1. Introduction

Anthropogenic atmospheric aerosols affect earth’s radiative balance by scattering and absorbing solar radiation. The climate ramifications of anthropogenic aerosol particles stem from disrupting the “natural” pattern of light absorption at earth’s surface and in the atmosphere. The climatic effects of atmospheric aerosols are believed to include decreasing surface temperatures and altering regional precipitation patterns [1].

Unfortunately, current estimates of the magnitude of aerosol radiative forcing are highly uncertain. Schwartz [2] has recently considered estimates of aerosol forcing and corresponding uncertainties that have been published in a recent Intergovernmental Panel on Climate Change (I.P.C.C.) report and has concluded atmospheric aerosols exert a climate forcing of -1.6 ± 1.1 W/m². This range of values suggests aerosols may offset greenhouse gas warming by ≈20% to 110%. Clearly, at present the exact magnitude of the aerosol’s radiative effect is highly uncertain. As such, the Intergovernmental Panel on Climate Change (I.P.C.C.) classifies the current level of scientific understanding of aerosol radiative forcing as “very low – low [3].”

Development and application of novel, in situ measurement technologies to detect and characterize aerosol optical properties in real time is an important step towards a more accurate, quantitative understanding of the aerosol layers influence on earth’s radiative balance. Aerosol scattering and extinction coefficient (b_scat and b_ext), and the ratio between these variables (single scatter albedo, ω)

are key optical parameters which, if known, allow improved estimation of direct climate forcing by aerosols dispersed in earth’s atmosphere. Precise measurement of these variables can prove challenging since aerosol scattering and extinction coefficients are typically < 100 Mm-1 in air masses not heavily loaded with particulates.

Historically, single scatter albedo (ω) is usually inferred from measurements of aerosol scattering coefficient (b_scat) and absorption coefficient (b_abs) made with separate instruments. Since aerosol extinction coefficient (b_ext) is simply the sum of aerosol absorption and scattering coefficients (b_ext=b_abs+b_scat), aerosol albedo can be determined through this approach. Aerosol absorption coefficient (b_abs) is often estimated using a particle soot absorption photometer (P.S.A.P.), aethalometer, or through photoacoustic spectroscopy while integrating nephelometers are commonly employed for determination of aerosol scattering coefficient (b_scat). Use of these instruments in concert allows estimation of aerosol albedo; however, both methods suffer from certain well-known measurement limitations. For instance, the P.S.A.P. instrument is a filter-based technique that fundamentally cannot be used to measure the effect of ambient relative humidity on aerosol absorption coefficient. Additionally, there is some scientific debate over the accuracy of absorption coefficient measurements made with filter based techniques since the microphysical properties of the aerosol necessarily change when collected on a filter. However, available empirical evidence suggests the P.S.A.P. offers acceptable performance under certain conditions of measurement [4, 5]. Recently, photoacoustic spectroscopy has also been applied to measurement of aerosol absorption coefficient [6]. This technique offers an attractive alternative to filter based techniques for the determination of b_abs since the sample can remain dispersed throughout the measurement.

While nephelometry for determination of aerosol scattering coefficient is a mature technique, all nephelometers are not able to collect light scattered from an aerosol sample at angles very close to 0° (forward scatter) and 180° (backscatter) due to geometrical constraints [7-10]. This truncated measurement can lead to underestimation of scattering coefficient, particularly for particles with dimensions that are large with respect to the wavelength of the probe beam. For commercial nephelometers, truncation angles of 5 – 15° are encountered and a significant fraction of light diffracted by particles with large size parameters goes undetected [9]. Correction factors can often be employed to help account for light scattered by the aerosol that is not detected. However, the magnitude of the error is dependent upon the particle size distribution of the aerosol. This requires prior knowledge (or estimation) of particle size distribution for this approach to work. Nonetheless, reducing this truncation error is an important step towards making more accurate measurements of aerosol scattering coefficient.

