2.1 Lidar retrievals of CCN and INP concentrations

The triple-wavelength polarization Raman lidar BERTHA (Backscatter, Extinction, lidar Ratio, Temperature, Humidity profiling Apparatus) described in Haarig et al. (2017a) measures the backscatter coefficient at three wavelengths (355, 532, and 1064 nm) and the extinction coefficient at 355 and 532 nm and in a new configuration also at 1064 nm (Haarig et al., 2016). The depolarization ratio is measured at 355, 532, and 1064 nm simultaneously. The inelastic (Raman) channels at 387 and 607 nm enabling the independent measurement of the extinction coefficient (Ansmann et al., 1992) can be used at night only. The sunlight causes an enhanced background noise level due to the broad interference filter (3 nm width). In order to have coincident observations of lidar and aircraft, daytime measurements had to be used in the present study. The closest nighttime observation (always from the same date) provided the extinction-to-backscatter ratio (lidar ratio) to constrain the daytime observations. For daytime measurements, the calculation of the backscatter coefficient from the elastic signals at 355, 532, and 1064 nm was performed via the Fernald–Klett method (Klett, 1981; Fernald, 1984).

The conversion from backscatter coefficient and particle linear depolarization ratio (PLDR) to particle number and surface-area concentration follows the method described in Mamouri and Ansmann (2015, 2016). The particle depolarization ratio is used to separate the contributions of different aerosol types to the backscatter coefficient: mineral dust (d) with a high depolarization ratio (around 0.3), marine aerosol (m) with a low depolarization ratio in the humid state (≤0.05), and continental aerosol (c) with a low depolarization ratio (≤0.05). By multiplication with an appropriate extinction-to-backscatter ratio (lidar ratio; Müller et al., 2007; Groß et al., 2013; Baars et al., 2016), the backscatter coefficients are converted into extinction coefficients (see also description in Marinou et al., 2019). These extinction coefficients were verified independently by BERTHA's Raman lidar measurements at night. Long-term AERONET observations (columnar particle number concentrations and aerosol optical depth – AOD;  Holben et al., 1998) are used to derive empirical conversion factors from extinction coefficients to particle number and surface-area concentrations (Mamouri and Ansmann, 2016; Ansmann et al., 2019). The respective equations and the conversion factors are listed in Table 1. The AERONET data are filtered for dust events (Ångström exponent AE < 0.3, AOT > 0.1, at 500 nm) and pure marine (0.25 < AE < 0.6, AOT < 0.07) and continental (AE > 1.6) conditions. It should be mentioned that the conversion factor for small continental aerosol particles ($n_{\mathrm{50},\mathrm{c},\mathrm{dry}}$; particles with radius >50 nm) is obtained using AERONET data from Leipzig (central Europe), but with a factor of 0.5 to best approximate the African rural aerosol conditions (Shinozuka et al., 2015).

Table 1List of abbreviations, formulas, and uncertainties for the lidar-derived input parameters to estimate CCN and INP number concentrations (Mamouri and Ansmann, 2016) and for the separation of fine- and coarse-mode mass concentration (Mamouri and Ansmann, 2017). For Saharan dust the updated conversion factors of Ansmann et al. (2019) are used. All conversion factors are given for a lidar wavelength of 532 nm. In the following, the indices d, c, and m represent the aerosol types dust (df – fine-mode dust – r<500 nm; dc – coarse-mode dust – r>500 nm) and continental and marine particles, respectively. The extinction coefficient is calculated as the product of the lidar ratio S_i (S_d=55 sr; S_c=50 sr; S_m=20 sr) and the backscatter coefficient β_i of the aerosol component i. NC and MC stand for particle number concentration and mass concentration, respectively. The density ρ_d of dust is 2.6 g cm⁻³.

^* For an extinction coefficient of 1 Mm⁻¹.

