Optical and physical properties of stratosphericaerosols from balloon measurements in the visibleand near-infrared domains. I. Analysis ofaerosol extinction spectra from the AMON andSALOMON balloonborne spectrometers
Gwenael Berthet, Jean-Baptiste Renard, Colette Brogniez, Claude Robert,Michel Chartier, and Michel Pirre
The effects of aerosols are important in atmosphericmodeling when heterogeneous reactions are takeninto account, especially for tentatively interpretingone kind of discrepancy between observations andmodeling of nitrogen species.1–3 This effect is oftenexpressed in terms of aerosol surface area density.
Various instruments such as balloonborne particlecounters,4–7 impactors,8–10 the balloonborne Radiom-
G. Berthet, J.-B. Renard �[email protected]�, C. Robert,M. Chartier, and M. Pirre are with the Laboratoire de Physique etChimie de l’Environnement�Centre National de la Recherche Sci-entifique, 3A Avenue de la Recherche Scientifique, F-45071, Or-leans cedex 2, France. M. Pirre is also with the Universite d’Orleans, Orleans, France. C. Brogniez is with the Laboratoired’Optique Atmospherique, Universite des Sciences et Technologiesde Lille, F-59655 Villeneuve d’Ascq cedex, France.
Received 18 March 2002; revised manuscript received 17 Sep-tember 2002.
etre Ballon �RADIBAL� instrument,11 sunphotom-eters,12,13 and aerosol lidars,14–16 based on varioustechniques have been used to measure and charac-terize atmospheric aerosols. Moreover, aerosol sizedistributions and aerosol surface areas can be in-ferred from satellite multiwavelength extinctionmeasurements by use of algorithms or methods de-veloped and tested for the various satellite data, i.e.,data from the Polar Ozone and Aerosol Measurement�POAM� II and the Stratospheric Aerosol Gas Exper-iment �SAGE� II,17–19 data from Halogen OccultationExperiment �HALOE�,20,21 data from the spaceborneOccultation RAdiometer �ORA� instrument,22 andsimulated data for SAGE III.23,24
Here we begin a series of papers that treat theoptical and physical properties of stratospheric aero-sols measured with balloonborne instruments andprovide a critical analysis of the data-reduction meth-ods used for the retrieval. This paper is focused onthe analysis of extinction measurements from the twoballoonborne spectrometers, Absorption par les Mi-noritaires Ozone et NO �AMON� and Spectroscopie
x
d’Absorption Lunaire pour l’Observation des Mi-noritaires Ozone et NOx �SALOMON�, and on a data-reduction procedure that uses the hypothesis of log-normal size distribution. In the second paper of ourseries25 we shall present an analysis of the validity ofthe method for retrieval of size distribution. Theanalysis is made through comparisons, as far as pos-sible, of properties such as effective radius and aero-sol surface area density values with data from othersources that use polarimetric and counting methodsof measurement that have been published already.
Remote optical measurements of stratosphericaerosol extinction were performed from 1992 to 2000in the ultraviolet–visible spectral domains, in addi-tion to measurement of nighttime stratosphericchemical species, by the two balloonborne spectrom-eters AMON and SALOMON, which use stars andthe Moon, respectively, as light sources. The twoadvantages of these instruments, namely, self-calibration and continuity of their spectral domains,permit accurate transmission spectra to be obtainedfrom 375 to 700 nm.26,27 The two instruments usesimilar data-reduction algorithms that provide ex-tinction coefficients derived after inversion of theaerosol optical depths. The 1992–2000 flights oc-curred at different times and locations �at mid lati-tude, at high latitude, inside the polar vortex� thatcorresponded to various aerosol conditions �back-ground aerosols, post-Pinatubo eruption aerosols, fro-zen particles�, which enabled us to conduct theanalyses for various situations and for different aero-sol spectral signatures.
It is possible to retrieve the aerosol size distribu-tions in stratospheric layers for which one performsthe measurements by searching for Mie-computedextinction spectra associated with the log-normal sizedistributions that best reproduce the measured con-tinuous extinction spectra. Refractive indices fortypical stratospheric sulfuric acid–water mixtures attypical stratospheric temperatures are assumed.Such a comparison method with Mie scattering the-ory, which was utilized, for example, in the works ofPueschel et al.,10 of Steele et al.,18 and Russell et al.28
to characterize post-Pinatubo aerosols is common forthe characterization of aerosols from optical mea-surements.
Here we present a description of the AMON andSALOMON instruments and of the data-reductionprocedure that leads to aerosol extinction coefficientprofiles obtained continuously in the ultraviolet–visible domains. The consistency of the aerosoloptical-property retrieval method is validated bycomparison with SAGE II and lidar data. Then wedescribe the various extinction spectral shapes ob-tained from the different flights, and the retrievedlog-normal size distributions.
2. Method of Observation from the AMON andSALOMON Instruments
The two balloonborne instruments, AMON andSALOMON, are designed to perform nighttime mea-surements of stratospheric trace-gas species, O3,
NO2, NO3, possibly OBrO, and OClO in the polarvortex, that present spectral features in theultraviolet-visible domain. These instruments haveoperated in the Northern Hemisphere since 1992�AMON� and since 1998 �SALOMON�. Detailed de-scriptions of the instruments, i.e., their structure,their pointing systems, and their performance, can befound in the papers of Robert,29 Naudet et al.,30 andRenard et al.26,27,31 We present below a summary ofthese descriptions.
A. AMON Instrument
The AMON instrument, which uses stars as its lightsource, is composed of a 20-cm Cassegrain telescopewith a 1-m focal length, a grating spectrometer, anda CCD detector of 385 by 578 pixels. The spectrom-eter covers, continuously, the ultraviolet–visiblespectral domains in five spectral bands �50 or 75 nmwide� from 375 to 675 nm, with a theoretical spectralresolution of 0.18 nm�pixel in the ultraviolet bandand of 0.14 nm�pixel in the four visible bands. Spe-cies absorption features are studied separately inthese five bands.
AMON observations are performed by the stellaroccultation method, which consists in analyzing, dur-ing the setting of a star, recorded spectra that areaffected by atmospheric absorption by the speciesthat are present in the lines of sight of the star. Ata balloon float altitude of 34–39 km, depending on theflight conditions, first a reference spectrum is re-corded when the star is a few degrees above the ho-rizon of the gondola in which the instrument is riding.Raw spectra are then recorded while the star is set-ting under the gondola’s horizon. Transmissionspectra are obtained by division of the star’s occulta-tion spectra by the reference spectrum. Stellar ab-sorption lines are removed from the transmissionspectra, so the only remaining features are atmo-spheric absorption bands. The measurement is self-calibrating because no assumption concerning theresponse of the instrument is needed. Then the op-tical depth spectra are deduced from the logarithm ofthe inverse transmission spectra.
It has been shown that the AMON spectrometer isable to observe spectral signatures with a minimumoptical depth of 10�3 in the spectral domain. High-precision raw spectra are obtained when the balloon’sfloat altitude variations are less than �100 m andwhen no star-pointing problems occur. The verticalresolution of this instrument is �0.5 km.
Flights of the AMON have occurred since 1992.We consider here three flights from the mid-latitudeAire sur l’Adour �France� launching base �24 May1992 at 0100 UT; 16 October 1993 at 0030 UT, and 24March 1994 at 2230 UT� and two flights from thehigh-latitude Kiruna–Esrange base �Sweden, Arcticpolar circle� inside the polar vortex �26 February 1997at 2200 UT and 12 February 1999 at 0030 UT�. Wediscuss how each flight had its own characteristics interms of signal-to-noise ratio, depending especiallyon the brightness of the star and its pointing accu-
racy, and on qualities of the reference spectra, de-pending on the float altitude.
B. SALOMON Instrument
The SALOMON instrument is composed of a stabili-zation unit for the gondola �the pivot�, a moontracker, and a SAOZ-type spectrometer32 that oper-ates continuously in the ultraviolet–visible spectraldomains with a theoretical spectral resolution of 0.34nm. The spectrometer was shifted in the 375–700-nm wavelength domain to cover the NO3 absorp-tion band at 662 nm.
