Harmattan, Saharan heat low and West African Monsoon circulation: Modulations on the Saharan dust outflow towards the north Atlantic

The outflow of dust from the North African continent towards the north Atlantic is stimulated by the atmospheric circulation over North Africa, which modulates the spatio-temporal distribution of dust source activation and consequently the entrainment of mineral dust into the boundary layer, as well as the transport of dust out of the source regions. The atmospheric circulation over the North African dust source regions, predominantly the Sahara and the Sahel, is characterised by three major circulation regimes: (1) the Harmattan (trade winds), (2) the Saharan heat low (SHL), and (3) the West African Monsoon cir5 culation. The strength of the individual regimes controls the Saharan dust outflow by affecting the spatio-temporal distribution of dust emission, transport pathways, and deposition fluxes. This study aims at investigating the atmospheric circulation pattern over North Africa with regard to its role favouring dust emission and dust export towards the tropical North Atlantic. The focus of the study is on summer 2013 (June to August), during which also the SALTRACE (Saharan Aerosol Long-range TRansport and Aerosol-Cloud interaction Experiment) field 10 campaign took place. It involves satellite observations by the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) flying on-board the geostationary Meteosat Second Generation (MSG) satellite, which are analysed and used to infer a data set of active dust sources. The spatio-temporal distribution of dust source activation frequencies (DSAF) allows for linking the diurnal cycle of dust source activations to dominant meteorological controls on dust emission. In summer, Saharan dust source activations clearly differ from dust source activations over the Sahel regarding the time-of-day when dust emission begins. The 15 Sahara is dominated by morning dust source activations predominantly driven by the break-down of the nocturnal low-level jet. In contrast, dust source activations in the Sahel are predominantly activated during the second half of the day when down-drafts associated with deep moist convection are the major atmospheric driver. Complementary to the satellite-based analysis on dust source activations and implications from their diurnal cycle, simulations on atmosphere and dust life-cycle were performed using the meso-scale atmosphere-dust model system COSMO-MUSCAT (COSMO: COnsortium for Small-scale MOdelling; 20 MUSCAT: MUltiScale Chemistry Aerosol Transport Model). Fields from this simulation were analysed regarding the variability of the Harmattan, the Saharan heat low, and the Monsoon circulation as well as their impact on the variability of the Saharan dust outflow towards the north Atlantic. This study illustrates the complexity of the interaction among the three major circulation regimes and their modulation of the North African dust outflow. Enhanced westward dust fluxes frequently appear 1 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-309, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 15 May 2017 c © Author(s) 2017. CC-BY 3.0 License.

jet (LLJ) was found to be the dominant driver for dust source activation over North Africa (Schepanski et al., 2009b), which was verified by several follow-up studies (e.g., Fiedler et al., 2013;Tegen et al., 2013;Fiedler et al., 2014). Besides the LLJ, deep moist convection embedded in the intertropical discontinuity region fosters the formation of dust fronts by cold pools in particular over the Sahel zone during the wet season (Bou Karam et al., 2009;Heinold et al., 2013;Fiedler et al., 2014). The advection of cooler air from the Mediterranean Sea into the Sahara, occasionally reaching the Sahel zone, may promote the 5 formation of a Soudano-Saharan Depression. This depression then migrates westwards, eventually turns anti-cyclonically into a northward direction and may be enhanced by Atlas lee-cyclogenesis ultimately forming a Mediterranean cyclone .
Various measurement campaigns and modelling efforts have been carried out investigating the atmospheric dust cycle and related processes in order to improve the general understanding of dust suspended in the atmosphere and related atmospheric 10 processes. With focus on Saharan dust and related dust radiation feedbacks and cloud formation processes, several projects focussed on the North African continent and its adjacent dust outflow regions during the last decade: The SAMUM-1 experiment (Saharan Mineral Dust Experiment, Heintzenberg (2009)), which was conducted in Southern Morocco during May and June 2006 and aimed for characterizing Saharan dust particles to quantify dust-related radiative effects; the SAMUM-2 experiment (Ansmann, 2011), which took place during January and February 2008 at Cape Verde in the so-called dust outflow region, 15 and focused on the optical properties of aged dust and dust aerosol mixed with biomass-burning aerosol; the DODO (Dust Outflow and Deposition to the Ocean, McConnell et al. (2008)) experiments aiming at determining dust radiative properties over the dust outflow region in 2006; the Fennec project (Ryder et al., 2015) aiming at improving the understanding of physical processes that control the Saharan climate system and conducting field experiments during June and July 2011 and 2012; and the SALTRACE experiment, which was conducted in June to July 2013 and investigated the dust properties related to dust 20 radiative effects and micro-physical processes of aged Saharan dust that has been transported to the Caribbean (Weinzierl, in press). Although many studies focus on the continental region including the adjacent dust outflow, measurements during SALTRACE were obtained at both sides of the tropical Atlantic: at Cape Verde relatively close to the North African continent and at Barbados situated in the Caribbean Sea.
This study focuses on North African dust source activations and their controlling on atmospheric circulation patterns that ul- 25 timately determine dust transport towards the Caribbean Sea. It was carried out in the framework of the SALTRACE project and contributes to the understanding on the origin of dust aerosol transported to and eventually measured in the Caribbean (Weinzierl, in press). The aim of this study is twofold. First, it discusses the spatio-temporal distribution of active dust sources over North Africa regarding implicit information on the meteorological elements driving dust emission and their predominance during a typical summer season. To achieve this, the study makes use of satellite observations for identification of active 30 dust source regions. Second, the study elaborates the atmospheric circulation with regard to its relevance for dust source activation and dust transport towards the tropical North Atlantic. For this, simulations from the dust-atmosphere model system COSMO-MUSCAT are analysed. The period of investigation is chosen to cover the entire summer season June to August 2013.
Transport-related processes occurring during dust transport over the North Atlantic are presented in a companion paper.
The manuscript is structured as follows: Following a general introduction to atmospheric controls on dust source activation in 35 Section 2, an overview on the data sets used for the identification of dust sources and characterization of connected atmospheric circulation patterns is given in Section 3. Section 4 presents the spatio-temporal variability in dust source activation and leads to the analysis of different atmospheric features modulating the North African dust burden in Section 6. Aspects addressed therein are discussed in Section 7 with regard to its impact on dust export fluxes, and summarized in Section 8.
2 Atmospheric controls on dust source activation 5