Two significant recent advances in the measurement of aerosol optical properties have been the application of cavity enhanced spectroscopy for extinction measurements [11-15] and the development of the integrating sphere nephelometer [16,17]. The highly sensitive cavity enhanced techniques offer the requisite sensitivity for measurement of b_ext as they typically are capable of detecting optical losses of only a few parts per million (or less) over a meter path length. The instrument footprint is also reduced to a bench-top scale allowing single point measurements and significant potential for portability. Conversely, the geometry of the integrating sphere nephelometer allows a significant reduction in nephelometer truncation angle when compared to commercial nephelometers. In this approach, a laser beam is directed through a hollow sphere which contains the aerosol sample. The inner surface of the sphere is coated with a diffuse reflector (often BaSO₄) with a high reflectivity (>0.95) in the spectral region of interest. A light detector is then placed on the surface of the sphere and a light baffle used to prevent light from being scattered directly into the detector’s field of view. Laser light scattered by particulates is collected by the sphere and a portion of it reaches the detector through multiple reflections off the sphere’s surface. The user can obtain the aerosol scattering coefficient through calibration with samples of known scattering coefficient. The main advantage of the integrating sphere geometry is that it can significantly decrease the angle of truncation. For instance, the nephelometer developed by Varma, Moosmuller and Arnott [16] at the Desert Research Institute has achieved a truncation angle of only 1° on average through use of strategically placed truncation reduction tubes and the integrating sphere design.

In this work, we have developed a new instrument for simultaneous measurement of aerosol scattering coefficient and extinction coefficient by combining the advantages of cavity enhanced extinction measurements and the integrating sphere nephelometer. In this approach, extinction measurements are made using cavity ring-down spectroscopy (CRDS) inside of a 46 cm diameter integrating sphere which contains the aerosol sample. The integrating sphere collects light that is scattered from the probe CRDS beam and the intensity of the scattered radiation is monitored during the ring-down event with a separate detector and used to determine scattering coefficient through a three point calibration with Rayleigh air, CO₂, and R-134a. The main advantage of the technique is the scattering and extinction measurements are made simultaneously on the exact same sample volume. This eliminates measurement errors that arise from separate measurement instruments having different particle sampling line losses. Additionally, the simultaneous measurements prevent the scattering and extinction measurements from being carried out at different temperatures. This is significant since small changes in temperature can lead to large changes in aerosol optical properties for hygroscopic particles, particularly at high relative humidity. A review of the literature reveals a similar method presented for making simultaneous scattering and extinction measurements on atmospheric aerosols within an optical resonator [18]. The instrument described herein differs from that work in wavelength of measurement, and offers improvements in truncation angle and collection of a larger fraction of scattered light.

2. Materials & methods

2.1 Description of the aerosol albedometer.

Figure 1 illustrates a schematic of the albedometer.

 

Fig. 1. Schematic of the Aerosol Albedometer. Laser light that is scattered from the CRDS beam is collected by the integrating sphere and measured with the scattering photomultiplier tube. Both the measured extinction and scattering originate from the exact same sample volume. Temperature/relative humidity probe and pressure sensor not shown for clarity.

Download Full Size | PDF

The device consists of a 46 cm diameter integrating sphere fabricated from 2 stainless steel hemispheres. An airtight connection between the hemispheres was achieved through use of an orbit flange at the equatorial position, an o-ring, and a series of clamps. The internal volume of the sphere was ≈50 L. All inner surfaces of the sphere were coated with Duraflect®, a proprietary diffuse reflectance coating applied by Labsphere, Inc. that exhibits a reflectance >95% between λ=350 – 1200 nm. A 0.635 cm hole is present at each pole of the sphere to allow passage of the probe laser beam. Two mirror mounts with gas purge inlets were centered on each of the polar holes and permanently affixed to the sphere. The gas purge lines had two purposes: 1) to fill the sphere with calibration gases when desired or 2) to purge the mirror housings with filtered air during aerosol experiments to prevent particulate deposition on the mirror surfaces [12, 13]. Sensors for pressure (MKS Instruments 722A13TCD2FA), and relative humidity/temperature (Vaisala, HMP50) were added to the sphere through air-tight ports machined on the sphere’s surface. Typical pressure inside the sphere was ≈700 mm Hg and temperatures ranged between 20 – 23° C. A 2.54 cm wide light baffle fabricated from stainless steel was placed in an arc within the interior of the sphere. The baffle was also coated with Duraflect coating. The baffle’s position was strategically chosen to block the laser beam from being scattered directly into the detector’s field of view. This assures all light collected by the scattering detector must have undergone at minimum 1 reflection off the sphere’s surface prior to being detected.

