The Stratospheric Aerosol and Gas Experiment (SAGE) II solar occultation instrument has been making measurements on stratospheric aerosols and gases continually since October 1984. Observations from the SAGE II instrument provide a valuable long-term data set for study of the aerosol in the stratosphere and aerosol and cloud in the upper troposphere. The period of observation covers the decay phase of material injected by the El Chichon volcanic eruption in 1982, the years 1988 - 1990 when stratospheric aerosol levels approached background levels, and the period after the eruption of Mount Pinatubo in 1991. The Mount Pinatubo eruption caused the largest perturbation in stratospheric aerosol loading in this century, with effects on stratospheric dynamics and chemistry. The SAGE II data sequence shows the global dispersion of aerosols following the Mount Pinatubo eruption, as well as the changes occurring in stratospheric aerosol mass and surface area. The downward transfer of stratospheric aerosols into the upper troposphere following the earlier eruption of El Chichon is clearly visible. Estimates have been made of the amount of volcanic material lying in the upper troposphere and the way in which this varies with latitude and season.

The Polar Ozone and Aerosol Measurements II (POAM II) has been developed by the Naval Research Laboratory and launched end of September 1993 on the french satellite SPOT 3. The instrument observes solar occultations at 9 wavelength channels. The inversion algorithm allows the retrieval of the extinction vertical profiles and the separation of the species contributing to this extinction: air molecules, aerosols, ozone, water vapor and nitrogen dioxide; furthermore the retrieval of aerosol extinction at 4 wavelengths provides an information on the aerosol size distribution. An inversion algorithm has been developed at the LOA and preliminary comparisons with the NRL algorithm are presented. Two balloon-borne instruments have been flown from Kiruna (Sweden) in coincidence with POAM II observations: one of the instrument (RADIBAL) measures the radiance and polarization diagrams in the near infrared during the balloon ascent; the other instrument (BALLAD) observes the earth's limb at 3 wavelengths from the balloon ceiling altitude. Both instruments are complementary and provide the aerosol profile and the aerosol size distribution. Preliminary comparisons with POAM II data are presented.

Stratospheric aerosols produced by the eruption of the Mount Pinatubo in the Philippines (6 June, 1991) have a detectable effect on NOAA AVHRR data. Following the eruption, a longitudinally homogeneous dust layer was observed between 20 degree(s)N and 20 degree(s)S. The largest optical thickness observed for the dust layer was 0.4 - 0.6 at 0.5 microns. The amount of aerosols produced by Mount Pinatubo was two to three times greater than that produced by El Chichon and the Stratospheric Aerosol and Gas Experiment (SAGE) on-board the Earth Radiation Budget Experiment was not able to give quantitative estimate of aerosol optical thickness because of saturation problem. The monthly composite Normalized Difference Vegetation Index (NDVI) (generally bounded between -0.1 and 0.6) has systematically decreased by approximately 0.15 two months after the eruption. Such atmospheric effect has never been observed on composite product and is related to the persistence and spatial extent of the aerosol layer causing the composite technique to fail. Therefore, long term monitoring of vegetation using the NDVI necessitates correction of the effect of stratospheric aerosols. In this paper we present an operational stratospheric aerosol correction scheme adopted by the Laboratory for Terrestrial Physics, NASA/GSFC. The stratospheric aerosol distribution is assumed to be only variable with latitude. Each 9 days the latitudinal distribution of the optical thickness is computed by inverting radiances observed in AVHRR channel 1 (0.63 microns) and channel 2 (0.83 microns) over the Pacific Ocean. This radiance data set is used to check the validity of model used for inversion by checking consistency of the optical thickness deduced from each channel as well as optical thickness deduced from different scattering angles. The deduced optical thickness and spectral dependence are compared to Mauna Loa observation from 1991 to end of 1992 for validation. Using the optical thickness profile previously computed and radiative transfer code assuming lambertian boundary condition, each pixel of channel 1 and 2 are corrected prior to computation of NDVI. Comparison between corrected, non corrected, and years prior to Pinatubo eruption (1989, 1990) NDVI composite, shows the necessity and the accuracy of the operational correction scheme. The same technique is applied to the afternoon satellite AVHRR archive (NOAA7,9,11) from 1981 to 1993. The stratospheric profile derived over ocean shows that the El Chichon eruption was of less importance than Pinatubo. The stratospheric aerosol optical depth distribution computed from AVHRR data during the El Chichon period compared well to latitudinal monthly profile based on SAGE observations.

