Ozone and water-vapor measurements by Raman lidarin the planetary boundary layer: error sources andfield measurements
Benoıt Lazzarotto, Max Frioud, Gilles Larcheveque, Valentin Mitev, Philippe Quaglia,Valentin Simeonov, Anne Thompson, Hubert van den Bergh, and Bertrand Calpini
The degradation of air quality is a serious environ-mental problem that affects urban and industrial ar-eas worldwide. Air pollution injures human healthand ecosystems, diminishes crop yield, and spoils pat-rimony and materials. The phenomena involved inair pollution are complex. Once they are emittedinto the atmosphere, ~primary! pollutants are trans-ported, dispersed, transformed by gas–solid phasechange and chemical reaction, and finally removed bydry and wet deposition.
Most challenging is the fact that the health andenvironmental effects of secondary pollutants~formed in the atmosphere! are frequently more se-vere than those of their precursors ~primary pollut-
B. Lazzarotto, G. Larcheveque, P. Quaglia, V. Simeonov, H. vanden Bergh, and B. Calpini [email protected]! are with the
idar Group, Laboratory for Air Pollution, Swiss Federal Institutef Technology, CH-1015 Lausanne, Switzerland. M. Frioud and. Mitev are with the Observatory of Neuchatel, CH-2000 Neu-hatel, Switzerland. A. Thompson is with the NASA Goddardpace Flight Center, Code 916, Greenbelt, Maryland 20771.Received 14 June 2000; revised manuscript received 23 January
ants!. This is so for ozone and other photochemicalpollutants, such as peroxyacetyl-nitrate, and for sec-ondary particles produced in the atmosphere by thephotooxidation of volatile organic compounds cata-lyzed by nitrogen oxides ~NOx!. Photochemical airpollution is a complex science because of the nonlin-earity of its response to changes in primary emission~see, for example, Ref. 1!.
Three-dimensional air quality models are used ashe most powerful tools for identifying effective strat-gies to improve air quality. With the mesoscaleulerian chemical transport model developed at thecole Polytechnique Federal Lausanne ~EPFL! wean simulate pollutant dynamics over cities such asthens and Milan and over regions with high traffic
oads in Switzerland and provide technical guidanceo air quality management agencies.2–4 The model
resolution is of the order of 1 km on the horizontalscale, with a vertical resolution of some tens of metersfor the lowest layer of the model, as much as 500 m forthe top layer, and a total height of 5 km above groundlevel. The domain size is typically 100 km 3 100km. Before the model results can be used with con-fidence, they must be validated against field mea-surements with similar spatial and time resolutions.
Most of the time, an air quality network in adensely urbanized region is built upon a set of ground-
20 June 2001 y Vol. 40, No. 18 y APPLIED OPTICS 2985
w
2
based stations equipped with point detectors. Trace-gas measurements are often influenced by localsources and thus are not representative of the averageconcentrations over the typical grid size of the model.Tropospheric lidar measurements, however, are basedon an integrated optical path of typically 50–500 m,depending on the trace-gas species. This spatial res-olution is in ideal agreement with the model resolu-tion. Therefore the lidar measurements can supplyessential information, including ozone and water-vapor vertical profiles, for validation of the model.
Ozone as a secondary pollutant is an ideal speciesfor comparison with its predicted values obtainedfrom the model. Because ozone is produced in themodel by photochemical reactions and transport ef-fects, good agreement between field measurementsand model values is an indicator of the model perfor-mance.
The water-vapor content of the atmosphere plays amajor role in the dynamics and climatology of theatmosphere and is also a clear tracer of the dailyevolution of the top of the planetary boundary layer~PBL!. Thus its continuous remote detection over aperiod of time of some days, the typical duration of anair pollution event, may also contribute significantlyto improvement of the model predictions.
In this paper we present the design and develop-ment of a Raman lidar system with optimized reso-lution in comparison with model results. Thisozone–water-vapor Raman lidar development is afollow-up of our recent research on an elastic ozonedifferential absorption lidar ~DIAL! technique.5–7
This technique is a method with higher sensitivitythan Raman lidar, and with well-established ozonemeasurements in atmospheric conditions in whichthe aerosol density is low enough, or homogeneous, assuch in the free troposphere. In the PBL, wherehighly variable aerosol concentrations are frequentlyobserved, elastic DIAL may fail or even no longer beused because of strong and unpredictable aerosol op-tical interference on the ozone retrieval.8,9 In par-ticular, the aerosol backscatter and its high butpoorly known wavelength dependence were pointedout by Volger et al.10
The PBL is also the atmospheric layer with the high-est vertical resolution in the model. This fact broughtus to the idea of developing an alternative instrumentto the elastic DIAL, an instrument much less per-turbed by the nonhomogeneous aerosol load conditionin the PBL and able to perform simultaneous ozoneand water-vapor measurements with high resolutionat low altitudes. This instrument is based on mea-surement of the Raman-shifted backscattered light in-duced by the most abundant molecular species in theatmosphere, i.e., nitrogen, oxygen, and water vapor,from an UV laser pulse emitted in the atmosphere.The water-vapor content is obtained by the classic Ra-man analysis.11,12 Ozone is calculated by a differen-tial absorption method that uses oxygen and nitrogenRaman backscatter as ON and OFF signals. The vari-ous Raman return signals are generated simulta-neously from a single laser source, thus probing the
986 APPLIED OPTICS y Vol. 40, No. 18 y 20 June 2001
same volume of air at a given time and essentiallyavoiding the problems related to pulse-to-pulse laserstability or to atmospheric turbulence that usually oc-cur for most elastic DIAL instruments in which twosuccessive pulses are emitted.13 The disadvantagecaused by the weaker Raman signals compared withthe elastic signals is compensated for by the high mo-lecular densities and the well-known values of the Ra-man cross sections. This idea refers to the pioneeringstudies of Melfi et al.11 and by Renault et al.12; thoseauthors published a first measurement of the verticalwater-vapor distribution in the PBL and later pro-posed to correct the water-vapor Raman return fortropospheric ozone attenuation.14 The same principlewas successfully applied for the stratosphere.15
Here we apply these earlier concepts to the devel-opment of an operational lidar instrument for day-time and nighttime measurements and for timeseries of some days to follow the vertical dynamicsand time evolution of an air-pollution episode. InSection 2 of this paper we present a model estimatefor determinating the critical system parameters.In Section 3 we define the experimental setup, and inSection 4 we give some typical results obtained forvarious time series of ozone and water-vapor concen-trations and comparisons with other instruments.
2. Raman Lidar: Principle and Predicted Sourcesof Error
Figure 1 is a schematic of the Raman lidar system.The ozone retrieval is based only on the nitrogen andoxygen Raman backscattered signals. For thewater-vapor retrieval we can use either nitrogen andwater-vapor or oxygen and water-vapor pairs of sig-nals. Each of the three Raman-shifted wavelengthscorresponds to a lidar equation:
PX~lXRaman, R! 5 PL~lL! KX
O~R!
R2 nX~R!bXRamanDR
3 expH2*0
R
@aL~r! 1 aX~r!
1 nO3~r!~sL 1 sX!#drJ , (1)
where the subscript X stands for ozone, nitrogen, orater vapor; PX~lX
Raman, R! is the Raman lidar powerbackscattered from species X at Raman-shifted wave-length lX
Raman and distance R; PL~lL! is the laseremitted power at wavelength lL; KX is the instru-ment constant at Raman-shifted wavelength lX
Raman;O~R! is the telescope active surface area; nX~R! is themolecular density of species X at distance R; bX
Raman
is the Raman differential backscattering cross sectionfor species X; aX~r! and aL~r! are the atmosphericextinction coefficients, without the ozone absorptioncoefficient, at Raman-shifted wavelength lX
Raman andat pump laser wavelength lL, respectively; and sXand sL are the ozone absorption cross sections atRaman-shifted wavelength lX
Raman and the pump la-
t
ssrof
s
Dil
ma
owtr
w
d
at
ser wavelength lL, respectively. The ozone concen-ration nO3
~R! can be calculated from the nitrogenand oxygen Raman signals by use of a modified DIALequation in the following form:
nO3~R! 5
1sO2
2 sN2
ddR Fln
PN2~R!
