Aerosol and cloud properties in the Namibian region during AEROCLO-sA field campaign: 3MI airborne simulator and sun-photometer measurements

. We analyse of the airborne measurements of above-cloud aerosols from the AEROCLO-sA field campaign performed in Namibia during August and September 2017. To improve the retrieval of the aerosol and cloud properties, the airborne demonstrator of the Multi-viewing, Multi-channel, Multi-polarization (3MI) satellite instrument, called OSIRIS, was deployed on-board the Safire Falcon 20 aircraft during 10 flights performed over 20 land, over the ocean and along the Namibian coast. The airborne instrument OSIRIS provides observations at high temporal and spatial resolutions for AAC and cloud properties, with well-defined uncertainties. OSIRIS was supplemented with the airborne multi-wavelength sun-photometer PLASMA2. The application of the algorithm developed for the POLDER spaceborne instrument in the visible range to the OSIRIS measurements allowed to characterise the Aerosol Above Cloud (AAC) properties. The variations of the aerosol properties are consistent 25 with the different atmospheric circulation regimes observed during the deployment. Airborne observations typically a show strong Aerosol Optical Depth (AOD, up to 1.2 at 550 nm) of fine mode particles from biomass burning (extinction Angström exponent varying between 1.6 and 2.2), transported above a stratocumulus deck (cloud top around 1 km above mean sea level) with Cloud Optical Thickness (COT) up to 35 at 550 nm. The above-cloud visible AOD retrieved with OSIRIS agrees within 10 % with the PLASMA2 sun-photometer 30 measured in the same environment. α (mean of observed above than the ones for These in the fine mode size are also from the analysis of the fine mode properties from inversions. Lower values of the volume mean radius of the fine mode particles are retrieved on and 12 September (r v = 0.18 and 0.15 µm, = 0.20 region. Nonetheless, many unresolved questions remain on the transport and ageing of biomass burning particles in the region. A synergistic analysis of multi-campaign data will benefit this analysis. The high spatial resolutions, 455 offered by the airborne polarimeter OSIRIS and Lidar LNG data acquired during the AEROCLO-sA campaign, also offer the opportunity to improve retrievals of the heating rate profiles, based on the satellite method described in Deaconu et al (2017). These retrievals would benefit from the aerosol-cloud interactions studies in the South- eastern Atlantic region.


Introduction
Aerosols from natural and anthropogenic sources directly impact the climate by interacting with solar and telluric radiations and indirectly through interactions with cloud properties (IPCC, 2013). According to their origin and 45 the atmospheric transport, aerosol particles are unequally distributed in the troposphere where they can reside for several days or weeks. As a consequence, their chemical, optical and microphysical properties also present a strong variability (Lagzi et al., 2014). on the direct radiative effect calculation. In the same region, Peers et al. (2016) reported disagreements between five AeroCom models and satellite observations of AAC from the Polarization and Directionality of Earth Reflectances (POLDER) instrument. These authors explained the differences in the models by the use of different parametrizations for the aerosol injection heights, vertical transport and absorption properties.
These considerations have motivated a significant number of intensive field campaigns between 1992 and 2018 70 . As a matter of fact, the South-East Atlantic Ocean indeed represents a unique opportunity to study aerosol-cloud-radiation interactions and the absorption properties of biomass burning aerosols, which are still debated. Pistone et al. (2019) obtain Single Scattering Albedo (SSA) values for biomass burning aerosols from both airborne in situ and remote sensing methods during the Observations of Aerosols above Clouds and their Interactions (ORACLES) airborne campaign performed close to the Namibian coast in August-September 2016.

75
From a sweeping view mode imager (the Airborne Multi-angle SpectroPolarimeter Imager, AirMSPI), mean SSA values at 550 nm were observed between 0.83 and 0.89 in August-September 2016. Mean SSAs ranging between 0.82 and 0.92 (from June to October 2006) were also retrieved at 550 nm over a large region centred on the Southeast Atlantic Ocean using POLDER (Peers et al., 2016). A previous study based on AERONET retrievals (Eck et al., 2013) has shown similar SSA values during the fire season. The latter study demonstrated that the seasonal 80 trend of SSA in this region was mainly due to a change on aerosol composition, and particularly on the black carbon fraction.
Because the simultaneous retrieval of aerosol and cloud properties is still challenging (Cochrane et al., 2019), the European Space Agency (ESA) and EUMETSAT developed a new spaceborne Multi-viewing, Multichannel, Multi-polarization Imager (3MI) to be launched in 2022 on-board the METOP-SG satellite. To evaluate the next 85 generation of retrieval algorithms, a 3MI airborne prototype, OSIRIS (Observing System Including PolaRisation in the Solar Infrared Spectrum; Auriol et al., 2008), has been developed at Laboratoire d'Optique Atmospheric (LOA, France). The total and polarized radiances sampled by OSIRIS between 440 and 2200 nm, and the new retrieval algorithms developed by Waquet et al. (2013) and Peers et al. (2015), allow to simultaneously retrieve the aerosol and the cloud properties in case of aerosols above clouds. Additionally, polarimetric measurements 90 constitute a promising opportunity for the simultaneous retrieval of the aerosol and the surface properties (Dubovik et al., 2011(Dubovik et al., , 2019. Therefore, the new observation capabilities proposed by the airborne instrument OSIRIS give an interesting opportunity to constrain climate models and satellite retrievals in a climate-sensitive region where high absorbing aerosol load coexist with a stratocumulus cloud deck (Mallet et al., 2019).

