Journal of Atmospheric Chemistry 23: 51-80, 1996. @ 1996 Kluwer Academic Publishers. Printed in the Netherlands.
51
Intercomparison of Instruments for Tropospheric Measurements Using Differential Optical Absorption Spectroscopy C. C. M. U. A.
CAMY-PEYRET’, B. BERGQVIST2, B. GALLE2, M. CARLEER3, CLERBAUX3, R. COLIN3, C. FAYT4, F. GOUTAIL’, NUNES-PINHARANDA’, J. P. POMMEREAU’, M. HAUSMANN’j, PLATT6, I. PUNDT6, T. RUDOLPH$ C. HERMANS7, P. C. SIMON7, C. VANDAELE7, J. M. C. PLANE8 and N. SMITH8
‘Laboratoire de Physique Mole’culaire et Atmospherique, Universite’ Pierre et Marie Curie, Paris, France, 2Swedish Environmental Research Institute, Gliteborg, Sweden, ‘Laboratoire de Chimie Physique Moltkulaire, Universite’ Libre de Bruxelles, Brussels, Belgium, ’ Universite’ de Mons-Hainaut, Mons, Belgium, ‘Service d’A&onomie du CNRS, Verrieres-le-Buisson, France, “Institutfiir Umweltphysik, Universitat Heidelberg, Heidelberg, Germany, ‘Institut d’A&onomie Spatiale de Belgique, Brussels, Belgium, ‘School of Environmental Sciences, University of East Anglia, Norwich, England
(Received:1 September1994;in final form: 26 June1995) The resultsof anintercomparison campaignof eightdifferentlongpathUV-visibleDOAS instrumentsmeasuringNO2, 03 and SO2concentrationsin a moderatelypollutedurban site are presented. For effectiveopticalpathlengthsof 230and780mtheoverallspreadof thesemeasurements (&lo) are5 x lOlo, 6 x 10” and1 x 10” molec. cmv3 (2.0,2.4, and0.4 ppb) for thesemolecules respectivelywhenall instrumentsuseda commonsetof absorptioncrosssections.The remaining differencesarenot completelyrandomandthe systematicdifferencesareattributedto the different retrievalmethodsusedfor eachinstrument. Abstract.
Key words: NOZ, 03, SOI, intercomparison, differentialopticalabsorptionspectroscopy(DOAS),
troposphere.
1. Introduction The study of the chemistry of the atmosphere, and hence of man-made air pollution, relies for a large part on precise spectroscopic measurements of minor atmospheric chemical species. The measurements require the use of reliable field instruments in order to obtain accurate and continuous data. During the last decade, many such instruments have been developed which are based on the absorption properties of the constituents in various regions of the electromagnetic spectrum ranging from the far infrared to the ultraviolet. Long absorption paths (100 m to 10 km) are generally used to increase sensitivity. These instruments are based on the application of the
52
C. CAMY-PEYRET
ET AL.
Beer-Lambert law which requires knowledge about the absorption cross-sections of the various molecules of interest. These cross-sections have been measured in the laboratory for a large number of these molecules, but not always with the accuracy and precision required by present-day instruments. Indeed, large discrepancies still exist between various sets of measured cross sections. The problems inherent to these field measurements vary according to the wavelength region under investigation, the light source (Sun, zenith sky or various artificial sources),the dispersion device (grating or interferometer) and the detectors used. In contrast to laboratory spectroscopy, the true intensity lo(X), as would be received from the light source in the absence of any atmospheric absorption, is usually difficult to determine. It would involve in principle removing the air from the absorption path. The solution to this apparent dilemma lies in measuring the socalled ‘differential’ absorption. This quantity can be defined as the part of the total absorption which is due to a given molecular species and which varies ‘rapidly’ with wavelength, in contrast to the background absorption which varies ‘slowly’ with wavelength. Accordingly the total intensity can be split into two parts: I(X) = I&)
* SI(X)
where IB(x) is due to the background absorption varying slowly with X and describing a general ‘slope’ and 61(X) shows rapid variations due to absorption lines or bands. The meaning of ‘rapid’ and ‘slow’ variation of the intensity as a function of wavelength is, of course, related to the observed wavelength interval and to the width of the absorption bands or lines to be detected. The extinction due to Rayleigh and Mie scattering can usually be assumedto be slowly varying with X. This slow variation with X is removed by the application of various filtering techniques, in order to leave the differential absorption peaks of the atmospheric trace gases.This filtering process is an important part of the DOAS (Differential Optical Absorption Spectroscopy) method; various methods are employed according to the types of spectrograph and detector used. In order to assessthe accuracy and advantages of several recently built DOAS instruments which use a wide variety of solutions to the above-mentioned difficulties, an international intercomparison campaign was organised. This campaign was arranged within the framework of TOPAS (Tropospheric Optical Absorption Spectroscopy), a subproject of the EUROTRAC! programme. It took place during two weeks at an urban site in Brussels and brought together eight different instruments. The list of the participating groups and their instruments is given in Table I. This paper will describe the campaign, the instruments and the results of a set of simultaneous measurements in the lower troposphere by all eight instruments. The UV and visible region allows many species to be monitored, but the present report will be concerned exclusively with SO*, NO2 and 03, measurements which were accessible to most of the participating groups. The wavelength regions used by each group to measure these species are given in Table II.
Swedish Environmental Research Institute Universit6 Libre de Bruxelles University of East Anglia Service d'A6ronomie du CNRS and Atmos Equipment Institut d' A6ronomie Spatiale de Belgique Universitfit Heidelberg
SERI1 SERI2 ULB
-
1.4 13.6
13.2 8.0 7.0
ULB
UEA CNRS
IAS UH1 UH2
PDA PMT + slotted disk PDA
Solar blind UV vacuum diode PDA PDA
PMT + slotted disc PDA
Detector c
ORIEL (MULTISPEC) Home-built CHEMSPEC
SPEX(1870B) Home-built FTS B R U K E R (IFS 120 HR) SPEX (1870B) JOBIN-YVON (CP200)
Spectrometer
1024
1024
1024 512
-
1024
EGG 1412 Hamamatsu NMOSS3901-512Q EGG/PI Hamamatsu 1462 EGGRL9024RS
EM19758QB Hamamatsu $2304-1024Q -
Detector type
2.7 2.2
100 100 Number of pixels/ channels
3.7
6.9 2.9
6.9 3.0 5.0
f number
125
500 190
500 225 -
Focal length (ram)
16 16 16
14 12
12
12 12
ADC resol. (bit)
concave concave
plane (600)
plane (1200) concave (360)
plane (1200) concave (1200)
Grating (grooves/ ram)
--45 20 -30
-20 20
20
20 5
Detector temp. (°C)
100 6O
100
200 100 2 mm (diam) 50O 25
Entrance slit width (~m)
a FWHM of a low pressure atomic line; u In the plane of the detector; c PMT: Photomultiplier tube, PDA: photodiode array.
1.6 4.0
SERI1 SERI2
UH1 UH2
IAS
Dispersion b (nrn/mm)
Organisation
Acronym
UEA CNRS
Instrument characteristics
TABLE I.
