Dust is one of the main components of the atmospheric aerosol load, representing about 75 % of the global aerosols injected into the atmosphere . Mineral dust emitted from the Sahara and surrounding regions contributes to more than half of the global dust emissions .

The African dust is transported westwards as a hot, dry, dust-laden elevated layer (the so-called Saharan Air Layer or SAL) over the North Atlantic for thousands of kilometres . These large amounts of Saharan mineral dust within the SAL are advected off the African continent to the Caribbean , the Amazon rainforest or the Mediterranean basin . A large number of studies in the literature have found that mineral dust causes a strong radiative impact on atmospheric vertical stability, thermodynamics and wind shear among others. Recently, some authors have shown that although mineral dust produces net radiative heating within the SAL in summer, the small amount of water vapour that this layer contains, significantly higher than that present in the clean free troposphere, represents the major radiative driver for the SAL in summer . Furthermore, many studies in the literature confirm that mineral dust within the SAL also plays an important role as cloud condensation nuclei (CCN), giant CCN (GCCN) and ice nuclei, affecting the properties of clouds and their indirect effect on radiative balance and consequently on climate .

Most of the studies on the SAL have been focused on dust transport to the Caribbean region through the northern equatorial Atlantic. Between the tropics and the Canary Islands, there is no place where the SAL can be analysed and characterised once it has left the African continent and has docked on the subtropical North Atlantic marine boundary layer (MBL), impacting the subtropical free troposphere (FT). Another problematic point of the current research is related to the preferential study of the dust leaving Africa at tropical latitudes in the summer season when the Saharan outbreaks over the North Atlantic are mostly confined in an elevated mixed layer . As showed, the SAL can be identified near the west coast of Africa all year round with a clear spatial seasonal cycle: between 5∘ S and 15∘ N in winter and 5–30∘ N in summer. Therefore, the Canary Islands lie on the northern edge of the dust belt in summer and are significantly affected by the SAL. The study of the SAL at subtropical latitudes will add important additional information since this location is representative of the almost pure Saharan convective boundary layer (CBL).

In winter, most of the dust in the eastern tropical North Atlantic is transported within the marine layer in a shallowest layer, up to 1.5–3 km . Winter dust mobilisation in the Sahel region is often associated with the Harmattan winds that transport dust southwestward from the Bodélé and adjacent areas at low levels . During this dry season, dust is frequently mixed with biomass-burning aerosols confined in the upper layers . In contrast, the atmospheric mechanisms that cause dust intrusions over the subtropical North Atlantic in winter are generally linked to baroclinic processes and are usually of short duration and limited geographic extension. These baroclinic processes are often located in the vicinity of the Canary Islands and affect Western Sahara, northern Mauritania and western Algeria. Therefore, this dust is free of biomass-burning aerosols. An outstanding example was the huge dust intrusion that occurred in February 2020 . Although it is not strictly accurate to talk about the SAL in winter since the intrusions on the ocean do not constitute an elevated layer, in this study we will refer to them as “winter-SAL”. The impact of the winter-SAL on the subtropical MBL, and more specifically on the Canary Islands, was analysed by .

The altitude to which the aerosols are initially lifted and in which they are subsequently transported will influence their lifetime and radiative effects . Consequently, there is a need for systematic ground-based mineral dust and vertical thermodynamic observations near dust source regions to understand the evolution of the 3D spatial and temporal dust distribution and the role played by desert dust in the effective radiative forcing (ERF).

Lidar offer the most efficient technique for the identification and characterisation of the vertical distribution of atmospheric aerosols. However, the quantitative data inversion in the case of elastic backscatter lidar involves knowledge of the relationship between aerosol backscatter (β) and extinction coefficients (α), that is, the extinction-to-backscatter ratio (LR =α/β or lidar ratio) . This presupposes knowledge of the vertical aerosol distribution, critical in the case of complex aerosol distributions such as in elevated layers. The uncertainty of this parameter determines the accuracy of the retrieved profiles . The accuracy of the retrievals can be improved by combining active and passive remote sensing aerosol observations. The synergetic aerosol monitoring of co-located photometer and lidar observations is a powerful tool for aerosol research .