Dust source characteristics
Dust emission is a threshold problem: Momentum provided by the wind is an essential prerequisite to mobilize soil particle from bare ground. Thereby, the minimum amount of momentum required for particle mobilization depends on surface characteristics such as soil texture and particle size distribution (Kok et al., 2012). However, dust mobilization may also be limited by the soil erodibility describing the susceptibility of the surface for wind erosion. The soil erodibility is determined by (1) 10 roughness elements such as vegetation, rocks or soil clods, (2) intrinsic characteristics such as soil texture, mineralogy, or soil content of organic matter, and (3) temporally varying characteristics such as soil moisture, soil aggregation, crusting and the availability of erodible material (Webb and Strong, 2011). All characteristics in concert determine the interparticle cohesive forces and thus the amount of momentum required for particle mobilization. Consequently, wind speed distribution and surface sediment characteristics together determine the dust emission flux. On a daily basis, in particular the supply of momentum 15 and thus the wind speed distribution is the dominant factor controlling dust source activation and subsequent dust emission, ultimately resulting into its diurnal cycle (Schepanski et al., 2009b). Additionally, precipitation and subsequent increase in soil moisture control dust emission in particular in the Sahel zone and the northern margins of the Sahara where convective precipitation associated with meso-scale convective systems (MCS) respectively cyclogenesis occurs (Fiedler et al., 2014;Bergametti et al., 2016). Vegetation cover and its changes contribute significantly to the dust variability, in particular at seasonal time 20 scales in regions with distinct vegetation periods.
The focus of this study is on the atmospheric controls on dust source activation and subsequent dust emission. It discusses different synoptic-scale features of the atmospheric circulation affecting the local wind speed distribution and thus inhibiting or forcing dust source activation, which ultimately result in a local variability in the occurrence frequency of dust source activation and consequently atmospheric dust concentrations downwind. 25

Atmospheric circulation over North African dust source areas
Besides the availability of sediments susceptible to wind erosion, the presence of sufficiently high wind speeds is the most limiting factor for dust emission. Although local wind speeds determine local dust mobilization and subsequent entrainment into the atmospheric boundary layer, the atmospheric circulation provides the general conditions for the state of the atmosphere at 30 local scale. For example, atmospheric turbulence, static stability, wind speed distribution and gustiness are affected by pressure gradients determining the prevailing geostrophic wind, and advection of humidity and temperature.
The atmospheric circulation over North Africa in boreal summer (June, July, August (JJA)) is dominated by three major features: (1) the Harmattan wind, (2) the Saharan Heat Low (SHL), (3) the West African Monsoon (WAM) circulation including the African Easterly Waves (AEWs), which are characterised by distinct troughs and ridges. The Harmattan names the north-easterly trade winds over North Africa that result from the continental scale pressure gradient between the subtropical 5 subsidence zone and the inter-tropical convergence zone (ITCZ). With regard to dust entrainment, these winds are of particular interest as they are involved in the development of the nocturnal LLJs, which are the dominant and frequent drivers for dust source activation over North Africa (Schepanski et al., 2009a, b;Fiedler et al., 2013;Schepanski et al., 2015a). As part of the WAM, cool and moist air is transported northward from the Gulf of Guinea into the North African continent. The resulting baroclinicity fosters the generation of nocturnal LLJ embedded in the monsoonal flow at the southern margins of the Saharan 10 heat low (Parker et al., 2005a;Bou Karam et al., 2008;Schepanski et al., 2009a;Fiedler et al., 2013). The atmospheric environment of moist monsoonal air further allows deep moist convection and the development of Mesoscale Convective Systems (MCSs), in particular along the AEJ that forms a wave-structure over the Sahel zone in summer. The SHL plays an important role for the strength of the monsoon trough (Thorncroft and Blackburn, 1999). Dust emission related to deep-moist convection also occurs along the Saharan side of the Atlas mountains, where orographically induced cold pools or lee-cyclogenesis result 15 into the generation of dust fronts Reinfried et al., 2009;Schepanski et al., 2009bBou Karam et al., 2010).