1.0 1.2
1.2
0.5 0.6 0.15 (at 300 nm) 0.8 0.4
SpectraP resolution (nm)
~o
o
o
03
:z
,.-]
O
tTZ
©
2: ,.-3
03
03
O
,--1
.-]
~,
54
C. CAMY-PEYRET
ET AL,
II. Campaign recording conditions
TABLE
Instrument
SERIl SERIZ ULB UEA CNRS IAS UHl UH2
Wavelength (nm) fitting range for NOz 03
SO2
422-435 331-371 260-370 344-379 332-369 260-600 424-435 470-540
293-307 293-307 260-370 278-307 260-600 295-350 -
281-289 278-290 260-370 309-345 258-271 -
Recording time (min)
Number of molecules included in one recording
20 6 45 8 1 5 20 2
Additional operation time’ (min)
Path used
Eb D D D Cb D Fb G
a Required between 2 recordings for background spectrum and/or dark current measurements; b For most measurements (see text).
2. Principle of Measurement 2.1. DOAS METHOD The measurements performed during the campaign were based on differential optical absorption spectroscopy (DOAS). This technique (Platt, 1994; Plane and Smith, 1994), used since the 1930’s for stratospheric total ozone (Dobson, 1931) and since the 1970’s for total NO2 measurements (Brewer et al., 1973; Noxon, 1975; Blamont et al., 1975), was pioneered by Platt and Perner (1983) for tropospheric composition monitoring. It consists of looking at the narrow spectral features of the absorbants that remain after removing the broad band absorptions and attenuations due to Rayleigh molecular and Mie aerosol scattering. The DOAS technique can provide concentrations of several atmospheric species simultaneously. The intensity 1(X) received by the detector can be written as I(x)
=
&j(x)
eXp[(-%iai(A)
-
kR
-
kA)d]
(2)
where I(X) is the measured attenuated intensity of the lamp (W . Sr-‘) la(X) is the incident intensity of the lamp (W . SC’) ni is the concentration of the ith species (cmp3) ai is the cross section of the ith species (cm2) kR the Rayleigh molecular extinction coefficient (cm-‘) kA is the Mie aerosol and particle extinction coefficient (cm-‘) d is the length of the optical path (cm). The principle of the DOAS method involves discriminating between the broad and the narrow spectral attenuation features. The total intensity may be written as
INSTRUMENTS FOR DIFFERENTIAL OPTICAL ABSORPTION SPECTROSCOPY
55
I(X) = IB (X) . SI( X), where 1~ (X) is the background intensity and SI( X) is the differential intensity. Similarly the cross-section is written as G(X) = erg (X) + SG( X), where o-g(X) is the broad band cross-section and &(A) the differential crosssection. By removing the broad band component by an adequate smoothing procedure, Equation (1) becomes: 61(X) = Slo(X) exp[-CniSai(X)d]
.
(3)
There are several mathematical methods used to derive Sfe( X) from experimental spectra. It is important that &(A) be derived with the same technique as the one used to derive 610 (X) . Once So(X) and S&(X) are known, a least-squares procedure can be applied to the differenital optical thickness, in order to determine the concentrations of the absorbing constituents. Each group participating in the campaign used its own least-squares procedure. 2.2.
ABSORPTION
CROSS-SECTIONS
The choice of the absorption cross-section values has a direct impact on the accuracy of the measurements. The DOAS technique requires cross-sections at the resolution of the instrument used, and preferably differential cross-sections measured by that particular instrument in order to eliminate systematic errors where possible. However, it was decided that all participants in the campaign should use the ‘same’ absorption cross sections. Because the spectral resolution of the different instruments varied between 0.15 and 1.2 nm, it was necessary to use high resolution absorption cross-sections and to convulute them with the actual spectral line shape of each instrument in order to obtain an appropriate ‘common set’ of cross-sections. Several absorption cross section measurements of SO2 at various resolutions can be found in the literature. However, except for the measurements of Thomsen (1990) and of Hear-n and Joens (1991), the wavelength intervals were always very restricted and no data were available below 290 nm. This justified a recent reinvestigation at high resolution by Carleer et al. (1993) using a FT spectrometer. The results of this latter investigation were used in this campaign. Several studies have also been carried out on NO2 absorption cross sections at various temperatures and resolutions and in several cases the measurements were corrected for the presence of the dimer N2O4. Divergences of the order of 8% and more are found between these measurements, which warranted that they be reinvestigated (Carleer et al., 1993). However, the values of Schneider et al. (1987) were used by all participants in this campaign. For 03 the recent measurements published by Daumont et al. (1992) were used. These are in close agreement at 254 nm (< 0.5%) with the standard ozone absorption proposed by Mauersberger et al. (1987).
56
C.CAMY-PEYRETETAL.
3. Description
of Participating
Instruments
Eight different instruments were employed during the TOPAS intercomparison campaign. Table I presents the main characteristics of each instrument. Recording time is defined as the time during which one spectrum is recorded. Several scans can be coadded during this recording time (e.g. in the case of the FT spectrometer 2000 scans are coadded during the 45-min recording time). The wavelength region covered by the spectrum may contain the signature of one or several molecules from which concentrations will be derived. Limiting the molecules under consideration to the species N02, 03 and SO2, Table II lists for each group a number of relevant parameters describing the recording conditions used during this campaign. Columns 2 to 4 give the wavelength region (nm) used to fit the spectrum of each of the three considered species. Column 5 gives the time (min) required to record one spectrum and column 6 gives the number of molecular species whose signature is contained in one spectrum. Column 7 gives the additional time required between two recordings; this time includes auxiliary operations such as dark current and wavelength calibration measurements. In order to synchronize the measurements of the four groups using the same path D (see Section 4.1), the ULB group performed one measurement every 60 min, instead of every 45 min. Finally, column 8 indicates the path used by each group (see Section 4.1). A more complete description of each set-up is given in the following sections. 3.1. THESWEDISHENVIR~NMENTALRESEAR~HINSTITUTE(SERI) INSTRUMENTS
Two different instruments 3.1.1.
The SERIl
were used during the campaign.
Slotted Disc Spectrometer
SERIl is a fully automated system optimized for background trace gas monitoring. This system has a 30-cm coaxial transmitting/receiving telescope (Galle et al., 1990) coupled to an optical fiber, specially designed to operate with retroreflectors on different paths. A computer controlled mirror alignment system makes automatic switching between different optical paths possible, as well as keeping the system aligned on long paths. A lamp shutter makes it possible to block the outgoing beam and record a background spectrum for subtractions of scattered sunlight. The dispersive element is a 0.5 m Czerny-Turner spectrograph with a computercontrolled grating. It has an effective resolution of 0.5 nm. Spectra are recorded at a repetition rate of 100 Hz using a slotted disc arrangement. By scanning the grating the instrument is sequentially tuned to different wavelength regions depending on which species is to be studied. This optimizes the detection limits of the components to the detriment of simultaneousity. Simultaneous measurements are, however, not necessary in background monitoring as the temporal variability of the most species is normally small.