PO2~R!G
2@aO2
~R! 2 aN2~R!#
sO22 sN2
. (2)
The water-vapor mixing ratio in g of H2Oykg of dryair is obtained from the ratio of the water-vaporRaman-shifted signal to either the Raman nitrogen-or the Raman oxygen-shifted signal. Using the ni-trogen channel yields the following expression:
nH2 O
nN2
~R! 5PH2 O
PN2
~R!KN2
KH2 O
bN2
Raman
bH2 ORaman
3 exp(*0
R H~sH2 O 2 sN2!nO3
~r!Çozone correction
1 @aH2 O~r!
2 aN2~r!#Jdr . (3)
Fig. 1. Schematic of Raman lidar. A pulsed laser beam ~PB! isemitted into the atmosphere via a beam expander ~33! and a set ofright-angle prisms. The elastically backscattered signal ~EBS!and the Raman-shifted backscattered signals ~RBSs! are collectedby a 200-mm Newtonian telescope, filtered at the entrance of apolychromator where they are spectrally resolved, and detected bythree photomultiplier tubes ~PMTs!. The PMT signals are storedin a transient recorder ~20-MHzy12-bit and photon counting! anda PC-based computer unit allows for real-time ozone and water-vapor retrieval ~raw data!.
)
As the oxygen signal is approximately three timesweaker than the nitrogen signal, the water-vapormixing ratio calculated by use of an oxygen Raman-shifted signal will be less accurate, and we use it onlyas an additional test of the consistency of the water-vapor result. Note that in Eq. ~3! the water-vaporconcentration retrieval depends linearly on the ratioof the Raman water-vapor signal to the Raman ni-trogen signal, whereas in Eq. ~2! the ozone concen-tration depends on the first derivative of the twoRaman lidar signals ~nitrogen versus oxygen!. Wehall see below that the various predicted errorources for ozone and water vapor will have a strongelation to this dependence and that in any case thezone differential retrieval will be much more af-ected than the water-vapor linear retrieval.
As the ozone retrieval does not depend on the in-trument constant KX, the method can be regarded as
self-calibrated. In contrast, for absolute water-vapor measurements, exact knowledge of KX or inde-pendent parallel measurements for instrumentcalibration are needed. The term ~sH2O 2 sN2
!nO3~r!,
denoted ozone correction in the exponential part ofEq. ~3!, reflects the ozone influence on the water-vapor and nitrogen Raman signals and can signifi-cantly affect the final results,12 as is emphasizedbelow.
A. Molecular and Aerosol Dependence of theRaman Lidar
From Eq. ~2!, the ozone retrieval does not depend onthe backscattering properties of the atmosphere.This is the major advantage of the Raman–DIALmethod compared with the conventional ~elastic!
IAL and may be of crucial importance, particularlyn the case of a strong and inhomogeneous aerosoload.10 The influence of atmospheric extinction on
ozone retrieval is expressed by the second term of Eq.~2!. This term also appears in the elastic DIAL for-
alism and can be split into its molecular and itserosol contributions.First, the molecular atmosphere contribution to
zone retrieval is estimated by use of the Rayleighavelength dependence for the extinction. One ob-
ains the following ~negative! correction to the ozoneelative content:
Dmol 5aO2
mol 2 aN2
mol
nairmol~sO2
2 sN2!
, (4)
here nairmol is the air’s molecular density. As the
numerator in Eq. ~4! also directly depends on the airensity, the result Dmol is given as an altitude-
independent value in this formalism. It correspondsto a correction of ;23 parts in 109 by volume ~ppbv!.
Let us now add a homogeneous aerosol layer char-cterized by two variables to the model, namely, theotal lidar ratio e, defined as the total extinction atotal
divided by the total backscattering btotal, and thebackscattering ratio b, defined as the total backscat-tering btotal divided by the molecular backscatteringbg. With the assumption of atmospheric molecular
20 June 2001 y Vol. 40, No. 18 y APPLIED OPTICS 2987
f
lTc
w
lvli
Table 1. Spectroscopic Data Used for a 266-nm Pump Laser Source Table 2. Experimental Inputs for the Raman Lidar Simulation
2
species with a mean diameter much smaller than thelaser wavelength, molecular backscattering bg is de-fined according to Collis and Russell16 by
bg 5 nairmol dsRayleigh~p!
dV5 nair
molS550lL
D4
5.45 3 10232,
(5)
where the molecular backscattering bg is given inunits of inverse meters per steradian, the laser wave-length lL is in nanometers, and the air density nair
mol isin inverse cubic meters. One may also assume apower law for the wavelength dependency of the ex-tinction that is caused by Mie particles ~aaer ; lg!,ollowing previous research17 in which g was shown
to have a range of g [ @21, 20.5#. With g 5 21, thefollowing estimation of the aerosol correction Daer tothe ozone concentration is obtained:
Daer
Dmol 5 ~lL!23~lO2
Raman!21 2 ~lN2
Raman!21
~lO2
Raman!24 2 ~lN2
Raman!24 S ebcRayl 2 1D , (6)
where cRayl 5 8py3 is the molecular lidar ratio ~Ray-eigh contribution!. With the parameters listed inable 1, this notation yields the following aerosolontribution to the ozone correction:
Daer > ~0.035 3 eb 2 0.3!Dmol. (7)
Here we have used for the molecular density and thewater-vapor content a model atmosphere defined inRef. 22 ~Model 3!, with “mid-latitude winter” condi-tions. The water-vapor vertical profile was takendirectly from this model atmosphere but was multi-plied by a factor of 3 for comparison with experimen-tal results presented below. The total extinctionand backscattering coefficients were defined for thefourth harmonic of a Nd:YAG laser source at 266 nm,with the effect of a constant aerosol vertical profiletaken into account. To simulate the most severeaerosol conditions, we tuned the aerosol’s opticalproperties to their maximum acceptable ~or worse!values for e and b, 40 and 1.8, respectively. With
988 APPLIED OPTICS y Vol. 40, No. 18 y 20 June 2001
these extreme aerosol conditions, the ozone shift Daer
that is due to the aerosol for this Raman DIALreaches a value of 6.7 ppbv @from Eq. ~7! for g 5 21#,
hereas for g 5 20.5 the ozone shift is Daer 5 3.4ppbv. In similar conditions the systematic relativeerror caused by differential aerosol backscatter influ-encing the elastic DIAL ~and not the Raman DIAL!would have exceeded 100%.16 This error is negativeand leads to nonphysical negative ozone concentra-tions. This difficulty in the case of elastic DIALmeasurements is impossible to overcome withouttheir being additional information about the back-scatter properties of the aerosol or without additionalassumptions about the aerosol properties.8
B. Ozone Interference in Water-Vapor Raman LidarRetrieval
Ozone absorbs in the Hartley and Huggins bands at220–350 nm and thus affects the signal at each of theRaman oxygen, nitrogen, and water-vapor wave-lengths. The effect of the ozone absorption on thewater-vapor retrieval defined in Eq. ~3! was simu-ated with a model atmosphere with various constantalues of ozone concentration, a homogeneous aerosoload ~e 5 40, b 5 1.8!, and the lidar parameters listedn Table 2. The ozone corrections DH2O for the
water-vapor retrieval are shown in Fig. 2 as the dif-ference between the water-vapor retrieval profile
Fig. 2. Ozone effect on water-vapor retrieval. Predicted error inthe water-vapor mixing ratio retrieved by Raman lidar as a resultof various ozone constant vertical profiles and a homogeneous aero-sol load. The horizontal scale is the difference between the water-vapor retrieval without the ozone absorption taken into accountand the same retrieval with ozone absorption.
