Differential absorption lidar system for routine monitoring of tropospheric ozone

Differential absorption lidar system forroutine monitoring of tropospheric ozone

J. A. Sunesson, A. Apituley, and D. P. J. Swart

A differential absorption lidar system for routine profiling of tropospheric ozone for daytime andnighttime operation is described. The system uses stimulated Raman scattering in hydrogen anddeuterium of 266-nm radiation from a quadrupled Nd:YAG laser. Ozone profiles from altitudes of 600 m

to approximately 5 km have been obtained with analog detection. Implementing corrections fordifferential Rayleigh scattering, differential absorption from oxygen, sulphur dioxide, and nitrogendioxide, and differential aerosol extinction and backscatter can reduce the total system inaccuracy to5-15% for a clear day and 20-30% for a hazy day, except at the top of the mixed layer. Photon counting

must be installed to increase the measurement range from 5 to 15 km. An example of an application ofroutine measurements of tropospheric ozone profiles is given.

Key words: Lidar, differential absorption lidar, ozone, UV, stimulated Raman scattering, troposphere,aerosol interference, photomultipliers, signal-induced noise, error calculation.

1. Introduction

This paper describes a differential absorption lidar(DIAL) system for the measurement of vertical ozone-concentration profiles in the troposphere, from 0-15km altitude. The system was developed at the DutchNational Institute of Public Health and Environmen-tal Protection (RIVM) within the framework of theEuropean Environmental Research Program EU-REKA-EUROTRAC. RIVM runs an ozone researchproject and participates in the EUROTRAC sub-projects Tropospheric Ozone Research' (TOR) andTropospheric Environmental Studies by Laser Sound-ing2 (TESLAS).

Ozone is one of the key trace gases in the atmo-sphere. In the troposphere, ozone plays an impor-tant role in chemical processes, whereas stratosphericozone protects the Earth from UV radiation andprovides for stratospheric warming. Furthermore,ozone is of importance in climate change. 2 3 In-dications are that the concentration of ozone atground level in Europe has doubled during the pastcentury. This increase is worrisome, because thebackground concentrations are now approaching lev-

The authors are with the National Institute of Public Health andEnvironmental Protection, P.O. Box 1, NL-3720 BA Bilthoven,The Netherlands.

Received 3 March 1993; revised manuscript received 9 March1994.

0003-6935/94/307045-14$06.00/0.© 1994 Optical Society of America.

els at which effects on people and ecosystems can beexpected. The TOR and TESLAS projects addressquestions concerning the tropospheric ozone budgetand use both measurements and models in the study.In TOR, the modeling and measurement of ozone arecoordinated. A European measurement networkmeasuring ozone and ozone-chemistry-related spe-cies and parameters is set up.

Most of the ozone monitors in the measurementnetwork are point monitors at ground level that yieldno information about the vertical ditribution of ozone,such as differences between the mixed layer and thefree troposphere. These differences can only be re-vealed with vertical profiling measurements and areconventionally obtained with balloon-borne sensors.However, this method provides only poor temporaland spatial resolution. Both can be improved withlidar techniques. On some TOR network stations anozone DIAL system will be installed. The develop-ment of these systems is coordinated within TESLAS,where researchers from France, the United Kingdom,Germany, Italy, the Netherlands, and Sweden cooper-ate. The system described in this paper is intendedfor the TOR station Kollumerwaard in the north ofthe Netherlands (530 20' N, 6° 16' E). The specificapplications defined for this system are

* Routine monitoring of vertical ozone profiles upto the lower stratosphere (0-15-km altitude),

* Investigation of exchange processes between thefree troposphere and the mixed layer, and

20 October 1994 / Vol. 33, No. 30 / APPLIED OPTICS 7045

* Investigation of exchange processes between thestratosphere and the troposphere (especially tropo-pause folding events).

The system development started in 1988, and alaboratory version capable of measuring in the mixedlayer and lower troposphere is now complete.4 5

This paper describes the system, and some of theresults achieved are discussed. Furthermore, ananalysis of error sources is made, and an example ofan application of routine ozone DIAL measurementsis given.

2. DIAL Technique for Ozone Measurements

The DIAL technique has been used extensively tomonitor trace-gas concentrations in the atmosphere.Species as diverse as Hg, NO, NO2 , SO2, and ozonehave been measured.6-'2 The technique uses twowavelengths, Xon and XAff, that are differentially ab-sorbed by the gas to be measured. A simplified formof the single-wavelength lidar return signal is6

P(R, AR, X) = C - (R, X)exp{ -2 J [o(X)n(r)

+ abs(r, X) + aext(r, X)]dr} X (1)

where P(R, AR, X) denotes the received power as afunction of distance R, spatial resolution AR, andwavelength X; C is a system-determined constant andf3(R, X) is the volume backscatter coefficient; ou(X) isthe absorption cross section of the gas measured andn(r) is its number density; and aabs(r, X) is the molecu-lar absorption by gases other than the species ofinterest and Oaext(r, X) is the extinction that is due toRayleigh and Mie scattering. The number densitycan be derived from the following expression, theso-called DIAL curve:

n[Pon(R)1 = ln(R) R

[off(R)= Ioff(R)_ [Arn(r)

+ Aaoabs(r) + Aoext(r)Idr, (2)

where Au denotesderivative of Eq.n(R)

[u(XO.) - u(Xoff)], etc. Taking the(2) and rearranging terms yields

2A aR ln7R)] (3a)

+1 a In P..(R)1 (3b)

2Au aR Po ~ff(R) (b

1- - Autabs(R) (3c)

Aou

1- A-ext(R). (3d)

Au

In general, each of the terms (3a)-(3d) must be

known. Usually the simplifying conditions hold,that is, the wavelengths Aon and XAff are close togetherand differential absorption from other gases is negli-gible. In that case the terms (3b)-(3d) become small,and the concentration profile can be obtained just byevaluating Eq. (3a), which takes only Au and themeasured quantity P(R) at Xen and XAff as input. Ifthese simplifications cannot be made, terms (3b)-(3d)must also be calculated. Usually the measurementsare first evaluated as if terms (3b)-(3d) are negligible,and correction terms are calculated afterward. Thederivative in Eq. (3a) can be approximated by adifferential quotient in a range interval AR = R2 -RI, which yields A(RI, R2) as the average concentra-tion within the interval.

To optimize the sensitivity of a DIAL measure-ment, the optical depth that is due to absorption ofthe species of interest should be close to unity overthe measurement range.'2 For ozone, in the UV, atconcentration levels encountered in the mixed layerand the free troposphere, this requirement is fulfilledonly between 260 and 300 nm. Because this ozone-absorption band lacks narrow, deep structures, Xonand XAff must be quite widely separated. Further-more, one wavelength pair is insufficient for anaccurate retrieval of the whole ozone profile up to 15km of altitude.' 2 Finally, the choice of wavelengthsresults from a trade-off between strong absorption(yielding high sensitivity, high accuracy, and limitedrange because of the absorption) and weak absorption(with low sensitivity, less accuracy, and greater range).Because the wavelength region suitable for tropo-spheric ozone measurements is solar blind, daytimemeasurements can be performed without narrow-band detection.

