1. Introduction

A new technique for profiling aerosols has been used at a coastal site. The technique is called camera lidar (clidar) and images a laser beam, from the ground to the zenith, onto a charge-coupled device (CCD) camera with a wide-angle lens (Barnes et al. 2003, 2007). Briefly, the clidar system consists of a 532-nm laser (pointed vertically); a quarter-wave plate used to alter the laser polarization from linear to circular; and a detection system comprising a 10-nm-wide laser line filter, wide angle optics, and a cooled CCD camera. The clidar is a bistatic lidar system in which the laser transmitter is spatially separated from the detector. The scattered laser light is imaged from the side at all altitudes at once on the CCD camera. No scanning is required. The wide-angle optics, which allow the entire beam to be imaged at once, yield a constant angular field of view of 0.2° per CCD pixel. This results in a variable-altitude resolution because only a small length of the laser beam is imaged in the near-ground region (typically less than 1 m depending on the separation between the laser and detector) while a large portion of the beam is imaged at higher altitudes. Because of the bistatic configuration of the clidar system, there is a different scattering angle for each altitude. Near the ground the scattering angle is approximately (depending on terrain) 90° while at higher altitudes the scattering angle approaches 180°. Altitude in the clidar system can be determined from geometry rather than requiring costly electronics to measure time of flight of the returned signal. In contrast, traditional monostatic elastic backscatter systems (in which the transmitter and detector are not spatially separated) require timing to determine altitude. Traditional systems also typically have constant-altitude resolution (e.g., 15-m resolution for micropulse lidar systems) and are affected by field-of-view overlap constraints in the near-ground region, which do not affect the clidar signal. Traditional monostatic systems have a single scattering angle of 180° (backscatter) for all angles. Therefore, to invert a traditional lidar measurement to obtain total scatter, an extinction-to-backscatter ration needs to be assumed for each altitude. For clidar measurement inversion, an extinction-to-side-scatter (where the angle of side scatter ranges from 90° to 180°) ratio at each altitude is needed. This may be obtained from either measured or assumed phase functions.

In June 2006 the clidar was used to measure aerosols in a marine boundary layer environment. The site, called Cape Kumukahi, is about 400–500 m from the shore and breaking waves (depending on the wind direction) on the eastern tip of the Big Island of Hawaii. The altitude is only a few meters above sea level. The site normally receives easterly trade winds and the air has been over the ocean, far from any point pollution sources, for many days. The site has a small building used by the University of Hawaii, the Scripps Institution of Oceanography, and the National Oceanic and Atmospheric Administration. The site has electrical service, and air sample lines have been installed on an adjacent Coast Guard lighthouse tower. The intake for the sample line is 25 m above sea level. On two evenings, 9 and 22 June 2006 (UT), the clidar was deployed for observations starting just before sunset. Also operating were a nephelometer (Radiance, 530 nm) sampling from the 25-m level, and a weather station. An aerodynamic particle sizer (TSI, model 3320) was also sampling from the 25-m intake. A portable nephelometer (DataRam, 880 nm) was placed on the roof of the building 7 m above sea level.

The second-generation clidar camera (Barnes et al. 2007) was used with the mechanical shutter. The camera was placed 122 m from the building and laser. A portable pulsed neodymium-doped yttrium aluminium garnet (Nd:YAG laser; Big Sky Ultra, 532 nm) was used as the light source. The power of the laser was 0.6 W (20 Hz at 0.03 J per pulse), which is considerably less than that used for previous observations made at Mauna Loa Observatory with the 20-W laser (Barnes et al. 2007). A cable linked the laser and the mechanical shutter on the camera to synchronize them. The shutter blocked background light in between the laser pulses. The lighthouse was operating at the time of the observations, but fortunately the rotating light is shielded from directly shining toward the land (and camera). There were cloud-free measurement intervals on both nights when the full clidar profile to the zenith was acquired. This is needed to normalize the profile, as was done in the standard lidar method (Fernald et al. 1972). The laser beam passed through a quarter-wave plate that changed the linear polarization to circular polarization. The circular polarization allowed the clidar measurement to be compared directly with the nephelometer measurements, which used unpolarized lamps as light sources. Aerosol scatter is dependent on the polarization of the light, but the short wavelength of the circularly polarized light integrated over many seconds effectively averages the polarization effects.