Download Print Version | Download XLSX

Table 2The INP parameterizations for immersion freezing, with their references and valid temperature intervals. In the case of immersion freezing, ice nucleation starts via an INP immersed into a liquid droplet.

Download Print Version | Download XLSX

In the next step, INP and CCN concentrations are retrieved. INP parameterizations were developed for the aerosol types dust, soot, and marine particles (see Table 2). The number concentration n₂₅₀ and the surface-area concentration s are the aerosol-relevant input parameters which are obtained by conversion of the lidar-derived particle extinction profiles. In the present study, we focus on immersion freezing, i.e., ice nucleation by an INP immersed into a liquid-water droplet. The parameterization by DeMott et al. (2010) is used for the dust and non-dust (continental or marine) particles with a radius of more than 250 nm, whereas the DeMott et al. (2015) parameterization is explicitly developed for dust particles. Harrison et al. (2019) developed a parameterization for the very ice-active mineral K-feldspar, which is part of the Saharan dust. Mineralogical measurements at Barbados show that only 1 % of the dust particles consist of K-feldspar (Kandler et al., 2018). Therefore, we assume that the K-feldspar parameterization by Harrison et al. (2019) is valid for 1 % of the total surface area of dust. The surface-area-based INP parameterization developed by Ullrich et al. (2017) leads to much higher values and is not shown in this study. McCluskey et al. (2018) developed a parameterization for marine aerosol with samples from the Atlantic Ocean. The INP parameterizations and input parameters are listed in Table 2.

To estimate the CCN number concentration n_CCN, Mamouri and Ansmann (2016) use a dry activation diameter of 200 nm for dust and 100 nm for continental pollution and marine particles at 0.15 %–0.2 % water supersaturation. An enhancement factor f_ss determined in Mamouri and Ansmann (2016) from various laboratory and field studies (activation diameter and supersaturation) is used to retrieve n_CCN for different supersaturation levels (Table 1). The supersaturation of 0.2 % with respect to water is motivated by the findings of Wex et al. (2016), who reported this as a typical value for trade wind cumuli in the Barbados region. CCN concentrations at the same supersaturation (0.2 %) were measured in situ with a CCN counter aboard the Falcon research aircraft.

Table 3Dry activation diameter d_act for various chemical compositions calculated with kappa-Köhler theory (Petters and Kreidenweis, 2007). The κ values are estimated from literature (Ko09 – Koehler et al., 2009; He09 – Herich et al., 2009; Ku11a – Kumar et al., 2011a; Ku11b – Kumar et al., 2011b; Pe&Kr07 – Petters and Kreidenweis, 2007; Pe09 – Petters et al., 2009; Kr12 – Kristensen et al., 2012). The uncertainty in κ can be considerable especially for Saharan dust and organics.

Download Print Version | Download XLSX

The use of the different activation diameters (100 nm for continental pollution aerosol and for marine particles; 200 nm for dust) is motivated by the following facts. Based on kappa-Köhler theory (Petters and Kreidenweis, 2007), we computed the activation diameter for 0.2 % water supersaturation and temperatures from −10 to 20 ^∘C for various materials and chemical compositions (Table 3), in which 10 ^∘C is the most realistic value within the SAL, as indicated by local radiosondes. Fresh Saharan dust mimicked by dry-generated dust is very hydrophobic (low-hygroscopicity parameter κ) so that the activation diameter is around 275 nm (at 10 ^∘C). Cloud-processed Saharan dust particles (mimicked by wet-generated dust samples) may have changed their hygroscopic properties (higher κ value) so that their CCN efficacy increased. Laboratory studies with wet-generated dust particles (in contrast to dry-generated fresh dust particles) reported higher κ values (Koehler et al., 2009; Herich et al., 2009; Kumar et al., 2011b). However, although the Saharan dust was transported over several thousands of kilometers across the Atlantic Ocean, observations suggest that the dust in the SAL remained nearly unprocessed (Lieke et al., 2011; Denjean et al., 2015; Weinzierl et al., 2017; Kandler et al., 2018). Therefore, κ should not change significantly during transport and be closer to the value for fresh Saharan dust which is taken from laboratory studies (Koehler et al., 2009; Herich et al., 2009; Kumar et al., 2011a, b). Twohy et al. (2009) found good agreement for their CCN measurements with a κ of 0.05 in the eastern Atlantic. Herich et al. (2009) concluded that the activation diameter for Saharan dust (dry generated) is most probably 200 nm at a supersaturation of 0.2 %. This is confirmed by studies of Shinozuka et al. (2015) and Lv et al. (2018). Following this discussion, we assume an activation diameter of 200 nm for Saharan dust at Barbados, which corresponds to a κ value of approximately 0.05.