As the Moon flux is greater than the star flux, theSALOMON can detect absorption features of 10�4,which is ten times better than can be accomplishedwith the AMON spectrometer. The SALOMON ismuch less sensitive to pointing problems than is theAMON, which allows the SALOMON to obtain opti-mized signal-to-noise ratios in its data more easilyduring strong-wind conditions. However, its coarsevertical resolution �1– 2 km� is the drawback of thisinstrument because of the apparent size of the Moon�0.5°�.
The SALOMON instrument has flown since 1998.Here we will consider the flight from the high-latitude Kiruna–Esrange base on 22 February 2000at 1930 UT �the conditions of measurement of theother flights did not allow us to retrieve the aerosolextinction coefficients�.
3. Retrieval of the Aerosol Extinction Coefficient
The algorithms used for data reduction, which aresummarized below, are similar for the two instru-ments.26,27,31,33 The steps that lead to a retrieval ofthe aerosol extinction coefficients are common to thetwo instruments.
A. Data-Reduction Method
Cosmic rays and dark current are removed from allspectra of the two instruments. The wavelengthscale is established by use of the H and He lines of astar for the AMON and of the strongest lunar linesand the atmospheric O2 line at 687 nm for theSALOMON. To minimize in the AMON spectra theeffect of chromatic scintillation that is due to themotion of air masses along the line of sight,34 a slidingaverage over three or five consecutive spectra is ap-plied to all spectra. The effect of chromatic scintil-lation is negligible in SALOMON spectra owing tothe apparent size of the Moon. Finally, spectra aresmoothed by a Gaussian filter to reduce noise.
In what follows, the spectral domain of the AMONstarts at 400 nm because of the low-level ultravioletflux of the star. Slant column densities are retrievedfrom the AMON optical depth spectra where the spe-cies absorption lines are strongest, i.e., in the 550–625-nm domain for O3 and in the 410–470 nm domainfor NO2. Because of the SALOMON’s one-band spec-tral domain, species are retrieved over a much largerabsorption range than with the AMON: from 430–680 nm and 400–550 nm for O3 and NO2, respectively.
Figure 1 summarizes the steps involved in the re-
trieval from the 22 February 2000 flight of theSALOMON for a Moon elevation of �1.7°. It can beseen from Fig. 1�a� that the strongest contributions tothe spectral domain were due to Rayleigh scatteringand O3, with the maximum O3 absorption occurringwhere the Rayleigh scattering tended to a minimum.First the effects of Rayleigh scattering were removed�Fig. 1�b�� by use of Bucholtz spectral cross sections35
and temperature and pressure profiles. These pro-files were measured during the balloon ascent forthe SALOMON flight and the 1999 AMON flight.For the 1997 AMON flights, temperature and pres-sure were measured by sounding balloons, and forflights before 1995 standard temperature and pres-sure profiles were used. Then O3 absorption wassought and removed �Fig. 1�c��. Figure 2 presentsthe optical depth spectrum of Fig. 1�b� smoothed over6 pixels in the 430–680-nm spectral domain and anO3 least-squares fit with the Bremen University203-K spectral cross sections as measured by Voigt etal.36 The quality of the fit was excellent, with astandard deviation of 3 � 10�3 between the observedand the fitted optical depth spectra. Thus one canconclude that there was no significant O3 residue inthe optical depth spectra. Finally, after removal ofother absorbents �NO2, NO3, and OClO at high lati-tude� by the same technique, the residual optical
Fig. 1. Example of retrieved aerosol slant optical depths�SALOMON observation on 22 February 2000 from Kiruna; Moonelevation, �1.7°�: �a� Raw optical depth spectrum, �b� spectrumafter Rayleigh scattering has been neutralized; �c� spectrum afterthe O3 contribution has been removed, and �d� residual spectrumobtained after the contributions of other species �NO2, NO3, andOClO� were removed and the spectrum was fitted by a third-orderpolynomial. The remaining slope is attributed to the spectrum ofthe aerosol optical depth.
depth spectrum could be attributed solely to aerosols�Fig. 1�d��. This residual was fitted by a third-orderpolynomial, which, by analogy to the other species,can be considered as a cross-section curve. The fitenabled us to eliminate any random variationscaused by noise. The standard deviation betweenthe residuum and the fit provides the error in mea-surement of aerosol optical depth. Riviere et al.37
and Renard et al.27 performed this fit by using asecond-order polynomial. Therefore we have madean improvement, because a third-order polynomial ismore appropriate for representing the nonmonotonicspectral shapes that can be encountered in some mea-surements �see Section 6 below�. This proceduremakes the best use of spectral domain continuity andpermits the detection of inflection points that cannotbe found from sparse wavelength data.
Spatial inversion of the optical depth profiles isachieved by use of the least-squares inversionmethod previously developed for trace-gas spe-cies.27,31 The calculations are made for the opticaldepth profile at each wavelength in 1-nm steps.Then the vertical aerosol extinction coefficient profileis obtained for each wavelength, which allows thewavelength dependence to be retrieved at each alti-tude, as shown previously by Riviere et al.37 For theupper stratospheric layers, typically above 27 km, thelow number of aerosols and the low density of the airmasses produce a weak and noisy aerosol signal.Thus above this altitude, an average of three or fivenoisy spectra is taken.
B. Error Calculations
The quality of the reference spectrum is critical foraerosol retrieval. Empirically, analysis of the tenflights of the AMON and SALOMON instrumentsrevealed that three conditions are necessary for theaerosol retrieval to be performed: the star or theMoon elevation must be greater than 1° above thegondola’s horizon, the float altitude of the gondola
must be as high as possible �typically above 30 km�,and only occultation measurements for which the airmass factor along the line of sight is high enough�unlike ascent measurements� must be used. Theseconditions were fulfilled for six flights.
One part of the error in the determination of aero-sol optical depth spectra corresponds to instrumentalnoise and uncertainty in cross-sections. The othererrors are caused by imprecision in the temperatureand pressure data used. This is especially so whenonly standard profiles of these variables are used,e.g., as was the case for flights before 1995. How-ever, such uncertainty leads to errors of only a fewpercent in determination of the amount of Rayleighscattering, which do not significantly affect the re-trieval of the wavelength dependence of the extinc-tion. The last source of errors is related to thedependence on temperature of the cross sections. Infact, the temperature varies by as much as a few tensof kelvins along the line of sight, whereas the crosssections used were taken at constant temperatures.Although O3 cross sections are not strongly temper-ature dependent, there could be small discrepanciesbetween measured optical depth spectra and fits.31
Therefore all these facts explain the small featuresthat still remain in the residual spectra �Fig. 1�d�� inaddition to the presence of two O2 lines at 628 and687 nm. The same value for the error was used forall extinctions in the wavelength range. Thus, therewere no weighted measurements, thus preventingspecific wavelengths from being favored. It can beseen that the polynomial inflections and increases inextinction �that we present and discuss in what fol-lows� remain even if we have used smaller spectraldomains than those presented here.
In Fig. 1�d� the standard deviation of the differencebetween the residual spectrum and the polynomial is3.5 � 10�3, which results in a signal-to-noise ratio�which we calculated from the ratio of the averagevalue of the polynomial to the standard deviation� of14.0, with a maximum value of 17.5 for this flight.Generally, for equivalent lines of sight, signal-to-noise ratios for the AMON data are lower because ofthe lower brightness of the star and the greater sen-sitivity of this instrument to pointing. However, forthe 1997 AMON flight the signal-to-noise ratio wasmaximized, with a maximum value of 16.0 at 0° to�3° of elevation, as a result of the presence of excel-lent pointing conditions and a high-quality referencespectrum. We must add that the 22 February 2000flight of the SALOMON offered ideal conditions foraerosol retrieval because of the high signal level ofthe raw Moon spectra, the high quality of the refer-ence spectrum, and the accuracy of the measuredtemperature and pressure variables. For both in-struments, we reject aerosol optical depth spectrawhen the signal-to-noise ratios fall below 2.