2.2 Sample inlet & flow system.

Aerosol samples were added to the sphere through an 80 cm length of 1.27 cm diameter stainless steel tubing aligned vertically to maximize particulate transport. A ball valve was placed in this transfer line to switch the aerosol flow on/off as desired. Linear flow velocity within the inlet tube was 50 – 80 cm/sec during aerosol sampling. Aerosol transport efficiency through this inlet tube was modeled using Deposition 2001a software [19] and measured using a real time aerosol monitor (Mie Inc.). Modeling suggested aerosol transport efficiencies >98% for particles smaller than 10 µm diameter, and this high transport efficiency was confirmed experimentally for polystyrene size standards between 0.1 – 2.1 µm diameter. Therefore no efforts to correct for particulate losses in sampling lines have been made in this study. The sample outlet was a Swagelok type fitting welded to the sphere that was connected to a 50 cm length of 0.95 cm diameter plastic tubing that was routed to a Radiance Research M903 integrating nephelometer. This nephelometer was used to provide reference measurements of aerosol scattering coefficient (b_scat) throughout our experiments. An additional length of tubing led from the nephelometer’s output port to a 47 mm glass fiber filter, mass flow controller and air sampling pump. During aerosol experiments, the mass flow controller was set to a flow rate of either 4 or 6 liters per minute (LPM). Gas calibration experiments were conducted by flushing the cell with test gases at several LPM for 20 – 60 minutes, and then slowing flow to 0.5 LPM to make optical measurements. During aerosol experiments, a continuous filtered air flow rate of 0.5 LPM was used through the mirror purge lines. This proved sufficient to eliminate particle deposition on the high-reflectivity mirrors.

2.3 Generation of aerosol samples.

Polystyrene, India ink, and ammonium sulfate aerosol samples were generated through a series of steps. First, a solution of either polystyrene microspheres (diam.=0.202, 0.771, 1.826 µm, Polysciences Inc.), India ink, or ammonium sulfate is prepared in filtered, deionized (18.2 MΩ/cm) water. Typical mixing ratios were 1 or 2 drops of the size standard solution or India ink per liter of water and typical concentrations of the ammonium sulfate solution was ≈0.2 mM. The resulting solution was then atomized with a TSI 9302 atomizer and the resultant aerosol passed through an electrically heated tube to de-solvate the aerosol. The sample was then passed through a diffusion dryer containing silica gel dessicant and into a 67 L mixing chamber where it could be diluted with a flow of clean, dry air if desired. The sample then passed through a second diffusion dryer (TSI 3062) and into the albedometer through the sample inlet line described earlier. It was often noted that room air pumped through the diaphragm pump used to operate the atomizer exhibited a residual absorption (approx. 5 Mm^-1) that was not observed in air from a compressed gas cylinder. The exact cause of this absorption has not been assigned. However, the average absorption coefficient was noted and subtracted from the observed aerosol extinction when particulates were present in our measurement chamber.

We have also made measurements on the aerosol plume resulting from combustion of dry pine needles. To generate this aerosol, pine needles were burned inside of a metal cylinder in the open atmosphere outside of our laboratory. The resultant aerosol plume was sampled through a window via a 0.95 cm diameter plastic tube and passed directly to the albedometer. To eliminate the influence of the ambient aerosol, background readings of the ambient aerosol were made prior to igniting the pine needles. Average background readings for b_scat and b_ext were tabulated and these averages subtracted from all subsequent values after ignition of the pine needles. No effort was made to control combustion conditions during our trials as this was deemed beyond the scope of this work.

2.4 Optical setup.

We have employed a frequency doubled (λ=532 nm) Nd:YAG laser (Continuum Lasers, Minilite I) pulsed at 15 Hz for our measurements. This laser is rated by the manufacturer for maximum pulse energies of 12 mJ contained within 3-5 ns pulses. We have found this laser emits considerable residual power at the 1064 nm line and this light was removed from the beam through use of a visible band-pass filter. The CRDS cell employed a symmetrical optical resonator formed from two 6 m radius of curvature, 2.54 cm diameter high reflectivity mirrors (R≈0.999935). The mirrors were placed 50.5 cm apart. The CRDS signal was monitored by a standard photomultiplier tube (931B, Hammamatsu) mounted in a light–tight metallic shield fastened to the output mirror housing. The photomultipler bias voltage was -912 V. In order to avoid saturation of the photomultiplier, the CRDS output beam was attenuated by a 3.0 O.D. neutral density filter prior to measurement.

Light scattered by the aerosol sample was collected by the inner surface of the integrating sphere and a fraction of this light was directed through a 532 ±10 nm band-pass filter. The intensity of this scattered light was monitored through use of a separate photomultiplier tube (931B, Hammamatsu) mounted in a housing (Newport, 70680) that was affixed to the surface of the sphere. This photomultiplier was biased at -800 V with a regulated power supply (Stanford Research Systems, PS325).