2. Conditions

The working model for interpreting the aerosol measurements is that of a well-mixed marine boundary layer with near-surface contributions from breaking waves on the shoreline added in. The two days 9 and 22 June 2006 (UT) are designated days 1 and 2, respectively, for this study. Day 1 was nearly cloudless and the trade winds averaged 2.4 m s⁻¹ near the surface. On day 2 there were scattered boundary layer clouds and the trade winds were stronger, at 4.8 m s⁻¹. A buoy operating in the region (University of Hawaii 2009) measured twice the wave height on day 2 (2 m) as that on day 1 (1 m), which was coming from an easterly direction on both days. The increased wave height resulted in increased aerosol near the surface, which will be quantified below. Given the wind directions and speeds, and distances from the instruments to the shoreline, the transport time for the breaking wave aerosol was about 160 s for day 1 and 100 s for day 2. Also, the boundary layer height was significantly lower on day 1 than on day 2, which will be quantified in the next section.

3. Data and analysis

When using in situ aerosol instruments there are always issues regarding how the samples are introduced into the instruments. For example tubing size, flow rate, temperature, and humidity can affect the particle size distribution and change the fundamental measurement. Because the clidar measures scattered light in the open air, the best comparison with the in situ instruments was to try to approximate the open-air conditions. The 25-m-high sampling tube (total length 35 m), used by the Radiance nephelometer and the TSI particle sizer, had a diameter of 10 cm, and the residence time was about 6 s. Sharp turns in the tubing were minimized. There was no humidity control or heating of the sample air, and there were no size cutoffs intentionally introduced. The absolute error of the Radiance nephelometer is not stated by the vendor, but the truncation error of the Radiance nephelometer was found to be significant (>30% above 1 μm) in a study by Heintzenberg et al. (2006), so its total scatter measurement would be reduced by the loss of large particles (Heintzenberg and Quenzel 1973).

Because the DataRam nephelometer operates at 880 nm it was important to obtain a conversion factor so measurements could be compared with the Radiance nephelometer (530 nm) and clidar (532 nm). For 1 h on day 1 the Radiance nephelometer was disconnected from the air sample line and placed on the building roof next to the DataRam. The sample lines on both nephelometers were a few centimeters. The stated accuracy of the DataRam is 5% of the reading or ±2 Mm⁻¹ for the test conditions. The total scatter agreed to ±2.9 Mm⁻¹ when an Ångström coefficient of 0.30 was applied to the DataRam measurements. This coefficient is similar to what was seen during a campaign at a similar Hawaiian coastal site (0.07 ± 0.14; Masonis et al. 2003) and at a Lanai, Hawaii, Aerosol Robotic Network (AERONET) site (0.76 ± 0.37; Smirnov et al. 2002). The total scatter often changed by 10–20 Mm⁻¹ in 1 or 2 min. These relatively quick changes could not be tracked by the clidar. Integration periods of 3 min were needed to get an adequate signal, but with a stronger laser and/or a better camera there is no reason that the clidar could not have subminute time resolution.