Laser pulse energy 120 mJTransmission efficiency 90%Receiver efficiency 20%Quantum efficiency 10%Telescope diameter 20 cmPMT gain 105
Impedance 50 VNumber of shots 5 3 4000Sampling rate 20 MHzFull overlap range 200 m
H
hsmsFri
trcT
without ozone interference and the same retrievalcorrected for the ozone absorption effect. The cor-rection values are set to 0 g of H2Oykg of dry air at200 m, at which a full overlap between the laser beamand the telescope’s field of view is assumed. Themodel shows that, even for moderate ozone concen-trations of 40 ppbv, the correction can exceed 1 g ofH2Oykg of dry air at 600 m and more than 2.5 g of
2Oykg of dry air at 1200 m, that is, ;20% of thewater-vapor content in the standard atmosphere.
The integrated ozone content that one needs to beable to take the ozone correction into account can beobtained either by independent ozone ~e.g., balloon!measurements or by additional use of the ratio of theoxygen and nitrogen Raman signals. The firstmethod requires additional absolute ozone concentra-tions but is not convenient for practical purposes andtherefore is more suitable for calibration and verifi-cation of the lidar system. The second method hasthe advantage of a self-corrected water-vapor mea-surement with the risk of additional uncertainty ow-ing to the contribution of each of the three Ramanbackscattered lidar signals. Both methods were ap-plied in our experiments.
In the second method, the integrated ozone contentto distance R can be calculated from oxygen and ni-trogen Raman signals as
expF*0
R
nO3~r!drG 5 FPO2
PN2
~R!G1y~sN22sO2
!
3 FKN2
KO2
nN2
nO2
~R!bN2
Raman
bO2
RamanG1y~sN22sO2
!.
(8)
Further, by replacing the integrated ozone content inEq. ~3! by Eq. ~8! and neglecting the differential ex-tinction that is due to Rayleigh and Mie scatteringprocesses in accordance with Ref. 12, we can deter-mine the water-vapor mixing ratio as
nH2 O
nN2
~R! 5PH2 O
PN2
~R!KN2
KH2 O
bN2
Raman
bH2 ORaman
3 FKN2
KO2
nN2
nO2
~R!bN2
Raman
bO2
RamanGsH2 O2sN2ysN2
2sO2
3 FPO2
PN2
~R!GsH2 O2sN2ysN2
2sO2
(9)
or as
nH2 O
nN2
5 Kcal
PH2 O~R!
PN2~R!FPO2
~R!
PN2~R!G
sH2 O2sN2ysN2
2sO2
, (10)
where the factor Kcal is the overall instrument con-stant value and must be established by calibration ofthe lidar.
C. Statistical ~Quantum! Noise
The statistical noise is a major perturbation in Ra-man lidar measurements17 because of the low Ramancross section, typically 4 orders of magnitude lowerthan the elastic cross section, and therefore becauseof the low Raman signal level. The statistical noisehas been modeled following the Poisson statisticswith the parameters defined in Table 2. In this casethe model study was performed in a purely molecularatmosphere ~no aerosol load!. Based on typical ex-perimental values, we estimate a number of photo-electrons per pulse at the photocathode, from adistance of 200 m and for one analog-to-digital con-verter ~ADC! channel of 7.5-m resolution, of some
undreds for both nitrogen and oxygen Raman lidarignals and some counts for water vapor. Thisodel simulation of the statistical noise is based on
eries of 100 runs for each similar initial condition.rom these series of runs the standard deviation withespect to the mean value is obtained, and it is shownn Fig. 3 relative to the range.
Figure 3~a! shows the ozone standard deviationhat is due to the effect of statistical noise on theetrieved ozone concentration for three profiles withonstant ozone concentrations of 0, 40, and 80 ppbv.hese results were achieved by averaging over five
Fig. 3. Statistical noise. Predicted standard deviations for ~a!ozone and ~b! water-vapor Raman lidar retrieval as a result ofstatistical noise ~Poisson statistics! with the experimental param-eters defined in Table 2.
20 June 2001 y Vol. 40, No. 18 y APPLIED OPTICS 2989
as
rwarnpAfvtsttmiatdttaash
pTaor
T
2
files of 4000 laser shots as defined in Table 2. Witha sampling rate of 20 MHz the ultimate range reso-lution was 7.5 m. Such a short optical integrationpath would nevertheless result in too large an uncer-tainty for the ozone retrieval. Thus we used an ef-fective range resolution of 90 m, or an equivalent binresolution of 12 ADC channels. Because the pres-ence of ozone itself significantly lowers the Ramansignals through extinction, higher ozone values di-rectly induce greater statistical error. The results inFig. 3~a! indicate that, on the one hand, an averagegreater than at least 20,000 shots is needed for suit-able ozone Raman DIAL retrieval. This means thatfor a laser source with a 10-Hz repetition rate theaveraging time will be approximately half an hour.On the other hand, above an altitude range of typi-cally 700 m above ground level ~AGL! the error sourcethat is due to the statistical noise reaches a valuehigher than 5% of the measured ozone concentrationin the case of a constant profile with 80 ppbv. Thisuncertainty is considered an upper acceptable limit.Note that even at O3 5 0 ppbv the statistical noisethat affects the oxygen and nitrogen Raman signalsalready induces an ozone standard deviation. Fromthis simulation of the statistical noise only, one canconclude that the maximum achievable range forozone with this Raman DIAL system will be of theorder of 700 m above the lidar site, with a spatialresolution of 90 m, a time resolution of half an hour,and a precision that is due to the statistics of 65% for
typical ozone concentration of 80 ppbv in the atmo-phere.A similar statistical analysis was performed for
etrieval of the Raman water-mixing ratio @Fig. 3~b!#,here again the three constant ozone profiles ~0, 40,nd 80 ppbv! were considered. In this case theange resolution was set at 22.5 m ~three ADC chan-els!, and reasonable water-vapor estimates wereredicted for an altitude range of as much as 1200 mGL. Note that the standard deviation is calculated
or a water-vapor mixing ratio with a referencedalue given at 200 m AGL ~full overlap condition! byhe water-content model atmosphere. In compari-on with the results obtained for ozone, at 700 m AGLhe water-vapor standard deviation in an atmospherehat contains 80 ppbv of ozone is less than 1% of theean water-vapor mixing ratio. This better result
s due directly to the linear dependence of Eq. ~3! onwater-vapor retrieval that is less sensitive to sta-
istical error than in the case of the ozone Ramanifferential analysis. Here one can conclude thathe Raman lidar instrument will yield an estimate ofhe water-vapor content in the atmosphere with anccuracy of better than 2% ~statistical noise only! forn altitude range of 1200 m above the lidar site, apatial resolution of 22.5 m, and a time resolution ofalf an hour.
D. Optical Cross Talk in the Detection Box
Spectral separation of the three Raman signals andrejection of the elastic wavelength are performed by agrating polychromator. For such a device we could
990 APPLIED OPTICS y Vol. 40, No. 18 y 20 June 2001
consider mutual optical cross talk among the Ramanchannels and a leak of an elastic signal that has notbeen totally suppressed into the Raman channels.These two types of optical cross talk are simulated:One is due to the elastically backscattered lightadded to each of the different Raman channels andthe other is due to the cross talk between adjacentRaman-shifted channels.