INSTRUMENTS
FOR DIFFERENTIAL
OPTICAL
ABSORPTION
SPECTROSCOPY
57
During part of this campaign the dual beam DOAS method was used, where spectra are recorded from two different paths successively. In this method spectra are recorded close in time from two paths alternatively. By rationing the spectra from the two different paths, an absorption spectrum representing the path between the two reflectors is obtained. This method suppresses instrument factors such as lamp spectral features, reflectivity losses and spectrometer anomalies, thus improving the detection limit. With the relatively high concentrations encountered during the campaign, the detection limit was determined mainly by mismatch between the measured spectrum and the reference spectrum, and no significant improvement in detection limit was obtained using the dual beam method. As the temporal variations of concentrations are large in urban air, higher temporal coverage was thought to be more important and thus the dual beam method was only used occasionally during the campaign. All the data reported here using SERIl were performed with path E. The system is controlled by an IBM AT computer. Data acquisition and slotted disc control is performed by means of a specially designed MCA card. With this card, a computer set number of spectra are digitized with a 12-bits ADC and summed in a 1024-channel memory, thus leaving the computer free for other tasks (evaluation, printing, etc.) during data acquisition. During the campaign, three wavelength regions were used, optimized for measurements of N02, 03 and SO2, respectively. For each wavelength region, 50 000 spectra (and 50 000 background spectra) were coadded during a 20-min period, so that one measurement of each of the three species was made per hour. The software code is written in Pascal and automates measurements, evaluation, background correction, amplification control, telescope alignment, path switching and grating setting. The evaluation procedure comprises background subtraction, Gaussian smoothing, polynomial short pass filtering and multiple linear leastsquares fitting. 3. I .2. The SERI2 CCD Spectrograph SERI2 is a system optimized for urban-air monitoring. In its normal operation mode this system is fibercoupled to a 20 cm diameter transmitting/receiving telescope similar in design to the SERII telescope. During this campaign however, and in order to facilitate comparisons, it was coupled to the UL,B optical path (path D), shared by three other groups (see below). The dispersive element is a home-built spectrograph with a curved holographic flat field grating. The detector is a cooled CCD array of 1024 elements covering the spectral range 270-370 nm with an effective resolution of 0.6 nm. In this campaign, the system used a fixed wavelength region where a number of species of interest (03, SO2, N02, H2CO and HN02) can be measured. This chosen wavelength region is not optimal with regard to the detection limit for all of the measured constituents, but makes it possible to perform simultaneous measurements of all
58
C. CAMY-PEYRET
ET AL.
five species. During the campaign this system was integrating spectra for 6 min (180 spectra) and thus measurements were made of all components every 6 min. The data acquisition and electronic control of the CCD is performed with the MCA-card mentioned earlier. To minimize dark-current variations the detector is kept at a constant temperature of 5 “C by means of a Peltier cooler. The software code and evaluation procedures are similar to SERIl and SER12 spectra are always ratioed against a lamp reference spectrum in order to minimize the influence of pixel to pixel sensitivity variations. 3.2.
THE
UNIVERSITI?
LIBRE
DE BRUXELLES
(ULB)
INSTRUMENT
The ULB system was developed in 1990 (Cat-leer et al., 1991) with the aim of testing the applicating of the technique of Fourier transform spectroscopy to the detection of minor constituents in the troposphere using visible and UV light. A long path tropospheric absorption system (path D, see below) was built on the ULB campus. This system consists of a light source situated at the focal point on a 30 cm diameter f/8 Cassegrain type telescope which collimates the light onto a very long focal length mirror, also 30 cm in diameter, placed on the roof of a building situated 387 m away from the laboratory. The returning light is received by a second 30 cm diameter Cassegrain telescope modified in such a way that its output beam is parallel and 5 cm in diameter. By means of plane mirrors, this beams is directed onto a 25 cm focal quartz lens which focuses the light on the 2 mm diameter f/5 entrance aperture of a BRUKER 120HR spectrometer. The folding mirror situated 387 m away from the laboratory is equipped with a driving system consisting of two stepper motors and their associated electronics, enabling the aligment of the mirror from the laboratory through a telephone line. The Fourier transform spectrometer is controlled by an Aspect computer which also digitizes the interferogram. The use of a high-speed vector processor Fourier transforms the interferogram into a spectrum in a matter of a few seconds, depending on the required resolution. The spectra taken during this campaign were recorded in the region 260-370 nm using an ‘ozone-free’ 250-W Xenon lamp as source and a solar blind UV vacuum diode as detector. This set up enables the simultaneous detection of SO2, NO2 and 03. The resolution was 16 cm-’ (- 0.15 nm) and the dispersion of the spectra is then 7.7 cm-l per spectral point. Interferograms are digitized and recorded during the forward and the backward travel of the moving mirror, in a double-sided mode, each spectrum is the result of the Fourier transform of the coaddition of 2000 interferograms. The time required to record a spectrum is about 45 min, which is mainly due to the frequency response of the detector-preamplifier combination used during the campaign (W vacuumdiode).
INSTRUMENTSFORDIFFERENTIALOPTICALABSORPTIONSPECTROSCOPY
59
3.3. THEUNIVERSITYOFEASTANGLIA(UEA)INSTRUMENT The spectrometer is a 0.5 m Czerny-Turner Spex model 1870B with a grating of 1200 grooves/mm, blazed at 400 nm and an f number of 6.9. The entrance slit of the spectrometer was set to a width of 500 pm, and a height of 0.2 mm for the duration of the campaign. A 1024 elements, EG & G model 1412 diode array detector is used to record the spectrum. The spectrometer disperses the received light onto the diode array detector which is cooled to -20 “C with a four-stage Peltier cooler. Typical exposure times vary from 50 to 500 ms, depending on the level of received light. In order to avoid nonlinear effects, which are present close to saturation, the exposure time is set so that the diode count reaches about 50% of its saturation value (16384 counts/diode) before being read out. The combination of spectrometer and diode array provides a spectral width of 35 nm and a maximum spectral resolution of 0.034 m&diode. Under the conditions used during the campaign, the FWHM was 24 pixels (0.816 nm) and spectra were taken at intervals varying from 1.5 to 8 min. Two spectra were recorded in the 365 nm region (for N02) followed by two in the 335 nm region (for N02, 03 and CHxO), and this sequence was then repeated. The UEA DOAS system is fully automated by means of a master computer which controls the scanning of the spectrometer to the wavelengths of interest, the recording of spectra by the diode array, the insertion of optical filters into the transmitter, and the operation of an optical by-pass in order to take spectra of the arc lamp directly. The atmospheric spectra are divided by the lamp spectrum in order to remove the spectral structure of the Xenon lamp, and to correct for etaloning on the front surface of the diode array detector and the variation in spectral response between diodes. The resulting processed spectrum is then smoothed using a fast Fourier transform (FFT) with a high-pass frequency filter, and the broad spectral trend is obtained from a low-pass filter FFT analysis. The logarithm of the latter spectrum divided by the former gives the differential optical density spectrum. The concentrations of the atmospheric molecules which absorb in the spectral region of interest are then obtained from a least-squares fit of their reference optical density spectra by means of a routine employing singular value decomposition (Plane and Nien, 1992). Although the UEA setup possesses its own Xenon lamp transmitter and retroreflector, for the purpose of the intercomparison in Brussels the telescopes and light source of ULB were used (path D). The return beam from the receiving telescope was split using plane mirrors, and steered into the slits of the spectrometer. This arrangement generally worked well. However, during some of the intercomparison periods the degree of alignment of the atmospheric beam and the direct by-pass beam from the Xenon lamp onto the spectrometer slits was sometimes not good enough to permit high quality atmospheric optical density spectra to be obtained.
60
CCAMY-PEYRETETAL.