2.5 Data acquisition & analysis.

Data on both the CRDS and scattering channel were simultaneously acquired by a 2 channel, 14 bit, 200 Msample/s digitizer card (Gage Applied, Compuscope 14200) installed in a 3.40 GHz Pentium D computer running Windows XP Professional. Software written in Labview (National Instruments, version 7.1) was used to process data from the fast DAQ card, and to collect temperature, relative humidity, and pressure data through a separate USB data acquisition device (National Instruments, NI USB-6008). Typically, 100 individual ringdown transients were averaged in the software to generate each measurement of extinction coefficient, scattering coefficient, and albedo. Acquisition of data was triggered by the rising edge of the Q-switch sync line of the Nd:YAG laser. Once acquired, the Labview software also fit the CRDS ring-down transients to a first order exponential decay on-the-fly using the non-linear least squares curve fitting feature of Labview. This allows for rapid extraction of the cavity time constant (τ) and determination of aerosol extinction coefficient as described by Eq. (4) below. In addition, the software is programmed to divide the scatter channel signal by the CRDS channel signal (I_S/I_RD) at every point along the ring-down curve. The average scattering/CRDS ratio is then computed over 1500 data points and returned to a front panel indicator in the software. This ratio was found to increase linearly with the value of the scattering coefficient (b_scat) for test gases and is the variable measured to determine scattering coefficient. Use of this ratio for measurement of b_scat also has the built-in advantage of accounting for pulse-to-pulse fluctuations in laser intensity.

2.6 Measurement of aerosol extinction (b_ext).

Cavity ring-down methods measure the time constant (τ) for the attenuation of a laser pulse after its introduction into a stable optical resonator. The attenuation follows first-order decay kinetics and is characterized by a time constant (τ), the time required for the beam intensity to reach 1/e of the original intensity. This time constant (τ) can be related to the extinction coefficient of the sample placed between the mirrors by:

where t_r is the round trip transit time for the beam, R is the mirror reflectivity, L is the cavity length, and b_ext is the sample’s total extinction coefficient. If cell geometry and mirror reflectivity are held constant, the aerosol sample extinction coefficient can be extracted through knowledge of the ring-down time constant observed for the sample (τ_sam), the ringdown time when filtered air fills the cell (τ_air), and the following expression:

It should be noted that Eqs. (2) and (3) yield the spatially averaged value of b_ext over the region between the CRDS mirrors. For the measured value of b_ext to be accurate, the sample must fill the region between the mirrors completely in a homogenous manner. When calibration gases are added to the sphere, the gas fills the sphere and the volume within the mirror mounts completely, and this condition is satisfied. However, when aerosol experiments are conducted the mirror housings are intentionally purged with filtered air to contain the aerosol within the sphere and prevent particulate deposition on the CRDS mirrors. Thus, the extinction coefficient reported through using Eq. (3) will be falsely small when compared with the true extinction of the aerosol sample. To correct for this effect we can modify Eq. (3) as:

where L_total is the distance between the CRDS mirrors, and L_sample is the distance over which the sample is contained. It should be noted that this approach yields results identical to those obtained through previous mathematical treatments of the same problem provided the extinction coefficient of the mirror housing purge gas is considered to be zero extinction [14]. The use of filtered air as the blank measurement in Eqs. (3), (4) sets this condition for our experiments and all extinction in excess of this must be due to aerosol extinction and/or gasphase absorption. Another consequence of this treatment is that all values of b_ext reported in this report have been air subtracted, that is air’s Rayleigh scatter contribution to the extinction coefficient has been removed.

2.7 Measurement of aerosol scattering coefficient (b_scat).

As previously mentioned, scattering coefficient was measured through monitoring the intensity of light scattered from the probe CRDS beam by a detector mounted on the surface of the integrating sphere. For quantitative analysis we have employed a method originally suggested by Strawa, et al. for making measurements of scattering coefficient within an optical resonator [18]. In this approach, the scattering coefficient (b_scat) is determined through measurement of the ratio between the scattering signal (I_S) and the cavity ring-down signal (I_RD) throughout the ring-down decay transient. Strawa and co-workers derived an expression to relate aerosol scattering coefficient to the intensity ratio:

where R is the reflectivity of the CRDS mirrors employed, L is the length over which the scattering medium is present, and K is an experimentally derived calibration constant which accounts for collection efficiency of the scattered light and any differences in response functions between the CRDS and scattering channel detectors. Since the mirror reflectivity (R), sample length (L), and the value of K are all constants for a particular instrument, these terms can be pooled to generate a new constant K’ and it becomes very clear that the magnitude of the ratio between the scattering and ring-down signals is a proxy of intra-cavity scattering coefficient (b_scat).