95
In this paper we present the OSIRIS measurements of AAC collected during the AErosols, RadiatiOn and CLOuds in southern Africa (AEROCLO-sA) field campaign in Namibia during the biomass-burning period in 2017 . The OSIRIS instrument was supplemented by the airborne sun-photometer PLASMA2 as well as the Lidar pour l'Etude des Interactions Aérosols Nuages Dynamique Rayonnement et du Cycle de l'Eau (LEANDRE, Pelon et al., 1990) Nouvelle Génération (LNG, Bruneau et al., 2001). The present paper reports on 100 optical and radiative properties of both aerosols and clouds retrieved during the AEROCLO-sA campaign close to https://doi.org/10.5194/acp-2020-992 Preprint. Discussion started: 16 October 2020 c Author(s) 2020. CC BY 4.0 License.
the Namibian coast, both over land and sea. Section 2 describes the flight trajectories and the main meteorological conditions encountered during the campaign. In Sect. 3, the OSIRIS, the airborne and ground-based sunphotometer and the airborne Lidar LNG retrieval methods are described. Section 3 reports the mean aerosol and cloud properties retrieved in the Namibian region. Finally, the results are summarised and discussed in Sect. 4.

Flights patterns and general atmospheric circulation description.
The AEROCLO-sA deployment comprised the measurements from the ground based station of Henties Bay, Namibia (22°6'S, 14°30'E; 20 m above sea level (a.s.l.)) from 23 August to 12 September 2017 .
The airborne component was conducted by the SAFIRE Falcon 20 aircraft from 5 to 12 September 2017. Ten 110 flights were performed over ocean and land in the area from 7°30'E to 20°E and from 17°S to 22°30'S ( Fig. 1). Two pre-campaign flights (not presented here) were also performed over the Mediterranean Sea during the summer 2017. Those flights correspond to pristine conditions over both clear and cloudy ocean scenes and were used for instrumental calibration.
The locations of the OSIRIS observations used in the present study are shown in Fig. 1a. Several filters are applied 115 to the OSIRIS measurements to ensure optimum conditions for the retrieval (i.e. homogeneous clouds fields and stable flight conditions), which slightly reduces the number of data (by 35%). As a first quality assurance, only stable flight conditions at high altitude (higher than 8 km a.s.l.) are selected. In addition, high altitude clouds and heterogeneous cloud scenes are rejected. In these conditions, a total of 2h15 of OSIRIS measurements were processed. Figure 1b also represents the PLASMA2 measurements used in the present study which corresponds 120 to low flight levels and stable flight conditions. Data from the Copernicus Atmospheric Monitoring System (CAMS) reanalysis (Flemming et al., 2017) allow to obtain the biomass burning plume trajectory at 6-hour resolution at a 0.75°x0.75° spatial resolution. In Fig. 2, the atmospheric circulation and the associated biomass burning plume during the field campaign are represented by the geopotential height and wind at 700 hPa and the biomass burning AOD at 550 nm. The plume path is displayed 125 by highlighting the wind vectors at grid points with AOD higher than the 90 th percentile of the AOD in the region (red arrows).
The regional atmospheric circulation on 8 September ( Fig. 2a) represents the mean circulation during this period with air masses coming from tropical Africa, moving westward until Ascension Island which are then deflected to the South-East due to the anticyclone centred over southern Africa. These conditions were observed during most 130 of the AEROCLO-sA campaign except on 5 September (Fig. 2b), when the anticyclone was centred over the Indian Ocean between the South African coast and Madagascar. During this specific day, air masses were mostly transported over the continent and dust emissions were reported from both the climate model and airborne the lidar measurements .