3.4. THEINSTRUMENTFROMTHESERVICED'A~RONOMIEDUCENTRENATIONAL DELA RECHERCHESCIENTIFIQUE(CNRS)
The instrument called SANOA (Systeme d’ ANalyse par Observation Active) consists of a measuring unit, its associated computer and a projector. The measuring unit is a diode array spectrometer based on a Jobin Yvon flat field spectrometer (CP200) with a focal length of 19 cm, having an f number of 2.9 and using a concave holographic grating of 360 grooves/mm. The detector is a 512 diodes Hamamatsu NMOS linear array (25 pm x 2 mm diodes) with its amplifier and a 12 bits A/D converter. The detector is uncooled. With an entrance slit of 25 pm x 2 mm, a spectral resolution of 0.4 nm is achieved, approximately constant within the whole spectral range of 200-375 nm. The optical entrance of the spectrometer is a stepper motor driven mirror oriented toward the projector, followed by a folding mirror at 45” and 30-mm diameter lens which focuses the light onto the entrance slit. A rotating disk, placed immediately in front of the slit, cuts out the light for dark-current measurements. The whole instrument is installed into a compact, clean, dehydrated, thermally insulated and tight fiber-glass container. The measurement unit is connected to a Hewlett Packard 362 computer which controls the measurement sequences, the duration of exposure and the orientation of the mirror. This also ensures the acquisition of the spectra, their analysis in real time and the visualisation of the results. The projector is a loo-mm diameter, 500 mm focal length telescope equipped with a 75 W Xenon lamp, protected by a frontal quartz window. During this campaign, the projector was placed at a distance of 230 m (path C). For each single measurement, the duration of exposure is first adjusted between 0.1 and 60 set, in order to get a signal amplitude of about two-thirds of saturation (4035 counts/diode). Afterwards, successive spectra of the same exposure are coadded during 60 sec. The dark current is then measured during 60 set with the same exposure and subtracted from the signal. The measurement sequence consists of two successive recordings: the first adjusted for the flux received by the detector at 375 nm and the second at 280 nm in order to enhance the UV signal. The cycle is repeated every 10 min, except during the comparison periods where a 5 min cycle was adopted. Spectra are analysed by a differential procedure after a wavelength alignment onto a reference spectrum measured in the laboratory and to which all absorption cross-sections are linked. The wavelength alignment is obtained by adjusting the position of the largest Xenon lines, as well as those of the largest atmospheric absorption features, in order to determine the shift and the stretch of the actual spectra compared to the reference one. The wavelength of the actual spectra are then corrected with a precision better than 0.05 pixel, or 0.015 nm. The amount of absorbant is calculated by a linear least-squares fit minimizing the residuals between the differential signal and the differential absorption cross-sections of the absorbents. A special feature of the SANOA procedure is the adoption of a
INSTRUMENTSFORDTFFERENTIALOPTICALABSORPTIONSPECTROSCOPY
61
sequential fit (one molecule at a time within its specific spectral range) during which the absorbent signatures are progressively measured and removed from the signal. The procedure is iterated three times which was proven to be enough to ensure a reasonable result. The differential residuals as well as the original signal are displayed at the end of the procedure. The peak to peak amplitude of the residual is 5 x 10e4. Besides N02, 03 and SO2, which are the primary targets of the comparison, the instrument measures also CH20, HNO2, CS2, (02)2, toluene, xylene and naphthalene; the signatures of the latter are subtracted from the spectra, The average standard deviation (1 cr) precision for N02, SO2 and 03 for an optical path length of 230 m used during the campaign are thus respectively of the order of 1,0.3 and 1 ppb or 3, 1 and 3 x 10” molec . cmV3. 3.5. THEIN~TITUTD'ABR~N~MIESPATIALEDEBEL~IQUE(IASB)IN~TR~MENT The spectrometer is a l/8 m crossed Czemy-Turner from ORIEL (type Multispec) with an f number of 3.7. This instrument has a flat focal field adequate for use with photodiode arrays (PDA) up to 25 mm long. The combination of an entrance slit of 100 pm with a grating with 600 grooves/mm and a 1024 pixel PDA provides a measured full width at half maximum (FWHM) of 1.2 nm. The spectral window extends from 270 to 600 nm. In that case, the sampling defined by the size of the pixel (25 pm) and the spatial resolution corresponding to the FWHM is 4 pixels. The detector is a EG & G (Reticon) PDA of 1024 pixels which can be cooled to ~$5 “C by means of a two-stage Peltier cooler in order to reduce the dark current. Such a low temperature can only be reached by running methanol at -5 “C in a cooling loop connected to an external cryostat. The read-out electronics were manufactured by Princeton Instrument Inc. and provide a dynamic range of 16 bits. The detector assembly is integrated in a vacuum tight container evacuated by a small pump to avoid water vapor freezing on the detector at low temperature. Filters can be placed in front of the entrance slit in order to exclude second-order overlap. The filter wheel has a closed position used to measure the dark count corresponding to each measurement integration time. The system includes a compatible IBM 386 AT computer, the ST-1000 detector controller and the spectrograph with a self-scanning photodiode array of 1024 pixels. The detector controller provides power, thermostating and timing signals to the detector head, coordinates data gathering with the experiment, sets exposure time, digitizes and averages data and transmits it to the computer. For a rapid DMA (Direct Memory Access) transfer, the ST-1000 has free access to theAT’s 16 bits bus providing a large dynamic range from 0 to 65536 counts. The data bus is used bidirectionally by the computer to send control information bytes to the ST-1000 and to receive raw data back. New software has been written in order to make instrument automation possible. Successive spectra can be taken with the filter wheel at different positions and with exposure times and numbers of scans automatically computed according to the current signal.
62
C. CAMY-PEYRET
For the campaign, were used (path D). 3.6. 3.6.1.
THE
INSTRUMENTS
ET AL.
the telescopes, the light sources and the optical path of ULB
USED
Coaxial Arrangement
BY
THE
UNIVERSITiiT
HEIDELBERG
(UH)
(UHl)
In contrast to the original set-up developed by Platt and Perner (1983) where the light source and receiving telescope with spectrometer were situated at different end of the light path, we used two coaxially arranged Cassegrain-type telescopes for both sending and receiving the measured light. Retroreflector arrays were mounted at the other end of the light path. A Xenon arc lamp is mounted at the focus of a parabolic mirror. The outer ring-shaped part of the parabolic mirror is used to send the light beam to the retroreflector prisms and the inner part serves for receiving the reflected beam (Axelson et al., 1990). In Brussels, one retroreflector array was set up on the roof of the ITT-building (path F) and one small array for the short distance (path B) which was used to determine the instrumental structure. The received light is focused into a quartz fiber bundle. In order to optimize the alignment of the beam, the whole telescope arrangement is mounted in a frame which can be rotated around the vertical and horizontal directions by two stepper motors. The spectrometer, designed in Heidelberg, used a holographic flat field grating (1200 grooves/mm) with 100 mm focal length. A slotted disc scanned the spectrum (N s 60 nm) and a photomultiplier was used as detector. The spectral resolution (defined as the observed full width at half maximum of the 546 nm Mercury line) was about 1 nm. Spectra in different wavelength regions were taken by changing the position of the slotted disc in the focal plane with the use of a stepping motor, The light entered the spectrometer via a quartz fiber bundle. To adapt the aperture of the fibers (f/7.5) to the spectrometer (f/2.7) the exit of the fiber bundle is imaged on the entrance slit of the spectrometer by a pair of quartz lenses. To avoid thermal drift, the temperature of the spectrometer was kept at 39 f 0.3 “C. The instrument is controlled by software running on a personal computer. A program developed in Heidelberg, is used for data acquisition, data processing and aligning the light beam. 3.6.2.