In addition, Eq. (7) suggests that the ratio (I_S/I_RD) scales linearly with scattering coefficient. Use of this approach to calibration is favorable for several reasons. For one, correction of the scattering signal intensity to account for source intensity fluctuations is built in to the equation. Additionally, it is possible to compute the ratio I_S/I_RD at many points along the decay transient leading to a signal averaging effect.

Unlike CRDS, measurement of b_scat requires a calibration at several points (at minimum 2) to determine the proportionality constant [1/K’ in Eq. (7)] that relates the measured ratio to scattering coefficient. In our experiments we have conducted this calibration by using 3 gases of known scattering coefficient (air, CO₂, and 1,1,1,2 tetrafluoroethane). Calibration is achieved by purging our measurement cell with calibration gases and then making measurements of the observed I_S/I_CRDS ratio. Averages are tabulated and plotted against each gases accepted b_scat value at our pressure and temperature. It should again be noted that both scattering by gases (Rayleigh) and by aerosols will contribute to the measured scatter signal. In order to express only the aerosol’s contribution we must subtract out air’s contribution to the scatter signal. We achieve this by assigning Rayleigh air a value of 0 Mm^-1 when preparing calibration lines (plot of observed ratio vs. b_scat). This practice is commonly encountered when calibrating integrating nephelometers. One important difference between this work and that of Strawa, et al. is that use of the integrating sphere allows collection of far more scattered light than the device described previously. The sphere approach also reduces truncation angle (described in section 3.1) when compared directly to the previous instrument.

3. Results & discussion

3.1 Estimation of truncation angle.

It is well established that integrating nephelometers cannot detect scattered light over all angles due to the optical geometry inherent to the device [7-10]. Rather, the collection of scattered light is truncated in the forward and backward direction. Truncation angles for commercial nephelometers often range between 5 - 15°. This incomplete collection of scattered light is of considerable importance given the scatter phase function of coarse particles is peaked in the forward direction, and a significant fraction of scattered radiation can go undetected, thereby underestimating scattering coefficient. Therefore, truncation angle is considered a key metric of a nephelometer and the truncation angle of our design must be evaluated.

A simple geometrical model allows calculation of truncation angle (θ) for a single direction for any particle present within our probe beam as:

where r is the radius of the axial hole in the sphere (r=0.3175 cm for our experiments), and d is the distance (cm) of the scattering particle from the hole of interest along the optical axis. From considering the geometry of the experiment and this equation it is clear the truncation angle in either direction varies with the particle’s position in the sphere. Figure 2 illustrates this concept graphically. The red and blue lines shown in Fig. 2 represent the truncation angle for a particle located at distance (d) from one of the axial holes. The blue line represents the truncation angle in the direction of the hole located at d=0 and the red line indicates truncation angle in the direction of the hole located at d=46 cm.

 

Fig. 2. Plot of truncation angle as a function of distance from an axial hole located at the pole of the sphere. Axial holes are located at d=0 and d=46 cm. As illustrated, truncation angle is highest for a particle which scatters light in the direction of an axial hole, from a location very near the same axial hole.

Download Full Size | PDF

As a scattering particle approaches one of the axial holes, truncation angle approaches 90° in the direction of the hole and approaches a minimum of 0.4° for the opposite direction. For particles contained within the central 30 cm of the sphere the truncation angle in either direction is under 3.2o with a minimum of ≈0.8° at the center. If we assume the aerosol is evenly dispersed throughout the sphere, the spatially averaged truncation angle in both directions is 2.4°. It is important to note that any computations of truncation error should not simply use the spatially averaged value of 2.4° but rather consider the actual truncation angle for particles at each location within the sphere. Nonetheless, the spatially averaged truncation angle compares favorably with the truncation angles obtained by commercial nephelometers (5-15°) so use of the integrating sphere geometry clearly offers a technical advantage in this regards. It should be noted that further improvement in truncation angle could be achieved if particulates were contained to the central region of the sphere through electrostatic or pneumatic means.