Let us consider first the cross talk between nitro-gen and oxygen Raman signals that bias the ozoneretrieval and the cross talk between water-vapor andoxygen Raman signals that bias the water-vapor re-trieval. Several model runs were performed forcross talk that ranged from 1026 to 1023 with a modelatmosphere with a constant ozone vertical profile ~80
pbv! and an aerosol layer with b 5 1.8 and e 5 40.he model results show that this cross talk has onlyn extremely small effect. The associated errors forzone were always less than 0.3 ppbv, whereas theyemained below 0.01 g of H2Oykg of dry air for the
water vapor in the worst case. As an example ofthese simulation runs, Fig. 4 shows the cross-talkeffect that is due to the Raman-shifted oxygen signalon water vapor relative to the altitude range. Be-cause the water-vapor Raman signal has by far theweakest signal intensity compared with the two otherRaman signals, mainly because of its comparativelymuch lower concentration, one could expect the stron-gest ~or worst! effect of the cross talk in this case.
he predicted shift DH2O in the water-vapor mixing-ratio retrieval presented in Fig. 4 is the differencebetween water-vapor retrieval with no cross-talk ef-fect and the same retrieval biased by the cross talk ofthe oxygen signal on water vapor. As these runs areperformed in an atmosphere with a high aerosol loadand a constant 80 ppbv of ozone, the simulationshows that the largest cross-talk effect is expectedover long range, where the signal-to-noise ratio is theweakest. For a cross talk of 1023 at a range of 200 mthe expected shift is ;21023 g of H2Oykg of dry air
Fig. 4. Optical cross talk. Predicted error in water-vapor Ra-man lidar retrieval owing to optical cross talk between water-vaporand oxygen Raman signals. The horizontal scale is the differencebetween the water-vapor mixing-ratio retrieval without the cross-talk effect and the same retrieval biased by the cross-talk effect.
22
a
hu
ttlpasavLbIt
Aova
t
~
vdf
Rtra
wv
~negative correction!, and it reaches a value of 210at 1200 m, or an effect that is higher by typically 1order of magnitude. This simulation is performedwith a constant detection efficiency relative to therange. In this sense the effect of incomplete detec-tion of the Raman signal at a short distance wherethe probed air volume image in the grating polychro-mator is the largest is not taken into account.
For the cross talk between the elastic backscat-tered light and the Raman channels, one can expect astronger effect because the elastic backscatter crosssection is as much as 4 orders of magnitude higherthan the Raman cross section. This result is ad-dressed in detail in the experimental layout pre-sented in Section 3 below. To prevent such strongoptical interference, additional filters were set at theentrance of the polychromator with a rejection ratiobetween the 266-nm light and the other Raman chan-nels of more than 5 orders of magnitude. Modelruns were performed with cross-talk values thatranged from 1029 to 1026 between the elastic back-scattered light and any of the Raman channels. Theassociated errors that are due to the elastic cross talkfor ozone always remained below 20.15 ppbv, andthose for water vapor below 20.1 g of H2Oykg of dryir in the worst case.In summary, the two types of optical cross talk
ave negligible effects if the wavelength separationnit permits a cross-talk level lower than 1023 to be
achieved for two adjacent Raman channels, and alevel lower than 1026 between the elastic backscat-tered signal at 266 nm and the closest Raman back-scattered signal, namely, oxygen at 277.5 nm.
E. Afterpulse Effect
Another instrumental effect that could strongly alterthe real lidar return signal is the afterpulse effect~APE! of the receiving photomultiplier.17 The after-pulses caused by internal processes within the PMTappear as secondary pulses that follow the genuinepulse. In modern PMTs, most afterpulses are as-sumed to be due primarily to positive ions ~eitherresidual ions from manufacture or atoms of helium,which diffuse through the glass envelope! that strikethe photocathode to release secondary electrons.The APE occurs mostly within 1–2 ms after the mainpulse.23,24
The time delay and the duration and the shape ofthe afterpulse depend on the ions involved and on thePMT configuration.25 The influence of the APE onhe lidar signal can be estimated as a convolution ofhe afterpulse produced by a short light pulse and theidar signal. The result is a bias that is superim-osed upon the original lidar signal. Because thefterpulse is shorter than the duration of the lidarignal, the influence of the APE can be presented asdelayed echo of the lidar signal with an integrated
alue that is proportional to the lidar signal itself.et us define the APE’s relative intensity as the ratioetween the APE signal and the Raman lidar signal.n our simulation we chose a time delay for the af-erpulse of 2 ms, or an equivalent range of 300 m after
the lidar maximum intensity, defined at 200 m farfrom the lidar experiment. The model atmospherewas chosen with 80 ppbv of ozone and the homoge-neous aerosol layer ~e 5 40, b 5 1.8!. The simulated
PE error is defined as the difference between thezone concentration profile @Fig. 5~a!# and the water-apor mixing ratio profile @Fig. 5~b!# with the APEnd the profile without the APE.The APE on ozone retrieval is shown in Fig. 5~a! for
hree values of the APE ratio. With an APE of 1023
the maximum predicted ozone error is 211 ppbv;14%!, whereas for an APE below 1024 this effect
remains within the acceptable limit of precision of themethod.
The same effect is studied in Fig. 5~b! for the water-apor mixing-ratio retrieval. The APE error in-uced in the water-vapor mixing ratio is less than 1%or an APE intensity of 1023. The difference in the
magnitude and the shape of the errors in ozone con-centration and water-vapor mixing ratio is again ex-plained by the fact that the ozone calculations in Eq.~2! include taking a derivative, which operation re-veals a characteristic perturbation in the signal. Inconclusion, if the photomultiplier units that one uses
Fig. 5. APE. Predicted error in ~a! ozone and ~b! water-vaporaman lidar retrieval owing to the APE. The horizontal scale is
he difference between the ozone and the water-vapor mixing-ratioetrievals with and without the APE. The model lidar signals arelso shown relative to their typical ADC voltage intensities ~log
scale! with the Raman oxygen and nitrogen signals for ozone andith the Raman water-vapor and oxygen signals for the water-apor mixing ratio.
20 June 2001 y Vol. 40, No. 18 y APPLIED OPTICS 2991
Onmc
mfmdwd
n
ctutmtrpar
lrTdTasficott
d
2
to acquire the various Raman signals are tested withan APE lower than 1024, this simulation will predictessentially no remaining effect on the signal analysis.
F. Additional Sources of Systematic Error in Raman Lidar
Systematic errors such as interference with othertrace-gas species in the atmosphere or in the spectralstability of the laser source must also be considered.
In Eq. ~2! ozone was the only absorbing speciestaken into account. The DIAL technique is sensitiveto the influence of any other absorbing trace-gas spe-cies in the wavelength region used. The main inter-ference in the UV for tropospheric measurementscomes from SO2 and NO2. Table 1 lists their respec-tive absorption cross sections compared with that ofozone for the three Raman wavelengths considered inthis study. The systematic error DO3 induced in theozone retrieval by any interfering gas ~IG! can bepresented as
DO3 5 2@sIG~lN2
! 2 sIG~lO2!#NIG
@sO3~lN2
! 2 sO3~lO2
!#, (11)
where NIG is the concentration of IG. The ratio ofthe differential absorption cross section for IG 5 NO2to ozone is 27.61 3 1023, whereas it reaches a valueof 21021 for SO2. In other words, the systematicerror DO3 in the ozone concentration caused by 100ppbv of NO2 is less than 21 ppbv, whereas 100 ppbvof SO2 will induce a systematic error of 210 ppbv.
ne should point out that such error sources affectot only the Raman DIAL but also any elastic DIALeasurements, but remain relatively small in most
ase studies.The systematic error induced by IG in water-vaporeasurements can be calculated by use of the same
ormalism as the one used for ozone correction. Theagnitude of this error will be proportional to the
ifferential cross section of the IG at nitrogen andater-vapor Raman wavelengths ~see Table 1!. Theifferential cross sections of NO2 and SO2 are, corre-
spondingly, 22.9 3 1022 and 21.17 3 1022 of theozone differential cross section. Inasmuch as the ex-pected NO2 and SO2 systematic errors represent aproportionally small part of the ozone correction, thelatter can be neglected. One should neverthelesspoint out that this correction term appears in theintegral term in Eq. ~3! and will play a more signifi-cant role in the calculation for a longer range of mea-surements.
