746 OPTICS LETTERS / Vol. 15, No. 13 / July 1, 1990

Measurement of atmospheric aerosol extinction profiles with aRaman lidar

Albert Ansmann, Maren Riebesell, and Claus Weitkamp

Institut ffir Physik, GKSS-Forschungszentrum Geesthacht GmbH, Postfach 1160, 2054 Geesthacht, Federal Republic of Germany

Received October 18, 1989; accepted April 11, 1990

A method is presented that permits the determination of atmospheric aerosol extinction profiles from measuredRaman lidar signals. No critical input parameters are needed, which could cause large uncertainties of the solution,as is the case in the Klett method for the inversion of elastic lidar returns.

The remote determination of aerosol concentrationprofiles has important applications in areas such asheterogeneous atmospheric chemistry, weather andclimate research, and pollution monitoring. Untilnow single- or multiple-wavelength backscatter lidarshave been used for this purpose. The appeal of thesedevices is their conceptional simplicity, ease of opera-tion, and high speed.

The basic formalism used to determine the aerosolextinction from elastic backscatter signals is the so-called Klett inversion method.1-3 This procedure,with all its subsequent modifications and improve-ments, suffers from the fact that two physical quanti-ties, the aerosol backscatter and extinction coeffi-cients, must be determined from only one measuredlidar signal. This is not possible without assumptionsabout the relation between the two and an estimate ofa boundary or reference value of the aerosol extinc-tion. These data are usually hard to assess and causelarge uncertainties in the aerosol extinction coeffi-cients that are determined.

In Raman lidar, the inelastic (Raman) backscattersignal is affected by aerosol extinction but not by aero-sol backscatter. Therefore analysis of the Raman li-dar signal alone permits the determination of the aero-sol extinction. There have been several descriptionsof Raman lidar applied to aerosol backscatter, visualrange, and optical depth,4 7 but until now a generalformalism for determining range-resolved aerosol ex-tinction coefficients from Raman lidar signals has notbeen given.

Let the Raman lidar equation be written as

P(z, XL XR) = BO2 (Z XL) XR)

X exp{-J [a(XL, t) + a(XR, t)]d4. (1)

Here P is the power received from distance z at theRaman wavelength XR if the laser pulse is transmittedat XL, O(z) is the overlap function between the laserbeam and the field of view of the receiver, a is thedepth-dependent total extinction coefficient at wave-lengths XL and XR, and B contains all depth-indepen-dent parameters. The backscatter coefficient ,B is

linked to the differential Raman backscatter cross sec-tion da/dQ of a gas of molecule number density N bythe relation

(Z, XL XR) = N(Z) dG(L XR, 7) (2)

The profiles of both nitrogen and oxygen can be usedsince their number densities are usually well known.With P(z) - P(Z, XL, XR), it follows from Eqs. (1) and (2)that

a(XL, Z) + a(XR, Z) =dz [ln OWzN(z)] (3)

O(z) is equal to unity if the path of the transmittedlaser beam is entirely within the field of view of thereceiver, which is the case above a certain minimumheight Zmin. The two extinction coefficients in Eq. (3)may be written as

a(XLR, Z) Smol(XLR, Z) + Saer(XLR, Z), (4)

where smol and Saer are the extinction coefficients dueto absorption and Rayleigh scattering by atmosphericgases and aerosol scattering and absorption, respec-tively. Equation (3) can then be written as

Saer(XLY Z) + Saer(R Z) = d1 n NW Iclz LZ2p(Z)J

-SmOI(L, Z) - Sm 0l((R, z)- (5)

Assuming a wavelength dependence of the aerosol ex-tinction such as

Saer(XL) = XR

Saer(XR) XL'(6)

we obtain from Eqs. (5) and (6) the unknown extinc-tion coefficient profile

Saer(XL, Z)

d n N[ (z) - SMOl(L, Z) - Smol(R, Z)=dz L z

2P(z) J 7

1 +XR

0146-9592/90/130746-03$2.00/0 © 1990 Optical Society of America

July 1, 1990 / Vol. 15, No. 13 / OPTICS LETTERS 747

This result holds for altitudes z > zmin for which theRaman lidar overlap function O(z) -- 1.

For an estimate of d/dz[ln N(z)] and Smol(L,R), theair density profile must be known. Height profiles ofthe atmospheric density can be assessed by using oneof several models available (such as the U.S. StandardAtmosphere) with measured values of the ground tem-perature and pressure, or they can be directly mea-sured with radiosondes. Light extinction by ozonecan also be taken into account. 8'9

The theory developed has successfully been appliedto profiles obtained with the GKSS Raman lidar.Since this system has been described in detail,10 onlythe more important features are repeated here. Theradiation source is a XeCl excimer laser, the transmit-ter and receiver are positioned side by side with theplanes of all curved mirrors horizontal, and the po-lychromator is a 17th-order echelle-type device with aholographic grating for cross dispersion. The techni-cal data of the system are given in Table 1.

Figure 1 shows an example of an aerosol extinctioncoefficient profile obtained with the Raman lidar us-ing the procedure outlined above. The statistical er,ror of the measurement, i.e., the error contributionfrom lidar signal noise, is given in the right-hand traceof Fig. 1. The number of laser shots (234,142) wasobtained in 19 min of measurement time. Before Eq.(7) was solved, the profile of the range-corrected signalP(z)z 2 taken with 60-m measurement resolution wassmoothed by forming a sliding average over 360 and1440 m for heights z < 3700 m and z > 3700 m, respec-tively. saer was then calculated with the same effectiveresolution. This must be kept in mind for the inter-pretation of the measured profile.