3.2 Calibration gas measurements.

Figure 3 illustrates the results of a typical calibration gas experiment for air, CO₂ and R-134a.

 

Fig. 3. Data traces obtained when calibration gases filled the measurement cell. The grey trace represents the CRDS data, the red trace is the scattered light signal, and the black trace is the I_S/I_CRDS ratio.

Download Full Size | PDF

The grey trace in this plot represents the CRDS decay waveform, the red data points denote the scattering channel signal, and the black data points represents the ratio between the scatter channel and the CRDS channel (I_S/I_CRDS) over 1500 data points. Several trends are apparent by considering this figure. First, the observed ring-down time constant (τ) decreases from 27.8 µs for Rayleigh air to 23.25 and finally 16.95 µs for CO₂ and R-134a. This trend is expected given the larger extinction coefficient of these gases as compared to air. An additional trend observed is the increase in the I_S/I_CRDS ratio from 0.145 for air to 0.201 and 0.320 for CO₂ and R-134a. This trend suggests the I_S/I_CRDS ratio is a suitable metric for determination of scattering coefficient of the sample placed within the sphere.

The calibration process also provides an excellent opportunity to assess the accuracy of the extinction measurement since the extinction coefficient is equivalent to Rayleigh scattering coefficient for non-absorbing calibration gases. The Rayleigh scattering coefficient for air at 532 nm, 700 mm Hg, and 22°C is 11.97 Mm^-1. Carbon dioxide has a scattering coefficient 2.61 times that of air, and R-134a has a scattering coefficient 7.25 times that of air [20]. After the multiplication and subtraction of Rayleigh air (setting air as 0), this yields expected effective extinction coefficients of 19.3 and 74.8 Mm^-1 for CO₂ and R-134a, respectively. Again, it should be noted that these values represent air subtracted extinction coefficients rather than absolute extinction coefficients for these gases. The average measured extinction coefficients observed over 15 measurements with our device were 19.3±3.5 and 75.9±1.3 Mm^-1 when CO₂ and R-134a filled our measurement cell. These values are reported as plus or minus 1 standard deviation (s). The expected extinction values for both gases fall within one standard deviation of our measured results.

3.3 Measurement of aerosols.

We have applied this instrument to the measurement of albedo for several test aerosols generated in the laboratory. For laboratory generated aerosols we have employed several different diameters of dielectric polystyrene particle size standards, pigmented India ink, and aqueous solutions of ammonium sulfate. Figure 4 illustrates the results obtained from a typical experiment when 1.826 µm polystyrene was the test aerosol.

 

Fig. 4. Plot of scattering coefficient, extinction coefficient, and albedo (second y-axis) over time during a monitoring experiment in which 1.826 µm polystyrene spheres were atomized and introduced into the measurement cell. The average albedo measured for these particles during this experiment was 0.99.

Download Full Size | PDF

In these experiments, filtered air initially filled our measurement chamber and background measurements were taken for 5 – 20 min. As previously mentioned, in many cases a residual extinction of (≈5 Mm^-1) was noted on the CRDS channel but not on the scattering channel when we used an air pump as the air source to run our atomizer. Given the observation of an increase in extinction without a corresponding increase in scattering we attribute this to molecular absorption. The exact cause of this absorption remains unclear, however we have subtracted the average extinction recorded when no particulates were present in the sphere from all subsequent data points for that run in order to report only aerosol extinction. After obtaining this initial baseline data, the atomizer containing the test solution was turned on and the aerosol was generated and added to the sphere as described in the methods section. This caused an increase in observed scattering and extinction coefficient as our measurement chamber filled with the aerosol sample as shown in the figure. After a sufficient number of data points were collected, the atomizer is turned off, and filtered air used to flush the measurement chamber. As can be seen in the figure, the measured values of b_ext and b_scat are continually reduced over 20 – 30 minutes to levels observed at the start of the experiment. One caveat of using the 50 L sphere as our sample chamber is 20 - 30 minutes can be required to completely fill or flush the chamber depending on flow rate. Therefore, the current design is not capable of tracking changes in b_ext, b_scat or albedo which occur on a time-scale faster than this.

An additional experiment aimed at demonstrating the ability to track changes in albedo over time is illustrated in Fig. 5.