The 3MI airborne prototype: OSIRIS
The OSIRIS imager provides both total and polarized radiances measurements. The airborne instrument is characterized by two optical systems: one for the visible and near infrared range (VIS-NIR, from 440 to 940 nm) with a wild field-of-view of 114° and one for the shortwave infrared (SWIR, from 940 to 2200 nm) with a field-140 of-view of 105°. The 2D detectors, which are respectively a CCD matrix of 1392x1040 pixels and a mercury cadmium telluride (MCT) focal plane array of 320x256 pixels allow to obtain very high resolution images with a spatial resolution of 20 m for the VIS-NIR detector and around 60 m for the SWIR one at an height of 10 km. This high resolution allows to record the same scene up to 16 times (at 10 km height) from different viewing angles. Polarized measurements are available at 440, 490, 670, 870, 1020, 1600 and 2200 nm. Measurements without 145 polarization capabilities are also performed in molecular absorption bands (763, 765, 910, 940, 1365 nm), and for a channel centred at 1240 nm, in addition to the channels previously listed.
The aerosol and cloud retrievals are performed using the OSIRIS measurements at 670 and 865 nm. A moving average is applied on measurements over few pixels before the retrieval is achieved. The related radiometric noise is then estimated to be lower than 5.10 -4 and 5.10 -3 for the total and polarized normalized radiances, respectively. 4 calibration are expected to be lower than 3% for these channels. The absolute calibration accuracy was improved for the visible radiances using in-flight calibration technics (Hagolle et al., 1999) applied to OSIRIS measurements acquired over the Mediterranean Sea.
The algorithm used to retrieve the AAC properties with OSIRIS is based on an Optimal Estimation Method (OEM) 155 developed for the POLDER instrument (Waquet et al., 2013). This method allows to simultaneously retrieve the aerosol and the cloud properties (Waquet et al., 2013 andPeers et al., 2015). Here, the aerosol retrieval is performed using the measurements in the solar plane of each image. The aerosol properties are then assumed to be spatially homogenous over the entire OSIRIS visible image (of about 20  20 km 2 ). Finally, the cloud properties are retrieved pixel by pixel over the entire image. As demonstrated in Waquet et al. (2013), this procedure increases 160 the sensitivity of the algorithm to the aerosol properties.
The algorithm mainly provides AOD, SSA, extinction Angström exponent () and the Cloud Optical Thickness (COT). SSA is defined as the ratio of the scattering to the extinction coefficient and primarily depends on the aerosol absorption (i.e. the imaginary part of the complex refractive index) and also the particles size (Dubovik et al.,1998 ;Redemann et al., 2001). The Angström exponent is indicative of the particles size (Reid et al., 1999;165 Schuster et al., 2006). The retrieved quantities are used to compute the instantaneous DRE over the solar spectrum, considering two main assumptions on the aerosol microphysics: (1) the complex refractive index of aerosols is assumed spectrally invariant; the imaginary part is retrieved and the real part value is fixed to 1.51, which is a reasonable average value for biomass burning aerosols (Dubovik et al., 2000) and (2) the particles size is retrieved only for the fine mode (particle diameters below 1 µm). The main source of uncertainties on the OSIRIS aerosol 170 retrievals is the assumption made for the real refractive index (Waquet et al., 2013a). The calculations were perturbed with different values of the real refractive index (1.51 ± 0.04) and additional error terms were added to the retrieval errors provided by the OEM (Waquet et al., 2013). Moreover, the polarized measurements acquired for scattering angles larger than 130° are sensitive to the 3D cloud geometry effects on radiative transfer (Cornet et al., 2018). Since clouds are assumed to be plane parallel in the simulations, the method is simply applied to 175 observations acquired for scattering angles smaller than 130°.

Figure 3
shows an example of the measured and modelled radiances after the convergence is reached. It shows that the method allows to robustly model the selected data within the measurements noise. For these values of scattering angles, the sensitivity of polarization to cloud microphysics is minimized and the cloud droplet effective radius is assumed to be equal to 10 microns, which is the mean value for the stratocumulus clouds observed over 180 the domain of interest (Deaconu et al., 2019).

The airborne Sun-photometer: PLASMA 2
PLASMA 2 (Photomètre Léger Aéroporté pour la surveillance des Masses d'Air version 2, Karol et al., 2013) is an airborne Sun-tracking photometer (referred as #950 on AERONET) which was on-board the SAFIRE Falcon 20 during the AEROCLO-sA campaign. The AOD of the atmospheric column above the aircraft is retrieved at 185 nine wavelengths (340, 379, 440, 500, 532, 674, 871, 1020 and 1641 nm) from the PLASMA2 measurements. PLASMA (versions 1 and 2) observations have been validated against in situ, AERONET and satellite measurements (Mallet et al., 2016;Rivellini et al., 2017;Torres et al., 2017;Formenti et al., 2018;Hu et al., 2018), indicating that the accuracy on the AOD retrievals is of the order of 0.01 regardless of the wavelength. During AEROCLO-sA, several low-level flights were performed, typically in overcoast conditions over the ocean to 190 retrieve the AAC optical properties and in clear sky conditions over land. The fraction of fine and coarse mode AOD is derived from the large spectral range of PLASMA2 using the GRASP algorithm (Dubovik et al., 2014, Torres et al., 2017. GRASP also allows to retrieve the volume size distribution using an assumed complex refractive index (i.e. 1.50+0.025) and assuming a bimodal lognormal particles size distribution with fixed modal widths.