Based on radiative transfer calculations it is studied whether Polar Stratospheric Clouds (PSCs) can be detected by the new Global Ozone Monitoring Experiment onboard the second European Research Satellite (ERS-2) planned to be launched in winter 1994/95. It is proposed to identify PSC covered areas by use of an indicator, the Normalized Radiance Difference (NRD), which relates the difference of two spectral radiances at 0.515 micrometers and 0.67 micrometers to one radiance measured in the center of the oxygen A-band at 0.76 micrometers . In presence of PSCs and under conditions of increasing solar zenith angles above O equals 80 degree(s) the NRD rapidly decrease to values clearly below those derived under conditions of a cloud free stratosphere. Calculations for O equals 86 degree(s) show that this method is successful independently from existing tropospheric clouds and by different tropospheric aerosol loadings or surface albedos. For solar zenith angles O < 80 degree(s) the PSC detection needs additional information about tropospheric clouds.

Radiative transfer simulation of the proposed MERIS channels were made for several aerosol types and aerosol optical thicknesses above water surfaces. The Medium Resolution Imaging Spectrometer will be launched onboard the Environmental Satellite in 1998. The model used for the radiative transfer calculations is based on the Matrix-Operator-Method and describes the ocean-atmosphere-system. Atmosphere and ocean properties were varied to study their influence on the retrieval of aerosol properties. Using three different statistic and semi-statistic retrieval methods for the aerosol type and optical thickness, it is shown to which properties these algorithms are most sensitive. For the aerosol retrieval the aerosol path radiances in three channels in the red and near infrared region of the spectrum were interpreted. These channels and their proposed bandwidths are: (lambda) equals 755 nm with (Delta) (lambda) equals 7.5 nm, (lambda) equals 870 nm with (Delta) (lambda) equals 10 nm and (lambda) equals 1022.5 nm with (Delta) (lambda) equals 25 nm. The results show that a crude aerosol characterization in maritime like and continental like aerosols and mixtures of both is possible using the spectral radiances in the three proposed channels. It is also shown how these results are affected by the accuracy of the knowledge about atmospheric and oceanic properties such as surface air pressure, scattering at suspended matter in the ocean and ocean roughness.

The accuracy of atmospheric corrections for VIS-NIR remote sensing data depends on how closely the actual state of the atmosphere matches the assumptions made about aerosol scattering in whatever radiative transfer model is being used to compute the atmospheric transmittance. An experiment was performed to measure the scattering properties of actual atmospheres so that an assessment could be made on the impact of using pre-calculated phase functions for the aerosol scattering in radiative transfer models rather than empirically obtained ones. High resolution spectra data (10 nm) were obtained with an airborne platform of the atmospheric radiance over a large low reflectance target (a lake) as a function of view angle, view azimuth and altitude on several different days (and atmospheric states). The observations were made over a viewing range of -60 to +60 degrees at nine angles in several azimuth planes. This resulted in a wide range of scatter angles over which the atmospheric volume scattering could be examined. The atmospheric states ranged from very clear to very hazy all within a period of a month. A comparison of the angular dependence of the volume scattering with standard phase functions used in radiative transfer models is made. The angular behavior of the volume scattering is found to be sensitive to altitude as well as the aerosol optical depth.

The satellite level radiance is affected by the presence of the atmosphere between the sensor and the target. The ozone and water vapor absorption bands affect the signal recorded by the AVHRR visible and near infrared channels respectively. The Rayleigh scattering mainly affects the visible channel and is more pronounced when dealing with small sun elevations and large view angles. The aerosol scattering affects both channels and is certainly the most challenging term for atmospheric correction because of the spatial and temporal variability of both the type and amount of particles in the atmosphere. This paper presents the equation of the satellite signal, the scheme to retrieve atmospheric properties and corrections applied to AVHRR observations. The operational process uses TOMS data and a digital elevation model to correct for ozone absorption and rayleigh scattering. The water vapor content is evaluated using the split-window technique that is validated over ocean using 1988 SSM/I data. The aerosol amount retrieval over Ocean is achieved in channels 1 and 2 and compared to sun photometer observations to check consistency of the radiative transfer model and the sensor calibration. Over land, the method developed uses reflectance at 3.75 microns to deduce target reflectance in channel 1 and retrieve aerosol optical thickness that can be extrapolated in channel 2. The method to invert the reflectance at 3.75 microns is based on MODTRAN simulations and is validated by comparison to measurements performed during FIFE 87. Finally, aerosol optical thickness retrieved over Brazil and Eastern United States is compared to sun photometer measurements.