The Open Path Multi-Rejlection
Cell Arrangement
(UH2)
The White system, which was used during the campaign, is a modified version of the improved system invented in 1976 (White, 1976; Ritz et al., 1992) which gives maximum stability against negative effects of vibrations or atmospheric fluctuations on the received beam. The base path length of the cell is 15 m, the total path length can be varied from 120 m (8 traversals) up to 2160 m (144 traversals). The dimensions of the optical elements were calculated in order to have the smallest possible front mirror with however, a sufficient stability and high
INSTRUMENTS
FOR DIFFERENTIAL
OPTICAL
ABSORPTION
SPECTROSCOPY
63
level of sensitivity in the cell and in the spectrometer. The front mirror is 250 mm in diameter (with a cut at about 70 mm from the center) and the two back mirrors are 150 mm in diameter. Prisms, which work with total reflection, are used instead of two plane mirrors. A third prism is added to double the stability. The focal spots on the front mirror and the prisms are 10 mm in diameter. For detecting the spectra, a commercial spectrometer (American Holographic, f/2.2) was used with a flat field holographic grating with 710 grooves/mm and a focal length of 100 mm. This grating allows a spectral range between 400 and 700 nm. The entrance slit has a width of 60 pm. To avoid spectral drift, the spectrometer was thermally stabilised to a temperature of 30 “C. The resolution (full width at half maximum of the 546 nm Mercury line) was 1.2 nm with a dispersion of 0.325 nm per channel. A home made detector (Ritz et al., 1992) was used with an EG & G 1024element random access diode array (RL1024RS). The pixel size is 25 pm in width and 2.5 mm in height. The temperature was fixed at -30 “C. The detector was filled with argon at atmospheric pressure to reduce etalon shift by deposition of vapours. In Brussels, a silver coating was chosen for the mirrors because of its reflectivity of about 98% between 470 and 700 nm without spectral structure. This allows measurement of NO2 and Nos. A Tungsten lamp was used as the light source in order to have no structures of emission lines within the spectra. The optical connections between the light source and the cell as well as between the spectrometer were performed by quartz fibre bundles. During the campaign, the reflectivity of the mirrors decreased to 94% because of air pollution. Nevertheless, measurement of NO2 was possible by using a shorter light path and increasing integration time: two path lengths (1200 and 240 m) were used to take spectra of a long path, a short path, as well as background spectra, in sequence. The optical path lengths were alternated by a stepper motor and another stepper motor was used to place a baffle between the lamp and the fiber bundle in order to take background spectra. The integration time per spectrum was varied from 20 to 150 set, the total integration time was 2-5 min. The diode array and the stepper motors were controlled by software running on a personal computer.
4. Description of the Campaign 4.1.
SITE
AND
OPTICAL
CONFIGURATION
The intercomparison campaign was held in Brussels on the campus of the Universite Libre de Bruxelles, which is situated south of the city, close to an extensive wooded area (Bois de la Cambre) free of dwellings and industrial activities. The first seven instruments described in Table I were installed in a laboratory situated on the 5th floor of a lo-floor building (building D of the Campus), at an altitude of 27 m above ground level and - 120 m from sea level. Through open windows all instruments
64
C. CAMY-PEYRET ET AL
I/ I.T.T. Tower
E.F
T
;
Fig. 1. Layout of the instruments and reflectors in the city of Brussels.
pointed in a NW direction towards a set of retroreflectors or a projector placed at various distances (see Figure 1). The two closest retroreflectors (A, B) belonging to SERIl and UHl were placed on the roof of the 18 m high ULB student house at a distance of 232 and 237 m from the laboratory respectively. The optical paths passed over the organic chemistry
INSTRUMENTS FOR DIFFERENTIAL OPTICAL ABSORPTION SPECTROSCOPY
65
building and over a private campus road open to car traffic. The third reflector (D) was the fixed, slightly concave, mirror of the ULB permanent installation; it is situated on the roof of the 13-floor Sociology building at a height of 50 m above ground level and 387 m away from the laboratory. Its direction is almost identical to that of the A and B reflectors. No additional source of traffic lies between this third reflector and the laboratory. The two furthest situated retroflectors (E and F) were installed by the SERIl and UHl groups on the roof of a 25-floor building (ITT Building; height 93 m) situated outside the ULB campus at a distance of 1007 m from the laboratory. To reach these two latter reflectors the light beams cross an additional public road (av. F. D. Roosevelt). This is a relatively important route from the south of Brussels into the center; morning and evening traffic is quite heavy. The CNRS instrument required that its light source (C) be placed at a distance of 200-400 m away from the spectrograph. It was placed on the student house close to the A and B reflectors at a distance of 230 m. During the campaign, the ULB, IAS, SER12 and UEA instruments used exclusively reflector D and the same Xe lamp. The CNRS group made most of its measurements with the optical setting C, but for comparison purposes, switched on several occasions to use the ULB Xe lamp and path D set-up. SERIl used mainly path E and occassionally path A, while UHl used mainly path F ad occasionally path B. Both these groups applied the dual beam DOAS method on a few occasions (see Section 3.1.1.), but these results will not be considered in this comparison. Each group added a few meters, corresponding to the distance of their instrument from the windows of the laboratory, to the respective path lengths listed above. In addition to these set-ups, the 15 m open White absorption cell and spectrometer (UH2) was installed on the roof (G) of the laboratory building (altitude 45 m). The average altitudes of paths A, B, C, D, E, F and G with respect to ground level were respectively 23,23,23, 39,59 and 45 m. 4.2.
ORGANIZATION
The campaign took place between 7 and 18 September 1992. were devoted to the installation of the various instruments. Most their measurements on the morning of 10 September. Prior to the campaign, two preparatory meetings were held. a number of general principles for the conduct of the campaign
The first two days instruments began At these meetings, were adopted:
(i) Apart from predetermined coordinated periods of measurements (see Section 4.3) during which all instruments measured simultaneously, each group was free to make measurements within the constraints imposed on those instruments sharing a common light source. Before and after each coordinated period, calibration runs of the light sources were performed. (ii) After each coordinated period, measurements were processed within 24 hrs and given to the Referee (C. Camy-Peyret) without knowing the results of
66
C. CAMY-PEYRET ET AL.
the other groups. These were then compared at briefings which followed each period. (iii) As already mentioned, several of the instruments can measure many atmospheric species. However, in order to maximise the efficiency of intercomparing the measurements, three target molecules (NO2,03 and SO2) were chosen in spite of the fact that some instruments could not cover all three species. (iv) The spectrum of a low pressure Hg lamp was recorded by each instrument, configured in an identical manner to that used to acquire atmospheric spectra (i.e. same slit size). This provided the ‘working’ resolution of each instrument. 4.3.
OVERVIEW
OF ATMOSPHERIC
CONDITIONS
The variations in concentration of the N02, 03 and SO:! species during the full ten days of measurements are presented in Figure 2. This figure gives averaged values of all measurements made during the campaign. As expected in an urban site, these variations were quite large and allowed the instruments to be tested in both high and low concentration conditions. The average concentrations for NO2 and for 03 were - 5 x lOI’ molec . cmp3, with the classical anticorrelation in their diurnal variability. For SOT, the average concentration was N 1 x 10” molec . cmp3 with a few occasional pulses reaching 4-5 x 1O’l molec . cmp3. For an urban troposphere, the level of pollution was reasonably low during the campaign period which was dominated by south western winds. These winds did not pass over the more industrial northern part of Brussels. The four coordinated periods are shown on Figure 2 and designated as I, II, III and IV Periods I and II were of 8 hrs duration, from 10.00 a.m. to 6.00 p.m. and took place on 10 and 11 September. The most important period was period III which lasted 24 hrs, starting at 4.00 pm on 14 September. Period IV lasted two days beginning at midnight on 16 September. Local meteorological data from the ULB greenhouse sensors were provided continuously during the campaign and general meteorological data provided by the Institut Royal MCtCorologique de Belgique were also available. Sunrise was at -07.00 hrs and sunset occured at -20.00 hrs. Barometric pressure was measured continously in the laboratory. On the whole, the weather during the two weeks of the campaign was fine; temperatures reached - 25 “C in the early afternoon and dropped to - 10 “C during the night. A few rain showers were experienced during the measuring periods, mainly on 11 September around 3.00 p.m. and on 14 September 8.30 p.m. Two instruments (CNRS and SERIl) were fully automated and measured continously, night and day, during the whole period of the campaign. The other instruments, which required manual operation of the shared light path, were confined to measure during the coordinated periods. Some instruments did not measure either 03 or SO2 (see Table II) and unfortunately, technical reasons prevented certain instruments from taking part in some of the coordinated periods.
