Evidence of the complexity of aerosol transport in the lower troposphere on the Namibian coast during AEROCLO-sA

The evolution of the vertical distribution and optical properties of aerosols in the free troposphere, above stratocumulus, is characterized for the first time over the Namibian coast, a region where uncertainties on aerosol–cloud coupling in climate simulations are significant. We show the high variability of atmospheric aerosol composition in the lower and middle troposphere during the Aerosols, Radiation and Clouds in southern Africa (AEROCLO-sA) field campaign (22 August–12 September 2017) around the Henties Bay supersite using a combination of ground-based, airborne and space-borne lidar measurements. Three distinct periods of 4 to 7 d are observed, associated with increasing aerosol loads (aerosol optical thickness at 550 nm ranging from ∼ 0.2 to ∼ 0.7), as well as increasing lofted aerosol layer depth and top altitude. Aerosols are observed up to 6 km above mean sea level during the later period. Aerosols transported within the free troposphere are mainly polluted dust (predominantly dust mixed with smoke from fires) for the first two periods (22 August–1 September 2017) and smoke for the last part (3–9 September) of the field campaign. As shown by Lagrangian back-trajectory analyses, the main contribution to the aerosol optical thickness over Henties Bay is shown to be due to biomass burning over Angola. Nevertheless, in early September, the highest aerosol layers (between 5 and 6 km above mean sea level) seem to come from South America (southern Brazil, Argentina and Uruguay) and reach Henties Bay after 3 to 6 d. Aerosols appear to be transported eastward by the midlatitude westerlies and towards southern Africa by the equatorward moving cut-off low originating from within the westerlies. All the observations show a very complex mixture of aerosols over the coastal regions of Namibia that must be taken into account when investigating aerosol radiative effects above stratocumulus clouds in the southeast Atlantic Ocean.

The ALS lidar measurements were carried out continuously between 22 August and 13 September, 2017. The data 120 coverage for aerosol study is low because of the quasi-ubiquitous presence of marine stratocumulus and fog during 121 a large part of the observation days. The fog opacity was often such that the laser beam was fully attenuated after 122 a few hundred meters. We therefore considered average profiles taken during periods when no low-level clouds or 123 fog events are observed, i.e. between about 1 and 4 hours on a given day (see Table 1). The description of the lidar 124 is given in Appendix A, together with the calibration and data inversion processing.  (Table 1). We mainly used the aerosol typing derived from CATS 167 measurements, which is similar to the one established for CALIOP. The correspondence between the aerosol 168 typing derived from CALIOP and CATS measurements are given in the Table 2. It should be noted that not all the 169 aerosol types are named exactly in the same way. An example of aerosol typing is given in Appendix A. We will only use data over the sea because Henties Bay is a coastal site affected by the sea breeze and bordered

186
The meteorological patterns are studied using Meteorological fields provided by the 6-hourly operational analyses    discrepancies on AOT may be also explained by the coarse spatio-temporal sampling of the model, which is 213 insufficient to highlight the sharp variation in AOT due to a very localized aerosol features during these 3 days.

214
As a result, even small differences in the simulation of the weather conditions could lead to substantial differences event was recorded during the field campaign, so that we can exclude any CAMS misrepresentation of wet deposition processes around Henties Bay. In addition, CAMS simulations show that the AOT is essentially due to organic matter (i.e. biomass burning aerosols), the contribution from non-biomass aerosol can then be excluded as 219 well. On 2 September a minimum in AOT is observed by the sun photometer which is not reproduced by CAMS 220 simulations (even though a local minimum in the CAMS AOT can be seen). During this day, the mid-tropospheric 221 circulation was characterised by a low-pressure system located offshore of Henties Bay, juxtaposed to a high-

222
pressure system over South Africa, resulting in a small river of smoke descending along the coast that CAMS is 223 simulating too far east over Henties Bay. On 7-8 September, the sun photometer-and MODIS-derived AOTs are 224 larger than the one computed from CAMS. This could be related to the presence of unscreened optically thin clouds 225 such as the ones observed in the ground-based lidar data on 8 September ( Figure A2d) and/or to the heterogeneity 226 of the meteorological field. Indeed, on 7-8 September, an elongated high pressure dominating over the continent,

227
led to the channelling of the smoke from the north-west that is slightly mis-located in the CAMS analyses.

228
In Figure  2

250
The increase in the lidar-derived column AOT (blue bars in Figure 2b) during P3 is also well correlated to the 251 increase of the partial column AOT in the 1500-3000 m AMSL.