In this study, we describe the seasonal evolution of the vertical atmospheric profiles over the subtropical North Atlantic region on Tenerife, Canary Islands, Spain, using 12 years of observations (2007–2018) with a micropulse lidar (MPL-3). In addition, we use radiosonde measurements in the same area. Sun photometer measurements allow us to minimise the uncertainties involved in the aerosol extinction retrieval from an elastic backscatter system. To assess the radiative effects, we use the two-layer approach presented in in which two-column lidar ratios (LR1 and LR2) are calculated as inputs, retrieved from the aerosol optical depth (AOD) information in two layers (marine layer and free troposphere). This technique yields a better description of the LRi than the standard one-layer approach.

Section  describes the test site. All the instruments and ancillary information are described in Sect. . In Sect.  and we characterise the lower troposphere under clean (dust-free) conditions (in four different seasons covering the whole year) and under Saharan dust conditions (focused on winter and summer seasons). Sections  and present the main results on the SAL's impact on the atmospheric vertical thermodynamic aerosol extinction profiles as well as the corresponding radiative impact of this dusty layer. Section  studies the potential role of the SAL in summer and winter on the formation of mid-level clouds as a result of the activation of heterogeneous ice nucleation processes. Finally, Sect.  presents the summary and the main conclusions from this work.

The Canary Islands, located in the subtropical North Atlantic, are characterised by a very stable and well-stratified lower troposphere with strong temperature inversions . These are modulated by quasi-permanent subsidence conditions as a result of the descending branch of the Hadley cell. made a comprehensive analysis of the temperature inversions in the lower troposphere of the Macaronesian region, including the Canary Islands. They found the presence of one or two sharp temperature inversions or transition levels below the 700 hPa level. The first transition level in the double-inversion structure coincides with the maximum extent of the surface-related turbulent mixing, which limits the dry adiabatic convection and caps the humid and well-mixed MBL. This transition level, denoted as the marine boundary inversion (MBI), is characterised by a strong vertical humidity gradient. The second transition level, denoted as the trade wind inversion (TWI), is located at the top of a highly stable layer, the trade wind layer (TWL), normally associated with synoptic-scale subsidence processes which limit vertically the quasi-permanent area of stratocumulus clouds frequently observed at this latitude. According to these authors, the percentage of occurrence of this double structure (normally found between 900 and 800 hPa) is low, with a maximum occurrence of 33 % in summer. Consequently, in this study we use a two-layer approach (one single temperature inversion) where the MBI and the TWI are the same temperature inversion capping the MBL, and separating it from the FT above. It is important to note that the study by on temperature inversions was carried out under no-dust conditions. However, the characteristics of the temperature inversions are modified by the impact of the SAL, which introduces a new temperature inversion in the lower troposphere. These aspects are also addressed in our study.

This study focuses on the observations and modelling results obtained over two locations in Tenerife: Santa Cruz de Tenerife station SCO (28.5∘ N; 16.2∘ W at 52 m a.s.l.) and the high-altitude Izaña Global Atmosphere Watch (GAW) Observatory IZO (28.3∘ N; 16.5∘ W at 2391 m a.s.l), horizontally spaced less than 30 km. These two sites belong to the State Meteorological Agency (AEMET) and are managed by the Izaña Atmospheric Research Centre (IARC) .

Atmospheric scenarios have been classified using two different and independent techniques: lidar extinction profiles, able to characterise the vertical distribution of aerosols, and atmospheric soundings necessary to characterise the atmosphere from a thermodynamic perspective. Only daytime lidar profiles and atmospheric soundings (launched at 12:00 UTC) have been included in this study. The emerging database is compiled with every lidar extinction profile and the corresponding daily sounding profile. We obtain consistent results of the MBL thickness retrieved with these two techniques as shown in Sect. S1 in the Supplement. This gives us confidence in the atmospheric layering performed in this study, both in terms of atmospheric aerosols and thermodynamic variables. Monthly average values of aerosol extinction at 523 nm (α) are shown in Sect. S2 (Figs. S1 and S2 and Table S1) giving us an overview of the dust vertical structure of the eastern subtropical North Atlantic atmosphere throughout the year. From this analysis, we can define the clean and Saharan scenarios as being predominant. Saharan scenarios present a different seasonality within the MBL, with maximum AOD values in winter, compared to the values shown in the upper layer with maximum AOD in July and August.

Clean air masses are defined as those with ARTI = 0 (i.e. non-African origin). Figure S3 shows how these clean scenarios predominate at both levels through the year. Saharan air masses might significantly impact SCO from November to January (74.0 % of the annual data), while IZO is mainly affected in July and August (29.5 % of the annual data). These results corroborate our findings based on lidar results presented in Sect. S2. Based on this information, clean scenarios are split into different seasons according to the ARTI monthly distribution, i.e. winter (November–January), extended spring (February–June), summer (July–August) and autumn (September–October). Otherwise, Saharan scenarios in winter (winter-SAL, confined to lower levels) are limited to the period November–January, while Saharan scenarios in summer (summer-SAL, as an elevated layer) are limited to July and August.