4. Comparison with Other Extinction Data
Aerosol extinction coefficient profiles at 532 nm werederived from aerosol lidar backscatter coefficients2,16
measured at the Observatoire de Haute Provence
Fig. 2. Example of O3 optical depth spectrum and the least-squares fit by the 203-K O3 cross sections measured at BremenUniversity.
�OHP; 43.9 ° N, 5.7 ° E� on 10, 13, 23, and 24 Marchand 5 May 1994. Figure 3 compares these resultswith an AMON extinction coefficient profile at 530nm for the 24 March 1994 flight �44.0 ° N, 0.0 ° E�.The figure shows very good agreement for the extinc-tion values of AMON and of the lidar sets of data,especially for the lidar profiles of 23 and 24 March.Particularly noteworthy is the high and fast variabil-ity of the profiles: from 10 March, within 3 daysincreases in extinction coefficients of more than 200%at 15 km, 100% at 20 km, and 250% at 25 km weremeasured by the lidar. Then, a return to back-ground conditions �which corresponds to a maximumextinction coefficient value at 525 nm below 2.4 �10�3 km�1 from the tropopause to 30 km if we con-sider the work of Thomason38� occurred in mid-May.
Moreover, for the 16 October 1993 AMON flight�0030 UT, 44.0 ° N, 0.0 ° E�, a coincidence with theOHP lidar arose. Extinction coefficient profiles ofaerosols at 532 nm were derived for 3, 20, and 21October 1993. Renard et al.26 showed that, oncemore, there was good agreement of the extinctionvalues for the two sets of data in spite of differencesin measurement locations and in the methods of ob-servation.
The aerosol extinction coefficients from the 16 Oc-tober 1993 AMON flight can also be compared withthose from SAGE II measurements �0715 UT, 49.1 °N, 12.2 ° W; 0538 UT, 49.3 ° N, 11.9 ° E� on the samedate �version 6.1 of SAGE data was used�. Figure 4shows the comparison of the extinction coefficientvertical profiles between the AMON spectrometer at400, 450, and 525 nm and the SAGE II experiment at385, 453, and 525 nm, for the two events in closecoincidence with the AMON measurement location.In Fig. 4�a� the SAGE II profile at 49.1 ° N, 12.2 ° Wreveals lower extinction values than the AMON doesabove 18 km, whereas the SAGE II profile at 49.3 °N, 11.9 ° E plotted in Fig. 4�b� reveals greater extinc-
tion values from 21 to 25 km and lower values below20 km. Even though SAGE II and AMON measure-ments were not made at the same geographic loca-tion, which can explain these differences in theabsolute extinction values, the comparison of the ex-tinction spectra of the two instruments in Fig. 5 re-veals correct agreement for the shape of thewavelength-dependent extinction, at least when theerror bars are taken into account.
5. Retrieval of Aerosol Size Distributions and SurfaceArea Densities
A. Input Parameters
We can estimate the aerosol size distributions bycomparing, in the continuous ultraviolet–visiblespectral domain, the measured extinction coefficientswith those that result from a Mie scattering model.
Fig. 3. Comparison of aerosol extinction coefficients profiles �at530 nm� measured by the AMON on 24 March 1994, along witherror bars, with several observations from the OHP lidar at 532nm.
Fig. 4. Comparison of aerosol extinction measured by the AMONon 16 October 1993, along with error bars, with two observationsfrom the SAGE II on the same date. The wavelengths for theAMON data �400, 450, and 525 nm� were chosen to match theSAGE II wavelengths �385, 453, 525 nm� as closely as possible.SAGE II observations located at �a� 49.1 ° N, 12.2 W at 0715 UTand �b� 49.3 ° N, 11.9 ° E at 0538 UT.
The model supplies the extinction spectrum associ-ated with a size distribution of spherical particlesthat reproduces at its best the measured extinctionspectral dependence. Although AMON andSALOMON spectral domains are not broad and donot include aerosol extinction in the infrared domain,using the 375–700-nm spectral domain for such astudy is advantageous because it represents a goodcompromise among Rayleigh scattering and O3 andwater-vapor absorption. In addition, Mie scatteringcalculations have shown that the visible domain isuseful for characterization of particles smaller than0.5 m �the extinction spectra become quite neutralin this spectral domain for larger particles�.
The wavelength-dependent aerosol refractive indi-ces used are those of an aqueous solution of sulfuricacid that corresponds to a H2SO4 mass percentage of75% for a typical stratospheric temperature of 223K.28 Any change in H2SO4 concentration related totemperature changes slightly affects the aerosol re-fractive index, but the effect on the extinction is weakand can be neglected.
For many years the log-normal distribution func-
tion �LND� has been found to be well suited for char-acterizing the various components of a mixture ofaerosols.39,40 Recently O’Neill et al.41 demonstratedthat the LND is a good reference for reporting mea-surements; thus it is used in what follows. A com-bination of p modes LND can be written as
dNd�r�
� i�1
pN0i
ri ln �i�2 exp�� 1
2�ln ri � ln r0i
ln �i�2� , (1)
where N0i is the number density, r0i is the medianradius of distribution, and �i is the geometric stan-dard deviation �i.e., the width of the distribution� ofthe ith mode.
We also use the effective radius, as defined by Han-sen and Travis42 and widely used28,43,44:
Reff �
�0
�
r3 dN�r�
drdr
�0
�
r2 dN�r�
drdr
. (2)
Fig. 5. Comparison of the wavelength dependence of aerosols extinction measured from the AMON flight �box rules, which represent theerror measurements� and from the SAGE II for the two coincidences: 49.1 ° N, 12.2 ° W at 0715 UT �triangles� and 49.3 ° N, 11.9 ° E at0538 UT �diamonds�. The numbers immediately to the right of the box rules correspond to the tangent altitudes for the AMON; thenumbers near the right-hand boundary of the figure correspond to SAGE II tangent altitudes.
For multimodal LND size distributions the effectiveradius is given by
Reff �
i�1
p
N0ir0i3 exp�9�2�ln �i�
2�
i�1
p
N0ir0i2 exp�2�ln �i�
2�
(3)
and the aerosol surface area density is given by
S � �0
�
4 r2 dN�r�
drdr , (4)
or simply, for LND distributions, as
S � i�1
p
4 N0i r0i2 exp�2�ln �i�
2� . (5)
B. Monomodal Log-Normal Size Distributions
Stratospheric background aerosols are representedmostly by monomodal log-normal size distributions6,8
with p � 1 in Eq. �1�. Therefore, for each flight, theaerosol size distribution in each stratospheric layerwas retrieved first by use of monomodal log-normalsize distributions in the Mie extinction coefficient cal-culation. Because the residual optical depth spec-trum attributed to the aerosol contribution isrepresented by a third-order polynomial �see Subsec-tion 3.A�, the aerosol extinction coefficient spectrumat each level is expected to follow a third-order poly-nomial. Thus the model extinction coefficients mustalso be fitted with a same-order polynomial. For acomparison of theoretical and measured extinctioncoefficients a normalization process is used in the375–700-nm spectral domain to eliminate the N0 pa-rameter.
We tested several normalized distributions to ob-tain the best fit of the normalized measured aerosolextinction spectra, with adjustable parameters �seeEq. �1�� r0 ranging from 0.005 to 3 m �in steps of 5 �10�3, 1 � 10�2, and 1 � 10�1 m, respectively, for r0values smaller than 0.05 m, from 0.05 to 0.2 m,and greater than 0.2 m� and � from 1.05 to 3 �insteps of 0.05� to cover a large number of various cases.
At a given altitude, for each tested distribution thestandard deviation of the difference over the wholespectral domain between measured and model ex-tinction spectra is calculated. Searching for theminimum standard deviation is the choice criterionthat provides the best fit of the measurement by themodel. The sampling steps used in the Mie modelsyield uncertainties in r0 and �. Nevertheless, thebest solution cannot be considered a unique solution.Therefore, to express the spread of the results weretain other solutions, provided that the associatedextinction spectral shapes are within the error bars ofthe measured shape.