Airborne lidar LNG
The vertical structure of the aerosol and cloud layers below the aircraft was obtained from the nadir-pointing airborne lidar LNG. The signal backscattered to the LNG system telescope at 1064 nm is range-square-corrected to produce atmospheric reflectivity. Total attenuated backscatter coefficient (ABC) profiles are derived from atmospheric reflectivity profiles by normalizing the atmospheric reflectivity above the aerosol layers to the 200 molecular backscatter coefficient profiles. Hence the slope of the lidar reflectivity above 6.5 km a.s.l. matched that of the molecular backscatter derived from dropsonde measurements of pressure and temperature. The vertical resolution of the ABC profiles is 30 m and profiles are averaged over 5 s, yielding a horizontal resolution of 1 km for an aircraft flying at 200 m s -1 on average. It is worth noting that ABC as observed with LNG is sensitive to both aerosol concentration and aerosol hygroscopicity. Indeed, relative humidity in excess of 60% modify the size 205 and the complex refractive index of aerosol, and hence their optical properties, enhancing the ABC (e.g. Randriamiarisoa et al., 2006).

Ground-based AERONET sun-photometer measurements.
To put the aircraft observations into context, we analysed observations form 15 August to 15 September 2017 from the 15 AERONET stations located in the South-eastern Atlantic region. The full column integrated properties 210 (i.e. AOD, complex refractive index, extinction Angström exponent, SSA and AOD fine mode fraction) have been averaged for each site and are shown in Fig. 4. Four AERONET stations are located in the AEROCLO-sA flight domain: Windpoort, Henties Bay, Gobabeb, as well as HESS, located in the southern part of the domain, 200 km south-East of Henties Bay, and outside of the flight tracks. During the campaign, and due to the persistent cloud cover, the measurements available at Henties Bay and Gobabeb were sparse. Therefore, the Windpoort station, 215 located at 250 km from the Namibian Coast, is a more suitable station to monitor the aerosol evolution during the campaign. Additional interesting comparison data are provided by the Namibe station, located in the northern part of the AEROCLO-sA region, and more influenced by biomass burning emissions from central Africa than the Windpoort site (therefore exhibituing higher AODs). Outside the AEROCLO-sA domain, and as shown by the atmospheric circulation patterns in Fig. 2, biomass burning plumes were often transported towards Ascension

220
Island. This remote location (i.e. 3000 km offshore the Angola coast) gives an opportunity to study the evolution of the biomass burning aerosols during their transport and aging (Zuidema et al., 2016;Mallet et al., 2019). The Aerosol Backscatter Coefficient (ABC) at 1064 nm on 12 September 2017 obtained from the airborne Lidar is shown Fig. 6a. The aerosol signal is mainly concentrated between the stratocumulus top at around 1 km and 6 km height. This vertical distribution well represents the general condition during the AEROCLO-sA campaign (Chazette et al., 2019). For this same day, Fig. 6b shows the spectral AOD measured by OSIRIS, PLASMA2 and 240 the ground-based Windpoort and Namibe AERONET stations. The OSIRIS above-cloud retrievals were performed using measurements at an altitude of about 9 km a.s.l., corresponding to the top of the descent in loop. Contrary to the configuration of OSIRIS on the SAFIRE Falcon 20 aircraft, PLASMA2 is an upward looking instrument. Therefore, the above-cloud aerosol properties from PLASMA2 were obtained from the measurements at the bottom of the descent in loop, above the cloud top. AOD retrieved from OSIRIS and PLASMA2 245 measurements agree within ± 10% at 670 nm (Fig. 6b, Table 1), and are in between the measurements of the two AERONET stations. Bias between PLASMA2 and AERONET AODs are around 70% for every wavelengths, whereas OSIRIS measurements agree with AERONET AODs from 20% with Windpoort measurements at 870 nm to 67% with Namibe measurements at 670 nm. Table 1 shows this same comparison from four loops, showing a very good agreement between OSIRIS and 250 PLASMA2 for variable AOD conditions (from 0.36 and 0.74 at 670 nm). This indicates that the OSIRIS abovecloud AOD retrieval is robust independently on the aerosol loading. Note that, on 5 September, the aircraft did not reach the cloud top level at the end of the loop, which causes a higher bias (30%) between the AOD from PLASMA2 and OSIRIS's.

