meaningful results during this particular study. Based on the average error stated for the σ_ep and σ_ap coefficients, the average absolute uncertainty in the ambient ω data obtained by combination of CAPS and MAAP was calculated to be approximately 0.06 for ω values comprised between 0.7 and 0.99, under the assumptions that σ_ap coefficient is not a function of RH (Nessler et al. 2005) and that its uncertainty does not change at ambient conditions. At ω values lower than 0.5, the uncertainty would increases significantly due to the relatively high error in the MAAP measurements. However, our ω values reported for on-road vehicular emission measurements either in Houston or in New York City compare well with previous studies that measured ω at rush hour conditions nearby major city highways (Massoli et al. 2009a).

4. CONCLUSIONS

We have presented laboratory and field measurements of the aerosol particle light extinction coefficient (σ_ep) using the recently developed CAPS-based extinction monitor. Our results showed that CAPS provides accurate, low-noise σ_ep measurements with fast time response. Measured detection limits (3σ) for σ_ep are on the order of 3 Mm^–1 at 1 second time resolution for center wavelengths of 445 nm and 632 nm. Measured absolute extinction cross sections of monodisperse, nonabsorbing polystyrene latex particles are in excellent agreement with Mie theory, and compare well with results from previous studies employing CRD-based techniques. Two different field deployments involving measurements on mobile platforms have demonstrated that the CAPS-based extinction monitor is a robust instrument; it is capable of operating and maintaining its optimal performance under non-ideal conditions and is therefore suitable for use in a variety of applications.

The development of a CAPS monitor which incorporates two cells each simultaneously operating at a different wavelength is underway: such a monitor could be used to effectively determine the size-dependent Ångstrom exponent, Å (Ångstrom 1929). The importance of measuring the Å parameter is known; in addition, the potential use of in situ measurements of Å to quantify the column fine mode fraction of atmospheric aerosols has been emphasized recently (Atkinson et al. 2009). Overall, the use of the CAPS monitor for the direct measurement of σ_ep would benefit ground-monitoring networks where aerosol light extinction is typically calculated from the sum of scattering and absorption measurements obtained using commercial nephelometers and PSAPs, respectively. Such an application would represent a dramatic step forward in providing low uncertainty measurements of particle single scattering albedo ω especially if a CAPS-based extinction measurement could be coupled on a systematic basis to a relatively precise light absorption instrument as the photo-acoustic spectrometer.

The authors thank Dr. Scott Herndon and Dr. John Jayne, respectively, for organizing the Aerodyne efforts at the SHARP campaign and the Queens College summer study. We also acknowledge Drs. Ezra Wood, Edward Fortner, Leah Williams, and Nga L. Ng for their help in the two field measurement campaigns. We also thank both the Department of Energy and the National Aeronautics and Space Administration for providing financial support through the Small Business Innovation Research program.