The activation diameter for continental aerosol particles (fine-mode pollution) depends on their chemical composition. Kandler et al. (2018) found sulfate particles to be a dominant contribution of continental pollution aerosol in the SAL, but the instrumentation was not suitable for detecting organics. Considering ammonium sulfate with a small contribution of less-hydrophilic organic particles to be continental aerosol within the SAL, a dry activation diameter of 100 nm at a supersaturation of 0.2 % is a suitable estimate and therefore used in this study.

We assume sea salt to be the dominant component of the marine aerosol and prescribe an activation diameter of 70 nm (Table 3; at 0.2 % supersaturation and 10 ^∘C). In contrast, Mamouri and Ansmann (2016) estimated a dry activation diameter of 100 nm based on literature. Going from 100 to 70 nm as activation diameter would increase n_CCN by a factor of approximately 1.5.

In conclusion, we used a dry activation diameter of 200 nm for Saharan dust and a diameter of 100 nm for continental and marine particles, assuming a supersaturation of 0.2 % in the SALTRACE studies.

The polarization lidar–photometer networking technique (POLIPHON) introduced by Mamouri and Ansmann (2014, 2017) delivers mass concentrations of fine- and coarse-mode dust, i.e, dust particles with diameter d<1 and d>1 µm, respectively (Table 1). The PLDR at 532 nm is used to separate the contributions of non-dust aerosol (PLDR = 0.05), fine-mode dust (PLDR = 0.16), and coarse-mode dust (PLDR = 0.35). The lidar-derived extinction coefficient of the fine- and coarse-mode dust component is converted into volume concentration using long-term AERONET datasets and finally to mass concentration using the mass density of dust (2.6 g cm⁻³; Mamouri and Ansmann, 2014, 2017).

2.2 Airborne in situ aerosol measurements

A full list and details of the instrumentation installed aboard the research aircraft Falcon of the DLR are given in Weinzierl et al. (2017). Information on size-resolved particle number concentrations is obtained from condensation particle counters and optical particle spectrometers. The condensation particle counters were operated at slightly different cutoff diameters around 10 nm. The spectrometer setup included an airborne version of the Ultra-High Sensitivity Aerosol Spectrometer (Cai et al., 2008; Brock et al., 2011; Kupc et al., 2018), a Grimm model 1.129 SkyOPC, and a Cloud and Aerosol Spectrometer (Baumgardner et al., 2001). The combination of these spectrometers covers the complete range of particle diameters, from about 70 nm to 50 µm. Particle number size distributions (NSDs) are derived from the entirety of these data using a consistent Bayesian inversion method (Walser et al., 2017). Here, the NSDs are approximated by trimodal log-normal distributions. In situ cloud condensation nuclei concentrations are measured with a Cloud Condensation Nuclei Counter (Roberts and Nenes, 2005; Lance et al., 2006) operated at a water vapor supersaturation of 0.2 %. These concentrations are corrected for losses of large CCN at the aircraft's isokinetic aerosol inlet.