In this paper we discuss the possible effect of nonsphericity of solid tropospheric aerosols on the accuracy of aerosol thickness retrievals from reflectance measurements over the ocean surface. To model light-scattering properties of nonspherical aerosols, we use a shape mixture of moderately aspherical, randomly oriented polydisperse spheroids. We assume that the size distribution and refractive index of aerosols are known and use the aerosol optical thickness 0.2 to computer the reflectivity for an atmosphere-ocean model similar to that used in the AVHRR aerosol retrieval algorithms. We then use analogous computations for volume- equivalent spherical aerosols with varying optical thickness to invert the simulated nonspherical reflectance. Our computations demonstrate that the use of the spherical model to retrieve the optical thickness of actually nonspherical aerosols can result in errors which, depending on the scattering geometry, can well exceed 100%.

Radiance measurements in the O₂A-band are sensitive to changes of optical depth of stratospheric and of tropospheric aerosols. There results a chance to use these measurements for separate estimation of optical depth of stratospheric and of tropospheric aerosols. First inversion results using simulated data are used to evaluate accuracy of stratospheric optical depth estimation. Possible applications are discussed.

POLDER (POLarization and Directionality of the Earth's Reflectances) is a new instrument devoted to the global observation of the polarization and directionality of solar radiation reflected by the Earth surface-atmosphere system. The instrument concept has been accepted on the Japanese ADEOS platform scheduled to be launched early 1996. The original capabilities of POLDER, compared to previous current radiometers are, (1) its polarized reflectance measurements in the visible and near-infrared range of the solar spectrum, (2) its capability to measure a surface target reflectance from about 10 directions during a single pass. A method for cloud phase retrieving from POLDER measurements is tested. Indeed, liquid water clouds could be discriminated from ice clouds provided they exhibit distinct polarization signatures. In the rainbow region (scattering angles of about 140 degree(s)), water droplets strongly polarize incident sunlight while ice crystals probably do not. This feature is examined on data acquired by the airborne POLDER instrument over cirrus and stratocumulus clouds during the EUCREX'94 (EUropean Clouds and Radiation EXperiment, April 1994) experiment. Moreover, over clouds, the polarized component of the reflectance at the wavelength of 443 nm and scattering angle of 90 - 100 degree(s) is sensitive to molecular optical thickness between the cloud top and the satellite altitude and, therefore, may be used for cloud altimetry. On the other hand, a method for cloud top pressure retrieval from POLDER measurements based on a differential absorption technique is presented. It makes use of the ratio of two radiances measured in the absorption A band of the oxygen (at 763 nm). The two different methods are compared on data acquired during EUCREX'94. Considering the main limitations of the instrument and the methods, the two mean retrieved cloud-top pressures are found to be in good agreement and are close to the expected true one.

SAGE II water vapor measurements covering the period (1986 - 1991) have been used to examine the time-periodic variations of water vapor in the stratosphere and the upper troposphere. The results represent an extension of our previous studies which were based on the observations from the first three years (1986 - 88). A statistical model containing quasi- biennial, seasonal, semi-annual oscillations, and a linear component was fit to the time series of monthly zonal mean water vapor mixing ratio. A preliminary comparison with the newly released UARS/MLS water vapor data set is also discussed here.

Measurements of stratospheric composition have now been made by the Halogen Occultation Experiment (HALOE) on board NASA's Upper Atmospheric Research Satellite (UARS) since October 1991. Amongst the parameters measured are water vapor, H₂O, and methane, CH₄. These species comprise the dominant components of the total hydrogen budget in the lower stratosphere, but not so at higher levels, where the molecular hydrogen, H₂, component is significant, and at high altitudes is dominant. This paper reports on measurements of the water vapor and hydrogen fields in the stratosphere and mesosphere, and on studies of the derivation of molecular hydrogen in the mesosphere.