Lidar backscatter or clidar single-angle scatter often needs to be converted to a more useful quantity such as extinction or total scatter. The absorption was not measured in this study but for marine and breaking wave aerosol the absorption is only a few percent of the total scatter (Smirnov et al. 2003). Thus, the extinction (absorption + total scatter) is very nearly equal to the total scatter. Because total scatter is the precise quantity measured by the nephelometers, it will be presented in this work. To convert the single-angle scatter of the clidar to total scatter an aerosol phase function must be used. This is very similar to the use of the extinction/backscatter ratio that must be used in the conversion from lidar backscatter to extinction (Fernald et al. 1972). In Fig. 1, two aerosol phase functions are plotted that are used in the clidar conversion. The phase functions are normalized to 2 (4π when both angles are included), and with the appropriate geometric factor they can be used to determine scattered light in any direction. Converting the clidar single-angle scatter to total scatter is described in detail in Barnes et al. (2007), but can be summarized as follows. The geometry of the camera and laser determines the scattering angles. The signal near the top of the beam (upper troposphere–lower stratosphere) is assumed to be pure molecular scattering and is used as the normalization range, similar to the standard lidar method (Fernald et al. 1972). Using a molecular density from a model or radiosonde, and the well-known molecular scattering dependence on angle, the molecular model is scaled to match the clidar signal within the normalization range. The molecular component of the signal is then removed for every altitude. The remainder is the aerosol scatter at that specific scattering angle (and thus at that specific altitude). This remainder is then multiplied by the aerosol phase function value at that angle, resulting in total scatter. The accuracy of the clidar is about 3% (one sigma) for a 3-min integration when only the signal statistics are considered. An additional 5%–10% error can be introduced by the molecular density profile used in the analysis (Russell et al. 1979). Changes in intrinsic aerosol properties with time or altitude will change the aerosol phase function and the lidar extinction/backscatter ratio, which can be a major source of error.

The National Aeronautics and Space Administration (NASA) AERONET phase function (Smirnov et al. 2003) is the result of an inversion of several wavelength channels of a sunphotometer (Dubovik and King 2000). The AERONET instrument (Holben et al. 1998) is capable of scanning modes that get many angles of scatter. With assumptions about the aerosols (both spherical and nonspherical particles, and coarse and fine modes), the data can then be inverted to get an aerosol phase function. This particular function is a long-term average from a coastal site at Lanai at 500 nm. The function should be representative of the average aerosol found in a marine boundary layer. Variation in the intrinsic aerosol properties in the column would be a source of error in the conversion.

The Porter phase function in Fig. 1 was measured with a polar nephelometer (Porter et al. 1998; Lienert et al. 2003). It was measured at a coastal site on Oahu, Hawaii, and should be representative of aerosols found in breaking waves. In Table 1, four parameters are calculated from the two phase functions. The asymmetry parameter and hemispheric backscatter fraction are indications of particle size (Andrews et al. 2006). The higher asymmetry and lower hemispheric backscatter fraction for the AERONET phase function indicate larger particles than that for the Porter function. The extinction-to-backscatter ratio, also known as the lidar ratio, is determined by the phase function value at 180°. In the clidar technique the laser light scattered near the ground is usually at 90°, so a new term, the extinction to side-scatter ratio, has been defined and is also listed in the table. In this study the camera was 122 m from the laser so that although the clidar scattering angle is about 90° near the ground, it approaches the regular lidar 180° backscatter angle as the upper boundary layer is measured.

4. Results

In Fig. 2 (upper), four cloudless profiles of the total scatter are shown for day 2. The Porter phase function has been used for the conversion from the clidar ratio to total scatter. The logarithmic altitude scale shows the high resolution that is possible in the atmospheric boundary layer (ABL) with the clidar technique. The top to the ABL, in terms of aerosols, is clearly measured and is seen to coincide with the sharp drop in relative humidity (RH) measured by the Hilo National Weather Service radiosondes. The sondes were launched 22 km from the site of the clidar measurements at times roughly 6 h before and after the measurements. An average ABL height would be about 2800 m. There is a strong negative gradient in the aerosol in the first 35 m and less of a gradient up to 150 m. The aerosol is then fairly constant from 200 to 800 m. This is a valuable result that helps to characterize the sampling environment for the in situ sampling.