3. System Description

A. Laser Choice

DIAL wavelengths are often generated with a tunabledye laser, pumped by an excimer or a Nd:YAG laser.This scheme has been used for ozone measure-ments" 1, 2 but is complex and requires frequent wave-length control and calibration, making it less suitedfor routine operation. An alternative scheme, origi-nally proposed by Uchino et al., 13 uses a fixed-frequency UV laser together with stimulated Ramanscattering (SRS) in gases such as H2, CH4, or D2.This scheme offers considerable technical advantagesover the use of dye lasers because the Raman-shiftingprocess is practically maintenance free. For the useof this scheme in tropospheric ozone measurementsthere are two main alternatives for the laser that ispumping the Raman-shifting process: a KrF ex-cimer or a frequency-quadrupled Nd:YAG laser.Together with the Raman active gases H2 and D2,these alternatives yield the wavelength sets listed inTable 1.

The Nd:YAG laser alternative was chosen. Stud-ies of the precision, accuracy, and time needed toobtain one ozone profile showed that precision andaccuracy would be similar for both alternatives.4 4

7046 APPLIED OPTICS / Vol. 33, No. 30 / 20 October 1994

Table 1. Wavelength Sets for FrF Excimer and Nd:YAG Lasers

Laser Type Wavelength (mm) Generation Methoda

KrF excimer 248 Fundamental277 S1, H2292 S2, D2313 S2, H 2

Nd:YAG 266 Fourth harmonic offundamental

289 S1, D2299 S1, H 2

aS1 denotes the first Stokes line andStokes for SRS in H2 and D2, respectively.

S2 denotes the second

For the Nd:YAG system the measurement time neededwould be 5 to 10 min. This would be a factor of 5-20shorter than the time needed for an excimer-laser-based system because of its higher pulse energy andpulse repetition rate, but this advantage is not criticalfor the purposes intended. A Nd:YAG-laser-basedsystem has the advantage that it needs less mainte-nance and is less complex in the wavelength-genera-tion scheme-only first Stokes components and pumpbeams are needed, whereas an excimer-based systemneeds second Stokes components, the successful opti-mization of which is more tedious.'5

B. System Overview

The system uses two lasers, one pumping a Ramancell filled with D2 , the other pumping a cell filled withH2. This way the first laser generates the wave-lengths 266 and 289 nm, and the second laser gener-ates the wavelengths 266 and 299 nm. Using twolasers provides more energy than a single laser andalso offers the possibility to separate the signalstemporally (see Subsection 3.C.4). The beams fromboth lasers are directed vertically into the atmo-sphere through two steering mirrors, which aremounted to the side of the receiving telescope in adouble-biaxial configuration. The steering mirrorsare positioned symmetrically with respect to thetelescope. The light received by the telescope isdirected toward a small polychromator through a UVmirror and a focusing system. Each of the threewavelengths is detected by a separate photomultipliertube. The entire system is supported by one(1.5 m x 4.5 m) optical table. Analog-to-digital con-version of the signals is done in a computer auto-mated measurement and control (CAMAC)-basedtransient recorder system. A minicomputer systemis used for overall system control, and data can beviewed on a graphics display and copied to a plotter.An overview of the system is shown in Fig. 1, and thesystem parameters are listed in Table 2.

C. Subsystem Description

1. Wavelength GenerationBoth lasers are equipped with a harmonic generator(Spectra-Physics HG-2) to produce a wavelength of266 nm in two stages: from 1064 to 532 nm and

subsequently from 532 to 266 nm. Each laser canprovide approximately 1 J/pulse at 1065 nm maxi-mum, but only 600-700 mJ/pulse is actually usedbecause the fourth-harmonic generation process be-comes unstable at higher pulse energies. The fre-quency doubling to 532 nm yields 280-300 mJ/pulse,and the energy finally obtained at 266 nm is 80-90mJ/pulse for the lasers and crystals used. Theunused energy at 1064 and 532 nm is disposed of withdichroic mirrors and beam dumps. The first fre-quency-doubling stage is kept in phase match by theuse of temperature stabilization of the crystal. Thesecond doubling stage is also thermostated but needsangle tuning during a warm-up period of 10-15 min,after which the pulse energy remains stable and canbe maintained for many hours. The tuning se-quence is reproducible. If the laser system is shutdown for a short period, of the order of a few minutes,the power level present before the shutdown can beregained without tuning the doubling crystals.

Using 266 nm as the pump, we obtain furtherwavelengths needed for the ozone measurement bySRS. The Raman cells are made of stainless steeland have been tested to withstand 80 bars of pressure.The entrance and exit windows are tilted with respectto each other to avoid backreflections into the laser.The pump beams were focused into the Raman cellwith high-energy antireflection-coated Suprasil lenses.The output from the cells was collimated with thesame type of lenses.

The SRS efficiency was measured with differentmixtures of H2 or D2 with Ar as a buffer gas, wherethe buffer gas was used to enhance energy transfer tothe first Stokes line through lowering of the Ramangain.15 In this setup, quantum efficiencies as high as70% for H2 and 40% for D2 have been achieved. 5

Using N2 as a buffer gas yielded comparable conver-sion efficiencies. However, SRS also occurs in N2,which yields extra output lines in addition to thedesired H2 and D2 lines. This complication is avoidedby the use of buffer gases that are not Raman active.Next to Ar, He was used as a buffer, but lowerefficiencies were found. The gas mixtures and pulseenergies used for the ozone measurements are listedin Table 2. The lifetime of a gas filling was limitedonly by gas leaks and the same shifting results couldbe obtained for many months with the same filling.

2. Receiver OpticsThe two composite laser beams sounding the atmo-sphere are symmetrically aligned with respect to thetelescope axis. With the given geometry, full overlapbetween the laser beams and the telescope field ofview can be reached, ideally at a minimum altitude ofapproximately 300 m. A negative lens with f = - 300mm is placed at the exit of the telescope to produce acollimated beam approximately 25 mm in diameter.The beam is folded 900 by a UV mirror toward a lenswith a f = 125 mm to focus the beam on the entrancepupil of the monochromator. The aperture ratio ofthe obtained beam is f/5, which is well suited to fill


IEEE - 488

Fig. 1. Schematic overview of the tropospheric ozone DIAL system. D's, dichroic mirrors; M's, UV mirrors.

the aperture of the monochromator. The lenses areantireflection coated to maximize throughput. Themonochromator is used to suppress nonresonantbackground and to separate the measurement wave-lengths. The UV folding mirror also reduces thenonresonant background. In the exit focal plane ofthe monochromator a mask is placed with threepupils where the wavelengths are focused. Thebeams are directed toward the detectors with smallaluminum, MgF2-coated mirrors for the 266- and299-nm signals, and the 289-nm signal is left unob-structed. Optical cross talk between 289 and 299nm is eliminated through temporal separation of thetwo laser pulses (see Subsection 3.C.4). Daylightbackground could only be detected at 299 nm andbarely exceeded the photomultiplier dark current.

