atmospheric aerosol profiling with a bistatic imaging lidar system

Atmospheric aerosol profiling with a bistatic imaging lidarsystem

John E. Barnes, N. C. Parikh Sharma, and Trevor B. Kaplan

Atmospheric aerosols have been profiled using a simple, imaging, bistatic lidar system. A vertical laser beamis imaged onto a charge-coupled-device camera from the ground to the zenith with a wide-angle lens(CLidar). The altitudes are derived geometrically from the position of the camera and laser with submeterresolution near the ground. The system requires no overlap correction needed in monostatic lidar systemsand needs a much smaller dynamic range. Nighttime measurements of both molecular and aerosol scat-tering were made at Mauna Loa Observatory. The CLidar aerosol total scatter compares very well with anephelometer measuring at 10 m above the ground. The results build on earlier work that compared purelymolecular scattered light to theory, and detail instrument improvements. © 2007 Optical Society ofAmerica

OCIS codes: 010.1100, 010.1110, 010.1120, 010.3640, 040.1520, 010.1310.

1. Introduction

Mapping the aerosol structure of the atmosphere overboth time and altitude provides important informationfor a host of atmospheric science research fields. Aero-sols play a role in climate change1 both directly, byaffecting Earth’s radiation balance, and indirectly, byaffecting cloud properties. Understanding their contri-bution to climate change will require active aerosolprofiling information.2 Both natural and manmadeaerosols impact local and regional air quality, leadingto human health and public policy implications. Aero-sols may be used as tracers for air mass transport andatmospheric dynamics studies. Aerosols also play arole in local precipitation patterns. Many of the effectsof aerosols depend strongly on aerosol properties thatare highly variable with altitude including aerosoltypes, chemistry, and concentrations. Thus aerosolprofiling fills an important need in the atmosphericscience community.

Aerosol distribution in the atmosphere is nonuni-form, with most aerosols occurring in the first fewkilometers of the atmosphere. This low-altitudeaerosol-rich region is typically colocated with theboundary layer. Measurements of boundary layeraerosols, while highly important, are often technicallychallenging. In situ aerosol measurements may bemade at a variety of altitudes by plane or balloon-borne instruments. Such measurements are oftenable to provide more specific aerosol information(e.g., chemical composition) than remote-sensing tech-niques. They are, however, expensive and typicallyprovide only a few profiles. Thus they may not map thetemporal evolution of the aerosol structure. Passive,ground-based, remote-sensing aerosol measurementsare often conducted with radiometers, which provideintegrated aerosol optical depth over the entire atmo-sphere. Advantages of these types of measurementsinclude their relatively low cost and their ability to berun continuously. A primary disadvantage is the in-ability to generate the altitude-dependent aerosol pro-file.

Lidar remote sensing of aerosol profiles offers capa-bilities for altitude and time-series monitoring of aero-sol conditions. The most frequently used lidar systemsfor this purpose are elastic backscatter lidars, whichemploy pulsed laser transmitters and collect backscat-tered light with a colocated time-binned detector.Often elastic lidar is used in conjunction with otherinstruments3 (e.g., Sun photometers) as one parameteris measured and two parameters (extinction and back-scatter) are to be determined. An alternate method4

J. E. Barnes ([email protected]) and T. B. Kaplan arewith the NOAA�Earth System Research Laboratory�Global Mon-itoring Division, Mauna Loa Observatory, Hawaii, 1437 KilaueaAvenue, Hilo, Hawaii 96720. N. C. Parikh Sharma is with theDepartment of Physics and Earth Sciences, Central ConnecticutState University, 1615 Stanley Street, New Britain, Connecticut06050, USA.

Received 19 July 2006; revised 7 November 2006; accepted 11December 2006; posted 22 January 2007 (Doc. ID 73181); pub-lished 1 May 2007.

0003-6935/07/152922-08$15.00/0© 2007 Optical Society of America

2922 APPLIED OPTICS � Vol. 46, No. 15 � 20 May 2007

uses the Raman scattered signal to get extinction.Aerosol scatter has also been measured with multipleor movable receivers in conjunction with lidar andused to derive aerosol properties under well-mixedconditions.5–7 Challenges faced by typical lidar instru-ments include possible afterpulse effects8 (which couldlead to detector saturation at the lowest altitudes),large dynamic-range requirements,8 and field-of-viewoverlap effects9,10 (in which the illuminated volume isnot entirely within the detector field of view at near-instrument ranges). Frequently data from elasticbackscatter lidars are not used within the overlaprange, which is precisely the near-ground range ofinterest in many cases. There are monitoring pro-grams using 500 m towers11 that could benefit fromwell-calibrated aerosol measurements throughoutthis region.