Due to the great importance of atmospheric water vapor for weather and climate, much effort is devoted to remote sensing of atmospheric water vapor. The detection over water is well established, while the situation over land surface is worse. Therefore, a new method is developed to derive the total atmospheric water vapor content over land surfaces even for higher aerosol contents with the aid of backscattered solar radiances. Numerous radiative transfer simulations with a matrix operator code of vertically backscattered solar radiance were carried out for different vertically stratified atmospheres. The resolution of 1.7 nm in the wavelength range from 700 to 1050 nm was adopted to the resolution of our multichannel spectrometer OVID (Optical Visible and near Infrared Detector). Various atmospheric conditions were chosen, which were defined by variable input parameters of: (a) vertical profiles of temperature, pressure, and water vapor, (b) total water vapor content, (c) aerosols, (d) surface reflectance, and (e) sun zenith angle. Clouds were not taken into account. From the evaluation of these theoretical calculations it can be concluded that this technique allows the detection of total atmospheric water vapor content over land surfaces with an error of less than 10%. This result is important with regard to future measurements planed with the MERIS imaging spectrometer on board the european satellite ENVISAT, which will be launched in 1998. In addition to these theoretical calculations also various aircraft measurements of the backscattered radiances in the wavelength range from 600 to 1650 nm were carried out. These measurements are done with the above mentioned OVID, a new multichannel array spectrometer of the Universities of Hamburg and Berlin. First comparisons of these airborne CCD measurements with calculated spectra are shown.

An attempt is made to construct a zonal and monthly mean ozone climatology for use in general circulation models, based on a combination of ozonesonde and satellite observations. One important advantage of such a climatology is a more realistic ozone distribution around the tropopause, where heating rates and climate forcing are most sensitive to changes in gas concentrations. Also, a linear trend study is performed, for the periods 1970 - 83 and 1980 - 93 separately, on concurrent ozone and temperature data obtained from a selection of ozonesonde stations. On average for northern polar- to mid-latitudes, these trends are insignificant for stratospheric ozone and temperature in the first period, but for the second period show a stratospheric ozone depletion and stratospheric cooling of around -0.5%/year and -0.15 K/year respectively. As for the troposphere in the same region, ozone shows an increase (approximately 1.5%/year) in the mid-troposphere but temperature trends are insignificant over the first period, versus no ozone trend but a clearly significant near-surface warming (approximately 0.2 K/year) in the second period. This average situation is however not representative for the separate regions it is composed of, i.e., Canada (4 stations), Japan (3 stations) and the U.S. (1 station). Above Syowa station at the Antarctic coast, the acceleration in stratospheric ozone depletion as well as stratospheric cooling over the past two decades is clearly evident: from hardly significant ozone and temperature trends in the first period to values of up to -4%/year and -0.4 K/year respectively in the second period. In regions where near-surface ozone increase is evident over the past two decades, it is often accompanied by a significant near-surface warming.

A 2D advection model is developed at KNMI to advect and assimilate total ozone using a windfield at a single pressure level. The Advection Model KNMI with a resolution of 110 X 110 km² describes the transport of total ozone, considering ozone as a passive tracer, using a simple linear advection equation. Ozone data measured by the TOVS instrument on the NOAA polar satellites and windfields from the MARS archives of ECMWF are used. By means of the AMK model the TOVS total ozone maps, which are hampered by missing data, can be replaced by global total ozone maps at a given time without gaps in the data. The windfield that must be used for transporting total ozone is, however, not obvious. In this paper it is shown that the 200 hPa windfield is the optimal windfield to choose for advecting total ozone. Ground-based data are often used for validation of satellite measurements. By advection of the satellite total ozone data to the same location and time as the ground-based measurements, validation can in principle be improved. In this paper, Brewer total ozone measured at De Bilt, TOVS total ozone and assimilated TOVS total ozone, for the pixel closest to De Bilt, are compared and the results are discussed.

Ultraviolet (UV) spectrometer of the atmosphere is becoming increasingly important for the determination of ozone concentrations from space (cf., e.g., the new satellite instruments GOME on board ERS-2 and SCIAMACHY on board Envisat-1). By measuring the shape of the ozone Huggins bands, which are situated from about 310 to 340 nm, accurate ozone column and profile detection is possible in principle. In the retrieval procedure, based on radiative transfer calculations of atmospheric UV reflectivity, polarization is usually neglected. However, this may introduce an appreciable error, because in the UV molecular (Rayleigh) scattering, which can yield a high degree of polarization, is the dominant scattering process. In this study the magnitude and spectral dependency of the reflectivity error in the Huggins bands due to neglecting polarization has been investigated. The maximum relative error in the nadir reflectivity is of the order of 5%. Furthermore, the error shows the structure of the Huggins bands, amounting to about 1% in the relative reflectivity. This has implications for the retrieval of ozone column and profile. The UV albedo has been found to be in error by at most about 2% when polarization is neglected. This is contrary to the usual assumption that angularly integrated quantities are not influenced by neglection of polarization in molecular scattering atmospheres.