3. Signal DetectionThe optical signals are detected by one photomulti-plier tube (PMT) for each wavelength. The PMT'swere factory selected to have similar high linearityand low dark current. The voltage-divider networkwas designed for pulsed applications with peak volt-ages into 50 n as high as 2-4 V. However, thePMT's were never exposed to light pulses that would

give output signals in this range, because spuriousresponse of the PMT's would result: overexposureof the PMT causes the output to saturate (i.e., itcauses a nonlinear response). But, even when aPMT is exposed to a short light pulse of moderateintensity, spurious response can result in the form ofafterpulsing and signal-induced bias (SIB). After-pulsing shows as a deformed and delayed echo of theoriginal pulse shape, whereas SIB is a slowly decayingnoise tail, also occurring after the true signal hasgone out. The mechanism for the occurrence ofafterpulsing and SIB is not well understood-theyseem to be the result of a complicated interactionbetween the exposure of the photocathode and thegain of the PMT (i.e., the structure of the dynodechain and the electrical network). Afterpulsing andSIB are often grouped under the same name, signal-induced noise, because they have the same source,but they are separated here because they are dealtwith differently.

The effects of nonlinearity, SIB, and afterpulsing,were studied with an artificial lidar signal 7 and, withthe results, simulation was used18 to quantify theproblem. This showed that the influence of nonlin-earity on the concentration calculation precludes the


Table 2. System Parametersa

SpecificationsSubstytem

EmitterLaser Nd:YAG, 10-Hz repetition rate

600-700-mJ/pulse at 1064 nm, 8-10-ns pulse80-90-mJ/pulse at 266 nm, 4-5-ns pulse

Raman cells 2-m lengthf = 1100-mm lensesUncoated Suprasil windows, 20 mm thick

Gas filling 289 nm (S1, D2): 10 bars D2, 25 bars Ar299 nm (S1, H2 ): 5 bars H2 ,23 bars Ar

Gas purity D2, 99.7%; H2, 99.999%; Ar, 99.999%Emitted 266-nm, 30-mJ/pulse, < 0.7 mrad divergence

energy289-nm, 25-mJ/pulse, < 0.6 mrad divergence299-nm, 40-mJ/pulse, < 0.6 mrad divergence

ReceiverTelescope Dall-Kirkham configuration

0 60 cm, f/ 12, 2-mrad field of view (maxi-mum)

MgF2 coatedWavelength Czerny-Turner configuration

separation 300-nm focal length, [/4.21800 line/mm grating, passband < 6.5 nm at

1.2-mrad field of view20% throughputAdapted for simultaneous detection of three

wavelengthsDetectors Thorn-EMI 9817QA, linear focused

S20 cathode (15-20% quantum efficiency)12 dynodes (BeCu), 3 x 106 gain

Digitizers LeCroy TR6810, 5 MHz, 12 bits409.6-mV maximum input sensitivity

Computers HP1000/A600+LeCroy 6010 CAMAC controller

aS1, First Stokes line.

use of signals larger than 400 mV into 50 fl for thePMT's that were used. Therefore neutral-densityfilters with 10% transmission had to be used to keepthe signal from exceeding this limit for low-altitudemeasurements.

Earlier publications'9-2' suggest that the afterpuls-ing is a function of the PMT gain alone or that bothgain and gain variations are important. So, afterpuls-ing can be minimized by ensuring voltage stability ateach dynode; however, this does not remove SIB.SIB seems to be a function of the height of theprimary pulse and the decay-time constant of thesignal. A way to correct for SIB is to perform a fit onthe background part of a lidar signal, after subtrac-tion of the dc level, and subsequently to subtract thenew-fitted-baseline, extrapolated to R = 0 fromthe signal (see Fig. 2). Observations by us andothers 8 ,22 confirm that SIB can be treated as anexponentially decaying tail. Therefore a linear fit onthe logarithm of the lidar signal was performed.This method is sensitive to noise, however, and betterresults can be obtained by fitting an exponentialfunction to the raw signal. An additional problem isthat the effects can vary substantially between indi-vidual PMT's.

4. Data Acquisition and ProcessingThe output from the PMT's is digitized by a CAMAC-based transient recorder for each wavelength. Themodules are placed in a LeCroy 1434A CAMACmainframe with a LeCroy 6010 GPIB-interface andCAMAC controller. The controller accumulates aspecified number of shots before the data are trans-ferred to the main computer for storage and process-ing. In a typical measurement 1000 lidar returnsare averaged. This averaging, which is needed toimprove the signal-to-noise ratio, also improves thedynamic range that can be detected by the system.In practice, however, there is a limit to the improve-ment, because of imperfections of the digitizers.This must be borne in mind when the weaker parts ofthe signals are interpreted. As a rule of thumb,signal values lower than 0.1 (least-significant bit)cannot be used.

One of the lasers serves as a master oscillator and isused to trigger both the analog-to-digital conversionof the signals originating from this laser (266 and 289nm) and the firing of the second laser through a delayof approximately 500 jis. This allows us to avoid anyoptical cross talk between signals received from thetwo lasers. The second laser triggers the digitizationof the signal at 299 nm. To avoid electrical cross talkfrom the laser to the detection electronics, by theQ-switch synchronization line, by the ground connec-tions, or by pickup by cables, optocouplers are used toisolate the lasers electrically from the rest of thesystem. The lasers also use separate power linessuch that no connection exists between the laserground and other ground connections in the system.Furthermore, care has been taken to avoid loopsthrough ground connections in the detection system.

The concentration was calculated with a linearleast-squares fit of a segment of the DIAL curve [Eq.(2)]. The slope of the fitted curve was used tocompute the concentration within the segment, andthe standard deviation was used to indicate the noiselevel ( 1 a). We obtained a full profile by stepwisefitting of subsequent intervals of the DIAL curve.So, in the profiles shown, subsequent points are notindependent, but they illustrate the shape of theprofile. The true resolution and uncertainty inter-vals are indicated by error bars. The length of theintervals could be chosen arbitrarily.

4. Lidar Signals and Ozone MeasurementsIn Section 4 the state of the tropospheric ozone DIALsystem is discussed and illustrated by showing resultsobtained between the spring of 1990 and the begin-ning of 1991.