An alternative bistatic lidar instrument has beendeveloped and tested for profiling a boundary-layeraerosol.12 The instrument employs the bistatic con-figuration in which the laser transmitter and thedetector are spatially separated. The new instru-ment, called CLidar, uses a CCD camera as its detectorand determines scattering altitudes from geometryrather than timing considerations. Compared withtraditional monostatic lidar, the new approach hassubstantially less stringent dynamic-range require-ments.12,13 Integration of the signal is performed onthe camera and data acquisition is the camera-supplied control electronics. Afterpulse saturation ef-fects are eliminated. The new instrument also providesuseful data from where the beam is first imaged withexcellent near-ground altitude resolution. As with allbistatic systems, the scattering angle changes withaltitude. Aerosol properties may be derived from bi-static signals given additional measurements or appro-priate assumptions.

2. Technique

The CCD camera-based lidar (CLidar) system imagesa vertical pulsed laser beam onto a CCD using wide-angle optics. The choice of wide-angle optics eliminatesthe need for scanning to cover the entire altituderange. This system operates only at night. The first-generation CLidar system, its altitude resolution, anddynamic-range requirements, are described in detailin Barnes et al.12 A second system has been designed toenhance the signal-to-noise ratio. A schematic of thegeometry is shown in Fig. 1.

The CCD camera and optics are located a distanceD away from the laser. The laser transmits light withenergy EL. Here EL represents the total laser energysummed over the image exposure time. The lightscatters off air molecules, aerosols and�or cloud par-ticles, and the portion of the scatter, which is in thedirection of the scattering angle � and is at the po-larization angle �, is collected by the wide-angle op-tics and imaged onto a CCD. The polarization angle isdetermined by the plane of polarization of the laserand the location of the camera.12 For each pixel in theresulting image, the altitude may be calculated frommeasurements of camera location in relation to the

laser and from the field of view per pixel. The inten-sity of the scattering is given by the recorded counts.The CLidar signal, Er, is made up of the photonsmeasured by a pixel with field of view d�, from ascattering event that took place at altitude z andrange R from the camera.12 The analysis here as-sumes that d� is constant, which requires the opticsto accurately (linearly in angle) image the laser beam.The signal Er is given as

Er � KELATatm zTatm R��, �, z�dz�R2, (1)

where dz is given by

dz � �R2�D�tan�d��2��1 � �z�D�tan�d��2�� R cos�d��2�� zR�D�sin�d��2��D��R sin�d��2�� 1�.

(2)

Here K is a calibration constant representing thesystem optical efficiency and A is the effective collect-ing area of the optics, Tatm z and Tatm R are the total(aerosol and molecular) atmospheric transmittancefrom the laser to altitude z and from altitude z alongthe slant path R to the CCD detector, respectively,and ��, �, z� is the scattering coefficient. In practice,dz�R2 in Eq. (1) is very nearly equal to d��D, whichmeans that Eq. (1) does not have the 1�R2 depen-dence of the familiar lidar equation.12,13 The altitudeimaged in each pixel grows with height in such a wayas to almost cancel out the typical R2 dependence oftraditional backscatter lidar signals, resulting in lowdynamic-range requirements. This also leads to analtitude resolution, which is excellent in the near-ground region for which the instrument is designed,but which degrades at higher altitudes.

It is often desirable to use the aerosol backscattersignal from a lidar quantitatively, for example, tocalculate extinction (total scatter plus absorption) oraerosol optical depth. This requires a relation betweenthe light scattered at a single angle and total scatter(extinction-to-backscatter ratio in monostatic lidar),which must be either calculated from assumed aerosolparameters or related to additional measurements.

Fig. 1. CLidar system schematic.

20 May 2007 � Vol. 46, No. 15 � APPLIED OPTICS 2923

This can lead to large errors in the quantitativeparameters since the angular dependence of aero-sol scatter, the aerosol phase function, varies widely,especially in the 90°–180° range. In Fig. 4 of Ansmannet al.,2 for example, the extinction-to-backscatter ratiovaried by a factor of 3 in the boundary layer measure-ments of extinction. The same can be true for somesatellite retrievals of aerosols that rely on single-anglescatter. Kalashnikova and Sokolik14 used microscopydata and discrete-dipole approximation calculations toestimate the phase functions for mixtures of mineraldust characterized by particular combinations of com-position, shape, and size. This study showed that aero-sol optical depth retrievals would be overestimated byforward-looking, ground-based measurements by asmuch as 1.5 times and underestimated by techniquesbased on backscatter measurements (such as lidarsand satellites) by �2 times if spherical particle in-stead of irregular shapes are assumed in the re-trieval.