The spectral stability of the laser source could aswell be a source of systematic error because a shift inthe emitted wavelength will induce a displacement ofthe lidar signal image at the output of the polychro-mator. Hence the spectral stability of the lasersource specified by the manufacturer is 1 cm21, or awavelength shift of less than 0.015 nm at 266 nm.With a grating spectral resolution of 0.51 nmymm thewavelength shift that is due to the spectral stabilityof the laser source will be 30 mm. This effect can be
eglected.In summary, interference with other trace-gas spe-
992 APPLIED OPTICS y Vol. 40, No. 18 y 20 June 2001
ies can also be ruled out in most atmospheric condi-ions. The remaining ~and by far the largest!ncertainty in this experiment will be the signal sta-istics. An accuracy of 5% for an altitude range of asuch as 700 m, a spatial resolution of 90 m, and a
emporal resolution of 30 min is expected for ozoneetrieval in typical atmospheric conditions with 80pbv of ozone, whereas 2% accuracy is expected for anltitude range of as much as 1200 m and a spatialesolution of 22.5 m for the water-vapor retrieval.
3. Raman Lidar Experimental Setup
The Raman lidar system shown in Fig. 1 is based ona frequency-quadrupled Nd:YAG laser ~Continuum,Inc. Powerlite-8000! used in the transmitter.26 Theaser output energy at 266 nm is 120 mJ at a 10-Hzepetition rate with a 7-ns FWHM pulse duration.he initial laser beam divergence of 0.5 mrad is re-uced to 0.17 mrad by a three-time beam expander.he beam is emitted into the atmosphere by a right-ngle prism mounted upon a piezoelectric controlledtage that simplifies the final alignment. The largeeld of view and the relatively small separation ~30m! between the transmitting and the receiving axesf the lidar make possible the full overlap betweenhe laser beam and the telescope’s field of view at aypical altitude of 150 m AGL.
The backscattered light is collected by a 20-cm-iameter, 60-cm focal-length Newtonian telescope ~ fy
3!. A diaphragm is used to adapt the telescope’sfield of view set at 5 mrad. Two custom-designedbandpass filters ~Omega Optical Company! at theentrance of the polychromator achieve an initial sup-pression of the strong elastic backscattered signal at266 nm. The filters are tilted at 18° for an optimumtransmission of the three Raman backscattered sig-nals and a maximum rejection of the elastic backscat-tered light. At this angle each filter has as much as80% transmission for the three Raman wavelengthsand an optical density of 2.6 at 266 nm. These spec-ifications were measured directly after the filterswere delivered. One year later the filters showed aloss of ;10% in transmission. The filters wereplaced in the parallel beam between two lenses thatfit the f-numbers of the telescope and the polychro-mator. A 500-mm Czerny–Turner polychromator~ fy4! was used for wavelength separation of the Ra-man signals as well to reject the daylight backgroundand the elastically backscattered light. The poly-chromator resolution with a 3600-grooveymm UV-enhanced holographic grating was 0.51 nmymm.For additional suppression of the daylight back-ground a solar-blind filter ~Corion 300F-430T! wasplaced at the polychromator entrance. The filter re-jects light throughout the visible ~l . 360 nm! ofbetter than 1024 and a 70% transmission in the 250–320-nm band.
The Raman signals from oxygen, nitrogen, and wa-ter vapor were detected simultaneously by threeHamamatsu H-5780-06 photosensor modules. Eachmodule is equipped with an optical diffuser and ashort-focal-length lens to improve the spatial unifor-
27
1
v
AGca
amonsrpba
mity of the photosensor module. Because a pulseduration of less than 0.65 ns is achieved by suchphotodetectors, both analog and photon-counting de-tection modes are possible. The signals were ac-quired by a Licel transient recorder that combinesphoton counting with a 12-bit 20-MHz ADC. Rawdata were averaged over 4000 shots and stored withan ultimate range resolution of 7.5 m.
A. Experimental Determination of the Optical Cross Talk
A careful analysis of the cross talk measured in ourexperimental setup was made. Even though thethree signals could affect one another, the influence ofthe nitrogen channel on the two others is strongerbecause the nitrogen channel is the central wave-length and has the highest intensity level. There-fore we measured the optical cross talk caused by thenitrogen signal in the oxygen and water-vapor chan-nels. For the measurements we used light producedby stimulated Raman scattering of a 266-nm laserbeam in 35 atm of nitrogen. A small fraction of thebeam’s first Stokes at 283.6 nm was injected by anoptical fiber into the lidar-receiving telescope, and theresultant cross-talk intensities in the oxygen andwater-vapor channels were measured by the respec-tive PMT. The intensity of the injected 283.6-nmlight was chosen such that the resultant cross-talkintensities were well above the photodetector’s noiselevel. The light intensity in the nitrogen channelwas measured with neutral-density filters that atten-uate the signal below the saturation level of the pho-todetector. The cross talk in the oxygen and water-vapor channels was calculated as a ratio between thecross-talk intensity in the respective channel and theintensity in the nitrogen channel, with the relativephotodetector sensitivity and the neutral-density fil-ter attenuation in the nitrogen channel taken intoaccount. The measured cross-talk levels were 2.2 3025 for O2yN2 and 5.2 3 1025 for H2OyN2. The
simulation study presented in Section 2 showed thatit is only with cross talk greater than 1023 that adetectable bias in ozone or water-vapor retrieval wasexpected. Our experimental cross-talk values weremuch lower, low enough that we could neglect theirinfluence. These values are also in good agreementwith the stray-light level specified by the polychro-mator’s manufacturer.
Furthermore, care was taken to suppress the re-sidual elastic signal and to measure the degree of thissuppression in the Raman channels. As this degreeis quite high, its direct measurement was impossible.Instead, suppression of the elastic signal by the band-pass filters and that by the polychromator were mea-sured separately. For the polychromator, thesuppression of elastic backscatter light was mea-sured in a way similar to that described above for thecross talk among the Raman channels, this time witha fraction of the 266-nm light injected directly intothe receiving telescope. The degrees of suppressionfor the various Raman channels were calculated as aratio of the intensity of the 266-nm light entering thepolychromator presented in Section 2 to the light
intensities detected in these channels. The mea-sured values were as follows: for the nitrogen chan-nel, 1.5 3 105; for the oxygen channel, 5.9 3 104; andfor the water-vapor channel, 3.5 3 104. The band-pass filters were measured separately, with a totalattenuation ratio of the elastic signal ~ratio beforeersus ratio after the addition of filters! of 1.58 3 105
times. Thus the total suppression of the elasticbackscatter signal at the receiver was always greaterthan 5 3 109. This value ensured that the residualpump beam in the three Raman channels was not thesource of systematic error, as was confirmed by themodel results. This high rate of rejection of the elas-tic signal was further confirmed by the fact that nodetectable echo from low-altitude dense clouds wasobserved in the three Raman channels.