4.1 CCN profiles

In Fig. 3, the lidar-derived number concentration of CCN for dust n_CCN,d (red line) and continental pollution particles n_CCN,c (olive line) are presented. The total CCN number concentration n_CCN (black line; lidar) can be compared with measurements of the Cloud Condensation Nuclei Counter aboard the Falcon aircraft (black squares) at the same supersaturation. In Table 5, the vertically averaged values are compared. The lidar-derived n_CCN values are up to twice as large as the in situ measured values. However, the lidar retrieval uncertainty is quite large (factor 2). The retrieval uncertainty results from the uncertainty in determining the extinction-to-number-concentration conversion factor for small particles (r≥50 or r≥100 nm) using AERONET-derived AOD and columnar number concentrations (n₅₀, n₁₀₀) as described in Mamouri and Ansmann (2016). Besides the large retrieval uncertainty, other uncertainty sources may have contributed to the systematic bias between the lidar and airborne in situ observations: (i) The lidar conversion factors are derived for AERONET stations close to the Sahara. These conversion factors may not be applicable to aged dust after long-range transport and may overestimate the occurring accumulation-mode dust particle number concentration and thus n_100,d. (ii) The used dust activation diameter (d_dry=200 nm) may have been too low, and the true one was much larger than 200 nm (see Table 3; d_dry=275 nm for dry-generated (fresh) dust), and thus fewer dust particles were activated in the Cloud Condensation Nuclei Counter aboard the Falcon than estimated by lidar. (iii) Horizontal and temporal inhomogeneities in the dust concentration along the flight tracks and over the lidar site may have also contributed to the differences found. (iv) Although the Falcon data are corrected for the particles losses at the inlets (Spanu et al., 2019), there are several uncertainty sources in the in situ CCN measurement that may have contributed to the bias found.

Figure 3Lidar-derived CCN number concentrations at 0.2 % supersaturation (black line) with contributions from dust (red line; critical dry diameter of 200 nm) and continental pollution aerosol (olive line; critical dry diameter of 100 nm) compared to coincident airborne in situ measurements (black squares) during SALTRACE-1. The error bars of the lidar profiles indicate an uncertainty of a factor of 2. The error bars of the in situ measurements indicate the 16th and 84th percentile.

Download

Table 5Layer mean CCN and INP concentrations (n_CCN; n_INP) in the upper (>3 km) and lower (2–3 km) part of the SAL from lidar and Falcon (n_CCN only). The standard deviation of the layer mean is given. The uncertainty range for the lidar retrieval is a factor of 2 for n_CCN and 3 for n_INP (not indicated). The immersion-freezing INP parameterization of D15d for dust at a constant temperature is used to give an estimate. CCN concentrations are given for 0.2 % water supersaturation (ss). n_CCN and n_INP values for the observed dust–smoke mixture and the pure marine conditions (INP from M18m) measured at Barbados on 3 March 2014 and 26 February 2014, respectively, are added.

^* INP concentration calculated with McCluskey et al. (2018) for marine particles.

Download Print Version | Download XLSX

Overall, the CCN number concentration for the three presented dust cases agrees within a factor of 2 between the in situ measurement and the lidar retrieval. As the behavior is the same for all three comparison studies, it is expected to be representative of Saharan dust episodes in the Caribbean.

4.2 INP-relevant aerosol profiles

In Fig. 4a–c, the profiles of the sum of n_250,d and n_250,c are compared with the integral values of the particle number size distribution for r_dry>250 nm measured aboard the Falcon aircraft. The in situ values are transformed to the pressure and temperature at the measurement altitude to be comparable with the lidar observations. As can be seen, the in situ and lidar values agree well, except on 22 June and 11 July in the lower part of the SAL, where horizontal inhomogeneities in the dust load (see Fig. 2) may have partly caused the differences between the two measurements. The contribution of continental smoke and pollution aerosol to n₂₅₀ was less than 3 % in the SAL during the strong dust outbreak on 10–11 July 2013 and about 10 % on 22 June 2013. In total, there were fewer than 40 particles (r_dry>250 nm) per cubic centimeter in all three cases over the remote Atlantic.