The CLidar technique does require an assumption ofan extinction-to-single scatter ratio for a range of back-scatter angles (90°–180°) when calculating quantitiessuch as aerosol optical depth, rather than just oneangle as in monostatic lidar. But given the alreadylarge uncertainty in this ratio, little additional un-certainty is added in the CLidar method. And sincescattered light is measured accurately at the groundlevel, in situ aerosol measurements of absorption andtotal scatter can be used to directly measure the ex-tinction to single-angle-scatter ratio for at least theone location.

3. Instruments

Two CLidar instruments have been constructed. Bothused a linearly polarized Spectra-Physics Nd:YAGmodel GCR-6 with 300–600 mJ�pulse operating at awavelength of 532 nm, a pulse width of 8 ns, and apulse repetition rate of 30 Hz. Both the first-generationand the second-generation CLidar instruments utilizea 10 nm full width at half-height (FWHH) laser linefilter between the camera and the lens to filter outbackground light.

Detailed specifications on the first- and second-generation systems are provided in Table 1. For sys-tem 1 the effective optical diameter is 1.98�4 � 0.50mm�pixel. This diameter was misprinted in Barneset al.12 The second-generation CLidar instrumentemploys an Apogee Alta camera model E260. Thesensor employs transparent gate technology, whicheliminates interpixel dead space where image infor-mation may be lost and also many fewer high-noisepixels. An undesirable side effect of these gates, how-ever, is that the two gate materials do not have equaltransparencies, which gives rise to intensity oscilla-tions for objects focused to the size of a pixel or less.15

Thus practical considerations require that the imagebe slightly defocused to eliminate this effect. This hasthe unfortunate consequence of introducing addi-tional noise as wider beam images include more back-ground noise due to the larger number of pixelssummed. The Coastal Optics lens is expected to raise

the second-generation system signal several timesthat of the first-generation system.

The second-generation CLidar system also incorpo-rates a chopper and custom-designed holder to reducebackground light. Since background light is collectedfor the duration of the image exposure time (typicallya few minutes), while the laser pulses for only a smallfraction of that time, the sky background may be sig-nificantly reduced by blocking the CCD sensor duringtimes in between pulses. The optomechanical chopperis a Thor Labs model MC1000 with phase-adjustmentcapabilities. The standard 10 slot blade was modifiedto leave only two openings. A housing was machined toallow the chopper to be placed between the lens andthe interference filter. The chopper is synchronized toa transistor–transistor logic (TTL) trigger from the la-ser flashlamp. A cable connection was used in thesemeasurements, but a wireless connection would bemore versatile. The phase of the chopper is adjustedrelative to the trigger to ensure the CCD chip is unob-structed for the duration of the pulse travel time (ap-proximately 1 ms). From a practical standpoint thismay be accomplished by adjusting the phase offsetuntil the laser-beam image brightness is constantwhen the beam is imaged with either edge of the lens.The chopper reduced the sky background by approx-imately a factor of 5. This is especially important formeasurements in urban areas where sky back-grounds may be large. Greater reductions may bepossible depending on the pulse-repetition fre-quency of the laser.

4. Image Analysis

In general, the image of the beam is not aligned per-fectly with a column or a row of the CCD. In Barneset al.12 the beam was aligned close enough to use asimple summation of beam pixels and subtraction ofbackground. In fact, it can be useful to align the beamdiagonally on the CCD, which allows a larger image.

Table 1. SBIG ST-237 and Apogee Alta Camera Specifications

Feature, Specification ST-237 (1) Apogee E260 (2)

Active area (mm) 4.86 � 3.66 10.2 � 10.2Pixels 657 � 495 512 � 512Pixel size (�m) 7.4 � 7.4 20 � 20Read noise (counts�pixel) 14 19.5Gain 1.6Dark (electrons�s�pixel) 5 at 0 °C 0.72 at �22.3 °CFull well capacity

(electrons)20,000 102,000

Analog-to-digital converter(bits)

16 16

Pixel digitization rate (kHz) 30 1300Bias (counts) 100 1578CCD sensor TI TC237 KAF-0261E-1Quantum efficiency at

532 nm47% 46%

532 nm interference filterFWHH (nm)

10 10

Wide-angle lens Kinoptic, f�4 Coastal Optics DigitalSLR, f�5.2

Lens focal length (mm) 1.98 4.88Effective optical diameter

(mm�pix)0.50 0.94

Angle per pixel (deg�pix) 0.194 0.171


In this study, a more general, versatile technique isused.