A. Lidar Return Signals

In Fig. 3 typical return signals for the three wave-lengths are shown. The returns are corrected forthe inverse-square-range dependence of the lidarsignal and are drawn on a logarithmic scale. A


-

to)

< ~~~Uncorrected

Corrected

0 2250 4500 6750 9000Altitude (m)

Fig. 2. Signal-induced-noise effect on the 266-nm signal. The data were taken at 11:09 a.m., 5 September 1990. Solid curves,

uncorrected and corrected signals; dashed line, fitted baseline. See also Fig. 7 below.

short-range limit of = 600 m has been achieved. ForDIAL measurements, it is very important that a fulloverlap-between the field of view from the telescopeand the area illuminated by the laser-is retainedover the whole measurement range. Partial overlapcannot be used because the beam profiles from differ-ent wavelengths are not exactly equal. This resultsin different overlap functions for each wavelength.Not even the 266- and the 289-nm beams generatedby the same laser-Raman cell have equal overlapfunctions. It cannot be verified whether an appar-ent absorption feature is indeed the result of absorp-tion or is caused by differences in the overlap function.This is particularly important in the region where fulloverlap is approached. An indicative method used to

5

x

.)

C

3

2

0

assess the short-range limit is shown in Fig. 4. Iffull overlap is reached at the normalization rangeRnorm, ln[P(R)R2 /P(Rnorm)Rnrm 2] should decreasemonotonically, provided that there is no sharp aero-sol gradient near Rnorm.

As mentioned in Subsection 3.C.3, PMT's havesome serious limitations. To reduce nonlinearity allmeasurements up to 3 km of altitude were made witha 90% optical attenuator in the beam path. Forhigher altitudes, measurements without an attenua-tor were made, but SIB often distorted these measure-ments. The fitting procedure to remove SIB thatwas described in Subsection 3.C.3 is illustrated in Fig.2. The procedure was not completely satisfying,however, because SIB was not always clearly visible

6000

Altitude (m)

Fig. 3. Typical range-corrected LIDAR return signals for 266,289, and 299 nn. The data were taken at 12:10 p.m.,22 Feb.1991.


0 1000 2000 3000 4000 5000

0

0

M

I

N1

to0 1125 2250 3375 4500

Altitude (m)

Fig. 4. Range-corrected LIDAR return signals at 289 and 299 nm,normalized at 630 m and shown on a logarithmic scale. Themonotonic decay of the signals shows that full overlap between thetwo beams and the telescope is reached above 630 m. The datawere taken at 11:17 a.m., 3 August 1990. See also Fig. 6, below.

and could not be used if afterpulses were also present.In practice both SIB and afterpulsing proved to belimiting factors in the maximum range that could beachieved even after corrective treatment of SIB.

B. DIAL Curves and Ozone Profiles

In Fig. 5(a) typical DIAL curves for the wavelengthpairs 266 and 289 nm and 289 and 299 nm are shown.Figure 5(b) shows the ozone profiles derived from theDIAL curves of the respective pairs. The rangeresolution and the uncertainty of the ozone concentra-tion are indicated. The ozone-absorption cross sec-tions used were taken from Refs. 23 and 24. Themeasurement shown in Fig. 6(a) clearly reveals thetop of the mixed layer at approximately 2 km ofaltitude. Different concentrations within and abovethe mixed layer were found [Fig. 6(b)]. Ozone valuescould be obtained up to an altitude of approximately 5km. There is also a notable difference in signalstrength inside and above the mixed layer, because ofdifferent aerosol content-the free troposphere haslow aerosol loading, so that there the signal is almostentirely caused by Rayleigh scattering.

The effect of SIB on the concentration retrieval isshown in Fig. 7. Toward the far end of the measure-ment interval the slope of the signal at 266 nmbecomes smaller, which translates to a lower ozoneconcentration. Ultimately the ozone values becomenegative. Correcting the 266-nm signal for SIB asdescribed in Subsection 3.C.3 yields some improve-ment in concentration retrieval. In Fig. 7(b) theslope of the log of the range-corrected lidar signal isshown, together with the expected slope for a moder-ate ozone concentration. For the 266-nm signal, inthe range from 1500 to 2000 m the values aresubstantially more trustworthy, but the weaker partsof the signal still suffer from residual SIB. The289-nm signal is also plotted in Fig. 7(b), and nosevere SIB is present below 2250 m in this signal.The resulting ozone profiles are shown in Fig. 7(c).

0

0

a: o

a

o0

N

0

0

Cu

0

I

00N

0

a)C_

0

0 7,0 1500

Altitude (m)(a)

2250 3000

0 750 1500 2250 3000

Altitude (m)

(b)

Fig. 5. (a) DIAL curves for the wavelength pairs 266 and 289 nmand 289 and 299 nm, (b) ozone profiles. 0, ozone derived from thewavelength pair 266/289 nm; x, wavelength pair 289/299 nm;horizontal bars, ±1(i uncertainty interval; vertical bars, rangeresolution (210 and 330 m, respectively). The data were taken at12:10 P.M., 22 February 1991.

5. Error Calculation

A. Statistical Errors

The error bars on the ozone profiles give one animpression of the statistical errors in the ozone valuesthat are due to noise. Additional statistical errors,not accounted for in the error bars, are the uncer-tainty of the differential cross section of ozone and itstemperature dependence, which have errors of ap-proximately 1% and 0.5%, respectively.23 The statis-tical uncertainty varies with altitude: for 1000 lasershots at 600 m of altitude the uncertainty is 1-5%; at1 km of altitude (provided the mixed layer extends tothat altitude) the uncertainty is 2-10% for a spatialresolution of 180 m; at the top of the mixed layer theuncertainty can reach 20%; in the free troposphere(3-5 km of altitude, 300-400-m range interval) a5-40% uncertainty was found.

The uncertainty depends on both the ozone concen-tration and the weather conditions (see Table 3).


289/299 nm

266/289 nm

I I I A

NCu

Cu

N

a: ,Ec

i:

I'

0

0

to(V00

0

o

C

S 000

0Ir

0 1 125 2250 3375Altitude (m)

(a)

0 1125 2250

Altitude (m)

(b)

3375

a)

CC

9L-cc02'0)

( u

In

0

04500

Altitude (m)(a)

0

E 0

Ei

ccE,

-a.

CuCu

0

Cu4500

Fig. 6. Measurement with a pronounced boundary between themixed layer and the free troposphere. The boundary is indicatedby an arrow. (a) Range-corrected LIDAR signals at 289 and 299nm and (b) ozone profile.

0 750 1500 2250 30C0

Altitude (m)(b)

ruCu

00

Because SIB forces the use of a 90% attenuation filterin the detection (see Subsection 3.C.3), the signal-to-noise ratio must be improved through longer averag-ing times. Increasing the averaging time by a factorof 10 or more will reduce the uncertainty by a factorof = 3. Larger range elements can also be used atthe expense of spatial resolution.