Figure 4Number concentration n₂₅₀ for particles with a radius of more than 250 nm (a–c) and surface-area concentration s (d–f) measured aboard the Falcon aircraft (black squares) and derived from the lidar measurements (red profiles; solid line – sum of dust and continental pollution particles above 2 km; dashed line – sum of dust and marine particles below 2 km). The three SALTRACE case studies are shown: 22 June (a, d, g), 10 July (b, e, h), and 11 July 2013 (c, f, i). INP concentrations (g–i) are given at −25 ^∘C for the immersion-freezing parameterizations of D10d + c (input $n_{\mathrm{250},\mathrm{d}}+n_{\mathrm{250},\mathrm{c}}$; above approx. 2 km), D10d + m (input $n_{\mathrm{250},\mathrm{d}}+n_{\mathrm{250},\mathrm{m}}$; below approx. 2 km), D15d (input n_250,d), and H19d (input s_d; see Table 2). The 1 % K-feldspar contribution was used for H19d. The uncertainty in the lidar-derived n₂₅₀ and s values is 30 %. For the INP concentration an uncertainty of a factor of 3 is indicated by the dashed lines for the D15d profile. The error bars of the in situ measurements indicate the 16th and 84th percentile.

Download

Figure 4d–f compares the profiles of the total surface-area concentration derived from lidar extinction coefficients and from the airborne in situ measured number size distribution. Here, the contribution of the continental pollution particles to s within the SAL is 4 %–6 % during the strong dust outbreak (10–11 July) and 20 % on 22 June 2013. The lidar values are considerably larger than the in situ values. The use of conversion factors that are too large (based on AERONET observations close to the Sahara) may be one of the reasons for the disagreement.

An example on INP profiling is given in Fig. 4g–i at a temperature of −25 ^∘C. The DeMott et al. (2010) and DeMott et al. (2015) parameterizations (including the correction factor of 3) are used with $n_{\mathrm{250},\mathrm{d}}+n_{\mathrm{250},\mathrm{c}}$ and n_250,d profiles as input, respectively. Furthermore, the Harrison et al. (2019) parameterization for K-feldspar was added with a 1 % contribution of K-feldspar to the dust surface-area concentration as indicated by Kandler et al. (2018). The uncertainty range (factor 3) is exemplarily indicated for the DeMott et al. (2015) parameterization by the dashed line. As can be seen, the SAL contains INP concentrations of 10–200 L⁻¹ at −25 ^∘C (Table 5).

4.3 Fine- and coarse-mode mass concentrations

As an additional feature to the CCN and INP profiles, the dust mass concentration can be derived from the lidar measurements separately for fine- and coarse-mode dust (Table 1). The comparison with airborne in situ observations is shown in Fig. 5. The mass concentrations are calculated from the lidar-derived and in situ measured volume concentration by assuming a dust mass density of 2.6 g cm⁻³. An excellent agreement is obtained for the coarse mode. This indicates that the Falcon measurements capture the large particles in the SAL well (Spanu et al., 2019). The coarse-mode mass concentration from POLIPHON is around 16 times higher than the fine-mode mass concentration, leading to a mass fine-mode fraction of 0.06. For the optical properties, such as the backscatter coefficient, the fine-mode fraction is 0.2. These mass (or volume) and backscatter fractions are in full agreement with simultaneous AERONET sun photometer observations of the fine-mode volume and AOD fractions at Ragged Point, Barbados. Again, for the fine-particle-dominated quantities, i.e., the fine-mode mass concentration, the lidar derives higher values than observed in situ. Uncertainties in the in situ aerosol measurements or in the lidar conversion factors might be the reason. However, a good agreement of the lidar products with AERONET observations is found and corroborates the quality of the lidar products.

Figure 5Mass concentration of fine-mode (r<500 nm; dashed line) and coarse-mode (r>500 nm; solid line) dust derived from airborne in situ measurements (black squares) and lidar observations (red profiles) for the three SALTRACE-1 days. The error bars of the in situ measurements indicate the 16th and 84th percentile.