B. Experimental Determination of the Afterpulse Effect
During the initial tests of the Raman lidar setup,classic-type glass-bulbyhead-on PMTs were used thatshowed APEs. An upgrade of the system was madewith a new type of metal package photomulti-plier.27,28 With these new PMTs, no APE was de-tectable, even when the photocathode wasilluminated directly with 7-ns laser pulses with highintensity. Such laser pulses induced PMT outputpulse amplitudes of as much as 1 V on a 50V loadwithout an APE. A possible explanation for this factis the short time delay of the afterpulses, which is dueto the very small ~less than 1-mm! distance betweenthe cathode and the first dynode and makes it diffi-cult to distinguish the afterpulses from the mainpulse. Another explanation is that the new metalpackage is less permeable than a glass bulb for at-mospheric helium, and helium diffusion is known tobe one of the main reasons for afterpulses.
4. Results and Discussion
Raman lidar measurements were performed bothfrom the EPFL site ~46° 319N; 6° 389E! in March–
pril 1999 and during a field campaign in Crete,reece, in May 1999. The different results are dis-
ussed here and will help to underline the advantagesnd the limitations of our instrument.The first results were obtained with the idea of
chieving a time series of Raman lidar measure-ents in daytime and nighttime conditions for both
zone and water-vapor vertical profiles simulta-eously. During this time series, additional mea-urements at 8 m AGL of ozone concentration,elative humidity, temperature, and pressure wereerformed. The ozone concentration was monitoredy an UV absorption analyzer ~Dasibi 1008 AH! withprecision of 62 ppbv.The Raman-shifted nitrogen and oxygen backscat-
tered signals were acquired in the analog mode by useof a 12-bit, 20-MHz ADC whereas the photon-counting method of detection was used for the water-vapor signal. The signals were acquired withLabVIEW-based software with a real-time display ofthe preliminary results and with posttreatment byMatlab software. The time series in Fig. 6 shows the
20 June 2001 y Vol. 40, No. 18 y APPLIED OPTICS 2993
gFn
tm1s
2
ozone vertical profile measured continuously over aperiod of 28 h. Based on the statistical noise anal-ysis presented in Section 2, we set the lidar verticalresolution at 90 m and the time resolution at 30 min.A sliding average was applied both for the altituderange and for the time scale. The white rectangle atthe bottom of Fig. 6 indicates the spatial range whereozone cannot be retrieved by lidar because of theincomplete overlap between the laser beam and thetelescope’s field of view. The ground-level ozone con-centrations measured by the point monitor are alsoshown in the figure for comparison with the lidardata.
In this time series the ozone’s diurnal cycle wasclearly seen, with higher values during the period ofhigh solar radiation and lower values at nighttime.Note also that this time series was taken in the hu-mid conditions that are associated with formation ofthin water cloud layers at night. But, even so, es-sentially no data rejection in the ozone retrieval wasneeded, with the exception of some data at 20–22 UTat an altitude range higher than 500 m AGL. Fur-thermore, during this period of observation of morethan 1 day, a strong variation in the height of theinversion layer was observed. The combined opticalinterference of cloud layers and the change in aerosolgradient caused by the change in height of the top ofthe planetary boundary layer would certainly affector even make impossible any elastic DIAL ozone mea-surements in similar conditions. Such effects haveoften been reported9,10 but did not affect this ozonetime series.
The water-vapor time series following the datatreatment in Eq. ~10! with the three ~nitrogen, oxy-en, and water-vapor! Raman signals is presented inig. 7. These measurements were taken simulta-eously with the ozone measurements, but because of
Fig. 6. Ozone Raman DIAL obtained in March 1999 for a timeseries of 28 h. The ozone concentrations measured at the groundare given by an UV absorption detector. They are measured byRaman DIAL from altitudes of 200 to 700 m AGL. The spatialresolution is 90 m, and the time resolution is 30 min.
994 APPLIED OPTICS y Vol. 40, No. 18 y 20 June 2001
echnical problems with the photoncounting photo-ultiplier the water-vapor results are shown after
6 h UT. As indicated above, as the instrument con-tant Kcal is unknown, an absolute profile of the
water-vapor mixing ratio in air may be obtained onlyif an absolute reference is available at a given alti-tude. As such is not the case here, we used theabsolute water-vapor mixing ratio measured atground level as the reference value of the first alti-tude achieved by the lidar, namely, ;270 m AGL.In a manner similar to that for the ozone time seriesin Fig. 6, the white rectangle in Fig. 7 covers thespatial range where the water-vapor mixing ratiocannot be retrieved by lidar. But in this case thevalues measured close to the ground are equivalent tothe reference values at 270 m AGL. This meansthat the present water-vapor mixing ratio time seriesshould be regarded as a time series with relativenumbers; below, we shall compare an absolute water-vapor mixing-ratio vertical profile with balloon mea-surement.
The measurements are presented with a verticalresolution of 22.5 m and a time resolution of 30 min.It is important to note that, even under daytime con-ditions, with a water-vapor Raman-shifted wave-length at 294.6 nm close to the solar-blind border, thesolar background essentially did not perturb thephoton-counting signal detection. This result indi-cates good rejection of the solar background achievedby the combination of the polychromator and thebandpass and solar-blind filters. This water-vaportime series shows the characteristic daytime–nighttime behavior of the water-vapor content in theatmosphere, with convection at daytime lifting up-ward air masses with higher water-vapor content.
In deriving the result in Fig. 7 we accounted for the
Fig. 7. Water-vapor mixing ratio retrieved by Raman lidar ob-tained in March 1999 for a time series of 24 h from altitudes of200–1200 m AGL. The water vapor relative humidity measuredat the ground is given by a standard meteorological station. Thelatter values are also used at 270 m AGL as reference values for thefirst altitude range of the Raman lidar profile. The spatial reso-lution is 22.5 m, and the time resolution is 30 min.
wsDvtvowsssa
WbK
uweaH
ozone differential absorption effect described by Eq.~10!. In this formalism we used the three Ramanbackscattered signals and were able to correct thewater-vapor mixing-ratio retrieval for the ozone ef-fect without having to retrieve explicitly the verticalozone profile. Whereas ozone was explicitly re-trieved up to 700 m AGL, as shown in Fig. 6, thewater-vapor mixing ratio corrected for the ozone ab-sorption can be retrieved up to 1200 m AGL. Basedon this formalism it is furthermore possible to ex-press the systematic shift in the water-vapor timeseries that is due to ozone differential absorption.Figure 8 illustrates the difference DH2O between thewater-vapor mixing-ratio retrieval uncorrected forthe ozone differential absorption effect and the sametime series but corrected. As has already beenshown as a model result in Fig. 2, higher ozone con-tent leads to larger water-vapor corrections with anadditive effect that is due to the integral term in Eq.~3!. In this sense Fig. 8 gives additional informationabout the ozone content in the air because low DH2Ovalues will be associated directly with low ozone con-centration. This correlation is nicely confirmed if wecompare Figs. 6 and 8 for an altitude range up to700 m AGL, while it is now possible to gain a firstestimate up to an altitude of 1200 m AGL of the ozonecontribution. See, for example, the DH2O verticalprofile at 4 a.m. ~28 h UT in Fig. 8! with values thatare essentially near zero. The values are confirmedby the low ozone concentration in Fig. 6 up to 700 mAGL measured at the same time. They are also in-dicative of low ozone concentration for the rest of thealtitude range to 1200 m AGL, where ozone was nolonger retrieved by Raman DIAL. This informationwas gained because in this case the ozone contribu-tion is integrated along the profile and is not range
Fig. 8. Difference in the water-vapor mixing-ratio retrieval DH2Oncorrected for the ozone differential absorption effect comparedith the corrected time series. Small values of DH2O along the
ntire vertical range are associated with low ozone content in thetmosphere. Note that negative DH2O values ~.20.1 g of2Oykg of dry air! are also indicative of the limit of precision of the
method ~62%!.
resolved. Values of DH2O up to 1.2 g of H2Oykg ofdry air are indicative of higher ozone content, as isthe case at 4 p.m. ~40 h UT in Fig. 8! on the secondday of continuous measurement. This shift in thewater-vapor retrieval corresponds to an effect of;15% for ozone values typically below 60 ppbv, aspartially shown in Fig. 6, and would be even worse foratmospheric conditions with higher ozone concentra-tions. Finally, one should note that some negativevalues for DH2O were obtained with magnitudes to aslow as 20.1 g of H2Oykg of dry air in the worst cases.These negative values correspond to less than 2% ofthe effective water-vapor mixing ratio. They are di-rectly associated with the limit of precision of ourdetermination of the water-vapor mixing-ratio re-trieval and are slightly higher than the predictedstatistical error defined by the model study.