The OH Interferometer Observations (OHIO) concept is an option for the generalized far infrared Fabry-Perot instrument, optimized for satellite-based measurement of the OH radical in the earth's stratosphere with the simplest possible instrument configuration. This paper gives refined design parameters for OHIO. The design presented uses entirely existing, demonstrated technology, does not require stored cryogens, and concentrates on thermal emission measurements of OH, the one stratospheric species which can be measured uniquely and well in the far infrared from a satellite. Measurements are of the F₁, 7/2⁺ yields F₁, 5/2^- transition at 118.455 cm^-1 (84.42 micrometers ), which has been demonstrated to be the best spectral feature for atmospheric measurements of OH. The current design parameters, including realistic values for Fabry-Perot transmission, detector performance, and filtering required to suppress radiation passed in the higher orders of the grating monochromator, are demonstrated to be within a factor of four of what is required for global measurements of OH. Thus, with a modest further improvements in detector performance and spectrometer design, we may soon be able to demonstrated a working concept for a potential satellite instrument.

The chemical composition of the lower stratosphere has been measured using a polarizing interferometer operating in the far infrared and submillimetric spectral region. The instrument was flown three times (in 1992, 1993 and 1994) from the NSBF balloon base (Fort Sumner, New Mexico) in coincidence with overpasses of the UARS satellite, for a total of about 50 hours of measurements. In this paper we report some of the results obtained from the data analysis made up to now.

We present results of a sensitivity analysis of CO and CH₄ observations in the infrared with the spectrometer SCIAMACHY. 2D simulations of trace gas dispersion are used to predict the concentration distributions of the two target gases. Changes in the concentrations are converted into changes in incoming flux to the instrument by simulating the photon scattering through the atmosphere by using several radiative transfer programs, and the HITRAN database. The incoming flux is transferred to measured detector pixel counts and noise employing the SCIAMACHY instrument simulation software as developed in SRON. The computed noise is translated to uncertainties in the determination of concentrations using the Cramer-Rao formalism. We show that for earth albedo values above 10% and realistic noise values the CH₄ concentrations can be determined with a precision better than 1%; for CO the error can be higher than 10% in unfavorable conditions. As a by-product, the sensitivities for measuring H₂O and N₂O are also given.

We study satellite polarization data of the clouds of Venus obtained by the Pioneer Venus Orbiter from 1978 through 1990. We present a new method for comparing these data to results of exact multiple scattering computations. This method has been applied to the analysis of a single disk distribution of the polarization at wavelengths 550 and 935 nm, using a simple model for the atmosphere of Venus. We find little variation in the cloud particle size distribution for the equatorial part of the disk. For this region, the effective particle radius is about 1.0 micrometers and the width of the size distribution decreases when approaching the terminator. However, our analysis of observations at higher latitudes suggests that for these regions a different explanation is needed. Here, an upper haze layer with smaller particles than those of the underlying cloud but having the same composition explains the observations well. The optical thickness of this haze is between 0.1 and 0.5 for an effective haze particle radius of 0.40 micrometers .

The use of accurate space-born polarimetric measurements to retrieve tropospheric aerosol characteristics is a promising remote sensing tool, but also imposes strong requirements on the atmosphere-ocean model in terms of its adequacy and on computational techniques in terms of their accuracy and efficiency. The present work is concerned with computing the reflection matrix of an atmosphere-ocean system within this context. We use the Ambartsumyan non- linear integral equation to obtain the reflection matrix for a semi-infinite homogeneous ocean body containing hydrosols. The reflection and transmission matrices of a statistically rough ocean surface are obtained using the standard Kirchhoff formulation, with shadowing effects taken into account. The reflection properties of the combined ocean body and ocean surface are obtained employing the adding method. We use the Fourier decomposition of the scattering matrices and separation of the first-order scattering to substantially reduce the computational burden. An atmospheric model containing aerosols and molecules is computed and added on the top of the ocean system using the adding/doubling method. We report preliminary computational data and discuss the variation of the degree of linear polarization of singly and multiply scattered radiation as a function of scattering geometry, surface roughness, and aerosol and molecular optical depth.