The aerosol number density can be calculated fromthe expression for extinction coefficient E:
E��� � �d��� N0 , (6)
where � is the wavelength and �d is the scatteringcross section of the aerosol. �d is provided by themodel; one can then calculate the ratio between thenonnormalized measured extinction and the modeledcross sections. The result is an estimate of the aero-sol number density N0 in the relevant stratosphericlayer.
For each altitude value, surface area densities aredetermined from Eq. �5� for all the retrieved distri-butions. The spread of the results yields the errorrange, and the average gives the final surface densityvalue for the relevant altitude.
C. Bimodal Log-Normal Size Distributions
Bimodal shapes were observed in measured aerosolsize distributions after the Mount Pinatubo eruptionand are usually observed in the presence of polarstratospheric cloud particles.7 Considering the al-most flat spectral shapes and the nonmonotonic spec-tral shapes of several aerosol extinction spectra �aspresented in Fig. 6�a� for the 1993 AMON flight�, thepresence of another size of particle much bigger thanbackground aerosols can be assumed. Russell etal.28 assumed only monomodal log-normal forms forthe retrieval of size distributions from post-Pinatuboextinction spectra because of the limited accuracy oftheir measurements and the few wavelengths used,preventing them from getting enough spectral infor-mation to retrieve bimodal log-normal distributionparameters. Because of the continuity of AMONand SALOMON spectral domains and the good opti-cal depth sensitivities of these instruments, we ex-pect to be able to retrieve the relative amounts of twopopulations. Thus we investigated our measure-ments to determine whether testing Mie-computedbimodal LND could improve the fits.
The parameters r01 and �1 of the first populationwere the same as those used for monomodal distri-butions. For the second population, six radii �r02 �0.3, 0.5, 0.7, 1, 3, 5 m� and six distribution widths��2 � 1.1, 1.2, 1.5, 1.8, 2.2, 2.7� were considered. Inaddition, we used second population number percent-ages that ranged from 0.001% to 10%, with two val-ues per decade. This distribution sampling wassufficient because it provides close enough extinctionspectra from successive Mie computations, whose dif-ferences are smaller than instrumental data errorranges. In this study we took the same refractiveindex for both populations to reduce the number ofunknowns, assuming that the two populations havethe same composition �the validity of such an as-sumption is discussed below�.
The aerosol bimodal log-normal distribution wasretrieved for all the spectra, and the procedure is thesame as that described in Subsection 5.A �numberdensity calculation, normalization, choice criterionfor the best fit and other solutions�. Surface densi-ties were calculated from Eq. �5�, and errors werecalculated in the same way as in Subsection 5.A.
6. Results from the AMON and SALOMON Flights andDiscussion
The flight conditions are listed in Table 1. In whatfollows, we show measured extinction spectra and theaerosol log-normal size distributions that give thebest fit. Such distributions are presented as illus-trations of the input parameters for the Mie scatter-ing calculations. The main results for theseparameters are given below. The consistency of thedistributions that assess the physical properties ofaerosols is described in the companion paper.25
A. 16 October 1993 Flight of the AMON Spectrometer
The presence of Pinatubo aerosols remained signifi-cant during a period that included the 16 October1993 flight of the AMON. First it may be noted thatRenard et al.26 showed, from this flight and from OHPlidar measurements, that the Pinatubo aerosol sig-nature was present above 30 km. Then, when weexamined aerosol extinction spectra from this flight,we could see that two kinds of spectral shape pre-dominate: monotonic and relatively flat spectrafrom 16 to 20 km and nonmonotonic spectra for sev-
Fig. 6. Example of a normalized aerosol extinction spectra measured by the AMON on 16 October 1993 at 20.2 and 17.5 km. �a�Spectrum fitted by the Mie model with a monomodal log-normal distribution at 20.2 km. �b� Spectrum fitted by the Mie model with abimodal log-normal distribution at 20.2 km. �c� Same as �a� but at 17.5 km, �d� same as �b� but at 17.5 km.
eral altitudes above 20 km. For the flat spectra theextinction coefficients are quasi wavelength indepen-dent; i.e., there is almost neutral extinction. Suchspectral dependence in the ultraviolet–visible spec-tral domain is usually attributed to big particles,namely, particles bigger than the usual backgroundaerosols or pre-Pinatubo aerosols.6,9 It is consistentwith the results shown by Russell et al.28 for the samealtitude range and for the period from approximatelyNovember 1992 to August 1993. The nonmonotonicspectra from 20 to 24 km tend to peak at 400–450nm. This spectral dependence was also reported byRussell et al. for measurements made in 1993.28
Flat extinction spectra are well reproduced bymonomodal log-normal distributions, for example, at17.5 km �Fig. 6�c��. For that case the best fit corre-sponds to a quite narrow distribution �� � 1.3� oflarge particles, with a median radius of 1 m �dashedcurve in Fig 7�b��. Other monomodal solutions thatreproduce the measured extinction spectra well mustbe considered. They correspond to lower median ra-dii �r0 near 0.2 m� but broader distributions �� near1.7�. In fact, we observed that these two types ofmonomodal distribution fit the measurement quitewell, demonstrating the nonuniqueness of the solu-tion. However, nonmonotonic extinction spectra arenot as well reproduced, for example, at 20.2 km �Fig.6�a��. Nevertheless, the Mie model succeeds in fit-ting the measurements by using a monomodal sizedistribution of medium-sized particles with r0 �0.18–0.30 m �dashed line in Fig. 7�a��. In thesecases, all the Mie-computed fits that can reproducethe measurements lead to similar size distributions,unlike for the flat spectra. To sum up, the modelalways attributes the presence of rather large parti-cles �r0 � 0.18 m� to these two spectral behaviors.
The tendency to have two kinds of monomodal dis-tribution �one broad population with a median radiusof �0.2 m and one narrow population of much big-ger particles� that are able to reproduce the flat ex-tinction spectra can be related to the fact that morethan one type of population is present in these strato-spheric layers. Indeed, using bimodal log-normaldistributions significantly improves the fits at mostaltitudes, especially for the relatively flat spectra�Fig. 6�d��, for which the standard deviation of thedifference between measurement and model can besix times smaller than previously �Fig. 8�. Below 20km, the median radius r01 of the first population iseither 0.16 or 0.2 m, and the median radius r02 of
Table 1. Flight Conditions of the Instruments Discussed in This Paper
Date�Day-Month-Year� Time �UT� Location of Instrument Experiment Altitude Range �km�
Fig. 7. Monomodal �dashed curves� and bimodal �solid curves�log-normal size distributions for the 16 October 1993 AMON flight.�a� At 20.2 km the monomodal distribution was obtained with r0 �0.18 m and � � 1.7; the bimodal distribution was obtained withr01 � 0.30 m, �1 � 1.1, r02 � 1.0 m, and �2 � 1.2. �b� At 17.5km the monomodal distribution was obtained with r0 � 1.0 m and� � 1.3; the bimodal distribution was obtained with r01 � 0.16 m,�1 � 2.2, r02 � 1.0 m, and �2 � 1.2.
the second population is 1 m. We notice a certainconsistency between best monomodal and bimodaldistributions with respect to the median radius andthe width of the distributions. Figure 7�b� describesthis consistency at 17.5 km: We see clearly that thenarrow monomodal distribution, centered on 1 m,corresponds well to the second population of the bi-modal distribution. Considering the flat shape ofthis spectrum �Fig. 6�c�� and the good quality of themonomodal fit, we can assume that the spectral con-tribution of the first population, which can be attrib-uted to that of small background aerosols, isdominated by the 5% of big particles that form thesecond population. Above 20 km, bimodal distribu-tions lead also to better fits of the nonmonotonic spec-tra than the monomodal distribution �Figs. 6�b� and8�. No clear consistency is observable between thetwo distributions �Fig. 7�a��. The first populationparameters are r01 � 0.2–0.3 m with �1 � 1.05–1.1�such a � value is highly unusual and must be con-sidered with caution�. We note that, unlike below20-km altitude, here the quality of the fit is less sen-sitive to the second population input parameters,meaning that the second mode cannot be well defined.