Later in the year, the same lidar instrument wasmounted upon a movable platform and transported toCrete, Greece, to participate in the PhotochemicalActivity and Ultraviolet Radiation Modulation Fac-tors II ~PAUR II! program.29 The measuring sitewas situated in the northwest part of the island ofCrete in Nopigia ~35° 519N 23° 729E!, and the lidar
as placed 5 m above sea level ~ASL!. Whereas theystem most of the time operated in the ozone elasticIAL mode, the chance for additional Raman water-apor measurements was offered for a short period ofime by direct comparison with an absolute water-apor profile measured by balloon. The balloonzone and temperature profiles were determinedith an electrochemical concentration cell ozone-
onde in combination with a Vaisala RS-80 radio-onde, H-type Humicap sensor. This widely usedensor measures water-vapor mixing ratios with anccuracy of 65% in the lower troposphere. Proce-
dures for sonde preparation and data acquisition aresimilar to those developed by the National Oceanicand Atmospheric AdministrationyClimate Diagnos-tics and Monitoring Laboratories.30 Data of 1-s du-ration were recorded and processed as described byThompson et al.31 The balloon was launched fromessentially the same place as was the EPFL lidartrailer.
Figure 9 shows a comparison of the vertical water-vapor profile obtained by Raman lidar and the bal-loon measurements. During this experiment thelidar system measured only the Raman nitrogen andwater-vapor channels. Thus we used the ozone datafrom the balloon directly to correct for the ozone in-terference effect on water vapor according to Eq. ~3!.
e used the water-vapor mixing ratio measured byalloon to determine the lidar calibration constantcal at 210 m ASL, where a full overlap between the
laser beam and the telescope’s field of view wasachieved. This absolute water-vapor mixing-ratiovertical profile retrieved by Raman lidar was ob-tained by averaging over five files of 4000 laser shots~total integration time of 30 min, from 5 to 5.30 a.m.UT! with a vertical resolution of 22.5 m and appearedto be in good agreement with the balloon data. Thedifference DH2O shown in Fig. 9 between the water-
20 June 2001 y Vol. 40, No. 18 y APPLIED OPTICS 2995
zoh
enCN
muvsnt~
2
vapor profile obtained by lidar and that by balloonwas typically below 1 g of H2Oykg of dry air over theentire range ~or less than 15% in relative units!, withthe highest discrepancy observed at the height of theinversion layer. This transition layer was observedbetween 850 and 950 m ASL in the balloon data,whereas for the lidar data it was defined above 950 mASL. This difference could well be explained by thedifference in time and air volume sampled by the twomethods. In particular, averaged values obtainedby lidar over a period of 30 min were compared withinstantaneous values measured by balloon that couldin turn be influenced by highly local air masses. Inthis example the balloon was launched at 5:11 a.m.UT with a vertical speed of ;5 mys and 1ydatapointys.
In Fig. 9 the water-vapor content obtained withoutconsideration of the ozone correction ~labeled Lidaruncorrected! is also shown. This profile was calcu-lated with the instrument constant determined aspreviously without consideration of the integral fac-tor in Eq. ~3!. In this case the error estimateDH2OLidar–lidar uncorrected reached differences as highas 15% for ozone concentrations of ;70 ppbv; this isan integrative effect and therefore the highest dis-crepancy is reached at a long range. This differenceis greater than the largest discrepancy between bal-
Fig. 9. Raman lidar water-vapor measurements compared withballoon soundings. The lidar data are shown with a spatial res-olution of 22.5 m and a time resolution of 30 min. They arecompared with 1-s balloon data recorded at an ascent speed of ;5
ys. The water-vapor balloon measurement at 210 m ASL issed as an absolute calibration for the lidar signal. The water-apor mixing-ratio vertical profile retrieved by Raman lidar ishown with the ozone concentration profile measured simulta-eously by balloon taken into account ~Lidar with ozone correc-ion!. The same lidar profile is also shown without this correctionLidar uncorrected!. The corresponding difference ~Lidar uncor-
rected & lidar! may be compared with results presented in Fig. 8.The relative difference ~Lidar & sonde! indicates values below 15%over the entire range of measurement. This comparison wasmade in May 1999 in Crete during the PAUR II experiment withthe Raman lidar on board the EPFL lidar trailer.
996 APPLIED OPTICS y Vol. 40, No. 18 y 20 June 2001
loon and water-vapor lidar measurements correctedfor the ozone effect. Note that at the calibrationheight at 210 m ASL DH2OLidar–lidar uncorrected is notero because the water-vapor shift that is due to thezone absorption between the ground and this heightas already been taken into account.
5. Conclusion
The principle and design of a single-wavelength ex-citation Raman–DIAL instrument for daytime andnighttime ozone and water-vapor measurements inthe planetary boundary layer have been demon-strated. Our objective was to perform measure-ments at low altitude in the PBL, an atmosphericlayer characterized by a high and inhomogeneousaerosol load, where elastic DIAL observations cannotin some cases be used because of unpredictable Mieinterference.
An error analysis based on the influence of theinstrumental and atmospheric effects that could de-crease the Raman lidar performance was made. Itconcerned the influence of the aerosol content on theozone retrieval and of the ozone concentration on thewater-vapor retrieval, the statistical noise, the after-pulse effect, the cross talk between Raman channels,and the influence of the residual laser light on Ramansignals. The results of this study were confirmed byRaman lidar results obtained for the measurementsof ozone and water-vapor vertical profiles in the PBL.The error induced in the water-vapor mixing-ratioretrieval by the ozone differential absorption effectrevealed that, if three Raman signals ~namely, oxy-gen, nitrogen, and water vapor! are measured simul-taneously, it is possible to extend the water-vaporretrieval range corrected for the ozone effect.
This lidar development is important because ozoneand water-vapor measurements are key criteria forcontrol and quality testing of the predictive resultsobtained by atmospheric transport chemical model-ing. These Raman lidar measurements were fur-thermore obtained with a resolution that fit both thetime and the space model resolution. A future com-bination of the powerful elastic DIAL technique ap-plied at higher altitude or in more-homogeneousaerosol conditions with the Raman DIAL method atlow altitude will essentially be able to encompass analtitude range in full agreement with the vertical gridresolution of the model.
This study was supported by the Swiss NationalScience Foundation under contract 21-50861.97 andthe Swiss Federal Office for Science and Educationunder contract 97.0377. The balloon sounding wasprovided by NASA’s ~Total Ozone Mapping Spectrom-ter ~TOMS! project. The authors thank the orga-izers of the PAUR II project for logistical support inrete and are grateful to Ed Browell and his team atASA Langley for stimulating discussions.
References1. B. J. Finlayson-Pitts and J. N. Pitts, Jr., Chemistry of the
Upper and Lower Atmosphere: theory, Experiments, and Ap-plications ~Academic, San Diego, Calif., 1999!.
2. A. Clappier, A. Martilli, P. Gross, P. Thunis, F. Pasi, B. C. 17. J. A. Sunesson, “Differential absorption lidar system, for rou-
2
2
2
2
2
Krueger, B. Calpini, G. Graziani, and H. van den Bergh, “Ef-fect of sea breeze on air pollution in the Greater Athens area.I. Numerical simulations and field observations,” J. Appl.Meteorol. 39, 563–575 ~1999!.