255
The 40-minute slot represents a transect of about 400 km along the coast-line (in yellow Fig. 1a) with a north cape. A slight variability is observed on the AOD with a south-north gradient. 30 minutes after the segment at high altitude, the collocated observations from PLASMA2 at low altitude are consistent with OSIRIS, with the AOD from PLASMA2 being larger by 10% than OSIRIS. Parallel measurements of the airborne lidar LNG demonstrate the small spatial variability observed on 12 September with homogeneous distribution of aerosols in the 260 troposphere with low aerosol signal (ABC below 1.10 -6 m -1 sr -1 ).

Angström exponent and particle size distribution
Figure 8a describes the volume particle size distributions retrieved from PLASMA2 measurements at different altitudes during the descent in loop on 7 September 2017. The AERONET particle size retrievals from the 270 Windpoort station are also shown. The size distribution is generally characterised by a dominant fine mode between 500 m and 5000 m. This is consistent with the dominant fine mode typically observed in previous studies at an altitude of 1 to 6 km in this region (Toledano et al., 2007;Russell et al., 2010;Kumar et al., 2013). Measurements show a rather constant fine-coarse mode ratios within the aerosol plume. For the descent in loop over cloud on 7 September 2017, fine mode particles contributed to 97 % of the total AOD at 670 nm between 275 1000 m and 4000 m. The mean Angström exponent value obtained from PLASMA2 measurements is about 1.9 with an accuracy of 0.1. According to AERONET measurements, during the campaign period, the amount of coarse mode particles is also extremely weak and does not exceed 5% of the columnar AOD (Fig. 4). The smallest particles (Angström exponent larger than 2) are generally observed in the northern part of the Namibian region (Fig. 4). is characterised by an extremely dry surface at this time of year. In these conditions, the particles size distribution retrieved by PLASMA2 over the South-western part of the flight, over the Etosha Pan, is characterized by a significant additional contribution of coarse mode particles (α440-870 = 1.55). Toward the North-Eastern part of the flight, the aerosol properties retrieved by PLASMA2 are less influenced by the dust emissions from the pan, and are more representative of the properties of fine mode particles from biomass burning (α440-870 = 2.27).

290
The values of Angström exponents retrieved above clouds from OSIRIS and PLASMA2 are around 2.0 ± 0.2. In agreement with the PLASMA2 analysis, the α490-870 histogram shown in Fig. 8e indicates different aerosol types. The maximum α490-870 is observed on 12 September (median of 2.15), and the minimum on 8 September (median of 1.75). α490-870 values are generally constant for every flight, as in the case on 12 September (Fig. 7a). The lowest α490-870 values, around 1.7, are found less than 50 km away from the Namibian coast and were observed during the 295 flight on 8 September. This behaviour might be explained by the influence of dust particles generated on the continent with higher coarse mode particle fraction. The hygroscopic growth of biomass-burning particles potentially occuring during their transport over the ocean might also explain the differences observed in the aerosol optical properties from one flight to another. The aerosols observed on 5 September were directly transported from Central Africa without a long transport over the Atlantic Ocean (Fig. 2b). This specific circulation might explain 300 the higher α490-870 (mean of 1.91) observed above clouds during this flight, which indicates smaller particles than the ones observed for instance on 8 September. These changes in the fine mode size are also suggested from the analysis of the volume fine mode properties obtained from PLASMA2 inversions. Lower values of the volume mean radius of the fine mode particles are retrieved on 5 and 12 September (rv = 0.18 and 0.15 µm, respectively) rather than on 8 September (rv = 0.20 µm). AERONET Angström exponents are around 5 % lower than PLASMA2 measurements and from 8 to 25 % lower than OSIRIS inversions ( Table 1). The bias between AERONET and OSIRIS Angström exponents can mainly be explained by the presence of coarse mode particles that are not explicitly taken into account in the POLDER algorithm when biomass burning layers are detected above clouds (Waquet et al., 2013a). This is particularly observed for the flight on 12 September due to a slightly less dominant fine mode compared to other flights (i.e. 310 mean fine mode fraction of 90% on 12 September instead of 95% on average for the field campaign). This neglected coarse mode could also explain why OSIRIS slightly underestimated AOD compared to PLASMA 2 on 12 September (Fig. 7a).