Finally, this study enables us to calculate the aero-sol surface area densities with their uncertaintiesbecause of the spread of the results in the distributionretrieval as given in Section 5. Figure 9 shows acomparison of aerosol surface area density verticalprofiles calculated with monomodal and bimodal dis-tributions. Monomodal and bimodal surface areadensity values are similar, and almost all fall withinthe uncertainty range. It may be mentioned thatthe maximum value of the profile �Fig. 16�b�, below�,7.5 � 0.5 m2�cm�3 �at 16.4 km�, is in agreementwith the value obtained from the estimated decay ofthe maximum surface area density with time as pre-sented by Lambert et al.45; extrapolation to the Oc-
tober 1993 measurement yields a value of less than 9m2�cm�3 near 16 km. This result could indicatethat there is no obvious bias in the retrieval methodused here for surface area density.
The consistency of the results obtained in thisstudy can be assessed due to the coincidence betweenAMON and SAGE II measurements made on 16 Oc-tober 1993. SAGE II extinction coefficients are mea-sured at only four wavelengths and do not allow oneto retrieve size distributions without large uncertain-ties or to determine the bimodal nature of the sizedistributions by using the constrained linear inver-sion method.17 Thus a direct comparison of size dis-tributions between AMON and SAGE II cannot bemade. However, integrated aerosol properties suchas the effective radius can be recovered from SAGE IImeasurements. Because the effective radius is auseful parameter for the comparison of data fromdifferent measurement techniques, it is provided inversion 6.1 of the SAGE II data. Zonal means werecomputed for October 1993 for the 40°–50 ° N latituderange. Figure 10 compares these monthly averagedresults with the effective radii derived from themonomodal log-normal distributions of the 16 Octo-ber 1993 AMON flight. When the errors bars aretaken into account, the values and the tendency forincreasing values with decreasing altitude agreewell. Even though the measured AMON extinctionspectra are not perfectly fitted by use of monomodaldistributions, leading to large error bars for twopoints, these monomodal distributions are consistentenough that integrated properties of the distribu-tions, such as the effective radius and the surfacearea density, can be recovered reliably.
B. 24 March 1994 Flight of the AMON Spectrometer
Below 20 km, nonmonotonic extinction spectralshapes predominate �not shown�. They have a
Fig. 8. Standard deviations of the difference between measuredaerosol extinction and fit for the 16 October 1993, flight of AMON.Bimodal distributions led to better fits than monomodal distribu-tions.
Fig. 9. Aerosol surface area density vertical profiles calculated forthe 16 October 1993 AMON flight. Good agreement between sur-face area densities calculated from monomodal and bimodal dis-tributions can be observed.
slight relative minimum at the shortest wavelengths�450 nm� and a slight maximum at longer wave-lengths �550–600 nm�, and then extinction valuesdecrease at the longest wavelengths. From 20 to 22km, spectra tend to be monotonic, i.e., quasi-neutralat 400–500 nm, and then tend to decrease with in-creasing wavelength �Fig. 11�c��. Above 22 km thespectra, again nonmonotonic, are characterized bymaximum extinction values near 450 nm and by astrong extinction decrease after this maximum �Fig.11�a��. None of these features of the spectral behav-ior is correctly reproduced by monomodal distribu-tions, whereas bimodal distributions improve the fitssignificantly, as shown in Figs. 11�a� and 11�c�. Thebimodal distributions tend to represent first popula-tions of small particles �r01 � 0.1 m�, with a geo-metric standard deviation that depends on thealtitude, and narrow second populations ��2 � 1.1�that correspond to more than 0.5% of big particleswhose radii vary according to the spectral shapesdescribed above. The most common values are r02 �1 m below 22 km �Fig. 11�d�� and r02 � 0.3 m above22 km �Fig. 11�b��.
So at least two different populations, a first con-sisting of small background aerosols �r01 � 0.1 m�and a second one of much bigger particles �r02 as largeas 1 m�, seem to be present in the atmosphericlayers that were studied. Sometimes the two modesare clearly separated �Fig. 11�b��, which looks ratherunrealistic. Even though monomodal retrievalseems to indicate the presence of particles biggerthan usual background aerosols �median radius, �0.2m or even 1 m for several solutions� to yield mostof the extinction shapes, no clear consistency is no-ticeable between monomodal and bimodal distribu-tions �Figs. 11�b� and 11�d��. Thus it is difficult for amonomodal distribution to fit such an extinction spec-tral dependence attributed to the influence of twostrongly separated particle populations.
The aerosol surface area density profile that hasbeen calculated with the bimodal distributions is il-lustrated in Fig. 16�c� below. It must be noted thatthis profile is more rigourous than the surface areadensity profile already estimated for this flight2 byuse of the Jager conversion factors.46
It was shown in Fig. 3 that an increase in theaerosol content at the measurement location betweenthe beginning and the end of March 1994 had beenobserved by the OHP lidar. This phenomenon,whose origin is unknown �but one can speculate thatit could be due to a tropical aerosol plume�, can ex-plain the greater surface density values from 19 to 24km from this flight than the 1993 values. In addi-tion, the presence of such a plume could justify theclaim that monomodal distributions are not adequateto characterize the measured extinction spectra.
C. 26 February 1997 Flight of the AMON Spectrometer
The 26 February 1997 AMON flight �67 ° N, 23 ° E�was performed inside the polar vortex. Extinctionspectra from this flight have various spectral depen-dencies, indicating the presence of different kinds ofparticle, depending on the altitude. However, thespectra are characterized chiefly by nonmonotonicshapes similarly to the spectral behavior describedfor the 1994 flight of the AMON, namely, a relativeminimum near 450 nm, a relative maximum from575 to 600 nm, and then a decrease until 675 nm, butthese features are much more marked for the 1997flight. These nonmonotonic spectra are badly fittedby monomodal distributions and are better fitted byuse of bimodal distributions �Fig. 12�a��. Althoughthe model does not perfectly reproduce the spectraldependence, it always associates a small quantity ofbig particles with these spectra: Most of the secondpopulation particle sizes are centered on 1 m, withpercentages of 0.01–0.5% �Fig. 12�b��, and again thetwo modes are much separated.
From 21 to 23 km, three consecutive extinctionspectra have stronger maxima at 575–600 nm, whichcannot be reproduced by any of the tested distribu-tions �Fig. 13�. Riviere et al.37 have pointed out bymodel calculations that the signature of solid sulfuricacid aerosols in these layers may indicate a previouspolar stratospheric cloud event, perhaps generatedby wave perturbations. Thus a predominantamount of nonspherical frozen particles in an atmo-spheric layer induces a particular spectral signaturethat prevents aerosol retrieval by use of the Mie scat-tering assumption. However, we can assume that,at other altitudes, where this signature is muchweaker, a small number of these large particles ispresent among background aerosols; thus the sizedistribution can be roughly estimated by use of a Mieassumption in which these particles are consideredas a second population.
Aerosol surface area densities were calculated insuch conditions �See Fig. 16�d� below�. Of course, novalue can be calculated at the altitudes of the polarstratospheric cloud event �21–23 km�. Some of thesevalues can be compared with surface area densities
Fig. 10. Comparison of vertical profiles of effective radius fromthe 16 October 1993 AMON flight and the SAGE II monthly dataaveraged in October 1993 in the 40°–50 ° N latitude range. Mono-modal distributions were assumed for the calculations.
derived from measurements of the University of Wy-oming balloonborne optical particle counter47 thatwere made at 68 ° N, 21 ° E inside the polar vortex on25 February 1997 during the Improved Limb Atmo-spheric Spectrometer �ILAS� correlative measure-ment campaign. During this flight large solidparticles were measured at altitudes of 16–22 km.Therefore comparisons can be made only at altitudesfor which there is no ambiguity in the nature of theparticles measured by the two instruments, that is to
say at 13 km and above 23 km. Particle-counterdata were averaged on layers, those data were com-pared with AMON measurements, and excellentagreement was observed between the two sets of val-ues, which correspond mostly to background aerosols�Fig. 14�.