252
We note a significant increase in terms of the lidar-derived thickness of elevated aerosol layer between the 3 253 periods (~1-2.5 km during P1, ~2.5-3 km during P2 and ~2.5-5 km during P3,      Table 3). The average LNG-derived 8a around 1000 UTC. Figure 9 shows the comparison between the dropsonde profiles of temperature, wind and  shown), except for an increase of RH between 5 and 6 km AMSL (by 20%, coherent with the appearance of clouds 361 as seen in Figure A2c) and of wind speed at 4.5 km AMSL (by 5 m s -1 ). Rather, the difference may be explained 362 by regional scale circulation in the mid troposphere across the area. Over the ocean, ERA5 data indicates stronger 363 northwest winds (~23 m s -1 ) at the location of the airborne lidar AEC profile compared to the wind over Henties Namibian coast and the ocean leads to differential advection within the BBA layer, and a different vertical structure 366 of the aerosol layer between the coastline and over the ocean.

383
The AEC profile '2' derived from LNG observations and obtained ~100 km north of profile '1' exhibits a different  (1) between 2.5 and 5 km AMSL increases from values below 10% to values in excess of 60% between P1 and P2,

426
which is most probably associated with the transport of BBA over Henties Bay. Likewise, the RH values between 427 5 and 6 km AMSL increases from 5% to ~70-80% between P2 and P3, which is an indication that the meteorology 428 has changed and that the origin of air masses may be different. Periods P2 and P3 are clearly separated by an episode 429 of very dry RH conditions on 2 September, the day also corresponding to a minimum of AOT over Henties Bay 430 ( Figure 2). In general, the location of the elevated aerosol layer in the vertical corresponds to the highest RH as 431 previously observed from airborne measurements. In the following, we designed back trajectories analyses to 432 investigate the origin of the air masses in the FT.

Period P2
between 1500 and 5000 m AMSL and over the ocean. The distribution of trajectories suggests that the BBA heights over Angola can reach 5 km AMSL, as suggested by the CALIOP and CATS observations (see Figure   466 A3). As for P1, we observed no significant aerosol contribution above 5 km AMSL (Figure 4).

Possible contribution to the AOT from South America during P3
is seen compared with the two other periods. Some of the aerosol layers observed during P3 between 5 and 6 km hypothesis is not necessarily verified during the studied period. Indeed, when trajectories cross the Atlantic Ocean, 504 they encounter more a baroclinic fluid than a barotropic fluid due to the presence of strong low pressure centres 505 such as the cut-off low. The potential temperature is therefore no longer necessarily a tracer of the air mass and 506 isentropic trajectories can quickly diverge towards higher altitudes. This is shown in Figure 16 on 6 September 507 (the same is true on 7 September). Nevertheless, some trajectories pass under 5 km AMSL over northern Argentina.

508
The same trajectory simulation conducted with an isobaric hypothesis on 6 and 7 September shows that all the 509 back-trajectories come from Argentina for altitudes that remain in the range of biomass burning injection heights 510 (~5 km AMSL). However, isobaric trajectories are not necessarily more representative than isentropic trajectories 511 (Stohl, 1998).

524
Over the following days, the cut-off low is seen to merge back with the westerlies while progressing eastward, and 525 the high-pressure system at 500 hPa is observed to also move over the Atlantic Ocean and merge with the St

526
Helena high on 5 September (Figure 17b). The mid-tropospheric westerly jet may transport the aerosols issued     in Table A1 where we have added the features of the LNG lidar for comparison.  in Figure A1. It can be considered that the correction of the overlap factor induces a relative error lower than 15%

881
for an overlap factor between 0.8 and 1 (Chazette, 2003), corresponding to a distance of 150 m from the emitter.

882
The molecular contribution is obtained from the Era5 pressure and temperature data at the horizontal resolution of

913
Bay site. All available CALIOP and CATS orbits passing over Namibia were analysed and the results in terms of 914 aerosol typing are given in Table 1 and Table 2. The correspondences in terms of LR are given in Table 2 for both 915 instruments.

916
In  Table 2). When there is no CALIOP or CATS overpasses we take the value of LR of the 922 nearest day also considering the shape of the AEC profile and the origin of air masses using back trajectories.

923
Values of 65-70±25 sr and 55±25 sr at 532 nm are used for the two main aerosol types sampled, namely smoke  Table 2 via a minimization of the difference of AOT between the ground-based lidar and the sun 929 photometer: the LR in the PBL is adjusted so that the AOT calculated from the lidar AEC profile matches best the 930 AOT from the sun photometer at 355 nm. The LR values obtained during the field campaign are associated with 931 clean marine air aerosols (i.e. 20-23 sr) and polluted dust (i.e. 55 sr). This was done for all days listed in Table 3,

932
with the exception of 8 and 9 September 2017. On those days, the sun photometer AOT could not be used to 933 constrain the inversion of the lidar measurements. This is likely due to the presence of unscreened clouds in the 934 sun photometer inversion (as logged by the ground-based lidar on 8 September, Figure A2d). For those two days,

935
we have used a LR of 20 sr in the PBL to be able to invert the lidar data. Note that the use of a value of 55 sr in 936 the PBL on those days (i.e. the value retrieved for the previous days) leads to an unrealistically high lidar-derived