Complex refractive index and single scattering albedo
The analysis of AERONET measurements over the biomass-burning period also provides us with a spatial 315 distributed view of the aerosol complex refractive index and single scattering albedo. Values of SSA at 675 nm (hereafter SSA675) are generally spatially homogeneous over land with a mean value of 0.85 (25 th and 75 th percentiles of 0.84 and 0.86, respectively). The highest SSA675 is observed at the Namibe station (Fig. 4) with a mean value of 0.87, while the lowest values are observed at the Bonanza and Mongu Inn stations (mean SSA675 of 0.82). In correspondence, a mean refractive index of 1.51 + 0.027i at 675 nm is retrieved. The real part of the 320 refractive index at 675 nm ranges from 1.41 at Henties Bay to 1.54 at Bonanza, and the imaginary part at 675 nm ranges from 0.008 at Henties Bay to 0.032 at Bonanza. The known environmental characteristics of Henties Bay, a coastal site with high content of sea salt and sulphate aerosols, frequent fog and a persistent as well as elevated relative humidity Klopper et al., 2020), support the low values of the real and imaginary refractive indices. Figure 4 also shows different behaviours on the spectral variation of the imaginary part of the 325 refractive index k in the South-Eastern region. k441 is higher than k675 (up to 43%) on the northern part of the region, and lower (up to 10%) on the southern part of the region. The largest ratio between k441 and k675 is observed at the Namibe station with k441 higher than k675 by more than 40%.
The SSA of AAC can also be observed from OSIRIS inversions for the entire AEROCLO-sA flight campaign.   (Peers et al., 2016). The SSA from OSIRIS are less than 1% different from the ORACLES-2016 AirMSPI observations. Both measurements are consistent with the multi-year average SSA from POLDER during the biomass burning period 340 (less than 2% difference), as expected as these estimates are all based on polarimetric measurements. A higher bias (about 3%) compared to OSIRIS inversion is observed with AERONET retrievals at Windpoort, where the SSA is lower irrespectively of the wavelength. On the other hand, the measurements at the Namibe AERONET station are consistent with our retrievals at 670 nm, but not at 440 nm, where the SSA is significantly lower. This trend does not appear in the ORACLES measurements from AirMSPI, which were performed farther from the 345 coasts over the ocean. According to Kirchstetter et al. (2004), a decrease of the SSA at 440 nm can be partly explained by the presence of light-absorbing organic carbon (brown carbon). Unlike at the Namibe site, the SSA at Windpoort, located farther from the fire sources, was generally found to decrease from 440 to 875 nm.

Integrated water content
The integrated water vapour can be derived from extinction measurements at 940 nm (Halthore et al., 1997). In 350 Fig. 10, one can note the linear relationship between the water vapour content and the AOD at 550 nm from PLASMA2, especially for the higher range of water vapour concentration. The highest column concentrations of water vapour (up to 2.4 g cm -2 ) are observed for the two flights of 8 September. This correlation could be explained by the meteorological conditions, which would be responsible for the simultaneous transport of aerosols and water vapour (Adebiyi et al., 2015, Deaconu et al., 2019, or by the wood moisture emitted in tropical area, biomass 355 burning can also generate water vapour (Betts and Silva Dias, 2010;Sena et al., 2013). The variability observed in the relationship between AOD and water vapour through PLASMA2 measurements could also be the result of https://doi.org/10.5194/acp-2020-992 Preprint. Discussion started: 16 October 2020 c Author(s) 2020. CC BY 4.0 License. different emission regimes observed during the various flights. On the opposite, no correlation is observed for the second flight of 12 September with lower water vapour content (below 1 g cm -2 ). This particularity could be explained by the in-land location of the measurements performed above the Etosha Pan associated with dryer air 360 masses.

Cloud properties
The OSIRIS measurements also provide us with the optical properties of clouds with a high spatial resolution. For each pixel of the OSIRIS CCD matrix, the Cloud Optical Thickness (COT) can be retrieved at a spatial resolution reaching 20 m for a flight altitude around 10 000 m a.s.l.. Figure 11a shows a retrieved field of COT on 7

365
September at 09:37 UTC at the top of the first ascent of the flight close to the coast. COT values at 550 nm range from 5 to 30 with a mean value of 16.
To analyse the full AEROCLO-sA dataset, we chose to select the central pixel of every OSIRIS CCD matrix. This selection allows us to get rid of the scene details of each OSIRIS measurements and is well representative of the COT distribution in the region. As all the flights were performed between 8:00 and 11:00 local time, measurements 370 refer to similar atmospheric thermodynamics and sun conditions. Figure 11b shows the distribution of COT values at 550 nm for the whole campaign. The COT at 550 nm ranges from 5 to 35. On 5 and 8 September, a mean COT of 11 ± 4 is obtained while the mean COT is 15 ± 4 on 9 September and 19 ± 6 on 7 and 12 September. Mean values are similar for each flight but particularly high values are observed on 7 and 12 September. As shown in Fig. 11c, COT can be related to the distance from the coastline. An increase from 10 to 30 is observed from the 375 Namibian coast up to 100 km off the coast. On 7 September, measurements were performed up to 350 km away from the coastline. For this flight, a maximum value of COT is observed around 100 km away from the coast and then, the COT decreases down to 10 at around 250 km. It was noted, during the field campaign, that the clouds were generally optically less thick in the vicinity of the Namibian coast and more difficult to forecast.