In addition to the statistical error as plotted in Fig.1, systematic errors must also be considered. Thecontribution from the molecular extinction uncertain-ty amounts to <0.01 km-' if the ozone concentrationdeviates by no more than a factor of 3 from the stan-dard ozone profile and the air density deviates by nomore than 5% from the standard atmosphere; this cor-responds to errors in the estimated value of tempera-ture and pressure of 3T < 10 K and 6p • 1 kPa. The

Table 1. GKSS Raman Lidar Technical Data

TransmitterPrimary wavelengthMaximum pulse energyMaximum repetition rateOptics

ReceiverOpticsDetectors

PolychromatorTypeMain dispersion gratingCross dispersion grating

Data acquisitionTypeWavelength channelsTime channelsMinimum time-bin widthMaximum count rate

308 nm270 mJ250 Hz400 mm, f/3.75

800 mm, f/3.75THORN EMI 9893 QB 350

Czerny-TurnerEchelle, 316 grooves/mmHolographic, 2400 grooves/mm

Photon counting31024100 nsec300 MHz

8000

H

6000

4000

2000

0 0.2 0.4 0.0 0 5 10 15

Saer, km-1

20

yS aer, %

Fig. 1. Atmospheric aerosol extinction coefficient saer andstatistical error 6

saer measured with a Raman lidar at 308-nmprimary and 323-nm Raman (02) wavelengths. The mea-surement was taken on May 17,1988, at 00:12:01 hours localtime in Geesthacht, Federal Republic of Germany. Mea-surement time is 19 min, and the total number of laser shotsis 234,142. Ground values for the standard atmospheremodel are 100.8 kPa and 150C. The discontinuity at 3700 mreflects the change from 360 to 1440 m of averaging height atthis level. The molecular extinction coefficient smol is alsoshown for comparison (the dashed curve).

term d/dz [ln N(z)] can contribute significantly to theerror of Saer when standard atmosphere conditions areassumed but a temperature inversion is present. Atemperature gradient dT/dz between 0 and 3 K/100 mleads to an error bsaer of between 0.015 and 0.025 km-1.Temperature inversions are mainly observed in thelower troposphere, especially at the top of the bound-ary layer. The error due to non-X-1 behavior of theaerosol extinction amounts to +1% if the right-handside of Eq. (6) is raised to the power k and k is allowedto vary between 0.8 and 1.2 Error considerationsthus show that, owing to the relatively large sensitivityof the aerosol extinction coefficient at 308 nm to errorsin the air density estimate, aerosol measurements inclean, near-Rayleigh atmospheres are only possible incombination with air density determinations as can beobtained from radiosonde measurements of tempera-ture and pressure.

The advantage of the Raman method over the Klettmethod lies in the closed-form solution of the formeras opposed to the recursive formula of the latter. Cat-astrophic instabilities as are common in the Klettmethod cannot occur in the Raman approach; thepropagation of errors of input parameters into theaerosol extinction coefficient is straightforward; andthe nature of the input parameters is such that a mea-surement, and thus a small error in Saer, appears feasi-ble.

In conclusion, it has been demonstrated that Ramanlidar measurements provide aerosol extinction datawith adequate accuracy and spatial and temporal reso-lution. The main shortcoming of the Raman methodis the greater complexity of the apparatus. The limi-tation of the present GKSS system to nighttime mea-surements, caused by the choice of the 308-nm radia-tion, can be alleviated by switching to shorter wave-lengths, although increased ozone absorption and

I- I h

I , II

. II

748 OPTICS LETTERS / Vol. 15, No. 13 / July 1, 1990

larger effects of air density uncertainties must be care-fully investigated.

The authors gratefully acknowledge valuable sug-gestions by L. Bissonnette.

References

1. J. D. Klett, Appl. Opt. 20, 211 (1981).2. F. G. Fernald, Appl. Opt. 23, 652 (1984).3. Y. Sasano, E. V. Browell, and S. Ismail, Appl. Opt. 24,

3929 (1985).4. S. H. Melfi, "Remote measurement of atmospheric

transmissivity," in Proceedings of the Fourth Interna-tional Laser Radar Conference (American Meteorologi-cal Society, Boston, Mass., 1972).

5. Y. A. Arshinov, S. M. Bobrovnikov, and U. E. Zuev, inProceedings of Twelfth International Laser RadarConference (Service d'A6ronomie, Centre National de la

Recherche Scientifique, Verribres-le-Buisson, France,1984), p. 63.

6. R. A. Ferrare, S. H. Melfi, and D. Whiteman, in Proceed-ings of Fourteenth International Laser Radar Confer-ence (Istituto di Ricerca Sulle Onde Elettromagnetiche,Comitato Nazionale per le Scienze Fisiche, Florence,Italy, 1988), p. 159.

7. L. V. Kravets, Atmos. Opt. 2, 146 (1989).8. L. Elterman, Rep. AFCRL-68-0153, Environmental Re-

search Papers No. 285 (Air Force Cambridge ResearchLaboratories, Bedford, Mass., 1968).

9. M. Griggs, J. Chem. Phys. 49, 857 (1968).10. C Weitkamp, M. Riebesell, E. Voss, W. Lahmann, and

W. Michaelis, in Remote Sensing of Atmosphere andOceans (Austrialian Defence Force Academy, Canberra,Australia, 1988), Vol.2, p. 66; Rep. GKSS 88/E/ (GKSS-Forschungszentrum Geesthacht, Geesthacht, FederalRepublic of Germany, 1988).