Download

5.1 Pristine marine conditions

Caribbean background cases without aerosol transport from Africa were found during SALTRACE-2. At the end of February 2014, pristine marine conditions prevailed at Barbados, as already discussed in Haarig et al. (2017b); 26 February 2014 was chosen for the present study, as the influence of dry marine particles (Haarig et al., 2017b) was less pronounced than the days before (23 and 24 February 2014). The results are presented in Fig. 6. The marine aerosol reached 2 km height (Fig. 6a). The low values of PLDR (≤0.03) shown in Fig. 6b increased at the top of the marine aerosol layer to values of 0.06, indicating the presence of dry marine particles with a non-spherical shape, as discussed in Haarig et al. (2017b). These particles are misclassified as a very small dust contribution, as can be seen in Fig. 6c for the CCN number concentration. Otherwise the CCN reservoir consists of marine aerosol only (up to 250 per cm³). The dashed line in Fig. 6c indicates the lidar-retrieved n_CCN from 3 March 2014 showing a similar behavior. The INP reservoir at −25 ^∘C (Fig. 6d) derived with the parameterization of McCluskey et al. (2018) consists of 0.002–0.1 INPs per liter, which is around 3 orders of magnitude lower than in presence of Saharan dust. These findings from a remote-sensing perspective show the influence of Saharan dust on the cloud properties in the Caribbean and are corroborated by previous helicopter-based in situ measurements in the framework of the CARRIBA (Cloud, Aerosol, Radiation and tuRbulence in the trade wInd regime over BArbados; Siebert et al., 2013) project. The observations of February 2014 without aerosol long-range transport from Africa may be representative throughout the year for the marine contribution in the marine aerosol layer (below the temperature inversion at around 1.5–2.0 km). If dust is present in the SAL above, especially in summer, dust particles are mixed downwards in the marine aerosol layer and added to the marine (background) particles, significantly influencing the CCN and INP reservoir.

Figure 7Dust–smoke mixture observed during the SALTRACE winter campaign on 3 March 2014, 22:30–23:20 UTC. (a) Time–height display of the 532 nm volume linear depolarization ratio (VLDR; only the first 40 min are averaged for the profiles in b–h), (b) particle backscatter coefficient (green line) including its dust contribution (red line) and particle linear depolarization ratio (black line) at 532 nm, (c) sum of dust and continental pollution extinction coefficient (green line) using a smoke lidar ratio of 50 sr and sum of dust and marine particle extinction coefficient (blue line) using a marine lidar ratio of 20 sr compared to the total extinction coefficient (black line) independently measured with BERTHA (Raman lidar method). Above the height of 1.6 km a dust–smoke mixture fits best; below this height, the dust–marine mixture (with a small contribution of smoke or pollution) agrees better with the Raman extinction solution. (d) Number concentration n_100,d for dust (red), n_50,c for smoke (green), and n_50,m for marine particles (blue); (e) CCN number concentrations at water supersaturation of 0.2 % for the three components and the total CCN concentration (black line) above 1.6 km for dust–smoke and below for dust–marine; (f) n₂₅₀ values (colors as before); (g) surface-area concentration (colors as before); and (h) immersion-freezing INP concentrations at −25 ^∘C for D10 cont. + dust, D15 dust, H19 dust, M18 marine, and U17 soot. For the INP concentration an uncertainty of a factor of 3 is indicated as dashed lines for the D15d profile.

Download

Figure 8Summer (10 July 2013; red) versus winter (3 March 2014; cyan) aerosol conditions in the SAL. (a) Total particle extinction coefficient (solid line) and relative dust contribution to the total particle extinction coefficient (dashed line). (b) Same as (a) except for dust mass concentration. (c) Same as (a) except for CCN concentration.