Direct radiative effect 380
The Direct Radiative Effect (DRE) calculations are performed over the solar spectrum (0.2-4 microns) with the radiative transfer code GAME (Dubuisson et al., 1996). We follow the procedure described in Peers et al. (2015), with the only differences that the DRE calculations are performed online and are based on OSIRIS retrievals in the visible and near-infrared ranges. Considering the time period and location, a tropical model was assumed for the vertical atmospheric profile (McClatchey, 1972).

385
The distribution of the observation-based, instantaneous and one-dimensional AAC DRE in the solar spectrum from OSIRIS measurements is shown in Fig. 12 for every flight during AEROCLO-sA. Calculations show positive DRE in agreement with above-cloud observations from Zhang et al. (2016) in regions influenced by biomass burning emissions. Keil and Haywood (2003) also estimated a mean value of above-cloud DRE of +11.5 W m -2 above cloud and of -13 W m -2 in clear sky conditions (mean SSA of 0.90 at 550 nm).

390
The above-cloud instantaneous DRE obtained from OSIRIS measurements ranges from + 10 to + 200 W m -2 with mean values between + 61 W m -2 on 12 September and +105 W m -2 on 7 September. The mean DRE for the full campaign is +85 W m -2 . De Graaf et al. (2019b) compared the observations from SCIAMACHY (Scanning Imaging Absorption Spectrometer for Atmospheric Chartography, Bovensmann et al., 1999), POLDER (Peers et al., 2015) and OMI-MODIS (De Graaf et al., 2019a) satellite retrievals in the South Atlantic region. For the larger 395 area (10°N-20°S, 10°W-20°E), these authors reported that the average August 2006 DRE was +39.5, +34.5 and +47 W m -2 , respectively. Meyer et al. (2015) also reported DRE through multi-spectral measurements of MODIS in the region. Their work highlights the significant effects of the assumed aerosol model in radiative retrieval by up to +10 W m -2 in the region. Hence, the DRE retrievals are mainly affected by the absorption property of the assumed aerosol model (Zhang et al., 2016). The larger DRE retrieved during AEROCLO-sA is thus correlated to 400 strong absorption properties, to high aerosol loading and to high cloud albedo observed during this period.

Conclusion
A new set of data of cloud and above-cloud aerosol properties was acquired during the AEROCLO-sA campaign over Namibia during the biomass-burning period of 2017 by the OSIRIS radiometer deployed during ten scientific flights from 5 to 12 September 2017. Five flights are selected for OSIRIS analyses presenting stable cloud 405 conditions. Aerosol and cloud properties at a 20-meter resolution with well-quantified uncertainties are retrieved by radiometric inversion. https://doi.org/10.5194/acp-2020-992 Preprint. Discussion started: 16 October 2020 c Author(s) 2020. CC BY 4.0 License.
Measurements were performed in an extreme environment with high loading of biomass burning particles transported above a semi-permanent stratocumulus cloud deck. The high aerosol load (above-cloud AOD around 0.7 at 550 nm) retrieved from OSIRIS measurements matches the direct AOD measurements of the airborne Sun-410 photometer PLASMA2 within 10%, which contribute to validate the application of aerosol above cloud algorithms on airborne polarimeter. Biomass burning aerosol layers are mainly composed of fine mode particles, with Angström exponents varying between 1.6 and 2.2, depending on the plume history and intrinsic properties. Ground-based AERONET measurements on the continent generally show slightly lower values of Angström exponent (close to 1.7 on average), which is likely due to the presence of additional coarse particles located in the 415 boundary layer. Another reason which could explain differences on Angström exponent values is the hygroscopic effects influencing particle size.
Biomass burning aerosols transported over the South-eastern Atlantic Ocean represent a high absorbing effect which significantly impacts the climate. Mean above-cloud SSA of 0.87 at 550 nm were obtained from OSIRIS measurements. Rather constant aerosol absorbing properties were observed in the Namibian region during the 420 campaign in the visible range of OSIRIS. This effect can be due to competition between different effects as the general atmospheric circulation, the emission index and ageing processes. AERONET retrievals indicate an increase of the aerosol absorption at 440 nm over some specific sites, which suggests a modification of the plume composition, in particular regarding the ratio between the light-absorbing organic (brown) and soot (black) carbon. Brown carbon should thus be considered for the modelling of the absorption properties of the smoke plumes 425 observed closer to the fire regions. Additional measurements acquired in the UV band by the airborne micropolarimeter Ultra-Violet (MICROPOL), also operated on the Safire Falcon-20 during AEROCLO-sA, will be included in the OSIRIS algorithm to further investigate this possibility. A significant part of climate uncertainty in the South-eastern Atlantic region is also explained by the lack of measurement of stratocumulus cloud properties. The characteristics of the airborne imager OSIRIS give the 430 opportunity to retrieve cloud properties with a high spatial resolution. During the AEROCLO-sA campaign, a mean COT of 10 at 550 nm was observed. Measurements up to 350 km off the coastline allow us to probe stratocumulus property in different conditions. Results showed a maximum COT of 30 reached at 100 km off the coastline. The polarized measurements acquired by OSIRIS at 1620 and 2200 nm could be included in the retrieval in order to further characterise the cloud microphysics under different conditions in the region.