B. Systematic Errors

The systematic errors in ozone DIAL measurementscan be divided into three categories: (1) differentialbackscatter errors [term (3b)], (2) differential extinc-tion errors [term (3d)], and (3) differential absorptionerrors [term (3c)]. Errors (1) and (2) are caused bythe large wavelength separation between the on andthe off signals; the assumptions that simplify theconcentration evaluation are no longer valid.3'4' 11' 1 2' 1 4'25

Error (3) arises from differential absorption by gasesother than ozone. Below, the magnitude of theerrors is examined and possible remedies are sug-gested. The errors examined are differential Ray-leigh extinction, 02 differential absorption, SO2 and

0 A

Uncorrected

0 750 I 500 2250 3000

Altitude (m)

(c)

Fig. 7. Example of the effect of signal-induced noise on an ozonemeasurement. (a) Corrected and uncorrected LIDAR signals at266 nm together with the signal at 289 nm. See also Fig. 2. (b)Slope of the log of the range-corrected lidar signal at 266 nm beforeand after SIB correction. The dashed lines indicate the expectedslope for a moderate ozone concentration (40 ppbv) at the displayedwavelengths. Also shown for comparison is the slope of the289-nm signal together with the expected slope. (c) Ozone profilesuncorrected and corrected for SIB. Data taken at 11:09 a.m., 5September 1990.


_ ~~~~~~~~~Top mixed layer

_ ~~I I II

Table 3. Summary of Statistical Errors Relative to an OzoneConcentration of 80 Rig/m3

Error (%)

Clear HazyAltitude Resolution Atmo- Atmo-

Error Source (km) (i) sphere sphere

Photon noise 0.6 180 1 51.0 180 2 10

PBLa top 180 5 203.0 400 5 205.0 600 40 100

Ozone-absorption 1.0cross sectionb

Temperature- 0.5dependencecross sectionb

aPlanetary boundary layer.bAtmospheric condition not specified.

NO2 differential absorption, and differential aerosolextinction and backscatter.

The errors are calculated as concentration offsetsfor both wavelength pairs. The errors are related toa homogeneous ozone concentration of 1018 m-3,which corresponds to 80 [Lg/m

3 or 37 parts in 109,volume (ppbv). This value is frequently encounteredwithin the mixed layer. In the free troposphere the

ozone concentration usually is somewhat lower, result-ing in relatively larger errors. The reported errorestimates depend on both the magnitude of thesystematic effect and the magnitude of the differen-tial ozone-absorption cross section. The error esti-mates are listed in Table 4. Negative terms are to besubtracted from the measured ozone values, andpositive terms are to be added.

A distinction is made between a clear, clean day anda hazy, polluted day to illustrate the influence of theatmospheric conditions. A clear day is characterizedby a visibility of 10 km, an aerosol gradient ofapproximately 1 order of magnitude across the top ofthe mixed layer, and SO2 and NO2 concentrationsequal to the 50th percentiles at the DIAL system'slocation. A hazy day is characterized by a visibilityof 2 km, an aerosol gradient 5 times larger than on aclear day, and SO2 and NO2 concentrations equal tothe 95th percentiles.

1. Differential Rayleigh ExtinctionThe Rayleigh (or molecular) extinction coefficientsare known for all wavelengths used, so the effect canbe corrected for with the help of a density profile.At present a standard pressure formula26 is used tocalculate the correction term. The correctionamounts to -9.0% for 266 and 289 and -20.0% for289 and 299 nm at low altitudes at 15 C and 1 atm

Table 4. Systematic Errors Relative to an Ozone Concentration of 80 zug/m3 a

WavelengthError Source Compartment Pair (nm) Relative Error (%) Uncertainty (%)

Errors independent of visibility or pollution levelRayleigh extinction 266/289 -9.0 0.2

289/299 -20.5 0.3Oxygen absorption 266/289 -8.9 1.8

289/299 -1.4 0.3

Errors varying with visibility and pollution levelb Clear Hazy Clear Hazy

SO2 absorptiond Mixed layer 266/289 0.6 1.9 0.25 0.8289/299 -5.6 -17.0 2.2 6.8

NO2 absorptione Mixed layer 266/289 0.4 0.8 0.2 0.3289/299 2.5 4.8 1.0 2.0

Aerosol extinctionf Mixed layer 266/289 -2.9 -14.7289/299 -8.3 -41.7

Top mixed layer 266/289 -10.6 -53.0289/299 -30.0 -150.0

Free troposphere 266/289 -0.4 -0.5289/299 -1.2 -1.5

Aerosol backscatter Mixed layer 266/289 - 1.0 -1.2289/299 -3.1 -3.5

Top mixed layerg 266/289 -11-+31 -6-+48289/299 -33-+94 -16-+141

Free troposphere 266/289 0.8 2.0289/299 2.9 7.2

aUncertainties will remain after correction and will add to the noise in the derived ozone values.bClear, visiblity 10 km, pollution 50th percentile; hazy, visibility 2 km, pollution 95th percentile.CThe uncertainty estimates for the aerosol errors are coupled and are included in Table 5, below.d5Oth percentile, 9.5 x 1016 m- 3 (10 pLg/m3 ); 95th percentile, 2.8 x 1017 m-3 (30 Lg/m

3 ).e5Oth percentile, 5.2 x 1017 m- 3 (40 pLg/m3 ); 95th percentile, 1.0 x 1018 m-3 (79 pLg/m3 ).fBackscatter-to-extinction ratio 0.028 sr-', wavelength dependence 0.9.gDeviations going out of the mixed layer. A negative error is followed by a positive deviation.


ground pressure. For higher altitudes the correc-tion becomes smaller as the density decreases.Using measured values of the temperature and pres-sure at ground level to correct the pressure formulawill bring the relative ozone-correction uncertaintydown to less than 0.3%.

2. 02 Differential AbsorptionThe uncertainty in the absorption cross sections of 02is quite high,2 72 8 because no accurate high-resolutionspectra of 02 are available at these wavelengths; theuncertainty is estimated to be 20%. With the approxi-mate values that were found (cr2 66 = 1.30 x 10-29 m 2

,

U2 89 = 4.46 X 10-3' m 2 , 29 9 = 1.86 x 10-31 m2 ), theoffset in concentration that is due to differential 02

absorption at low altitudes was calculated. The off-set that was found is not negligible: -9.0 ± 1.8% for266 and 289 nm and - 1.4 ± 0.3% for 289 and 299 nm,so a correction should be made, at least for the firstwavelength pair. A density profile as used for theRayleigh scattering can be used for the correctionover the whole altitude range because 02 is uniformlymixed.