INSTRUMENTS
FOR DIFFERENTIAL
-
1 o/9
OPTICAL
ABSORPTION
6’7
SPECTROSCOPY
-
1119
1219
13/9
1419
1519
1619
1719
I ai
Time (Day/Month) Fig. 2. Average temporal variations of the concentrations of NOZ, 03 and SO2 during the campaign from the 10th to the 18th of September 1992. The four coordinated periods of measurements are indicated.
68
C. CAMY-PEYRET
ET AL.
5. Results and Discussion 5.1.
ACCURACY
OF DOAS
MEASUREMENTS
Due to the fact that this campaign was held in an urban site, in which rapid fluctuations of the pollutants can occur, and also because the various instruments used different optical paths and different observation times (see Table II), a very careful analysis of the data was needed. The results presented here in detail concern mainly those obtained during the 24 hrs period III and the 48 hrs period IV The results obtained during period III, shown in Figure 3, will be examined first: eight instruments performed NO2 measurements (top panel), five obtained 0s measurements (middle panel) and six measured SO2 concentrations (bottom panel). Due to technical problems, the UHl and UH2 instruments made only a few or no measurements during this period but were operational during period IV. Using exclusively the results of period III for the fast instruments (all but ULB and SERIl), relative standard deviations were calculated for time intervals of 6 min. Only instruments which performed at least one measurement every 60 min throughout most of this period have been retained: five for N02, four for 0s and four for SO2. It should be noted that only short path (C and D) results are presented in Figure 4 which also shows the average NO2 concentration variation throughout this period. The relative standard deviations are smaller for high concentrations (&5%) and larger for small concentrations (&lo%); they are roughly proportional to the inverse of the concentration. A similar situation is observed for 03 and SO2, but in the latter case the deviations are smaller by a factor of 2. In order to reduce any differences due to rapid fluctuations and also to allow comparisons with the slower ULB Fourier instrument (also using path D), the measurements made during period III by the fast instruments were integrated over an observation time of 4.5 min. The large number of results of SERIl and the few results of UHl, which both used a long path (E or F), were excluded from this treatment for reasons which will become clear later. The integrated values are presented in Figure 5 where the difference of each integrated value to the average value is shown as a function of time. The comparison shows that rapid fluctuations in concentrations are not responsible for the discrepancies between instruments using almost the same paths (C and D), because the overall accuracy for each molecule has not improved with respect to the comparison of the fast (non-integrated) measurements presented above. The absolute accuracy does not vary in spite of a fairly large concentration variation of the three species investigated during period III (by factors of N 7, - 8 and - 6 for NOz, 03, and SO2, respectively). It may, therefore, be stated that the dispersion of the results is mainly due to the instruments themselves including the use of individual algorithms for the retrieval of the data. The dispersion (&la) of the DOAS measurements on N02, 03 and SO2 performed by this large variety of UV-visible instruments using a path length of a few hundred meters is:
INSTRUMENTS
FOR DIFFERENTIAL
16
20
OPTICAL
ABSORPTION
24
4 Local
Fig. 3. N02,03 as obtained by:
l
Time
69
SPECTROSCOPY
8
12
16
(hour)
and SO2 concentration measurements during the 24-h coordinated period III ULB, + CNRS, n SERIl, 0 SERI2,O IASB, V UEA, A UHl, UH2.
70
C. CAMY-PEYRET ET AL,
Relative Standard Deviation (%) o 0
0
0
0
O
0
I
~ I''"-~__~
I
O
'If
'
I
' ~-"~
0
0
-r,
0
I
I
I
0 Z •,e-
0
bO
r.~ ©
l
;
I
I • "
l
I
[
l
I
-
i
I
-
I
'~ ~
~
~
~
r
II
:,
[
I I
I~
! 'I
~
iI
I i
I
I ;
I
o t
i~
,'1
11 iij~J!
i i~l
li
1 i i~ !
'l'iit~
~
~
II I
g
,m o,i
z
I
©
~'.~ o cN
,
¢0
I
I
p,..
"~..~ 0
o
(
uJo'oel ouJ
¢"
0 I. ) uo!leJ~.ueouoo
~,L
~8
O0
o
o
o
i.i.
o~
:
i'+l
<+
i*
I
+
@
ill.
I
Oi
I,'° I *
i °'+°ii i" ~: i °
~
~:
I
°i i~*i
aI
o~ El I
El
n:
+
"
°
i.--,o.: i . ,-,+.+ °
*
~1
9 ~
I
!o.: °!
;'r
e
° i ° i"
a o o:
i,o
'Io'
b. i o l ,
b
b
' I-
0
,
i~
0
,
i
b
"i
°
•
°
I
oi-
~°i
I .i~ I
!'~ 9
i* ..'°i
~ a J.
¶
"
o
•
< •
I
if,
b
0
i
b~
*'~
o
I-1
1
b
9
b' :o-I
I
i,+i I ~,
b '1• i
i" i°
!a
o
I'i :'i*oi
"i
i, I'
b
o
L~
°i ~
I
do :o
i'* I
~:)~
;|
i~ !o
I ~:
i" ~..o
I
+ i" I':* i o
~ i°
.Joiok,
~ <:o ol
i"
,o i
+'~ ~ • +!~+.
~:°
ol~
II b
L~ '
O
Z
i
L~
1 , 1 +
i
b
Fig. 5. Differences between 45-min integrated concentrations of NO2, 03 or SO2 and the average value obtained during period III as a function of time. The dotted-dashed lines represent the standard deviation (1<7) of the data from average value. • ULB, 0 CNRS, [] SERI2, 0 IASB, V UEA.
v
O
11)
--t
tO
I-3
O
,
,i
Concentration Differences ( l d 1 m o l e c . c m a )
© ,-4
©
('3 -1
Z
-1
©
>
O
Z
0
:Z
-1
72
C. CAMY-PEYRET
ET AL.