D. Other AMON Flights
For the 24 May 1992 flight of the AMON spectrome-ter, some problems, i.e., faintness of the star, inaccu-
Fig. 11. Example of normalized aerosol extinction spectra and Mie model fits for the 24 March 1994 AMON flight: �a� spectrum fittedat 23.6 km. �b� Monomodal and bimodal size distributions at 23.6 km. The monomodal distribution was obtained with r0 � 0.20 m and� � 1.5; the bimodal distribution was obtained with r01 � 0.025 m, �1 � 1.1, r02 � 0.3 m, and �2 � 1.1. �c� Same as �a� but at 21.6km. �d� Same as �b� but at 21.6 km. Monomodal distribution with r0 � 0.2 m and � � 1.7, bimodal distribution with r01 � 0.06 m,�1 � 2.2, r02 � 1.0 m, and �2 � 1.1.
racy in the measurement of temperature andpressure conditions prevailing during the flight, andflight conditions that did not permit measurements tobe made below 25 km, made aerosol retrieval difficult.Extinction coefficients were tentatively retrieved; theyindicate in any case that aerosols were present above25 and even above 30 km. Figure 16�a� below showsthe inferred aerosol surface densities.
The 12 February 1999 flight of the AMON tookplace inside the polar vortex. Pointing problemsand the low flux of the star led to low signal-to-noiseratios; therefore only three apparently acceptable ex-tinction spectra from 18 to 26 km were obtained.Bimodal distributions significantly improve the fitsin comparison with monomodal distributions. Twoclearly separated populations appear in the three re-
Fig. 12. Example of normalized aerosol extinction spectrum andMie model fits for the 26 February 1997 flight of AMON: �a�spectrum fitted at 15.5 km, �b� Monomodal and bimodal size dis-tributions at 15.5 km. Monomodal distribution with r0 � 0.3 mand � � 1.2, bimodal distribution with r01 � 0.05 m, �1 � 1.3,r02 � 1.0 m, and �2 � 1.1. Note that the extinction y-axis scaleof Figs. 12, 13, and 15 differs by a factor of 2 from the scale of Figs.6 and 11.
Fig. 13. Extinction spectral behavior attributed to the presence ofnonspherical frozen particles after a polar stratospheric cloudevent for the 26 February 1997 AMON flight. None of the distri-butions can fit this spectral dependence.
Fig. 14. Comparison of aerosol surface area densities derivedfrom the 26 February 1997 AMON flight and from the Universityof Wyoming optical particle counter �UW OPC�. Boldface sym-bols, values corresponding mostly to background aerosols; goodagreement can be observed between the two sets of values. Light-face symbols, values corresponding to aerosols resulting from apolar stratospheric cloud event; significant disagreement standsout because the nature of aerosols can differ for the two flights.
trieved bimodal distributions: one that can be asso-ciated with background aerosols �r01 � 0.1 m� and asecond, with much bigger particles �0.7 m � r02 �1.0 m�, in a small quantity ��0.05%�. As lowstratospheric temperatures were detected during thisflight, it may be speculated that the second popula-tion is similar to those measured during the 1997AMON flight. Deduced aerosol surface densities areplotted in Fig. 16�e�.
E. 22 February 2000 Flight of the SALOMONSpectrometer
The 22 February 2000 SALOMON flight was per-formed during a moonrise on the edge of the polarvortex. Extinction spectra were obtained at 17–27km.
From 22 to 26 km, extinction spectra are suffi-ciently well reproduced by monomodal size distribu-tions �Fig. 15�a��. These distributions correspond tovery narrow populations of particles �� � 1.05� withmedian radius values that oscillate from 0.16 to 0.17m, which is evidence of the dominant presence ofbackground aerosols �Fig. 15�b��. A second popula-tion seems unnecessary because there is no obviouslarge particle signature.
At 19 and 17 km, bimodal distributions improvethe fits significantly in comparison with monomodaldistributions �Figs. 15�c� and 15�e�� and correspond totwo clearly separated populations that are inconsis-tent with the best monomodal solution �Figs. 15�d�and 15�f ��: a first solution with r01 � 0.1 m and asecond one with r02 � 1 m and a very low percentage�0.01% at 17 km and 0.05% at 19 km�. At 19 km thefit presented in Fig. 15�c� leads to an effective radiusof 0.18 m, which is in agreement with the pre-Pinatubo effective radius value of 0.17 � 0.07 mnear 19 km given by Russell et al.,28 which corre-sponds to background aerosols. At 17 km the bestbimodal solution is far from perfectly reproducing themeasurement �Fig. 15�e��. The spectral shape isslightly similar to that of the 1997 AMON flight be-low 20 km, which could emphasize the contribution oflarge particles. The deduced aerosol surface densi-ties are presented in Fig. 16�f �. Values inferredfrom monomodal distributions are retained at alti-tudes at which bimodal distributions are unnecessary�in that case, similar values for surface density areasare obtained�.
7. General Discussion and Conclusions
From the continuous optical depth spectra measuredby the two balloonborne spectrometers AMON andSALOMON since 1992, aerosol surface area densitieshave been retrieved. Figure 16 summarizes the re-sults. We used a method of comparison of measuredand modeled extinction coefficients that included ei-ther monomodal or bimodal log-normal distributions,depending on the spectral signature of the measuredaerosols. Moreover, the measurement method aswell as the data-reduction procedure that led to de-termination of extinction coefficients on theultraviolet–visible spectral domains were validated
by comparison of aerosol integral properties withproducts from satellite and lidar instruments. Thisstudy has led us to offer the following comments andconclusions.
We have shown that several conditions must besatisfied for the extinction coefficient and the aerosolsize distribution retrievals to be optimized: i.e., ahigh-quality reference spectrum, accuracy in themeasured temperature and pressure values prevail-ing during the flight �so the effects of Rayleigh scat-tering can be effectively neutralized�, and, for theAMON instrument, pointing accuracy and suitablestar brightness.
For each spectrum, in addition to the best solution,several other distributions from among the greatnumber of tested distributions properly reproducethe measurements and thus have to be retained.This nonuniqueness of the solution, which is partic-ularly obvious for quasi-neutral spectral behaviors�cf. the flat extinction spectra from the 1993 AMONflight�, allows one to estimate the uncertainties of theretrieved aerosol surface densities.
The signature of background aerosols inultraviolet–visible extinction spectra is characterizedby a strong regular decrease with increasing wave-length. The aerosols are adequately represented bymonomodal log-normal distributions such as in the22–26-km layers of the SALOMON flight, wherethere was no need for two aerosol modes.
We have seen that each flight had its own aerosolpeculiarity �Pinatubo aerosols in the 1993 flight,post–polar stratospheric cloud event particles in the1997 flight, etc.� and that most spectral shapes differsignificantly from typical background aerosol spec-tral signatures. In such cases the use of bimodallog-normal distributions is more appropriate for re-producing the measured extinction spectra and re-sults in excellent fits for most of the spectra. Thusthe spectral dependence is assigned to one populationof background aerosols and to a second populationthat corresponds to bigger particles �r02 � 0.2 m� invarious concentrations that depend on the flight con-sidered: For example, only a few hundredths of apercent of particles in the second population werefound for the 2000 SALOMON flight. Even thoughthe measured extinction spectra are, in most cases,better reproduced by use of bimodal distributions,monomodal distributions provide consistent values ofthe effective radius and of the surface area density, orat least a good first approximation of them.