AERONET AOD and AE have been used to better categorise the Saharan scenarios, in addition to the previous criterion (ARTI > 0). The winter-SAL scenario is defined as those conditions in November–January with ARTI > 0, AOD at SCO ≥ 0.15 and Angström Exponent (AE) ≤ 0.5, according to the results from Guirado (2015). Similarly, the summer-SAL scenario is defined by ARTI > 0, AOD at IZO ≥ 0.10 and AE at IZO ≤ 0.6 .

In order to understand the impact of the SAL on the subtropical atmospheric vertical structure, we have displayed in Fig.  the differences in T, r and α profiles between the summer and winter Saharan scenarios and their corresponding clean scenarios. The magnitude of the previous differences depends on the seasonal values of these parameters in the subtropical troposphere in addition to their seasonal values of the advected Saharan air mass.

In the case of the summer-SAL, these differences are readily appreciated, up to 6 K warmer at lower levels (Fig. a), in response to both the SAL radiative impact (Sect. ) and the warm advection of dusty Saharan air masses. Negative temperature differences of up to 2 K are found at 5.2 km. This cooling effect results in an inversion at the summer-SAL top corresponding to the upper limit of the dust layer. In contrast, we can describe the winter-SAL, when dust transportation is confined to lower levels, as a layer with negligible impact in terms of T at lower levels (Fig. d). However, the T profile from 2 km up is roughly 3 K cooler than in clean conditions.

Our results show that the SAL also has a strong impact on atmospheric humidity (Fig. b and e), similar to the results presented by . Our results present the SAL as a dry layer at lower levels in both summer and winter. We observe negative r anomalies (Fig. b) at the SAL's base in the case of the summer-SAL (Δr∼ -3.6 g kg-1) and slightly higher α anomalies (Δα∼ +48 %) compared to clean summertime conditions. Regarding higher levels, positive and rather constant r anomalies of about 2 g kg-1 are found in the summer-SAL from the dust peak up to 5 km. These r anomalies, although small as absolute values, represent a high relative variation in comparison to the extremely dry FT under clean conditions (Δr up to +332 % at 5.3 km, roughly coinciding with the altitude of the maximum cooling). As a consequence, the SAL appears as a relatively moist layer at mid-levels. The anomaly in α peaks at 2.6 km, where maximum extinction values are commonly retrieved (Fig. c). The winter-SAL at lower levels is characterised by mixing ratio anomalies up to ∼ -3.0 g kg-1 (dry anomaly) below the SAL (Δr of -38 % at 1 km) (Fig. e), similar to those observed in the summer-SAL at similar levels. High α anomalies were also observed in the MBL (maximum values of Δα∼ +118 % located in the dust peak level) (Fig. f), with no significant anomaly in terms of r or α at higher levels.

The fact that the SAL in summer presents a significant and positive anomaly in water vapour content in the subtropics, as shown in this study, is consistent with previous results i.e.among others. This result shows that, despite being a dry layer, the SAL in the subtropical region is more humid than the clean FT around it. This fact is relevant since water vapour is a gas strongly absorbing longwave radiation. Consequently, moisture carried on the SAL can play a key role in the atmospheric radiative balance, depending on its vertical distribution, and therefore in the temporal evolution of the vertical structure of the SAL itself, as discussed in Sect. .

Several papers in the literature have been focused on determining the SAL radiative effects in summer . These authors found that under clear sky conditions the net radiative effect, a combination of LW and SW interactions, results in maximum net heating located slightly below the maximum dust concentration level and also a minor net heating observed near the surface. The existence of significant heating rates (∼ 1 K d-1 between 500 and 1000 hPa) indicates the important role of dust in stabilising the atmosphere, the same as, for example, the enhancement of a pre-existing trade wind temperature inversion found by as a result of the SAL layer above. In addition, cold anomalies above the SAL are hypothesised as the result of the strong vertical ascent of warm air due to the net warming within the SAL, yielding to an adiabatic cooling in these upper layers . However, other authors attributed an important role in the aforementioned vertical radiative heating rates to the enhanced atmospheric water vapour within the SAL. As a result, both dust and dry anomalies play an important role in heating of the SAL in the lower troposphere within the MBL .