NO2 : f5 x 10” molec . cmw3 or 2.0 ppb 03
: *6 x lOlo molec . cmP3 or 2.4 ppb
SO2 : fl
x 10” moiec e cmP3 or 0.4 ppb
The lower value for SO2 was expected because the signature (cross-section) of this molecule is stronger than that of the two others. It should be pointed out that the results presented here were greatly improved with respect to initial measurements by using a common set of cross-sections for all instruments and introducing minor hardware improvements to several instruments. The results obtained during period IV include short (CNRS), long (SERIl and UHl) and the folded path (UH2) measurements. Since the UH2 instrument only performed NO2 measurements, the comparison of the results concerning this period will be restricted to this species; they are presented in Figure 6. The interruption in SERIl data during this period was due to an attempt to measure Nos. The longest measuring time for one species (20 min) is that of SERIl and therefore all other measurements made during this period were integrated over this time interval. Again the results are presented (Figure 7) in the form of individual differences with respect to the average value. These results present some interesting features. It should be pointed out once again that these results concern different paths. The UH2 results were obtained using the folded path set-up G situated on the roof of the laboratory building at an altitude of 50 m from ground level. SERIl and UHl measurements were made with the long E and F paths which cross a busy road (av. F. D. Roosevelt) at an average altitude of 59 m and the CNRS measurements with the short path C within the university campus and situated roughly 200 m away from the road and at an altitude of 23 m. The results of Figure 7 clearly show a marked separation between CNRS and UH2 measurements during the time period extending from 4.00 p.m. to 8.00 p.m. on the 16th of September. This interval corresponds precisely to a peak traffic period during which the NO2 concentration increases very rapidly as can be seen in Figure 6. On that day, a slight (1 m/set) wind came from the North-West direction, bringing the NO2 plume rapidly to the location of path C at an average altitude of 23 m but not to location G at an altitude of 4.5 m. On the basis of these observations one can conclude that under these meteorological conditions it took roughly 4 hrs for the plume, produced at ground level, to reach an altitude of 45 m. A similar situation can be seen to begin to appear the next day at the same time, although the UH2 measurements were discontinued for technical reasons round 5 p.m. Another important conclusion can be drawn from the examination of the results or ~&,jd IV: when the NO2 concentration is more or less constant, i.e. outside of the peak traffic hours, all the measurements fall within a calculated (peak hours excluded) error limit (&la) of *5 x lOlo molec . cme3 (2.0 ppb). This error limit, which does not include the points where only one measurement was made, is identical to that found for this species during period III which involved other instruments all using the shorter paths. Thus, when the portion of the troposphere
Fig. 6.
0
0
2
4
8
0
•
,m
~b
.o
O
o
I
4
m
o.
.
8
I--
.%o
"~ O11, o
NO 2
o
12
I
o
o
o
°
I
16
°°°~ %-%%
•
co
o
• •
•
•
0
i100 Oo
.
I
24
"...
I
- -
oo
•
4
Local Time (hour)
20
1
0
omo~
.~,'%
A
•
I
8
R
I
12
o
"
16
I
@•
~A
I
20
24
NO2 concentration measurements during the 48-h coordinated period IV as obtained by: z~ UH1, O UH2, • SERI1 and 0 CNRS.
r(D (J E
v tO
E
0
"-'10 ~'E 0 dQ)
12
14
-.-a t.a.)
© r/l Ul ©
£3
z
©
>
>
t" ©
Z ..-1
0
Z
C. CAMY-PEYRET ET AL.
-** -
l l
A
A
0
l 00
l
0
0
0
t L
I
I
I
I
I
I
I
I
I
I
I
I
0
4
8
12
16
20
24
4
8
12
16
20
24
LocalTime(hour) Fig. 7. Differences between 20-min integrated concentrations of NO2 and the average value obtained during period IV as a function of time. The dotted-dashed lines represent the standard deviation (la) of the data from the average value excluding both 16 hrs to 20 hrs periods. (For symbols: see Figure 5.)
can be considered homogeneous, either because the concentration does not vary with time or that the instruments are sampling the same portion, there remains systematic differences between the measurements of the various instruments. This result indicates that these residual differences found between the instruments are due to differences between the instruments themselves. This point will be addressed in the following section. 5.2.
COMPARISON OF DOAS INSTRUMENTS
In order to examine the performance of each instrument using the shorter paths C and D, it was decided to compare the individual results to the average values. In Figures 8-10 are presented the 45 min integrated concentrations of N02, 03 and SO2 obtained by each instrument using paths C and D, during all the coordinated periods (I-IV), plotted as a function of the average value obtained by all instruments. For each graph, a least-squares fit was made. Each instrument is therefore characterized, for each molecule, by a number of parameters: a correlation factor T which evaluates its consistency to follow the average value, a root mean square value s which indicates the scatter from linearity, an origin which indicates, if compared to average, that the instrument is measuring higher or lower values at low concentration and finally a slope which, combined with the previous parameter,
INSTRUMENTS FOR DIFFERENTIAL OPTICAL ABSORPTION SPECTROSCOPY
75
gives some indication of whether it is measuring too high or too low at higher concentrations. These parameters are given on the graphs. From the examination of the graphs it appears that all the instruments performed very well individually against the instrument averages, although with some evidence of significant systematic deviations. For example, the Fourier transform spectrometer (ULB) appeared to measure consistently lower values at low concentration and higher values at high concentration than the grating spectrometers. In addition to these comparisons, it was decided to examine a similar comparison for NO2 and SO2 between UHl and SERIl whose measurements were both made on the longer paths. This latter comparison, which was basedon a 20 min integration time, is presented in Figure 11. This mutual comparison, which also contain data taken throughout the whole campaign, shows agreement of the same order as the previous comparisons. From the various comparisons made above it is clear that the small discrepancies between the instruments are not entirely random. It is difficult at present to attribute these differences to particular aspects of the hardware and software of the instruments, which have such different designs. The large number of variables involved such as the technique for adapting the resolution of the cross-sections to that of the instrument, the number of molecules measured simultaneously, the type of regression used, the need for a dark current spectrum and the retrieval procedure used are all examples of the different approaches to the treatment of the raw data. 5.3.
CONCLUSIONS
This first intercomparison campaign of DOAS instruments operating in the visible and UV region, conducted in the framework of the subproject TOPAS of the EUROTRAC programme, brought to light several fundamental issues which have contributed to improve this technique in its aim to study the composition of the troposphere. Significant progress has been made by all the participants due to the synergy of this type of campaign. After a first comparison of results made without any constraints, it was obvious that a common set of absorption cross-sections had to be used by each group. This has also encouraged improved cross-section measurements (Vandaele et al., 1994). The campaign also served to exchange various instrumental and software improvements between the groups. The measurements made during the campaign by eight different instruments were carefully analysed and yielded an absolute accuracy limit for N02, 03 and SO2 measurements made by the DOAS technique using an absorption path of a few hundred meters in a relatively unpolluted urban troposphere. There remain nevertheless differences between the results obtained from the various instruments. These are most certainly linked to the various retrieval methods used although it was concluded that these retrieval methods are well optimized and well adapted to each instrument.
76
C. CAMY-PEYRET ET AL.
NO* I
I
I
I
I
y=-6.76elO+l.l7x I I I I I
I
I
I
I
I
r= 0.97 s = 3.56eiO y= 2.69elO + 0.86 x I I I I I
10
- 2 0
s = 2.45elO y= 2.82elO + 1.02 x
s = 2.55elO y= 1.89e09 + 1.05 x 0
2
4
6 Average
8
i 10
Concentration
( 10”
molec.cmJ
)
Fig. 8. NO2 (45 min integrated) concentration measurements by 5 instruments during all coordinated periods compared to the average values.
INSTRUMENTS
FOR DIFFERENTIAL
OPTICAL
ABSORPTION
77
SPECTROSCOPY
I
I
I
I
- 15
SERl2 . 12 ,9 /
r = 0.99 s = 3.73elO y=-2.70elO+ 1.11 x I I I
I- = 0.99 s=2.53elO y=6.76elO +0.92 x
/ I
I
,
I
.3 ,O
/
6
,6
3 0
I
0
3
r = 0.99 s = 3.76elO y = -2.59elO + 0.95 x I I I
6
9
12
,3 / 0
,O 3
6
9
12
15
Average Concentration ( 10” molec.cm3 ) Fig. 9. 03 (45 min integrated) concentration measurements by 4 instruments during all coordinated periods compared to the average values.