Some spectral anomalies �nonmonotonic spectraldependence� appear, even without the presence ofany Pinatubo aerosols �26 February 1997 AMONflight and 22 February 2000 SALOMON flight�.These extinction spectra are often reproduced incor-rectly �or not at all� by the model, whatever the aero-sol size distribution is. Tests have revealed that thismisinterpretation is not due to a lack of suitable test-ing distributions, in particular to a lack of secondpopulation median radii and distribution widths.Also, this fit problem cannot be the result of the pres-ence of an amount of O3 that remains in the aerosol
Fig. 15. Example of normalized aerosol extinction spectra and Mie model fits for the 22 February 2000 SALOMON flight: �a� spectrumfitted at 25.5 km, �b� monomodal size distribution at 25.5 km with r0 � 0.17 m and � � 1.2, �c� spectrum fitted at 19 km. �d� Monomodaland bimodal size distributions at 19 km; monomodal distribution with r0 � 0.035 m and � � 2.0, bimodal distribution with r01 � 0.06m, �1 � 1.5, r02 � 1.0 m, and �2 � 1.1. �e� Spectrum fitted at 17 km. �f � Monomodal and bimodal size distributions at 17 km;monomodal distribution with r0 � 0.06 m and � � 1.8, bimodal distribution with r01 � 0.05 m, �1 � 1.3, r02 � 1.0 m, and �2 � 1.1.
Fig. 16. Aerosol surface area density vertical profiles calculated for the various AMON flights and the 22 February 2000 SALOMONflight. Surface density values from monomodal distributions are plotted when bimodal distributions are unnecessary. Surface densityvalues are removed at altitudes where the measured extinction spectra are not suitably fitted. Profiles for �a� the 24 May 1992 AMONflight, �b� the 16 October 1993 AMON flight, �c� the 24 March 1994 AMON flight, �d� the 26 February 1997 AMON flight, �e� the 12 February1999 AMON flight, and �f � for the 22 February 2000 SALOMON flight.
spectra, because these anomalies are located in par-ticular layers and not in others. It must then beconsidered that these spectral structures need fur-ther analysis.
In this paper we have validated our retrievalmethod of aerosol integral properties derived frommeasurements whose originality and advantages arebased on the self-calibration of the two instruments,unlike aircraftborne sunphotometers, and on the con-tinuity of the spectral domains, unlike those used inmeasurements by satellites.
It is now necessary to compare and discuss the sizedistributions, and the validity of assumptions usedfor data reduction, with those obtained by other bal-loonborne instruments and with other aerosol mea-surement techniques such as optical particlecounting and polarimetry. These comparisons anddiscussion are presented in the companion paper byRenard et al.25
The authors thank all the members of the CentreNational d’Etudes Spatiales launching team at Airesur l’Adour and Kiruna, and the Geneva Observatoryteam that operates the AMON gondola. The au-thors also thank T. Deshler for his data and LaRC forSAGE II data. This research is supported by theFrench Program of Atmospheric Chemistry and byvarious European Commission contracts.
References and Notes1. D. J. Lary, R. Toumi, A. M. Lee, M. Newchurch, M. Pirre, and
J.-B. Renard, “Carbon aerosols and atmospheric photochemis-try,” J. Geophys. Res. 102, 3671–3682 �1997�.
2. J.-B. Renard, M. Pirre, C. Robert, F. Lefevre, E. Lateltin, B.Noziere, and D. Huguenin, “Vertical distribution of nighttimestratospheric NO2 from balloon measurements: comparisonwith models,” Geophys. Res. Lett. 24, 73–76 �1997�.
3. J.-B. Renard, F. G. Taupin, E. D. Riviere, M. Pirre, N. Huret,G. Berthet, C. Robert, and M. Chartier, “Measurements andsimulation of stratospheric NO3 at mid- and high-latitudes inthe Northern Hemisphere,” J. Geophys. Res. 106, 32,387–32,399 �2001�.
4. D. J. Hofmann and T. Deshler, “Stratospheric cloud observa-tions during formation of the Antarctic ozone hole in 1989,” J.Geophys. Res. 96, 2897–2912 �1991�.
5. T. Deshler, D. J. Hofmann, B. J. Johnson, and W. R. Rozier,“Balloonborne measurements of the Pinatubo aerosol size dis-tribution and volatility at Laramie, Wyoming, during the sum-mer of 1991,” Geophys. Res. Lett. 19, 199–202 �1992�.
6. T. Deshler, B. J. Johnson, and W. R. Rozier, “Balloonbornemeasurements of Pinatubo aerosol during 1991 and 1992 at41 ° N: vertical profiles, size distribution, and volatility,”Geophys. Res. Lett. 20, 1435–1438 �1993�.
7. T. Deshler and S. J. Oltmans, “Vertical profiles of volcanicaerosol and polar stratospheric clouds above Kiruna, Sweden:winters 1993 and 1995,” J. Atmos. Chem. 30, 11–23 �1998�.
8. J. Goodman, K. G. Snetsinger, R. F. Pueschel, G. V. Ferry, andS. Verma, “Evolution of Pinatubo aerosol near 19 km altitudeover western North America,” Geophys. Res. Lett. 21, 1129–1132 �1994�.
9. R. F. Pueschel, K. G. Snetsinger, J. K. Goodman, O. B. Toon,G. V. Ferry, V. R. Oberbeck, J. M. Livingston, S. Verma, W.Fong, W. L. Starr, and K. R. Chan, “Condensed nitrate, sulfate,and chloride in Antarctic stratospheric aerosols,” J. Geophys.Res. 94, 11,271–11,284 �1989�.
10. R. F. Pueschel, P. B. Russell, D. A. Allen, G. V. Ferry, K. G.Snetsinger, J. M. Livingston, and S. Verma, “Physical andoptical properties of the Pinatubo volcanic aerosol: aircraftobservations with impactors and a Sun-tracking photometer,”J. Geophys. Res. 99, 12,915–12,922 �1994�.
11. C. Brogniez, J. Lenoble, R. Ramananaherisoa, K. H. Fricke,E. P. Shettle, K. W. Hoppel, R. M. Bevilacqua, J. S. Hornstein,J. Lumpe, M. D. Fromm, and S. S. Krigman, “Second EuropeanStratospheric Arctic and Midlatitude Experiment campaign:correlative measurements of aerosol in the northern polar at-mosphere,” J. Geophys. Res. 102, 1489–1494 �1997�.
12. P. B. Russell, J. M. Livingston, E. G. Dutton, R. F. Pueschel,J. A. Reagan, T. E. Defoor, M. A. Box, D. Allen, P. Pilewskie,B. M. Herman, S. A. Kinne, and D. J. Hofmann, “Pinatubo andpre-Pinatubo optical-depth spectra: Mauna Loa measure-ments, comparisons, inferred particle size distributions, radi-ative effects, and relationship to lidar data,” J. Geophys. Res.98, 22,969–22,985 �1993�.
13. E. G. Dutton, P. Reddy, S. Ryan, and J. J. DeLuisi, “Featuresand effects of aerosol optical depth observed at Mauna Loa,Hawaii: 1982–1992,” J. Geophys. Res. 99, 8295–8306 �1994�.
14. H. Jager, O. Uchino, T. Nagai, T. Fujimoto, V. Freudenthaler,and F. Homburg, “Ground-based remote sensing of the decay ofthe Pinatubo eruption cloud at three Northern Hemispheresites,” Geophys. Res. Lett. 22, 607–610 �1995�.
15. P. Chazette, C. David, J. Lefrere, S. Godin, J. Pelon, and G.Megie, “Comparative lidar study of the optical, geometrical,and dynamical properties of stratospheric post-volcanic aero-sols, following the eruptions of El Chichon and Mount Pina-tubo,” J. Geophys. Res. 100, 23,195–23,207 �1995�.
16. P. Di Girolamo, G. Pappalardo, N. Spinelli, V. Berardi, and R.Velotta, “Lidar observations of the stratospheric aerosol layerover southern Italy in the period 1991–1995,” J. Geophys. Res.101, 18,765–18,773 �1996�.
17. H. M. Steele and R. P. Turco, “Retrieval of aerosol size distri-butions from satellite extinction spectra using constrained lin-ear inversion,” J. Geophys. Res. 102, 16,737–16,747 �1997�.
18. H. M. Steele, J. D. Lumpe, R. P. Turco, R. M. Bevilacqua, andS. T. Massie, “Retrieval of aerosol surface area and volumedensities from extinction measurements: application to POAMII and SAGE II,” J. Geophys. Res. 104, 9325–9336 �1999�.