Atmospheric heating rates have been calculated using the radiative transfer model LibRadtran (see Sect. ) in order to detect the vertical radiative impact of the dust layer in summer and winter. As we have shown in the previous section, this dust-laden layer acts by significantly modifying the thermodynamic profiles in the lower middle troposphere.

Diurnal-averaged heating rates have been retrieved for two different cases. The first case, representative of the real atmospheric conditions under the presence of the summer-SAL or winter-SAL, includes the median T, r (or RH) and α representative of each scenario (Sect.  and , respectively). A second case has been considered including the same median thermodynamic soundings as in the previous dusty scenarios but using the median α profiles representative of clean conditions (retrieved in Sect. ). These scenarios, called summer-clean* and winter-clean*, are not real but can be used to separate contributions of mineral dust and water vapour to the radiative forcing in the atmospheric column. This approach is similar to that carried out in previous studies .

Simulations have been done including in LibRadtran spectral information of α, SSA and g extracted from MOPSMAP (see Sect. ). In the case of both summer-SAL and winter-SAL scenarios, a mixture of maritime clean and desert dust aerosol types have been considered to form the MBL, and only desert dust aerosols are considered to form the atmosphere above this layer. In the case of the summer-clean* and winter-clean* scenarios, only maritime clean aerosols have been considered. MOPSMAP spectral resolution of these optical aerosol properties depends on the spectral range, being modelled between 400 and 1000 nm with a spectral resolution of 10 nm, between 1000 and 3000 nm with a resolution of 500 nm, and finally with a spectral resolution of 3000 nm in the range 3000 nm to 40 µm. Spectral AE retrieved from MOPSMAP has been used to calculate the α profiles at the other wavelengths by applying the Ångström power law once the vertical profile from the lidar at 523 nm is known.

Figure a shows T, r and α profiles for the summer-clean and the summer-SAL scenarios, while Fig. b presents the comparison of the modelled output heating rates (in K d-1) for the same scenarios. The maximum heating rate due to mineral dust (differences between the summer-SAL and the summer-clean* atmospheres) is +0.7 and -0.15 K d-1 for SW and LW (dotted lines in Fig. b), respectively, located at about 2.5 and 3 km, near the dust concentration peak. Following , the net radiative effect near the surface is a balance between absorption heating, extinction cooling, emission cooling, and adiabatic heating and cooling. In our case, we have found net heating near the surface dominated by the SW forcing, attributed to the net effect of mineral dust (SW absorption and LW cooling). This result is consistent with previous studies in the literature that found an expected and dominating dust-induced SW radiative warming slightly below the dust maximum. Below this dust peak, a weaker SW effect is expected, dominated by dust extinction processes. We have actually observed that LW dominates below the dust peak through upwards and downwards emission processes . The mean net heating rate due to mineral dust observed in the first 1 km height is +0.34 and +0.13 K d-1 for SW and LW, respectively. The LW warming found at surface level has been also observed by , who attributed it to the combined effect of dust LW absorption, reflection and re-emission processes at lower levels. Above 6 km the heating rate due to aerosols does not show any change due to the absence of dust. However, in view of Fig. b, one can see how the real summer-SAL radiative impact is characterised by a strong cold anomaly near the SAL's top that cannot be attributed to the solely radiative effect of mineral dust. This cold anomaly, also observed by other authors , is consistent with the temperature difference of about -2 K (Fig. a) and also with the 190 m descent from iso-zero height level (Tables  and ). Figure b quantifies the net summer-SAL heating rate to be -0.6 K at the SAL's top. Maximum LW cooling (-3.6 K d-1) is observed at 5.9 km, while maximum SW heating (3.5 K d-1) is observed slightly below, at 5.6 km, both located near the maximum RH peak within the SAL (see Fig. S4). Net warming is observed to occur within the SAL (from 2.1 to 5.7 km), with relatively constant heating rates of about 0.8 K d-1 in an altitude range between 3.5 and 5.1 km. Atmospheric water vapour absorbs part of the SW radiation, with a maximum effect found in the position of the RH peak within the SAL (5.6 km). LW cooling due to water vapour is attributed to the LW emission without absorption (adiabatic cooling) in the top of the SAL (maximum cooling found at 5.9 km), induced by the strong vertical ascent of air near the STI, acting as a thermodynamic layer of discontinuity between two different air masses.