Here the goal is to determine the existing correlation between aerosol mass loading M and extinction (tau) _a((lambda) ) and in particular the possible improvement using the spectral dependence. The ratio M/(tau) _a obviously depends on the aerosol size distribution model and so we first carried out a theoretical analysis to observe this dependence. Two functions were selected, the Junge and the lognormal function. Also, ground-based sun photometer measurements under variable atmospheric conditions were made during two campaigns in April and May in both 1986 and 1987 at M'bour, 80 Km south of Dakar, Senegal. The spectral dependence of the aerosol optical thickness is used to derive the columnar aerosol size distribution, its mass loading and hence the M-(tau) _a relationship. Good correlations expressed by a power law have been established which can be used to estimate desert aerosol content within an acceptable margin of error. This associated error oscillates between +/- 8% and +/- 15%. However errors as high as 40% are reached in the estimation of M using the simple ratio formula. This improvement could be performed as well by satellite using the good spectral coverage of sensors like MODIS (EOS satellite) or MERIS (Envisat).

We propose to use MODIS Airborne Simulator data acquired during the SCAR-A experiment, that took place in July 1993 in the Eastern US, to validate the MODIS operational atmospheric correction algorithm. The first step is to calibrate the visible, near-infrared and middle- infrared channels using Rayleigh scattering in the visible for the absolute calibration, high clouds or sun glint views for the inter calibration. Calibrated data are compared with simultaneous observations from AVIRIS. The second step is to investigate an aerosol retrieval method using the 2.14 micrometers band. The split window technique applied to thermal channels at 11 and 12 microns is also used to derive the water vapor content over land. After applying atmospheric corrections, we present some of the directional features exhibited by selected targets.

POLDER is a new instrument devoted to the global observation of the polarization and directionality of solar radiation reflected by the Earth surface-atmosphere system. The original capabilities of POLDER, compared to previous current radiometers are, (1) its polarized reflectance measurements in the visible and near-infrared part of the solar spectrum, (2) its capability to measure a surface target reflectance from about ten directions during a single pass. The instrument concept has been accepted on the Japanese ADEOS platform scheduled to be launched early 1996. One of the scientific objectives defined for the POLDER satellite mission concerns the cloud characteristics. The main parameter that influences the cloud radiative properties if the optical thickness. The usual method which allows to retrieve this parameter is based on the plane-parallel approximation. It is here tested on data acquired by the airborne version of POLDER during the ASTEX (Atlantic Stratocumulus Transition EXperiment) campaign. Two cloud models are considered: the homogeneous plane-parallel model and on the other hand, a horizontally heterogeneous plane-parallel model. Bidirectional Reflectance Distribution Functions derived from these two models are compared with POLDER measurements obtained for a stratocumulus cloud situation.

The Polar Ozone and Aerosol Measurements II (POAM II) has been developed by the Naval Research Laboratory and launched end of September 1993 on the french satellite SPOT 3. The instrument observes solar occultations at 9 wavelength channels. Two channels are devoted to the retrieval of the water vapor profile, one at 935 nm in the water vapor absorption band, and a nearby channel at 920 nm, almost free of water vapor absorption. The two channels are used in a differential mode to separate the aerosol extinction from the water vapor absorption. The major difficulty to retrieve a water vapor profile from the transmission data, is due to the line structure of the absorption spectrum. Line-by-line models are used as basic benchmarks, and the GEISA and HITRAN models are compared. However the line-by-line models are much too complex to be run at each step in an inversion algorithm; a simple parameterization has been sought, following the method used for SAGE II. Results are presented.

The Stratospheric Aerosol and Gas Experiment II (SAGE II), that was launched in October 1984, has monitored the stratospheric aerosol layer after the Pinatubo's eruption. Two flights of the balloon-borne experiment RADIBAL (RADIometer BALloon) were performed in June 1992 and May 1993 in coincidence with SAGE II events. Because of the large aerosol loading, the inversion of the balloon measurements (consisting in radiance and polarization diagrams) was impracticable. A code taking into account the multiple scatterings has then been used to calculate theoretical diagrams for an aerosol model deduced from SAGE II data. The obtained diagrams have been compared satisfactorily to the experimental ones.