To analyze an image, the best fit to the beam iscalculated based on image intensities. A second-degreepolynomial was found to work well enough. Thebeam is then divided into equal, pixel length seg-ments (altitudes) from the lowest end of the beam tothe topmost. A perpendicular line to the beam at theposition (altitude) to be analyzed is calculated usingthe beam-fitting function, and then the intensitiesalong this perpendicular are used to fill an array. Atwo-dimensional interpolation routine is used toprovide intensities anywhere along the perpendic-ular. A function consisting of a Gaussian plus aconstant,

f�x� � A0 exp��x � A1�2��2A22�� A3 (3)

is fit to the array. Here x is the distance from thebest-fit curve along the perpendicular line, A0 repre-sents the Gaussian height, A1 is the actual beamcenter, A2 represents the Gaussian width, and A3gives the sky background per pixel. The CLidar beamsignal at each altitude is the area under the Gauss-ian, which is simply A0A2��2��. INTERACTIVE DATA

LANGUAGE (IDL) software was used for handling theimages, interpolating, and curve fitting.

Signals are corrected for molecular extinction bydividing the CLidar signal by the two-way molecu-lar transmittance calculated from the molecular-scattering cross section. The molecular density istaken from a time-dependent model based on sev-eral years of the Hilo radiosondes. The signal is alsocorrected for estimated aerosol extinction by usingan iterative approach. The aerosol profile is calcu-lated and used to estimate the aerosol extinction.These steps are repeated until the total extinction(molecular and aerosol at each altitude) converges.Details of the calculation are described below.

5. Experimental Results

The first-generation CLidar system has been shownto produce excellent agreement with theory undermolecular-scattering conditions.12 Experiments wereconducted using the first- and second-generationCLidar systems to assess system performance (signal-to-noise) and to assess system capabilities undermolecular-scattering and aerosol-scattering conditions.The experimental site for these studies was MaunaLoa Observatory (MLO), a high-altitude climate-monitoring observatory located at 19.54 °N and155.58 °W at an altitude of 3397 m above sea level(asl). Exposures were taken for 332 (10,000 lasershots) with cameras cooled to �25 °C.

First, optical performance of the first- and second-generation systems were compared under low aerosol-scattering conditions. For this experiment, thelaser-beam polarization was altered from linear tocircular by the insertion into the laser beam of aquarter-wave plate with a fast axis at 45° to the laserpolarization. As the direction of the laser polarizationrotates swiftly, unpolarized light is mimicked over

the time spans of interest for this experiment. Usingcircularly polarized light rather than linearly polar-ized light makes the azimuth angle (polarizationangle) unimportant in signal analysis and thus sim-plifies the interpretation.

Data were taken on 18 April 2006 UT (18:30 to24:38 local time) at MLO. The camera was located139 m south and upslope of the laser. Data weretaken simultaneously with the first- and second-generation CLidar systems to compare performancecharacteristics. The signal chopper of the second-generation system was not used so a more directcomparison could be made with the first-generationsystem, which does not have a chopper. The resultingscattering profiles taken on a low-aerosol night atMLO are shown in Fig. 2. The molecular-scatteringmodel is overlaid with the data. The better quality ofthe second-generation camera data shown is appar-ent. System 2 has �2.5 times more signal than sys-tem 1, and there are fewer spikes due to bad pixels.There are several altitudes in the system 1 profilewhere the signal dips below the molecular model,resulting in a negative aerosol-to-molecular ratio.These bad pixels cause the negative ratios to occur ineach profile. Further data processing could partiallyremove this effect, but this has not been done so thata clearer comparison can be made between the twosystems.

Results show the shape of the effectively unpo-larized Rayleigh scattering phase function with its�1 � cos2 �� dependence on scattering angles. Thereis aerosol present in the first 286 m (bin 350 for sys-tem 2) with an aerosol-to-molecular ratio between 0and 0.2. Nephelometer (TSI Model 3563) readingsduring these measurements recorded total aerosolscattering of 2.9 Mm�1 at 10 m above the ground.