Our results highlight the important radiative role played by the water vapour on the atmospheric column in the summer-SAL scenario, not only in the cold anomaly near the STI but also at lower levels, counteracting the warming effect of aerosols in the SW. This is a relevant aspect to consider, especially in atmospheric modelling. Dust models might certainly introduce an overestimation of the aerosol radiative effect within the SAL if the water vapour radiative effect is obviated.

The same analysis in terms of modelled output heating rates (in K d-1) has been performed for the winter-clean and winter-dust scenarios (Fig. ). As in the previous case of summer, the radiative effect of dust has been estimated by means of the heating rate differences between winter-SAL and winter-clean* atmospheres. We have observed a dust-induced radiative effect dominated by SW heating (maximum heating of +0.7 K d-1 at 1.5 km, near the dust peak), while below the dust peak the SW effect becomes lower. LW cooling is also maximum near the α peak (-0.13 K d-1). However, the real winter-SAL radiative impact presents maximum SW heating rates up to +3.5 K d-1 at 0.8 km (near the MBI), while the maximum LW cooling is found near the surface (-3.1 K d-1) (in Fig. ). The net effect within the surface layer (MBL and SAL) implies a small and rather constant SW warming of about +0.7 K d-1 in the 0.6–2 km layer. This net heating near the surface is attributed to the combined effect of dominant dust SW heating and the predominant effect of water vapour SW absorption at these levels. Above the SAL, the net radiative effect is dominated by the dry anomaly of the winter-SAL in comparison to clean conditions. This removal of water vapour in the first 2 km reduces the greenhouse trapping of LW radiation, with an LW cooling effect of about +0.3 K d-1 . As a result, dry anomalies in the first levels have the consequence of thermal cooling (decrease in the net upward longwave flux) from the top of the SAL upwards. In this winter case, the SAL also impacts the atmospheric column, from the dry anomaly at lower levels to the thermal cooling at higher levels.

The previous findings, in addition to those on Sect. , allow us to assess the role of the thermal advection on the differences observed on the vertical temperature profiles taking place within the summer-SAL and winter-SAL. According to Fig. a, a temperature difference between the Saharan and the clean scenarios greater than 3 K occurs between 0.8 m and 2.5 km height, registering the maximum temperature increase in the presence of SAL at ∼ 1 km height, with a ΔT= +6 K. This height is precisely where the maximum net radiative forcing cooling is recorded in the lower troposphere (-0.5 K d-1 as shown in Fig. ). This result might lead us to conclude that the warm air advection from the Sahara is the major driver of the temperature increase observed in this altitude interval. An analysis based on composites of average potential temperature from the NCEP/NCAR Reanalysis in summer (July and August) in the time period 2007–2018 has served us to estimate the thermal advection at the Canary Islands from the warmest region of the Sahara, located over western central Algeria around the Greenwich meridian. This analysis (shown in Supplement Fig. S5a) estimated an increase in temperature of up to +9 K at 1 km height due to this warm advection. Thus, we can say that the advection dominates between 800 m and 2.5 km height, but the radiative processes are more important below 800 m in the MBL (heating), between 2.5 and 6 km (heating), and above 6 km (cooling). In winter, there is practically no temperature gradient between the ocean and the African continent in the troposphere (Fig. S5b), and thus temperature advection processes are not relevant. A more realistic separation of the effects of temperature advection from that of radiative processes would be possible with dust model simulations that allow for the activation and deactivation of the radiative processes of mineral dust and water vapour, a capability that is out of the scope of this paper.