In Fig. 2 (lower) a corresponding plot is shown for day 1, again using the Porter phase function. The total aerosol and gradient behavior are similar to that of day 2, but the top of the ABL is noticeably lower, at 1500 m. Another difference from day 2 is that the top of the ABL is descending during the time between the radiosonde flights.

Roughly 2 h of data from day 2 are plotted in Fig. 3, showing a comparison of the clidar total scatter to that of the nephelometers. Five altitude bins of clidar data, which correspond to two vertical meters centered on the nephelometer intakes, have been averaged for each comparison. The clidar data have been scaled with a multiplier to obtain the best match with the nephelometers. The first three clidar points at 7 m, which disagree with the nephelometer, are included to show the effect of high levels of background light shortly after sunset (left arrow). After those points the clidar tracks the changes in aerosol seen by the nephelometer quite well. There is more aerosol at 7 m, and the fluctuations are also larger probably resulting from the breaking waves. It can be seen from the plot that both the 7- and 25-m levels cannot both be scaled with the same multiplier. This reflects a difference in the aerosol properties at the two levels, which would result in a difference in the aerosol phase function. The difference in the aerosol could be size distribution, composition, or shape, or a combination of all three. In this case the 7-m level is probably more influenced by the breaking waves than the 25-m level. The Porter phase function used works well for the 7-m level because the clidar multiplier is nearly 1 (0.98). The 25-m comparison requires a much larger correction (0.69) and probably reflects a higher fraction of marine boundary layer aerosols in the mix.

In Fig. 4 (upper) the average of all the profiles taken on day 2 is shown, which is analyzed with both aerosol phase functions. The cloud base is clearly seen at 600 m. The average nephelometer values for this time interval are also shown. There is a substantial difference between the two profiles in the total scatter. This is simply due to the difference in phase functions shown in Fig. 1 between 90° and 180°, and it emphasizes the large potential error that can be introduced in converting single-angle scatter to total scatter. This is analogous to the challenge faced by elastic backscatter lidar when converting 180° scatter to total scatter. The natural variability during the 2 h is plotted as one-sigma bars, indicating larger fluctuations in the breaking-wave region (first 100–200 m). In Fig. 4 (lower), the corresponding data are shown for day 1. The profile is very similar to day 2 up to the 600-m level at which point there is a fairly constant decrease in aerosol until the background level above the ABL is reached. Day 1 was largely free of clouds.

There is a larger mismatch at 7 m for day 1 than for day 2 between the clidar and nephelometer. The multiplier needed for day 1 is 0.65. This indicates that the Porter phase function is less representative of the aerosol. This may reflect the difference in transport times for the breaking waves mentioned above. The aerosol during day 2 arrived at the sampling location in about 100 s while on day 1 it was 160 s. The particle size data indicated larger particles on day 1 (4.06 ± 0.22 μm median aerodynamic diameter) when compared to those on day 2 (3.34 ± 0.18 μm mean diameter). Coagulation during the longer transit time might explain the increase in size, but that would have to be more significant than the settling and deposition, which would work toward smaller particles. The clidar–nephelometer combination is essentially measuring the aerosol phase function at a single angle (altitude). Other experiments can be envisioned to determine more points of the phase function with multiple nephelometers and/or cameras on a tall tower or on an aircraft.

In Fig. 5 all profiles are plotted for day 2. The base to the cloud region is seen to start at about 600 m. This is quite similar to the result from the Shoreline Environment Aerosol Study (SEAS) experiment (Clarke and Kapustin 2003; Porter et al. 2003), which occurred in a very similar coastal environment on Oahu. While the base of the cloud region is fairly constant in altitude, the cloud top is seen to change by more than 1 km. Measuring cloud top with a clidar or lidar is strongly affected by extinction and cannot be relied on, but the cloud-top changes seen in the data roughly agree with the visual observations before sunset. Two sharp features at 80 (6.6 h) and 60 (7 h) m may have been the plume from a passing cruise ship that was observed during the study. The arrival of the plume was consistent with the wind speed and direction, and the distance and position of the ship.