Signal-to-noise comparisons were conducted for thetwo systems for a 332 s integration. The signal countsare calculated from the area under the Gaussian fitcurve as described above. The sky, dark, and readoutnoise are measured per pixel. For signal-to-noise es-

Fig. 2. Comparison of the first- and second-generation CLidarsystems using circularly polarized light. Data were taken on 18April 2006 at 18:30 to 24:38 local time. Molecular models areplotted (dashed curves) with the signals.


timates, the noise per pixel is multiplied by A2��2��,which serves as an estimate of the number of pixelsin the beam. For system 1 in this experiment, a 100count software bias, which does not contribute tonoise, is removed. System 1 beam counts averaged15,030. Average background noise with the cameracooled to �25 °C was 242 counts per pixel. Sky back-ground averaged 162 counts per pixel. The camera hasa read noise of 25 counts per pixel. The beam wasdefocused to a Gaussian half-width of 3.5 pixels. Sys-tem 1 has a gain of 0.72 electrons�count, while system2 has a gain of 1.6 electrons�count. The signal-to-noiseratio (SNR) for the first-generation system in terms ofphotons is thus

SNRfirst generation � 15,030�0.72�0.47��15,030�0.72�0.47�� 2��0.5�3.5��0.72�0.47��162 � 242� 25��0.5 � 136. (4)

The system 2 camera has a hardware bias of1598. Beam counts averaged 40,140. The system 2camera cooled to �25 °C has average dark noise of19 counts�pixel. The sky background using the sys-tem 2 averaged 399 counts per pixel. The readoutnoise was 12.2 counts per pixel. The beam was defo-cused to a Gaussian half-width of 1.5 pixels. Thesecond-generation system thus has an SNR in termsof photons of

SNRsecond generation � 40,140�1.6�0.46��40,140�1.6�0.46�� 2��0.5�1.5��1.6�0.46��399 � 19� 12.2��0.5 � 366. (5)

The second-generation system ratio is 2.7 timeshigher than the first generation. These ratios shouldbe upper bounds to the actual ratios seen in Fig. 2since natural variability and CCD defects would in-crease the noise and lower the ratio. The ratio ofsignal to standard deviation was calculated for thepeak regions of Fig. 2 (bins 420–440 for system 1 andbins 460–480 for system 2). The ratios are 80 and205, which would represent lower bounds on the ac-tual performance. They have approximately the same2.7 factor between systems.

Next, system 1 performance was studied underaerosol-scattering conditions using linear laser polar-ization. System 2, unfortunately, was not operatingduring this time. The location of the camera was thesame as for the previous data, at an angle of 90° to thelaser polarization. The Rayleigh scattering phasefunction for polarized light causes the received signalto be proportional to the quantity �sin2 � � cos2 �cos2 ��, where � is the scattering angle and � is thelaser polarization angle. By selecting a site located at90° polarization �� 90°� the molecular-scatteringsignal remains almost constant in magnitude for sev-eral kilometers. (At high altitudes the molecular den-sity drops rapidly.)

The scattering experiments were conducted on 21December 2005 from 4:26 to 7:40 UT (18:26 to 21:40

local time). Over this time span, an integrating neph-elometer and a particle and�or soot absorption pho-tometer (Radiance Research), located at the stationwith its intake stack at an elevation of 10 m abovethe surface, provided simultaneous total aerosol-scattering and absorption measurements. Nephelom-eter total aerosol-scattering values for the green�550 nm� channel ranged from 0.8 to 37 Mm�1. Thesemay be taken to represent the total (integrated overall angles) aerosol-scattering coefficient at the alti-tude of the intake stack. To separate the aerosol andmolecular contributions to the CLidar signal, a mo-lecular model was fit to the CLidar signal betweenaltitudes of 5.07 and 12.2 km asl, which were as-sumed to represent purely molecular scattering. Themodel was based on molecular densities from a two-dimensional (time and altitude) spline fit to radio-sonde data from Hilo, Hawaii (59 km from the MLO).The molecular contribution was subtracted from thesignal, and the remainder of the signal was taken torepresent aerosol scattering. Signals were correctedfor molecular extinction and aerosol-extinction esti-mates. Aerosol extinction was estimated iterativelyfrom the estimated aerosol-scattering coefficient,which was calculated as the product of the CLidaraerosol- to molecular-scattering ratio, the molecular-scattering coefficient, and the ratio of the molecularphase function to the aerosol phase function. (Theazimuthal dependence of the phase functions was notmodeled here for extinction estimates.) The CLidarsignal was divided by the total aerosol and moleculartwo-way transmittance, a new molecular model wasfit, a new aerosol ratio was obtained, and the processwas continued until the total transmission stabilized.The average CLidar signal versus altitude duringthis time is shown in Fig. 3 overlaid with the fitmolecular signal. The previously mentioned negative-going spikes due to pixel sensitivity are again present.Most are in the same bins as in Fig. 2 (420, 380, 245,205).