3. S 2 and NO2 Differential AbsorptionSO2 and NO2 are present at high concentration levelsin the mixed layer only, where these pollutants areemitted. In the free troposphere the concentrationsare much lower, and their influences are negligible.The errors are dependent on the absorption crosssections of the two gases and on their vertical concen-tration profiles. The influence of these gases wascalculated assuming a uniform mixing ratio of thepollutants in the mixed layer and using the crosssections for SO2 and NO2 given by Thomsen 2 9 andBass,30 respectively. For SO2 and NO2 concentra-tions the 50th and 95th percentiles for Bilthoven, TheNetherlands, in 1989 were used. 3 '

Moderate concentrations of SO2 can have a substan-tial influence on the 289 and 299 nm measurement,whereas the influence of NO2 on this wavelength pairis smaller. The SO2 influence is below +5% for allpercentiles for 266 and 289 nm and up to -35% for289 and 299 nm. For NO2 the values are up to+ 1.5% for 266 and 289 nm and up to +9% for 289 and299 nm. It is worth noting that the relative errorwill be smaller if the ozone concentrations are high,as is often the case during episodes when NO2 concen-trations are high. The largest relative error can beexpected under winter smog conditions, when SO2concentrations can reach high levels without causinghigh ozone concentrations.

To correct for interferences from SO2 and NO2 ,concentrations measured with point sampling moni-tors at ground level could be used, assuming uniformmixing throughout the mixed layer.

4. Differential Aerosol Extinction and BackscatterIn the presence of aerosols two simultaneous effectscause systematic errors: (1) the wavelength depen-dence of the extinction [term (3d)], which gives rise to

a differential extinction error in the same way as doesmolecular scattering, and (2) regions of inhomoge-neous aerosol concentration that cause a differentialbackscatter error [term (3b)] that is due to thewavelength dependence of the backscatter.

Aerosol loading is most prominent in the mixedlayer, and a strong gradient in aerosol concentrationoften exists at the top of the mixed layer, so thebackscatter error will be most prominent there. Themagnitude of the extinction and the backscattererrors in the mixed layer and through the top of themixed layer were calculated from the error terms interms (3b) and (3d). The aerosol profile used had adrop in backscatter of approximately 1 order ofmagnitude at the top of the mixed layer. 432 Adistinction is made between a clear day (visibility 10km) and a hazy day (visibility 2 km). The errorsfound show that the 266 and 289 nm wavelength pairis much better suited for ozone measurements in themixed layer than in the 289 and 299 nm pair, wherethe extinction and backscatter errors reach 40-80%,even for clear weather. Under hazy conditions theaerosol-related errors for the 266 and 289 nm wave-length pair also become large. It is clear that acorrection should be made for these errors.

To correct the ozone DIAL measurements for theaerosol errors, backscatter profiles or extinction pro-files are needed in addition to backscatter-to-extinc-tion ratios for the wavelengths that are used. Likethe molecular scattering, the aerosol (or Mie) scatter-ing and extinction is wavelength dependent.26 The Xdependence and also the backscatter-to-extinctionratio vary with the type of aerosol. Lidar measure-ments yield no information on the wavelength depen-dence unless a larger number of wavelengths is used.Furthermore, the inversion of lidar data unambigu-ously into extinction or backscatter profiles is impos-sible because both extinction and backscatter occursimultaneously in the lidar equation, whereas theirinterrelation is not known a priori. Several methodshave been proposed142 5 to remove the aerosol influ-ence on the ozone profiles. These methods are usedto obtain vertical aerosol extinction and backscatterprofiles at one wavelength and extrapolate them toother wavelengths, making use of assumptions aboutthe aerosol properties. Using the Klett method, thiscan be done best at the least-absorbed wavelength.25

Alternatively, the N 2 Raman signal could be used todeduce the backscatter-to-extinction ratio profile.'4

Actual application of the aerosol corrections isoutside the scope of this paper. How well the correc-tion can be performed has been estimated fromsimulations by others.' 4 The estimates of the uncer-tainties remaining after correction are based on thispaper, and are listed in Table 5. Since Papayannis etal. 14 calculated the remaining uncertainty after correc-tion of both (coupled) aerosol effects, the remaininguncertainties could not be split into an extinctionterm and a backscatter term. The uncertainties for289 and 299 nm are not mentioned in Ref. 14,therefore it is assumed that the uncertainties remain-


Table 5. Errors for an Unoptimized and an Optimized System

Unoptimizeda Optimizedb

Systematic UncertaintyWavelength Noise (%) Error (%) Noise (%) (%)

Compartment Pair (nm) Clear Hazy Clear Hazy Clear Hazy Clear Hazy

Mixed layer 266/289 5-10 5-20 -2.9 -13.2 3-5 3-10 4 15289/299 5-10 5-15 -14.5 -57.4 3-5 3-5 15 30

Top mixed layer 266/289 5-20 15-100 -20.0-20 -56--5 3-10 3-50 15 -289/299 5-20 10-30 -70-60 -180--20 3-10 5-15 40 50

Free troposphere 266/289 20-100 - - - 10-50 ---289/299 10-40 20-100 1.7 5.7 10-20 10-50 5 5

aErrors caused by Rayleigh extinction and oxygen absorption have been removed. Noise figures are for 1000 laser shots, with 90%attenuator and analog detection.

bAll errors are removed. Noise figures are for 4000 laser shots, with 90% attenuator and analog detection.

ing after correction will be of the same order ofmagnitude or larger. However, because the 289 and299 nm wavelength pair is less suited for measure-ments in the mixed layer, where the aerosol errors arelargest, the magnitudes are of less importance. Theaerosol error correction presents a formidable chal-lenge, and in urban atmospheres the errors canbecome large. It is possible that more than threewavelengths are needed.

C. Summary of Errors

In Table 5 we present a comparison of the errors withand without optimization with respect to statisticaland systematic errors (i.e., longer accumulation andapplication of corrections). The uncertainties of thecorrections add to the statistical error. For theunoptimized case the Rayleigh extinction and the 02absorption are already accounted for, adding uncer-tainties of 2.0% and 0.6% for 266 and 289 nm and 289and 299, respectively, which is mainly because of theuncertainty in the 02 absorption cross section. Inthe optimized case, 4000 laser shots are assumed, andcorrections are applied for the systematic errors.

Table 5 shows that under clear conditions the 289and 299 nm wavelength pair is much less suited formeasurements in the mixed layer than the 266 and289 nm wavelength pair. For hazy conditions theerrors become even larger for the 289 and 299 nmwavelength pair and the errors for the 266 and 289nm wavelength pair also become large. In the freetroposphere, the 266 and 289 nm wavelength pair canno longer be used-at least in analog mode-becauseof SIB and low signal values.

All the errors mentioned above are present in theozone profiles shown in this paper. Correctionschemes were applied with the methods suggested,except for the aerosol-related errors. This meansthat the profiles presented have preliminary status.