As we have shown in Sects.  and , both SAL's dust and humidity anomalies profoundly change the vertical radiative fluxes. The relative frequency of cloud types from MPL data for both Saharan and clean scenarios for winter and summer has been calculated using the methodology described in Sect.  in order to explore the effect of SAL radiative impact on cloud formation. The cloud type distribution is shown in Fig. a and b for winter and summer, respectively. Analogous plots from cloud climatological and total sky camera observations at 13:00 UTC are shown in Fig. S6 of the Supplement. A notable reduction of low clouds in the case of the Saharan scenario (winter and summer) in comparison to clean conditions is observed using MPL data (Fig. a and b). On the contrary, a higher occurrence of mid- and high-level clouds under the summer-SAL scenario are observed in Fig. b (23 % under Saharan conditions versus 4 % in clean conditions). This pattern is corroborated, at least qualitatively, using direct and independent cloud observations (Fig. S6). Although the reduction in low-level clouds is observed for both winter-SAL and summer-SAL, the reduction is more prominent in summer, when the probability of occurrence of low clouds is reduced from 40 % in summer-clean conditions to 5 % under SAL conditions (Fig.  b). In winter, this drop in low clouds is less pronounced, with a reduction from 25 % (winter-clean conditions) to ∼ 4 % (winter-SAL conditions). This observed reduction is explained taking into account that the SAL transports dryer air at lower levels in comparison to the MBL's air humidity (both in summer and winter), and therefore both the lifting condensation level and the level of free convection rise, increasing the energetic barrier to convection . This low-level warming affects both summer (1.4 K d-1) and winter (0.8 K d-1) radiative fluxes, inhibiting the occurrence of convection, causing vertical changes or dissipation of clouds. However, there are some differences between cloud suppression in the summer-SAL and winter-SAL. In summer, SAL carries not only drier air but also warmer air masses, compressing the MBL in the first kilometre because of the enhanced subsidence processes at this time of the year. In winter the occurrence of low-level clouds is lower than in summertime due to a weaker TWL . Consequently, the SAL effect is expected to be lower in winter. The result is a Saharan air mass with a strong suppressive influence on cumulus convection above the moist trade wind layer, with a more prominent effect in summer , which can cause dust to be transported into the MBL due to entrainment at the base of the SAL much more rapidly than dust vertical transport caused by gravitational settling.

We can interpret some of the previous results in terms of the impact of dust on heterogeneous nucleation and the role of dust as ice nucleation particles (INPs) . We know that nucleation of ice particles in the atmosphere occurs by homogeneous freezing conditions (at T≤-37 ∘C) in the absence of particles to catalyse ice formation and also by heterogeneous nucleation processes due to the presence of ice-nucleating particles. This last process can occur at any temperature below 0 ∘C and involves typically middle clouds composed of ice crystals and supercooled water droplets . Saharan dust is known to be capable of glaciating clouds at temperatures that are unusually warm, between -5 and -9 ∘C , being this characteristic layer, typically consisting of altocumulus and altostratus, commonly found at the top of the near-adiabatic part of the profile, i.e. in the upper region of the SAL . In our work, clouds observed under the summer-SAL influence are predominantly middle clouds (Fig. b), mostly located between 6 and 7 km height, coinciding with SAL's top (STI). Conversely, we have not observed any significant feature in the case of the comparison between clean and dusty scenarios in winter at mid-levels. These middle clouds under summer-SAL conditions are detected above the peak in RH (45 %–50 %) associated with the SAL layer (see Fig. S4) and also at altitudes higher than the iso-zero level (between 5 and 7 km). Temperatures at these altitudes are relatively warm, approximately between -9 and -15 ∘C according to the median temperature profile in Fig. a, which is in agreement with previous studies .

As a result, we have found an indication of a possible effect of dust on heterogeneous ice nucleation, resulting in modestly supercooled middle clouds at altitudes where moisture and temperature conditions are favourable, i.e. near the STI. This enhancement of middle clouds when dust is present may have a significant impact on the atmospheric radiative balance.