 

Fig. 5. Plot of scattering coefficient, extinction coefficient, and albedo (second y-axis) over time during a monitoring experiment in which (NH₄)₂SO4, india ink, and (NH₄)₂SO₄ were sequentially added into the measurement chamber. Despite large changes in b_ext and b_scat as the aerosols are introduced, one can observe a change in albedo as the switch is made between a non-absorbing and absorbing aerosol.

Download Full Size | PDF

In this experiment, we have changed the solution which is placed in the atomizer over the course of the trial. Initially, an aqueous ammonium sulfate solution is atomized, then the solution in the atomizer is switched to diluted India ink, and finally switched back to ammonium sulfate. Thus, this process introduces three distinct plumes of aerosol into our albedometer over time. Over the course of the experiment we have monitored aerosol extinction and scattering coefficient and calculated albedo from these results. The introduction of each plume is clearly visible in Fig. 5 as large changes in aerosol scattering and extinction. The albedo of the first ammonium sulfate plume is high as expected, but when the absorbing India ink aerosol enters the sphere the albedo drops over 10 – 15 minutes and finally levels off. Then, as the third plume of ammonium sulfate enters the chamber the measured albedo rises again. This result suggests our device can be used to track changes in aerosol albedo over time.

Table 1 summarizes the mean albedo values measured over several trials for the different test aerosols. The data presented in this table was generated from all data points reflecting values of b_ext and b_scat typically between 30 – 100 Mm^-1 for a particular trial. One exception was for the ammonium sulfate aerosol in which an upper limit of 130 Mm^-1 was used. The lower limit of this range was determined through the empirical observation that albedo values obtained when b_ext and b_scat was below 30 Mm^-1 were imprecise (discussed more later). The upper limit for our experiments (100 Mm^-1) was set by our desire to make measurements in an atmospherically relevant extinction regime and only near the range of extinction coefficients encountered during calibration. The albedo values reported in this table are the means of the stated number of trials (N) plus/minus the standard deviation (s) of the measurements.

Table 1. Measured Albedo of Several Test Aerosols

View Table

As expected, the dielectric polystyrene and ammonium sulfate particles yielded albedo values very close to 1. The strongly absorbing India ink particles were observed to have an albedo of ≈0.5. The particulates produced from the combustion of pine needles yielded an average albedo of 0.87. However, it should be noted that the visual appearance of the combustion plume clearly was dependent on combustion conditions (flaming vs. smoldering). This phenomenon has been observed previously and is described in much more detail by Chen, et al. [21]. As we have made no effort to carefully control burn conditions, the albedo stated in this work may not be representative of pine needle combustion plumes under all conditions. Nonetheless, these results confirm our ability to measure and monitor the albedo of dispersed aerosols with our device.

3.4 Atmospheric measurements.

We have also explored monitoring atmospheric b_scat, b_ext, and albedo with our device. Figure 6 illustrates changes in aerosol extinction and scattering over a two day period in October 2007 at Kearney, NE. As can be seen in this figure, the observed scattering coefficient and extinction coefficient varied significantly over the sampling period. Despite the large variation, the single scatter albedo (green) remained relatively constant during this time. The average albedo measured over this time was 0.96. Relative humidity in the cell ranged from 32.7-48.4%, temperature varied between 17.1 – 22.8°C and pressure varied between 690.8 – 694.8 mm Hg over the sampling period. Another point of note is the good correlation between the reference nephelometer (M903) measurement and that obtained with our device. The inset of Fig. 6 illustrates a plot of scattering coefficient measured with the albedometer versus scattering coefficient measured with the reference nephelometer for this data set. The linear relationship observed illustrates that the response of both instruments exhibit the same trends in aerosol scattering over time. One potentially important difference is that the reference nephelometer yields scattering coefficient values which are on average 11% lower than the values obtained with our integrating sphere device. There are several possible explanations for this. First, the reference nephelometer is placed downstream of the albedometer and particles could be lost in the sample transfer line connecting the devices. However, modeling aerosol transport through the tube connecting our albedometer and the reference nephelometer with the Deposition 2001a program mentioned in the methods section suggests particles below 5 µm diameter should be transported very effectively (i.e. 99%) into the nephelometer for measurement.

 

Fig. 6. Aerosol extinction coefficient (b_ext), scattering coefficient (b_scat), and albedo (ω) at Kearney, NE over 2 two days in October 2007. Blue data points represent extinction coefficient as measured by CRDS, the red circles represent scattering coefficient as measured with the integrating sphere, the black circles represent scattering coefficient measured with the reference nephelometer, and the thin green trace represents measured albedo as plotted on the second y-axis. All times reflect local time at Kearney, NE. The inset shows the correlation between scattering coefficient measured by the albedometer and the M903 reference nephelometer.