19. C. E. Randall, R. M. Bevilacqua, J. D. Lumpe, K. W. Hoppel,D. W. Rusch, and E. P. Shettle, “Comparison of Polar Ozoneand Aerosol Measurement �POAM� II and Stratospheric Aero-sol and Gas Experiment �SAGE� II aerosol measurements from1994 to 1996,” J. Geophys. Res. 105, 3929–3942 �2000�.
20. M. E. Hervig, T. Deshler, and J. M. Russell III, “Aerosol sizedistributions obtained from HALOE spectral extinction mea-surements,” J. Geophys. Res. 103, 1573–1583 �1998�.
21. M. E. Hervig and T. Deshler, “Stratospheric aerosol surfacearea and volume inferred from HALOE, CLAES, and ILASmeasurements,” J. Geophys. Res. 103, 25,345–25,352 �1998�.
22. D. Fussen, E. Arijs, D. Nevejans, F. V. Hellemont, C. Brogniez,and J. Lenoble, “Validation of the ORA spatial inversion algo-rithm with respect to the Stratospheric Aerosol and Gas Ex-periment II data,” Appl. Opt. 37, 3121–3127 �1998�.
23. J. Anderson, C. Brogniez, L. Cazier, V. K. Saxena, J. Lenoble,and M. P. McCormick, “Characterization of aerosols from sim-ulated SAGE III measurements applying two retrieval tech-niques,” J. Geophys. Res. 105, 2013–2027 �2000�.
24. G. K. Yue, “Retrieval of aerosol size distributions and integralproperties from simulated extinction measurements at SAGEIII wavelengths by the linear minimizing error method,” J.Geophys. Res. 105, 14,719–14,736 �2000�.
25. J. B. Renard, G. Berthet, C. Robert, M. Chartier, M. Pirre, C.Brogniez, M. Herman, C. Verwaerde, J.-Y. Balois, J. Ovarlez,H. Ovarlez, J. Crespin, and T. Deshler, “Optical and physicalproperties of stratospheric aerosols from balloon measure-
ments in the visible and near-infrared domains. II. Compari-son of extinction, reflectance, polarization, and countingmeasurements,” Appl. Opt. 41, 7540–7549 �2002�.
26. J.-B. Renard, M. Pirre, C. Robert, G. Moreau, D. Huguenin,and J. M. Russell III, “Nocturnal vertical distribution of strato-spheric O3, NO2 and NO3 from balloon measurements,” J.Geophys. Res. 101, 28,793–28,804 �1996�.
27. J.-B. Renard, M. Chartier, C. Robert, G. Chalumeau, G. Ber-thet, M. Pirre, J.-P. Pommereau, and F. Goutail, “SALOMON:a new, light balloonborne UV–visible spectrometer for night-time observations of stratospheric trace-gas species,” Appl.Opt. 39, 386–392 �2000�.
28. P. B. Russell, J. M. Livingston, R. F. Pueschel, J. J. Bauman,J. B. Pollack, S. L. Brooks, P. Hamill, L. W. Thomason, L. L.Stowe, T. Deshler, E. G. Dutton, and R. W. Bergstrom, “Globalto microscale evolution of the Pinatubo volcanic aerosol de-rived from diverse measurements and analyses,” J. Geophys.Res. 101, 18,745–18,763 �1996�.
29. C. Robert, “Realisation d’un spectrometre stellaire multicanalembarquable sous ballon stratospherique,” Ph.D. dissertation�University of Orleans, Orleans, France, 1992�.
30. J.-P. Naudet, C. Robert, and D. Huguenin, “Balloon measure-ments of stratospheric trace species using a multichanel UV-visible spectrometer,” in Proceedings of the 14th ESASymposium on European Rocket and Balloon Programs andRelated Research, ESA document SP-355, �European SpaceAgency, Paris, 1994�, pp. 165–168.
31. J.-B. Renard, M. Pirre, and C. Robert, “The possible detectionof OBrO in the stratosphere,” J. Geophys. Res. 103, 25,383–25,395 �1998�.
32. J.-P. Pommereau and J. Piquard, “Ozone, nitrogen dioxide andaerosol vertical distribution by UV–visible solar occultationfrom balloon,” Geophys. Res. Lett. 21, 1227–1230 �1994�.
33. G. Berthet, J.-B. Renard, M. Chartier, M. Pirre, and C. Robert,“Analysis of OBrO, IO and OIO absorption signature in UV–visible spectra measured at night and at sunrise by strato-spheric balloon-borne instruments,” J. Geophys. Res. �to bepublished�.
34. J.-B. Renard, F. Dalaudier, A. Hauchecorne, C. Robert, T.Lemaire, M. Pirre, and J.-L. Bertaux, “Measurement of strato-spheric chromatic scintillation by the balloon-borne spectrom-eter AMON-RA,” Appl. Opt. 40, 4254–4260 �2001�.
35. A. Bucholtz, “Rayleigh-scattering calculations for the terres-trial atmosphere,” Appl. Opt. 34, 1227–1230 �1995�.
36. S. Voigt, J. Orphal, and J. P. Burrows: information availableat their Internet site, www.iup.physik.uni-bremen.de�grup-pen�molspec�index.html.
37. E. D. Riviere, N. Huret, F. G. Taupin, J.-B. Renard, M. Pirre,S. D. Eckermann, N. Larsen, T. Deshler, F. Lefevre, S. Payan,and C. Camy-Peyret, “Role of lee waves in the formation ofsolid polar stratospheric clouds: case studies from February1997,” J. Geophys. Res. 105, 6845–6853 �2000�.
38. L. W Thomason, “Observations of a new SAGE II aerosol ex-tinction mode following the eruption of Mt. Pinatubo,” Geo-phys. Res. Lett. 19, 2179–2182 �1992�.
39. G. A. d’Almeida, P. Koepke, and E. P. Shettle, AtmosphericAerosols, Global Climatology and Radiative Characteristics�Deepak, Hampton, Va., 1991�.
40. J. H. Seinfeld, and S. N. Pandis, Atmospheric Chemistry andPhysics, from Air Pollution to Climate Change �Wiley, NewYork, 1998�.
41. N. T. O’Neill, A. Ignatov, B. N. Holben, and T. F. Eck, “Thelognormal distribution as a reference for reporting aerosol op-tical depth statistics: empirical test using multi-year, multi-site AERONET sunphotometer data,” Geophys. Res. Lett. 27,3333–3336 �2000�.
42. J. E. Hansen and L. D. Travis, “Light scattering in planetaryatmosphere,” Space Sci. Rev. 16, 527–610 �1974�.
43. J. Lenoble and C. Brogniez, “A comparative review of radiationaerosol models,” Beitr. Phys. Atmos. 57, 1–20 �1984�.
44. R. G. Grainger, A. Lambert, C. D. Rodgers, and F. W. Taylor,“Stratospheric aerosol effective radius, surface area and vol-ume estimated from infrared measurements,” J. Geophys. Res.100, 16,507–16,518 �1995�.
45. A. Lambert, R. G. Grainger, C. D. Rodgers, F. W. Taylor, J. L.Mergenthaler, J. B. Kumer, and S. T. Massie, “Global evolu-tion of Mt. Pinatubo volcanic aerosols observed by the infraredlimb-sounding instruments CLAES and ISAMS on the UpperAtmosphere Research Satellite,” J. Geophys. Res. 102, 1495–1512 �1997�.
46. H. T. Jaeger, T. Deshler, and D. J. Hofmann, “Midlatitudelidar backscatter conversion based on balloon borne aerosolsmeasurements,” Geophys. Res. Lett. 22, 1729–1732 �1995�.
47. T. Deshler, B. Nardi, A. Adriani, F. Cairo, G. Hansen, F. Fierli,A. Hauchecorne, and L. Pulvirenti, “Determining the index ofrefraction of polar stratospheric clouds above Andoya �69 ° N�by combining size-resolved concentration and optical scatter-ing measurements,” J. Geophys. Res. 105, 3943–3953 �2000�.