To further investigate the role of dust in cloud formation, a dust event that impacted Tenerife in summer has been simulated by the dust regional DREAM atmospheric model (see Sect. ). The selected event on 19–24 August 2015 is associated with mid- and high-level clouds (> 5 km) as it is observed in the MPL-3 profiles (see Fig. a). MPL-3 displays a dust plume constantly influencing the station from the second half of 19 August until the first half of 22 August. Dust is definitively detected up to 6 km, with the presence of clouds observed during the entire event from 5 to 11 km (also observed from the Cloud–Aerosol Lidar with Orthogonal Polarization – CALIOP – total attenuated backscatter in Fig. S6). INPC from the DREAM model (see Fig. ) is calculated based on the U17 parameterisation described in Sect.  . The model predicts that most of the dust mass is in a layer between 2 and 6 km, in agreement with the MPL-3. Maximum concentrations (> 100 µg m-3) were reached on the afternoon of 20 August (see Fig. S8). The lidar quick look indicates subsidence of the dust layer and a boundary layer intrusion. Lowering of the height of peak concentration was predicted by the model as well. However, at the top of the layer, at heights above 6 km, where temperatures are below -10 ∘C, dust concentrations of around 5 µg m-3 can serve as a reservoir for INPs. The INPC model captures some of the features of the clouds forming above the station. Due to the temporal resolution of model outputs (3 h), some features of cloud development are not seen. The predicted INPC values range from 1 to 10 L-1, reaching values of 100 L-1, in agreement with results of previous remote sensing and modelling studies . The clear overestimation of dust concentrations by DREAM (Figs. S7 and S8 in the Supplement) suggests an overestimation in INPC as well. found that overestimation by 100 % in dust concentration results in an order of magnitude overestimation in INPC prediction. Low humidity and temperature limit at which deposition nucleation parameterisation is used indicate that this otherwise significant dust concentration does not contribute to INPC. Relatively warm middle clouds (T> -5 ∘C) observed by the lidar between 4 and 6 km (Fig. a) on 20 August were not predicted by DREAM because parameterisations are done for temperatures lower than -5 ∘C. The observed middle clouds were well captured by the model and were explained by indications of a local maximum in the INPC. The cirrus clouds present on 20 and 22 August are consistent with INPC values of 1–10 L-1 and above 10 L-1. DREAM also predicts INPC on August 22 (from noon) and 23, but the low cumulus clouds prevent observations of middle and high clouds. However, observations of clouds at Izaña with an all-sky camera proved the presence of cirrus from 00:00 to 05:00 UTC (not shown here). These high clouds predicted by DREAM at the beginning of 23 August are also slightly discernible from the lidar observations between the low clouds (see Fig. a). The deposition nucleation parameterisation requires supersaturation with respect to ice, a condition that is not reached above 9 km. Therefore, the predicted INPC is mostly due to immersion freezing mode according to the model. The convergence between lidar observations and dust model predictions allows us to link, at least in this one instance, summer-SAL conditions with the formation of mid- and high-level clouds. However, an extended observation period will be required to firmly establish this relationship.

There is a vast literature that describes the evolution of the SAL as dust outbreaks emerge from Africa and move into the North Atlantic. However, these studies lack a focus on the vertical characterisation of the Saharan layer using robust long-term datasets, a factor that limits our understanding of the atmospheric processes involved and hinders the validation of atmospheric numerical models. In addition, little attention has been paid to the seasonal variations and the radiative impact that the SAL exerts on the vertical atmospheric profiles.

Here we use long-term cloud-free lidar and vertical sounding observations performed in Tenerife during 2007–2018 to characterise the vertical structure of the lower and middle troposphere over the eastern subtropical North Atlantic. Tenerife is a key location for Saharan dust studies because of its proximity to the most important mineral dust sources. In addition, Tenerife experiences the full seasonality of dust transport at this latitude, ranging from clean normal flow to extreme dust-loaded Saharan air mass outbreaks that affect the lower atmospheric levels in winter and higher levels in summer. The study of the SAL in the Northern Hemisphere subtropical region adds important additional information since this location is representative of the almost pure Saharan dust. The results of this study have revealed that, in the case of clean conditions, the MBL is relatively well mixed with α values below 0.036 km-1, while significantly clean conditions are observed in the free troposphere (α< 0.022 km-1). We have observed a readily visible inversion layer, the MBI, with a secondary α maximum, which separates moist and marine aerosol-laden MBL from the clean FT above.

A clear influence of dust is observed in the summer-SAL scenario. An important increase in α and a decrease in r was found in the MBL, with α values 48 % higher and r values 44 % lower relative to clean conditions. The subtropical FT was observed to be significantly affected by the CBL. The SAL layer appears as a well-stratified layer according to the neutral or slightly positive θ and θe vertical gradients, with a peak in α> 0.066 km-1 at ∼ 2.5 km. Under SAL influence, the CBL reaches altitudes normally characterised by clean free troposphere conditions, being Saharan air masses more humid than FT at the same levels with a peak in r near the SAL’s top (Δr∼ +332 % at 5.3 km). It is noteworthy the decrease in the iso-zero level in this Saharan scenario, in comparison to the clean scenario in summer, indicating the presence of colder air at higher levels under the presence of SAL. Its top, the STI, has been located at ∼ 6.0 km. It is important to highlight the reinforcement in the MBI as a result of the SAL effect, with a strength of 3.1 K in the clean scenario in summer and 5.0 K in the same season with dust influence. In the winter-Saharan scenario, in which dust is only present at lower levels, we have found two different layers in the MBL but compressed in the first 2 km height: a dry (Δr∼ -38 % in comparison to the clean scenario) and well-mixed MBL and the SAL with extinction coefficients up to 0.049 km-1 (Δα maximum of +118 %) in the SAL's peak (∼ 1.3 km height). CFT conditions were found from 2.3 km on.