The clidar data, interpreted with the AERONET aerosol phase function, overestimates the scatter in the breaking wave region observed by the nephelometer. However, the data used in the AERONET inversion represent an average of the entire marine boundary layer, which often extends beyond 2 km (Porter et al. 2003); thus, the difference is not unexpected. A more appropriate comparison using the AERONET phase function–derived total scatter is that of aerosol optical depth (AOD). The clear-sky clidar profiles can be integrated from the ground through the troposphere to get the total AOD. No extrapolation to the ground is needed because it is in the case of a vertically pointed lidar. The result is 0.045 and 0.087 for days 1 and 2, respectively. The lower value for day 1 reflects the lower top of the ABL as well as lower extinction in the profile. Smirnov et al. (2003) found a yearly average AERONET AOD for Lanai of 0.078 at 500 nm, and they found that the AOD was between 0.020 and 0.100 about 88% of the time.

By using the AOD from the NASA Terra Moderate Resolution Imaging Spectroradiometer (MODIS) satellite instrument (Barnes and Salomonson 1993), an average extinction/backscatter ratio (lidar ratio) for the profile can be directly calculated. There were overpasses just off of the eastern side of the study location for both days 1 and 2. The MODIS measurements are limited to daytime so an upwind region was used that would have been transported over the study site during the clidar observations. The AOD for days 1 and 2 were 0.079 ± 0.015 and 0.136 ± 0.22. The ± values are the variability in the MODIS pixel values (52 for day 1 and 100 for day 2) averaged for the AOD. The error beyond the pixel variability in the MODIS AOD is about ±0.035 (Remer et al. 2005). Using these MODIS values, the extinction/backscatter ratios needed to match the clidar and MODIS AODs are 59 and 53 sr for days 1 and 2, respectively. The ratios differ significantly from the ratio of 33.7 sr determined from the long-term AERONET inversion of the Lanai data as well as Raman lidar observations of clean marine aerosol, which were 20–26 sr (Mueller et al. 2007). The difference for day 1 is not significant considering the combined MODIS and clidar error, but the day 2 difference is. A possible explanation is that the MODIS AOD measured under the partly cloudy conditions of day 2 is too high. Other studies with MODIS (Wen et al. 2006; Koren et al. 2007) have shown that there can be large enhancements of AOD at kilometer distances from clouds. It would be informative to repeat this type of experiment with a collocated AERONET instrument to directly compare the measured and inversion-determined ratios.

5. Conclusions

A new aerosol-profiling technique, called camera lidar (clidar), has been demonstrated at a coastal site and has provided aerosol profiles from the ground through the top of the boundary layer. The submeter resolution near the ground allows direct comparison with in situ instruments. Gradients in aerosols helped quantify the tower-sampling environment. The combination of the clidar (single-angle scatter) and nephelometer (total scatter) measured the aerosol phase function at a single angle (altitude), which varies from 90° to 180°, depending on the altitude. This is a useful constraint on calculated phase functions, which are often used for satellite and lidar retrievals.

The clidar profiles showed detailed aerosol information in various regions of the boundary layer, such as the cloud base and the top of the boundary layer. The combination of MODIS aerosol optical depth and the integrated clidar profiles resulted in average extinction/backscatter ratios of 59 and 53 sr for the two days studied.

The clidar equipment used in this study is simpler and less expensive than corresponding lidar instruments. The most expensive element was the laser. However, advances in CCD cameras and lenses now can provide significantly more signal than the ones used in this study. This means a much smaller and less expensive laser could be used. Application of the clidar technique to daytime would be a significant improvement. The main difficulty is filtering out the background light for the wide field-of-view needed for the optics. There are possible configurations currently being investigated for daytime operation.