To calculate a quantity such as extinction from a li-dar backscatter profile, the extinction-to-backscatter

Fig. 3. CLidar signal and molecular model for 21 December 2005from 4:26 to 7:40 UT (18:26 to 21:40 local time). Each bin repre-sents an angle of 0.194°.


ratio at 180° must be assumed or measured. Due to thebistatic nature of the CLidar, this ratio is needed from90° to 180°. The full description of the ratio from 0° to180° defines the aerosol phase function. For this anal-ysis, phase functions were obtained from one of theNASA�Aeronet16 instruments located at MLO. TheAeronet instrument17 measures solar radiation at mul-tiple angles and wavelengths. The aerosol phase func-tion is then calculated at four wavelengths (441, 673,873, 1022 nm) for a fine and course mode, as well asan overall total. For this study, 11 separate measure-ments were analyzed to get an average total aerosolphase function. The measurements were taken on theday before and the day after the CLidar measure-ments (all 21 December 2005 UT). The 441, 673, and873 nm data were interpolated with a power law toget functions at 532 nm, the CLidar wavelength.The average and the two functions having the mostextreme (maximum and minimum backscatter) devi-ations from the average were identified. These threefunctions are plotted in Fig. 4. The extreme cases areused to indicate the range of results possible withreasonable variations in the phase function.

Once an aerosol phase function is chosen, severalimportant quantities are defined. Table 2 contains fourof these, which have been calculated for the threephase functions in Fig. 4. The hemispheric-backscatter

fraction and the asymmetry parameter are integratedquantities18 and usually vary less than the extinction-to-backscatter ratio, which changes by over a factor of2 from the minimum function to the maximum. Theuncertainty in knowing this extinction-to-backscatterratio can dominate the errors in calculating the ex-tinction or aerosol optical depth from a lidar back-scatter profile. The extinction-to-sidescatter ratio,calculated at 90°, varies slightly less than a factor of2 for these particular phase functions. The Aeronetaverage hemispheric-backscatter fraction (daytime) of0.126 is similar to the nephelometer-measured value of0.115 during the period the CLidar data were taken(nighttime), suggesting the phase function does repre-sent the aerosol during the CLidar measurements. Thevalue of 0.115 is typical of aerosols at MLO.19

The total scatter from the CLidar signal with alti-tude is shown for the time series in Fig. 5. There are33 individual profiles, taken every 5.6 min, formingthe plot. This captures a typical transition at theMLO from the daytime upslope winds to the night-time downslope winds. The Aeronet measurementswere made during the daytime. The upslope air isoften influenced by the local biosphere and towns onthe island. At approximately 7:00 UT, the transi-tion to downslope, free tropospheric air is seen, andthe total scatter decreases. The aerosol layer is�300 m thick. The horizontal lines are due to lowersensitivity pixels on the camera mentioned previ-ously. From approximately 6.1 to 6.6 UT there is apeak in the total scatter at �60 m. This peak illus-trates the ability of the CLidar data to help interprettower in situ data. At MLO, there are 10 m (aerosol)and 40 m (trace gas) sampling towers. In this case,the aerosol peak would have been missed by the 10 mtower.

Fig. 4. Aerosol phase functions from MLO NASA�Aeronet instru-ment for 21 December 2005 UT. Minimum and maximum back-scatter refer to the region between 90° and 180°.

Fig. 5. CLidar total scatter �km�1� measured on 21 December2005. Local time is UT-10 h and the altitude is measured above thestation.

Table 2. Quantities Calculated from the 11 Aeronet Phase Functionsfor 21 December 2005 UTa

HemisphericBackscatter

Fractionb

AsymmetryParameter

g

Extinctionto 180°

BackscatterRatio(sr)

Extinctionto 90°

SidescatterRatio(sr)

Min 0.074 0.708 75.5 56.7Avg 0.126 0.599 51.0 35.8Max 0.168 0.522 35.0 29.9

aMin, Avg, and Max refer to the backscatter portion of the phasefunctions shown in Fig. 4.


Comparison with nephelometer measurements ispossible at the altitude of the nephelometer intakestack �10 m�. Figure 6 shows the nephelometer to-tal scatter with the CLidar total scatter obtainedby averaging bins 8–12 �9–11 m�. The CLidar datatracks the nephelometer exceedingly well and demon-strates the near-ground capability of the method. Theaverage total scatter during the overlap period is11.54 Mm�1 for the nephelometer �550 nm� and11.49 Mm�1 for the CLidar �532 nm�. The error barsin Fig. 6 show the CLidar total scatter calculatedwith the two extreme (maximum and minimum)phase functions. The corresponding averages are 9.84and 17.52 Mm�1 for these cases, respectively, reflect-ing the differences in the phase functions shown inFig. 4 at 90° and also listed in Table 2.