In order to improve the accuracy of the presently available type of instruments, further work is required. In particular, it would be interesting to compare results obtained by the DOAS technique to those obtained by other techniques. In order to do so, it would be useful to conduct a campaign in a more stable environment, where concentrations vary less than in an urban site. This would reduce the difficulties encountered here related to the various integration times used by the different instruments. Also, cell measurements should be performed by all instruments; although such measurements poorly represent the difficulties associated with real atmospheric measurements, they would probably reveal more clearly differences in results due to the retrieval methods used. Finally, the atmospheric pollution
78
C. CAMY-PEYRET ET AL
I
1
I
SERl2
.4 /
,3
2
1
r = 0.99 s = 7.34e09 I
=-1.35eiO I
r= 1.00 s = 3.73e09
+ 1.13 x I
0 IAS
CNRS
3 0
2
0-/
s = 6.97e09 y=l.S3e10+0,98x
0
i
I
I
1
2
3
1
I- = 0.98 f = 1.31elO y = -554e08 + 0.96 x
r = 0.99
0 0
1
2
3
4
Average Concentration ( 10” molec.cm3 ) Fig. 10. SO2 (45 min integrated) concentration measurements by 4 instruments during all coordinated periods compared to the average values.
conditions encountered in this urban campaign did not allow the detection limits of each experimental set-up to be reached. All these points will be addressed in a forthcoming campaign scheduled to take place in September 1994 at the Weyboume Atmospheric Observatory situated in a pristine area on the coast of East Anglia (U.K.). Acknowledgements Each participating group wishes to acknowledge financial support from its national agency which has funded the TOPAS-EUROTRAC programme. The participation of two representatives of the Hoffmann Messtechnik Cie. (Germany) to part of the
INSTRUMENTS
FOR DIFFERENTIAL
OPTICAL
ABSORPTION
SPECTROSCOPY
79
16
= 530elO+
0
1.02x
8
16
r = 0.95 s=3.53elO y=1.04elO+l.O3x
0
4
UHl Coticentration
8 ( 1CP molec.cm3 )
Fig. 11. Comparison of NO2 and SO? measurements performed by SERIl and UHl throughout the whole campaign.
campaign, and the financing of the closing banquet, were much appreciated. We thank Dr. A. Perez-Aranda from the Laboratoire de Physiologie Vegetale (ULB) for providing regular meteorological data; Imofo SA (Brussels) for permission to install reflectors on the ITT building; and Mr. J. P. Walgraeve and Mr. R. Petrisse of the Laboratoire de Chimie Physique Moleculaire (ULB) and Mr. P. Hulin (CNRS) for much of the technical assistance needed during this campaign.
80
C. CAMY-PEYRET
ET AL.
References Axelson,H., Galle,B., Gustavson,K., Ragnasson, P.,andTrivet, N., 1990,A transmittingreceiving telescopefor DOAS-measurements usingretroreflectortechnique,Tech.Dig. Ser. 4,641-644. Blamont,J., Pommereau, J. P., andSouchon,G., 1975,C.R. Acud. Sci., Ser. B, 247-252. Brewer,A. W., McElroy, C. T., andKerr, J. B., 1973,Nitrogendioxideconcentrations in the atmosphere,Nature 246, 129-133. Carleer,M., Colin, R., Vandaele,A. C., and Simon,P.C., 1991,Detectionof minor tropospheric constituents usingFouriertransformspectroscopy, in: Optical Remote Sensing of the Atmosphere, OpticalSocietyof America,Tech. Dig. Ser. 18,pp. 278-280. Carleer,M., Colin, R., Vandaele,A. C., and Simon,P. C., 1993,Measurementof NO2 and SO2 absorptioncross-sections, in: P. M. Borrell (ed.), Proc. EUROTRAC Symp. ‘92, SPB Acad. Publ.,The Hague,The Netherlands, pp. 419-422. Daumont,D., Barbe, A., Brion, J., and Malicet, J., 1992,OzoneUV Spectroscopy:Absorption cross-sections at roomtemperature, J. Atmos. Chem. 15, 145-155. Dobson,G. M. B., 1931, A photoelectricspectrophotometer for measuring atmospheric ozone,Proc. Phys. Sot. 43,324-328.
Fayt, C., Dufour, P., Hermans,C., Van Roozendael,M., andSimon,P. C., 1993,Instrumentand softwaredevelopmentfor DOAS measurements of atmospheric constituents,in: P.M. Borrell (ed.),Proc. EUROTRACSymp. ‘92, SPBAcad.Publ.,TheHague,TheNetherlands, pp.23l-233. Galle,B., Axelson,H., Gustavsson, K.,Ragnarsson, P.,andRudin,M., 1990,Atransmitting/receiving Telescope for DOASmeasurements usingretroreflectortechnique,in: Proc. OSA Conf on Optical Remote Sensing of the Atmosphere, Incline Village,Nevada,U.S.A. Goutail,F., Pommereau, J. P.,andNunes-Pinharanda, M., 1993,Ambient air monitoringby differential optical absorptionspectroscopy:the SANOA instrument, Proc. 3rd Conf Atmospheric Spectroscopy Applications, Reims,France,pp. 55-58. Hearn,C. H. andJoens,J. A., 1991,The nearUV AbsorptionSpectrumof CSr andSO2at 700K, J. Quant. Spectr. Radiat. Transfer 45, 69-75.
Laville, P.,Pommereau, J. P.,andGoutail,F., 1990,Evaluationof alongpathdiode-arrayspectrometer for 03, Nor, SO2andwatervapormonitoring,in: P.M. Borrell (ed.),Proc. EUROTRAC Symp. ‘92, SPBAcademicPubl.,The Hague,The Netherlands. Mauesberger, K., Hauson,D., Barnes,J., andMorton, J., 1987,Ozonevapourpressure andabsorption crosssectionmeasurements, introductionof anozonestandard, J. Geophys. Res. 92,848&8490. Noxon, J. F., 1975,Nitrogendioxidein the stratosphere andtroposphere measured by ground-based absorptionspectroscopy, Science 189,547-549. Plane,J. M. C. andNien, C. F., 1992,Differential optical absorptionspectrometer for measuring atmospheric tracegases,Rev. Sci. Instrum. 63, 1867-1876. Plane,J. M. C. and Smith, N., 1995,Atmosphericmonitoringby differential optical absorption spectroscopy, in: R. E. HesterandR. J. H. Clark(eds.),Spectroscopy in Environmental Sciences, Wiley, London. Platt, U., 1994,Differential optical absorptionspectroscopy(DOAS) in M. W. Sigrist (ed.),Air Monitoring by Spectroscopic Techniques, Wiley, London. Platt, U. andPerner,D., 1983,Measurements of atmospheric trace gasesby long path differential UV/visible absorptionspectroscopy, Springer Ser. Opt. Sci. 39,95-105. Ritz, D., Hausmann, M., andPlatt, U.: 1992,An improvedopenpath multi reflectioncell for the measurement of NO2andNOs,in: H. I. Schiff, andU. Platt (eds.),Proc. Znt. Symp. Environmental Sensing Optical Methods in Atmospheric Chemistry, Berlin, Vol. 1715,pp. 200-211. Schneider,W., Moortgat, G., ‘Qndall, G., andBurrows,J., 1987,Absorptioncrosssectionsof NO2 in the UV andvisible region(200-700 nm) at 298 K, J. Photochem. Photobiol. 40A, 195-217. Thomsen,O., 1990,GKSS 9O/E36, GKSSForschungszentrum, Hamburg. Vandaele,A. C., Simon,P C., Guilmot,J. M., Carleer,M., andColin,R., 1994,SO2absorptioncross sectionmeasurement in the UV usinga Fouriertransformspectrometer, J. Geophys. Rex 99, 25599-25602. White, J. U., 1976,Very longopticalpathsin air, J. Opt. Sac. Am. 66,411-416.