6. Application of Routine Ozone Measurements

In Fig. 8 the evolution of the ozone ground-levelconcentration over the day is shown together withDIAL measurements at different altitudes. TheDIAL values are averages over an altitude interval of

180 m and over a time interval of 1 h (typically 2-6measurements of 1000 laser shots/h). Correctionsfor Rayleigh extinction, 02, and SO2 and NO2 differen-tial absorption were made with ground-level data atthe system's location. Also shown in Fig. 8 is theheight of the mixed layer as determined from Mielidar data (at 1064 nm).

On 27 June 1990 [Fig. 8(a)], the ground-levelconcentration rises during the morning. The concen-trations at the altitudes 600, 800, and 1200 m alsorise but delayed in time. Until the mixed layerreaches a certain altitude the concentration at thataltitude is lower than the mixed-layer concentration.This points to production in the mixed layer. TheDIAL values are a combination of the two wavelengthpairs.

On 3 August 1990 [Fig. 8(b)] an episode of summersmog is present, and the mixed-layer height (or aresidual layer) is at approximately 2 km of altitudeduring all measurements. All the measurement alti-tudes are within the mixed layer, and their concentra-tion development more or less follows that of thedevelopment measured with the ground monitor.Measurements at higher altitudes (see also Fig. 6)show that the concentrations above the mixed layerwere lower. Only the 289 and 299 nm wavelengthpair was used to obtain the ozone values.

It is interesting to note the differences in thebehavior of the ozone concentration at certain alti-tudes between the days with different meteorologicalconditions: on one day there seems to be mixingfrom lower to upper altitudes and production ofozone, whereas on another day the mixing procesesare not discernible. Such mixing processes cannotbe derived from ground-level data alone. Even if theozone values are preliminary, the examples show thatthe ozone DIAL system can be a powerful tool inatmospheric research.

7. Discussion and Future Plans

The large dynamic range of the signals cannot behandled by the detectors: the strong backscatterfrom the close ranges has a detrimental effect on thesignal at far ranges through signal-induced noise.


4 6 8 10 12 14 16 18 20 22

Time (M.E.T.)

-- LIDAR(1200 m)

-- Ground level

-4J--- UDAR(1500 m)

-.- LIDAR ----- LIDAR(800 m) (600 m)

.-- x---- PBL(a)

16 18 20 22

Q LIDAR(600 m)

-. UDAR(1 125 m)

-- Ground level ----x---- PBL(b)

Fig. 8. Application of LIDAR measurements of ozone. (a) 27 June 1990, (b) 3 August 1990. M.E.T., define; PBL, planetary boundary

layer.

This limits the maximum range that can be coveredby the system. We plan to improve the handling ofthese problems through further studies of the PMT'sand by changing the configuration of the system.The studies should include both theoretical studies ofSIB and careful measurements to characterize thePMT's.

A fast shutter can be added to the configuration toprotect the detectors from exposure to large signalsfrom close range. To obtain very short transitiontimes, an electro-optical device could provide anelegant solution. Some progress has already beenmade in implementing such a device.5 Furthermore,to increase the range of the device from 5 to approxi-mately 15 km it is necessary to use photon counting.Because low-altitude measurements are made withan attenuator, installation of a second, smaller tele-scope is planned. In this way, analog and photoncounting channels can be split to have their ownreceiver sections.

The system performance must be checked against

in situ monitors and other ozone DIAL systems inintercomparison campaigns. These campaigns canprovide valuable information, in particular on thequestion of how many and which wavelengths areneeded to retrieve ozone profiles in the lower tropo-sphere under realistic influence of aerosol and pollut-ants. Improved error-correcting algorithms may bedeveloped with the help of these data.

Because the range of lidar measurements is re-stricted by a cloud base, the representativeness ofozone DIAL profiles should be tested. A study with along-term data set collected by balloon-borne sondeshas been performed.3 3 In that study statisticallysignificant differences in ozone-concentration profileswere found between soundings performed under clear-sky and cloudy-sky conditions. It was concludedthat, to obtain a nonbiased data set of ozone profilesperformed by ozone DIAL, routine measurementsshould be performed with a cloudiness of up to 6 octas(75% cloud cover). This implies that the ozone DIALsystem should also be operated under fairly cloudy


180

160

140

c 120

' 100

a 800

o 60

40

20

0

2500

2000

1500 .'

1000 -jco

500

0

240220 200 -180 -

E 1605 140as 120:o 100 No 80-

60 -4020-

0

' ~~~~~....,x--- x

- x,

x.. . . . x-

.6000

54004800

4200 E

-3600 5

3000 m2a

2400 m'co

1800 a

1200

600

-0

4 8 10 12 14

Time (M.E.T.)

.

conditions, and ozone profiles should be obtained byshooting in between the clouds. The measurementprocess should be automated as much as possible tobe able to do this.

8. Conclusions

A tropospheric ozone lidar system for routine mea-surements has been constructed. The system usesSRS of a fixed-frequency laser at 266 nm to firstStokes components in H2 and D2 to generate thewavelengths necessary for ozone measurements.The Raman-shifting results show that very highconversion efficiencies can be obtained for H2 andsomewhat lower efficiencies can be obtained for D2.

The detection and analog-to-digital conversion sys-tems have been tested and showed that the detectorsare susceptible to signal-induced artifacts. Someprogress has been made in understanding and correct-ing the artifacts and the nonlinearity. Further study,e.g., in a shutter system, and the modeling of detectorresponse are clearly warranted. On the whole, signal-induced noise effects have been found to be a limitingfactor in the maximum achievable range of the mea-surements, at least when transient recording is used.

Ozone profiles have been obtained from approxi-mately 600 m up to approximately 5 km altitude,depending on weather conditions and ozone concentra-tions. The total inaccuracies estimated and listed inTable 5 can be reduced with longer averaging timeand corrective algorithms for the aerosol influence.The remaining uncertainties are also listed in Table5. The precision of the measurements at higheraltitude will improve further if photon counting isused. Moreover, photon counting will extend themeasurement range to approximately 15 km of alti-tude.

The ozone-monitoring lidar system described inthis paper can be a powerful tool in atmosphericresearch, especially in providing data for use inatmospheric modeling.

References1. D. Kley, I. S. A. Isaksen, and S. A. Penkett, "TOR, tropo-

spheric ozone research," EUREKA subproject proposal (EURO-TRAC scientific secretariat, Garmisch-Partenkirchen, Ger-many, 1987).

2. R. Barbini, M. J. T. Milton, J. Pelon, and C. Weitkamp,"TESLAS, Joint European Program for the TroposphericEnvironmental Studies by Laser Sounding," EUROTRACsubproject proposal (EUROTRAC scientific secretariat, Gar-misch-Partenkirchen, Germany, 1987).

3. I. S. McDermid, "Ground-based lidar and atmospheric stud-ies," Surv. Geophys. 9, 107-122 (1987).

4. J. A. Sunesson, "RIVM tropospheric ozone lidar: feasibilityand definition," RIVM-Rep. 222201002 (National Institute ofPublic Health and Environmental Protection, Bilthoven, TheNetherlands, 1990).