Aerosol extinction is the sum of the total aerosolscatter and the aerosol absorption, and the absorp-tion is also measured at MLO. However the absorp-tion during the CLidar measurements �0.04 � 0.31Mm�1) is below the detection threshold of the instru-ment (approximately 1 Mm�1). The long-term aver-age of the fraction of absorption at MLO is 11% of theextinction,19 which would be near the threshold dur-ing the CLidar measurements. Neglecting the ab-sorption and assuming the phase function representsthe aerosol throughout the profile, the aerosol opticaldepth (AOD) can be estimated. The average CLidarAOD �532 nm� during this period is 0.0035, 0.0051,and 0.0081 for the maximum, average, and minimumphase functions, respectively. This range in AOD re-flects the spread in the phase functions between 90°and 180°. There were 61 measurements of AOD�500 nm� by the Aeronet instrument for the sameperiod as the phase functions, the day before and theday after. The average was 0.0048 with one standarddeviation of 0.0017. The standard deviation suggeststhe aerosol conditions were relatively uniform andthe Aeronet phase functions represent the aerosolsmeasured by the CLidar.

The CLidar measurements were taken at the sametime as the National Oceanic and Atmospheric Ad-

ministration (NOAA) aerosol-temperature–water-vapor lidar20 was operating. The aerosol channels ofthe lidar are designed for the stratosphere and uppertroposphere, which is where the CLidar’s resolution islow. So there is not a reliable altitude range to comparethe magnitude of the aerosol with molecular-scatterratio. But the altitudes of distinct layers of aerosol canbe compared. The CLidar data taken from 8:44 to 8:50UT (10:44 to 10:50 local time) show a layer with a peakaerosol-to-molecular-scattering ratio at 532 nm of 0.16.The layer extends over an altitude range from 4.6 to5.7 km with a peak at 4.8 km. The station altitude is3.4 km asl. At these altitudes, the CLidar scatteringangle is approximately 178° therefore the CLidar dataare readily comparable with 180° backscatter datawithout strong phase-function influences. The NOAAlidar’s infrared 1064 nm channel has a 300 m altituderesolution. The NOAA lidar infrared channel showsthe layer at 8:46 a.m. UT extending from 4.6 to 6.2km with a peak at 4.9 km. At this altitude, the over-lap correction to the NOAA lidar IR signal is approx-imately a factor of 47, while the CLidar data requireno overlap correction. This region is inaccessible tothe NOAA lidar green channel due to gating anddynamic-range considerations. The NOAA lidarwater-vapor channel indicates a drying of the atmo-sphere in the region of the layer, implying that thelayer represents an aerosol rather than a thin cloud.

6. Conclusions

The first- and second-generation CLidar systems havebeen shown to be capable of operating under molecularscattering, and low-to-moderate aerosol scattering con-ditions in both circularly polarized and linearly polar-ized configurations. The technique should also workunder conditions dominated by aerosol scattering aslong as signals are kept below the maximum pixelcount level. The second-generation system offers apeak signal (in photons) 2.5 times greater than thefirst-generation system and approximately the sameincrease in signal to noise. Comparison of retrievedaerosol-to-molecular scattering ratios from the first-and second-generation systems show excellent agree-ment. For the data of 18 April 2006, retrieved ratiosfrom the first- and second-generation systems show anaverage difference of approximately 0.01 over the ma-jority of the altitude range. The CLidar aerosol-to-molecular ratio has been shown to track ground-basednephelometer readings over aerosol ranges from 0.8 to37 Mm�1. When Aeronet measured phase functionsare used to calculated total scatter from the CLidardata, agreement was found. The CLidar data alsodemonstrate the system’s ability to accurately detectatmospheric layers that may occur within the overlaprange of traditional backscatter lidar systems.

The authors acknowledge the National Oceanic andAtmospheric Administration (NOAA) Global Monitor-ing Division aerosol group for providing aerosol data.The aerosol phase functions and aerosol optical depthsused were provided by the NASA�Aeronet group. The

Fig. 6. Comparison of total scatter of the CLidar (9–11 m) andcolocated nephelometer �10 m� for 21 December 2005 UT. Theerror bars indicate the range of total scatter when the minimumand maximum phase functions are used in the CLidar analysis.


work was supported by the National Science Founda-tion and NOAA.

References1. R. J. Charlson, S. E. Schwartz, J. M. Hales, R. D. Cess, J. A.