Download

5.2 Dust–smoke mixture

A pronounced outbreak of aerosol from Africa reached Barbados in the beginning of March 2014. The trajectories ending at 2000 m a.g.l. on 3 March 2014 (not shown) point to the Sahara as the dust source and to western Africa (Senegal and Guinea) as the source region for biomass-burning smoke. The transport across the Atlantic Ocean took around 2 weeks. We use the opportunity of the dust–smoke aerosol mixtures to highlight the strong impact of smoke on the CCN conditions. The transport of biomass-burning smoke from Africa towards South America and the Caribbean during wintertime has been previously reported (Ansmann et al., 2009; Baars et al., 2011). An indication for the strong smoke contribution to the measured backscatter signal was the relatively low particle depolarization ratio (≤0.17). Fine-mode smoke does not depolarize laser light (PLDR ≤ 0.05). Figure 7 gives an overview of the measurements on 3 March 2014. A lofted layer (1.6–3.1 km height) of dust and smoke was found above the marine aerosol layer, reaching 1.6 km height. The vertical profiles in Fig. 7b and c show mean values for the time interval from 22:30 to 23:20 UTC. The particle backscatter coefficient (Fig. 7b) is separated into a dust component and a non-dust component using the PLDR separation technique as described in Sect. 2. To estimate whether the non-dust component is of marine or continental origin, the extinction coefficient was calculated from the different contributions to the backscatter coefficient as previously described in Sect. 2 and in Ansmann et al. (2017). The dust-related backscatter coefficient was multiplied by the dust lidar ratio (S_d=55 sr), and the non-dust backscatter coefficient was multiplied by the lidar ratio for marine particles (S_m=20 sr; contributing to the blue curve in Fig. 7c) and for continental pollution particles (S_c=50 sr; contributing to the green curve in Fig. 7c). The sum of the extinction coefficient (dust plus marine and dust plus continental) is then compared with the total extinction coefficient (black curve in Fig. 7c) derived independently with the Raman lidar method (Ansmann et al., 1992). As can be seen, the lofted aerosol layer obviously contains a mixture of dust and smoke, whereas the layer below is dominated by marine particles.

In the next step, n_100,d, n_50,c, and n_50,m (Fig. 7d) are computed, and the resulting n_CCN (Fig. 7e) at 0.2 % supersaturation is calculated. The continental pollution contribution to the CCN number concentration is 4 times stronger than the one from the dust aerosol. Thus, in the winter half-year with significant smoke contribution from Africa, rather different CCN conditions are found across the Atlantic, leading to likely changes in trade wind cumulus cloud microphysical properties compared to the summer months when dust particles dominate the CCN reservoir.

In contrast, n₂₅₀ is dominated by mineral dust (Fig. 7f). The lidar-derived contributions to the surface-area concentration (Fig. 7g) of dust and smoke are equal. The INP concentration at −25 ^∘C estimated in Fig. 7h shows a weak contribution of marine particles (McCluskey et al., 2018), with less efficiency than the dust particles in the lofted layer by 3–5 orders of magnitude. The immersion-freezing INP parameterizations based on n₂₅₀ (DeMott et al., 2010, 2015) lead to values around 10 L⁻¹. The results are added in Table 5.

Figure 8 highlights the sensitive impact of smoke aerosol on the CCN concentration. The dust contribution to the optical properties (Fig. 8a) is almost 100 % in summer during strong dust outbreaks and around 50 % during the biomass-burning season, which is in full agreement with AERONET observations. Dust dominates the aerosol mass concentration in the SAL (Fig. 8b) throughout the year, disregarding summer or winter conditions. In strong contrast, the smoke CCN concentration (Fig. 8c) derived with lidar strongly varies between summer and winter. CCN levels of 200–300 cm⁻³ are derived during the strong dust outbreaks in summer (dust contribution around 80 %) but are close to 500 cm⁻³ in the March 2014 event, with a strong contribution of smoke particles (80 %).