435
As a result of the elevated above-cloud AODs, the strong absorption by aerosols, and the high cloud albedo, significant positive instantaneous aerosol direct radiative effects in the solar spectrum were observed close to the Namibian coast. The mean AEROCLO-sA instantaneous DRE value is +85 W m -2 for AAC. Our observations are in agreement with previous studies indicating a strong positive aerosol forcing over the region (De Graaf et al., 2019b with possible feedbacks on cloud development due to both aerosol and water vapour combined 440 radiative effects (Deaconu et al., 2019). However, the aerosol DRE on the AEROCLO-sA region is higher than the mean DRE observed in the southern Atlantic region, mainly because of the exceptional atmospheric conditions sampled during the flights (high loads of absorbing aerosols and high cloud albedo). As demonstrated by Cochrane et al. (2019) based on the ORACLES campaigns in 2016 and 2017, the DRE also strongly depends on the cloud scene, in particular on the cloud albedo. Field campaigns in the southern Atlantic region between 2016 and 2018 445 have shown a high variability of the cloud albedo due to the heterogeneity of the cloud fraction, the cloud droplet size and the cloud optical thickness. Spatial resolution used by observations also have significant impacts on the retrievals of cloud properties and consequently on radiative properties (De Graaf et al., 2019a).
In conclusion, the airborne multi-viewing, multi-channel, multi-polarisation measurements in the region allow us to improve the definition of above-cloud aerosol properties in the region and estimate climate response to the 450 presence of high absorption particles above cloud. Mean aerosol and cloud properties established from airborne remote sensing measurements during the AEROCLO-sA campaign are valuable to constrain climate models and satellite retrievals of the aerosol-cloud-climate interactions in this region.
Nonetheless, many unresolved questions remain on the transport and ageing of biomass burning particles in the region. A synergistic analysis of multi-campaign data will benefit this analysis. The high spatial resolutions, 455 offered by the airborne polarimeter OSIRIS and Lidar LNG data acquired during the AEROCLO-sA campaign, also offer the opportunity to improve retrievals of the heating rate profiles, based on the satellite method described in Deaconu et al (2017). These retrievals would benefit from the aerosol-cloud interactions studies in the Southeastern Atlantic region. AC acknowledges additional financial support provided by the Programme national de Télédétection Spatial (PNTS, grant n° PNTST-2020-06).
Airborne data was obtained using the aircraft managed by SAFIRE (www.safire.fr), the French facility for airborne research, an infrastructure of the French National Center for Scientific Research (CNRS), Météo-France and the 470 French National Center for Space Studies (CNES). The AEROCLO-sA database is maintained by the French national data center for atmospheric data and services AERIS.
The strong diplomatic assistance of the French Embassy in Namibia, the administrative support of the Service Partnership and Valorisation of the Regional Delegation of the Paris-Villejuif region of the CNRS, and the cooperation of the Namibian National Commission on Research, Science and Technology (NCRST) were 475 invaluable to make the project happens.

Author contributions
AC performed simulations, analyses of airborne and AERONET data, and write the manuscript under the supervision of FW. PF, FW, CF, and MM, designed the original AEROCLO-sA observational concept, and coled the 5-year investigation. FA, LB, CD, RL, and JMN developed, calibrated and assured high quality 480 measurements of the OSIRIS instrument during the campaign. LB assured PLASMA2 calibration, settings and data processing. CD assured PLASMA2 measurements during the campaign. FW and FPe developed POLDER algorithms. PG and FPa co-led the development of the OSIRIS and PLASMA2 instruments. OD and BT developed and allow access of the GRASP algorithm. MG established CAMS simulations. CF provided LNG extinction profiles. Every co-authors contributed to the scientific analysis and to the writing of the manuscript.

Competing interests
PF is guest editor for the ACP Special Issue "New observations and related modelling studies of the aerosolcloud-climate system in the Southeast Atlantic and southern Africa regions". The remaining authors declare that they have no conflicts of interests.

490
We thank the AERONET PIs Brent Holben, Jens Redemann, Carlos Ribeiro, Nichola Knox, Stuart Piketh, and their staff for establishing and maintaining the AERONET sites used in this investigation.
The AEROCLO-sA data are available in the BAOBAB platform available on https://baobab.sedoo.fr/AEROCLO/ and maintained by the French national data center Data Terra/AERIS. In particular, the dataset used in this paper are: OSIRIS (