Coakley, J. E. Hansen, and D. J. Hofmann, “Climate forcing byanthropogenic aerosols,” Science 255, 423–430 (1992).

2. A. Ansmann, D. Althausen, U. Wandinger, K. Franke, D.Miiller, F. Wagner, and J. Heintzenberg, “Vertical profiling ofthe Indian aerosol plume with six-wavelength lidar duringINDOEX: a first case study,” Geophys. Res. Lett. 27, 963–966(2000).

3. J. Welton, “NASA micropulse lidar network,” http://mplnet.gsfc.nasa.gov/.

4. A. Ansmann, M. Riebesell, and C. Weitkamp, “Mearurementsof atmospheric aerosol extinction profiles with a Raman lidar,”Opt. Lett. 15, 746–748 (1990).

5. J. A. Reagan, D. M. Byrne, and B. M. Herman, “Bistatic lidar:a tool for characterizing atmospheric particulates: part 1-theremote sensing problem,” IEEE Trans. Geosci. Remote Sens.GE-20, 229–235 (1982).

6. J. A. Reagan, D. M. Byrne, and B. M. Herman, “Bistatic lidar:a tool for characterizing atmospheric particulates: part 2-theinverse problem,” IEEE Trans. Geosci. Remote Sens. GE-20,236–243 (1982).

7. E. J. Novitsky and C. R. Philbrick, “Multistatic lidar profilingof urban atmospheric aerosols,” J. Geophys. Res. 110, D07S11(2005).

8. J. R. Campbell, D. L. Hlavka, E. J. Welton, C. J. Flynn, D. D.Turner, J. D. Spinhirne, V. S. Scott III, and I. H. Hwang,“Full-time, eye-safe cloud and aerosol lidar observations atatmospheric radiation measurement sites: instruments anddata processing,” J. Atmos. Oceanic Technol. 19, 431–442(2002).

9. T. Halldorsson and J. Langerhoic, “Geometrical form factorsfor the lidar function,” Appl. Opt. 17, 240–244 (1978).

10. J. Harms, “Lidar return signals for coaxial and noncoaxialsystems with central obstruction,” Appl. Opt. 18, 1559–1566(1979).

11. D. F. Hurst, P. S. Bakwin, R. C. Myers, and J. W. Elkins,“Behavior of trace gases mixing ratios on a very tall tower inNorth Carolina,” J. Geophys. Res. 102, 8825–8835 (1997).

12. J. E. Barnes, S. Bronner, R. Beck, and N. C. Parikh, “Boundarylayer scattering measurements with a charge-coupled devicecamera lidar,” Appl. Opt. 42, 2647–2652 (2003).

13. K. Meiki, K. Yamaguchi, X. Li, Y. Sato, and A. Nomura,“Range-resolved bistatic imaging lidar for the measurement ofthe lower atmosphere,” Opt. Lett. 21, 1318–1320 (1996).

14. O. V. Kalashnikova and I. Sokolik, “Importance of shapesand compositions of wind-blown dust particles for remote sens-ing at solar wavelengths,” Geophys. Res. Lett. 29, 10.1029/2002GL014947 (2002).

15. X. Zhou, W. Sun, Z. Jiang, J. Ma, X. Zhang, Y. Byun, W. Chen,and J. Chen, “Modeling the CCD undersampling effect in theBATC photometric system,” Pub. Astron. Soc. Pac. 117, 86–93(2005).

16. B. Holben, “NASA aeronet,” http://aeronet.gsfc.nasa.gov/.17. O. V. Dubovik, T. V. Lapyonok, and S. L. Oshchepkov, “Im-

proved technique for data inversion: optical sizing of multi-component aerosols,” Appl. Opt. 34, 8422–8436 (1995).

18. S. F. Marshall, D. S. Covert, and R. J. Charlson, “Relationshipbetween asymmetry parameter and hemispheric backscatterratio: implications for climate forcing by aerosols,” Appl. Opt.34, 6306–6311 (1995).

19. A. McComiskey, E. Andrews, D. Jackson, A. Jefferson, S. W.Kim, J. Ogren, P. Sheridan, and J. Wendell, “Aerosol andradiation,” in Climate Monitoring and Diagnostics LaboratorySummary Report No. 27, R. Schnell, A. M. Buggle, and R. M.Rossen, eds. (Department of Commerce, 2004), p. 62.

20. J. E. Barnes and D. J. Hofmann, “Variability in the strato-spheric background aerosol over Mauna Loa Observatory,”Geophys. Res. Lett. 28, 2895–2898 (2001).


atmospheric aerosol profiling with a bistatic imaging lidar system

Documents