5. J. A. Sunesson and A. Apituley, "RIVM Tropospheric OzoneLidar: system description and first results," RIVM-Rep.222201006 (National Institute of Public Health and Environ-mental Protection, Bilthoven, The Netherlands, 1990).

6. R. M. Measures, Laser Remote Sensing (Wiley, New York,1984).

7. H. Edner, G. W. Faris, J. A. Sunesson, and S. Svanberg,"Atmospheric atomic mercury monitoring using differentialabsorption lidar techniques," Appl. Opt. 28, 921-930 (1989).

8. K. W. Rothe, U. Brinkmann, and H. Walther, "Remotemeasurement of NO2 emission from a chemical factory by thedifferential absorption technique," Appl. Phys. 4, 181-182(1974).

9. J. G. Hawley, L. D. Fletcher, and G. F. Wallace, "Ground-based ultraviolet differential absorption lidar (DIAL) systemand measurements," in Optical and Laser Remote Sensing,D. K. Killinger and A. Mooradian, eds. (Springer-Verlag,Berlin, 1983), Chap. 3.

10. H. Edner, J. A. Sunesson, and S. Svanberg, "NO plumemapping by laser-radar techniques," Opt. Lett. 13, 704-706(1988).

11. E. V. Browell, A. F. Carter, S. T. Shipley, R. J. Allen, C. F.Butler, M. N. Mayo, J. H. Siviter, Jr., and W. M. Hall, "NASAmultipurpose airborne dial system and measurements of ozoneand aerosol profiles," Appl. Opt. 22, 522-534 (1983).

12. J. Pelon and G. Megie, "Ozone monitoring in the troposphereand lower stratosphere: evaluation and operation of a ground-based lidar station," J. Geophys. Res. 87, 4947-4955 (1982).

13. 0. Uchino, M. Tokunaga, M. Maeda, and Y. Miyazoe, "Differ-ential-absorption-lidar measurement of tropospheric ozonewith an excimer-Raman hybrid laser," Opt. Lett. 8, 347-349(1983).

14. A. Papayannis, G. Ancellet, J. Pelon, and G. Megie, "Multiwave-length lidar for ozone measurements in the troposphere andthe lower stratosphere," Appl. Opt. 29, 467-476 (1989).

15. D. Diebel, M. Bristow, and R. Zimmerman, "Stokes shiftedRaman laser lines in KrF-pumped hydrogen: reduction ofbeam divergence by addition of helium," Appl. Opt. 30,626-628 (1991).

16. A. Luches, V. Nassisi, and M. R. Perrone, "Improved conver-sion efficiency of XeCl radiation to the first Stokes at highpump energy," Appl. Phys. B 47, 101-105 (1988).

17. G. Kunz and F. Swart, "Light source for dynamic testing ofphoto detectors," TNO-Rep. FEL 1989-69 (TNO, Den Haag,The Netherlands, 1989).

18. C. N. de Jonge, D. P. J. Swart, and J. B. Bergwerff, "DIALtechnique for N0 2-detection: system quality and simula-tion," RIVM-Rep. 222701001 (National Institute of PublicHealth and Environmental Protection, Bilthoven, The Nether-lands, 1991).

19. Y. Likura, N. Sugimoto, Y. Sasano, and H. Shimizu, "Improve-ment on lidar data processing for stratospheric aerosol measure-ments," Appl. Opt. 26, 5299-5306 (1987).

20. H. Sang Lee, G. K. Schwemmer, C. L. Korb, M. Dombrowski,and C. Prasad, "Gated photomultiplier response characteriza-tion for DIAL measurements," Appl. Opt. 29, 3303-3315(1990).

21. R. E. W. Pettifer, "Signal induced noise in lidar experiments,"J. Atmos. Terr. Phys. 37, 669-673 (1975).

22. I. S. McDermid, S. M. Godin, and D. T. Walsh, "Lidarmeasurements of stratospheric ozone and intercomparisonsand validation," Appl. Opt. 29, 4914-4923 (1989).

23. A. M. Bass and R. J. Paur, "Ultraviolet absorption cross-sections of ozone. I. The measurements," in AtmosphericOzone, C. S. Zerefos and A. Ghazi, eds. (Reidel, Dordrecht, TheNetherlands, 1984), pp. 606-610.

24. A. M. Bass and R. J. Paur, "Ultraviolet absorption cross-sections of ozone" (National Bureau of Standards, Washing-ton, D.C. 20234, personal communication, 10 August 1992).

25. E. V. Browell, S. Ismail, and S. T. Shipley, "Ultraviolet DIALmeasurements of 03 profiles in regions of spatially inhomoge-neous aerosols," Appl. Opt. 24, 2827-2836 (1985).


26. E. J. McCartney, Optics of the Atmosphere: Scattering byMolecules and Particles (Wiley, New York, 1976).

27. M. W. P. Cann, J. B. Shinn, and R. W. Nicholls, "Oxygenabsorption in the spectral range 180-300 nm for temperaturesto 3000 K and pressures to 50 atm," Can. J. Phys. 62,1738-1751 (1984).

28. E. Trakhovsky, A. Ben-Shalom, U. P. Oppenheim, A. D. Devir,L. S. Balfour, and M. Engel, "Contribution of oxygen toattenuation in the solar blind UV spectral region," Appl. Opt.28, 1588-1591 (1989).

29. 0. Thomsen, "Messung des Absorptionswirkungsquerschnittsvon Schwefeldioxid im Wellenlangebereich von 265 bis 298nm," Ph.D. dissertation (GKSS Forschungszentrum, Geest-hacht, Germany, 1990).

30. A. M. Bass, A. E. Ledford, and A. H. Lauffer, "Extinction

coefficients of NO2 and N2 0 4 ," J. Res. Natl. Bur. Stand. Sect. A80, 143-166 (1976).

31. Laboratory of Air Research, "Luchtkwaliteit, jaarverslag 1989(air quality, annual report 1989)," RIVM-Rep. 222101006(National Institute of Public Health and Environmental Pro-tection, Bilthoven, The Netherlands, 1990).

32. F. X. Kneizys, E. P. Shettle, W. 0. Gallery, J. H. Chetwynd,L. W. Abreu, J. E. A. Selby, S. A. Clough, and R. W. Fenn,"Atmospheric transmittance/radiance: computer codeLOWTRAN 6," Rep. AFGL-TR-83-0187 (U.S. Air Force Geophys-ics Laboratory, Bedford, Mass., 1983).

33. H. de Backer, E. P. Visser, D. de Muer, and D. P. J. Swart,"Potential for meteorological bias in lidar ozone data setsresulting from the restricted frequency of measurement due tocloud cover," J. Geophys. Res. 99, 1395-1401 (1994).


Differential absorption lidar system for routine monitoring of tropospheric ozone

Documents