3. P. Grossi, P. Thunis, A. Martilli, and A. Clappier, “Effect of seabreeze on air pollution in the Greater Athens area. II. Anal-ysis of different emission scenarios,” J. Appl. Meteorol. 39,546–562 ~1999!.
4. S. Perego, “A numerical mesoscale model for simulation ofregional photosmog in complex terrain: model descriptionand application during POLLUMET 1993 ~Switzerland!,” Me-teorol. Atmos. Phys. 70, 43–69 ~1999!.
5. B. Calpini, V. Simeonov, F. Jeanneret, J. Kuebler, V. Sathya,and H. van den Bergh, “Ozone LIDAR as an analytical tool ineffective air pollution management: the Geneva 96 cam-paign,” Chimia 51, 700–704 ~1997!.
6. L. Fiorani, B. Calpini, L. Jaquet, H. van den Bergh, and E.Durieux, “A combined determination of wind velocities andozone concentrations for a first measurement of ozone fluxeswith a DIAL instrument during the Medcaphot Trace cam-paign,” Atmos. Environ. 32, 2151–2159 ~1998!.
7. L. Schoulepnikoff, H. van den Bergh, B. Calpini, and V. Mitev,“Tropospheric air pollution monitoring lidar,” in Encyclopediaof Environmental Analysis and Remediation, R. A. Meyers, ed.~Wiley, New York, 1998!, pp. 4873–4909.
8. E. V. Browell, S. Ismail, and T. Shipley, “UV DIAL measure-ments of O3 profiles in regions of spatially inhomogeneousaerosols,” Appl. Opt. 24, 2827–2836 ~1985!.
9. P. Quaglia, G. Larcheveque, R. Jimenez, V. Simeonov, G. An-cellet, H. van den Bergh, and B. Calpini, “Planetary boundarylayer ozone fluxes from combined airborne, ground based li-dars and wind profiler measurements,” Eur. J. Anal. Chem. 27,305–313 ~1999!.
10. P. Volger, J. Bosenberg, and I. Schult, “Scattering properties ofselected model aerosol calculated at uv-Wavelengths: impli-cations for DIAL measurements of tropospheric ozone,” Beitr.Phys. Atmos. 69, 177–187 ~1996!.
11. S. H. Melfi, J. D. Lawrence, Jr., and M. P. Mc Cormick, “Ob-servation of Raman scattering by water vapor in the atmo-sphere,” Appl. Phys. Lett. 15, 295–297 ~1969!.
12. D. Renault, J. C. Pourny, and R. Capitini, “Daytime Raman-lidar measurements of water vapor,” Opt. Lett. 5, 233–235~1980!.
13. A. J. Sedlacek and M. D. Ray, “Raman DIAL: application toareas characterized by varying aerosol burden,” presented atthe 19th International Laser Radar Conference, Annapolis,Md., July 1998.
14. D. Renault, and R. Capitini, “Boundary-layer water vaporprobing with a solar-blind Raman lidar: validation, meteoro-logical observation and prospects,” J. Atmos. Oceanic Technol.5, 585–601 ~1988!.
15. T. J. Mc Gee, M. Gross, R. Ferrare, W. Heaps, and U. Singh,“Raman DIAL measurements of stratospheric ozone in thepresence of volcanic aerosols,” J. Geophys. Res. Lett. 20, 955–958 ~1983!.
16. R. T. H. Collis and P. B. Russell, “Lidar measurement of par-ticles and gases by elastic backscattering and differential ab-sorption,” in Laser Monitoring of the Atmosphere, E. D.Hinkley, ed. ~Springer-Verlag, Berlin, 1976!, p. 89.
tine measurement of tropospheric ozone,” Appl. Opt. 33, 7045–7058 ~1994!.
18. W. K. Bishel and G. Black, “Wavelength dependence of Ramanscattering cross-sections from 200–600 nm,” in American In-stitute of Physics Conference Proceedings No. 100: ExcimerLasers, C. K. Rhodes, H. Egger, and H. Pummer, eds. ~Amer-ican Institute of Physics, New York, 1983!, pp. 181–194.
19. L. T. Molina and M. J. Molina, “Absolute absorption crosssections of ozone in the 185 to 350 nm wavelength range,” J.Geophys. Res. 91, 14,501–14,508 ~1986!.
20. A. C. Vandaele, P. C. Simon, J. M. Guilmot, M. Carleer, and R.Colin, “SO2 absorption cross-section in the UV using a Fouriertransform spectrometer,” J. Geophys. Res. 99, 25,599–25,605~1994!.
21. W. Schneider, G. K. Moortgat, G. S. Tyndall, and J. P. Bur-rows, “Absorption cross-sections of NO2 in the UV and visibleregion ~200–700 nm! at 298 K,” J. Photochem. Photobiol. A 40,195–217 ~1987!.
2. G. P. Anderson, S. A. Clough, F. X. Kneizys, J. H. Chetwynd,and E. P. Shettle, “AFGL atmospheric constituent profiles ~0–120 km!,” Tech. Rep. AFGL-TR-86-0110 ~U.S. Air Force Geo-physics Laboratory, Hanscom Air Force Base, Mass., 1986!.
3. P. B. Coates, “A theory of afterpulse formation in photomulti-pliers and the prepulse height distribution,” J. Phys. D 6,1862–1869 ~1973!.
4. T. Antonioly and P. Benetti, “Study of afterpulse effects inphotomultipliers,” Rev. Sci. Instrum. 54, 1777–1780 ~1983!.
5. P. B. Coates, “The origins of afterpulses in photomultipliers,”J. Phys. D 6, 1159–1166 ~1973!.
6. B. Lazzarotto, V. Simeonov, P. Quaglia, G. Larcheveque, H.van den Bergh, and B. Calpini, “A Raman differential absorp-tion lidar for ozone and water vapor measurement in the lowertroposphere,” Int. J. Env. Anal. Chem. 74, 255–261 ~1999!.
27. V. Simeonov, G. Larcheveque, P. Quaglia, H. van den Bergh,and B. Calpini, “The influence of the photomultiplier tubespatial uniformity on lidar signals,” Appl. Opt. 38, 5186–5190~1999!.
28. H. Kyushima, Y. Hasegawa, A. Atsumi, K. Nagura, H. Yokota,M. Ito, J. Takeuchi, K. Oba, H. Matsuura, and S. Suzuki,“Photomultiplier tube of new dynode configuration,” IEEETrans. Nucl. Sci. 41, 725–729 ~1994!.
29. A. Kylling, A. F. Bais, M. Blumenthaler, J. Schreder, C. Zere-fos, and E. Kosmidis, “The effect of aerosols on solar UV irra-diances during the PAUR campaign,” J. Geophys. Res. 103,26,151–26,060 ~1998!.
30. W. D. Komhyr, S. J. Oltmans, P. R. Franchois, W. F. J. Evans,and W. A. Matthews, “The latitudinal distribution of ozone to35 km altitude from ECC observations, 1985–1987,” in Pro-ceedings of the Quadrennial Ozone Symposium and Tropo-spheric Ozone Workshop, R. D. Bojkov and P. Fabian, eds.~Deepak, Hampton, Va., 1989!, pp. 147–150.
31. A. M. Thompson, B. G. Doddridge, J. C. Witte, R. D. Hudson,W. T. Luke, J. E. Johnson, B. J. Johnson, S. J. Oltmans, and R.Weller, “A Tropical Atlantic paradox: shipboard and satelliteviews of a tropospheric ozone maximum and wave-one inJanuary–February 1999,” Geophys. Res. Lett. 27, 3317–3320~2000!.
20 June 2001 y Vol. 40, No. 18 y APPLIED OPTICS 2997