Significant anomalies in comparison to clean conditions were observed in the lower middle troposphere as a consequence of the warm and moist advection attributed to the summer-SAL. This layer appears as a warmer layer (ΔT up to +6 K within the SAL) and colder at its top (ΔT up to -2 K near the SAL’s top); drier at the base (Δr∼ -44 %) but more humid within the SAL (Δr∼ +332 % at 5.3 km). In the winter-SAL scenario, a net cooling of ∼ 3 K is observed above 2 km height. In this case, the SAL appeared as a drier (Δr∼ -38 %) and dustier (Δα maximum of +118 %) layer, in comparison to winter-clean conditions.

We have delved into the previous thermal and moisture anomalies by means of radiative transfer calculations using a 12-year climatology of lidar and thermodynamic profiles. The results revealed the important role that both dust and water vapour play in the heating rates within the summer-SAL and the winter-SAL. The dominant dust-induced SW radiative warming (up to 0.7 K d-1) is found slightly below the dust peak. However, we have confirmed that atmospheric water vapour has a dominant radiative impact on the net SW warming observed within the summer-SAL (from 2.1 to 5.7 km) in addition to the strong cold anomaly near the SAL's top (-0.6 K d-1) (LW adiabatic cooling). The higher water vapour content found to be carried on the summer-SAL (∼ 2 g kg-1) is very low but represents a high relative variation in comparison to the very dry free troposphere under clean conditions (Δr∼ +332 % at 5.3 km, roughly coinciding with the altitude of the maximum cooling). These results confirm the important radiative role played by water vapour on the atmospheric column, both at higher levels near the STI and at lower levels counteracting the warming effect of aerosols in the SW. This is a relevant aspect to be well thought out in atmospheric modelling since ignoring the radiative effect of water vapour would overestimate the aerosol radiative effect. In the case of the winter-SAL, the radiative effect has been attributed to the combined dust and moisture effect. This is the case of the SW heating found within the SAL (maximum near the peak observed in r), the dry anomaly at lower levels (Δr∼ -38 % at 1 km) and the thermal cooling (∼ 0.3 K d-1) from the STI upwards. Thermal advection processes were found to dominate in summer from 800 m to 2.5 km height as a consequence of the strong temperature gradient between the ocean and the African continent in the lower troposphere. In contrast, radiative processes are found to dominate below 800 m in the MBL which results in heating, between 2.5 and 6 km producing heating, and above 6 km producing cooling.

Finally, our results suggest the strong impact of the SAL in summer and winter on low clouds and the activation of heterogeneous ice nucleation processes under favourable moisture conditions in the presence of the summer-SAL. As a result, we have detected (1) a lower occurrence of low clouds due to drier and warmer conditions at lower levels causing a strong suppressive influence on cumulus convection, with a more prominent effect in summer, and (2) a higher occurrence of middle and high clouds under SAL influence in summer. These mid- and high-level clouds are mostly located near the SAL's top above the iso-zero level (between 5 and 7 km height) at relatively warm temperatures (between -9 and -15 ∘C) and are likely to have a big impact on the surface energy budget. A case study of a 5 d dust event with MPL-3 data and DREAM dust model INPC predictions was used to give some evidence of the role of dust and the water vapour carried within the SAL in ice nucleation processes. The convergence between lidar and dust model predictions has served us to verify, at least in the first instance, the impact of dust on cloud formation.

The results presented here introduce a robust characterisation of the eastern subtropical North Atlantic atmosphere and contribute to our understanding of the vertical and seasonal characteristics and radiative impact of the SAL and its role in cloud nucleation. This information is crucial to improve the current simulations and parameterisations of atmospheric processes in aerosol models. Our use of a long time series of observations (12 years) concurrently carried out with independent and reference-type measurements gives us confidence in our ability to closely characterise the SAL and the factors that impact its behaviour and the effects on the regional environment. However, we raise a number of issues that will require further studies to better assess the radiative effects of these modestly supercooled mid-level clouds as well as to incorporate the presence of SAL-induced clouds in the evaluation of the radiative impact. Furthermore, given the high complexities inherent to ice nucleation processes, which are strongly dependent on T, RH and dust concentration in a way not yet fully understood, an extended observation period should be used to elucidate the role of dust in ice initiation processes. All of these analyses are out of the scope of this paper and are planned to be carried out in future studies.