Download Full Size | PDF

This result suggests particulate losses may not be responsible. A second explanation could be that the smaller truncation angle of the integrating sphere nephelometer allows for collection of more scattered light when compared to the M903. Light scattered within 10-15 degrees of forward or reverse is not collected by the M903, and this could lead to a systematic underestimation of scattering coefficient. Further investigations and comparison measurements are warranted in order to definitively quantify any underestimation by the M903.

3.5 Estimation of uncertainty in albedo.

As previously stated, our instrument determines single scatter albedo through simultaneous measurement of aerosol extinction and scattering. Both measurements are subject to random error leading to an uncertainty in their values, and therefore, also in measured albedo (ω). We can write:

where δ is the uncertainty in each variable. A useful measure of uncertainty in a measurement in the absence of determinate error is the standard deviation (s) of the measurement. Propagation of standard deviation through the albedo calculation can be achieved through the generally accepted method [22]:

where the s terms represent the standard deviation of each measurement, and ω, b_scat, and b_ext represent measured albedo, aerosol scattering coefficient, and aerosol extinction coefficient. Clearly, the relative uncertainty in albedo is limited not only by the degree of precision achieved in measurement of b_scat and b_ext but also the magnitude of these variables. The form of this equation suggests that uncertainty in albedo would be maximal at small values of b_scat and b_ext. During our experiments, the average measured standard deviations (s) for air blanks were s_scat=0.9 Mm^-1 and s_ext=0.2 Mm^-1 on the scatter and extinction channels. Figure 7 illustrates the propagated relative uncertainty in albedo (s_ω/ω) as a function of measured extinction coefficient for albedo values of 0.8, 0.9, and 1.0 using our measured standard deviations as input parameters.

 

Fig. 7. Propagated relative uncertainty in albedo vs. measured extinction coefficient for ω=0.8, 0.9, and 1.0.

Download Full Size | PDF

As observed in the figure, when aerosol extinction coefficient is below 20 Mm^-1 relative uncertainty in albedo is >5%. This graphic is in agreement with empirical observations that aerosol extinction coefficient should be above 30 Mm^-1 to obtain reliable albedo data with our instrument. Additionally, Fig. 7 demonstrates that relative uncertainty in albedo changes slightly with the albedo itself. This is simply a result of b_scat being a smaller number for any given b_ext value when albedo is less than 1. It is also of interest to note that for our experiments uncertainty in the scattering measurement alone was found to account for >90%of the propagated relative uncertainty in albedo. Therefore, improving the precision of the scattering measurement is crucial for minimizing uncertainty in albedo at lower values of extinction. If the scatter measurement could achieve a level of precision similar to that of the extinction channel (e.g. s=0.2 Mm^-1) relative uncertainty in albedo should approach 5% at extinction coefficients of only 5 Mm^-1. Nonetheless, the current albedometer offers adequate precision for use in a number of laboratory studies and field measurements at a variety of locations.

4. Summary

A new device for measurement of aerosol albedo at 532 nm has been constructed and initial proof-of-principle measurements made. Our results indicate the method described herein is suitable for the measurement of albedo for dispersed aerosol particles. The limit of detection on the extinction channel was 0.61 Mm^-1 while the scattering channel exhibited a L.O.D. of 2.7 Mm^-1. Our limit of detection for the CRDS measurement falls within the 0.3 – 8 Mm^-1 range of values previously reported for aerosol measurements by CRDS [11-15]. The 2.7 Mm^-1 detection limit observed on the scattering channel is poorer than that achieved by most commercial nephelometers (manufacturer reported ranges of 0.2 – 1 Mm^-1) nor is it as good as the 0.1 Mm^-1 reported by Fukagawa, et al. using the integrating sphere design without the CRDS measurement [17]. The poorer detection limit is likely a consequence of the much larger electronic bandwidth of detection required to collect individual ring-down transients. Improvements in precision (particularly on the scattering channel) may be necessary to make very precise albedo measurements on the ambient aerosol at locations in which the aerosol mass loading is low. The spatially averaged 2.4° truncation angle achieved through use of the integrating sphere design offers an improvement of at least a factor of two over commercially available devices. This instrument should find use in laboratory studies of aerosol optical properties, or in field studies of aerosol optical properties.