The Saharan Heat Low as connecting element in between the Harmattan and Monsoon circulation
The Saharan heat low (SHL) is a thermal low formed by intense solar heating of the desert land surface. The heated land 20 surface re-emits thermal energy into the lower atmosphere which subsequent leads to heated air layers. The heated air ascends due to its buoyancy and a zone of low pressure forms at near surface levels. As the SHL is a heat low, there are no fronts associated. As proposed by Lavaysse et al. (2009), the low-level atmospheric thickness (LLAT) can be applied as a measure to identify the geographical position, the extent, and strength of the SHL. Thereby, the LLAT is related to the mean temperature of the atmospheric layer confined by the 700 hPa and 925 hPa and thus represents the heat-induced dilatation of the lower 25 atmospheric levels.
The SHL is suggested to act as a key element of the atmospheric circulation over North Africa. During summer, it is situated over West Africa between the Atlas Mountains to the north and the Hoggar Massif to the east (Lavaysse et al., 2009). Due to its geographical position, its cyclonic circulation affects the strength of the Harmattan winds and the monsoon flow: Along its eastern flank the south-westerly monsoon flow is increased, along its western flank the north-easterly Harmattan winds are 30 increased (Lavaysse et al., 2009;Parker et al., 2005b). The SHL may even affect the position of the AEJ, and as the temperature gradient between the hot and dry Sahara and the cold and moist Gulf of Guinea impacts on the state of the AEJ. The SHL may even interact with the development of AEWs and the position of the monsoon front (Thorncroft and Blackburn, 1999;Chauvin, 2010;Lavaysse et al., 2010). The SHL also determines the ventilation of the North African continent through cooler, maritime air masses that are advected to the Saharan desert due to increased pressure gradients. On its eastern flank, cooler air masses originating from the Mediterranean Sea are transported into the North African continent eventually contributing to the atmospheric moisture budget over the Sahel (Vizy and Cook, 2009). On its western flank, air masses originating from the Atlantic ventilate the continent driven by the pressure gradient between the subtropical high (Azores high) and the SHL (Grams et al., 2010). Furthermore, Roehrig et al. (2011) found a modulation of the structure of the Azores high by the so-5 called pulsation of the SHL. The pulsation of the SHL characterizes the migration of the centre of the SHL from a so-called eastern phase to a so-called western phase (Chauvin, 2010;Roehrig et al., 2011). The western phase is associated with higher surface temperatures over the coastal region of Morocco and Mauritania, and a general south-westward propagation of this temperature signature. Lower surface temperatures are evident over the region between Sicily and Libya; this signature is propagating in south-easterly direction. The eastern phase is associated with the opposite temperature signature. As outlined 10 above, the position (eastern phase versus western phase, Chauvin (2010)), the depth and extent of the SHL generally interacts and affects the large-scale circulation over North Africa. In particular, the Harmattan and monsoon flow modulates the nature of the SHL and vice versa (Lavaysse et al., 2009).
Although dust source activation and subsequent dust emission is driven by local wind speed conditions, large scale circulation determines the local atmospheric conditions fostering or inhibiting dust emission. Modulated by and interacting with Harmattan 15 and Monsoon circulation, the SHL is an atmospheric element that on the one side can be characterised by its depth and extent, and on the other side which reflects implicitly both, the Harmattan and the Monsoon. In this study, we use the method proposed by Lavaysse et al. (2009) to identify the position, extent and depth of the SHL from COSMO-MUSCAT geopotential and temperature fields at 700 hPa and 925 hPa level. The area of the SHL is refined to the 90th percentile of the LLAT values.

20
Dust transport towards the tropical Atlantic and across to the Caribbean Sea is determined by the dust conditions over West Africa, the atmospheric circulation advecting the dusty air mass to the coastal regions off West Africa, the entering of the Atlantic region and the formation of the so-called Saharan Air Layer across the Atlantic ocean (e.g., Schepanski et al., 2009a).
Although dust emission is the vital element of the dust's journey across the Atlantic, the release of dust-loaded air from the North African continent to the tropical Atlantic is not inevitable. Dust may also be deposited over the North African continent. 25 Airborne dust circulating over Western Africa may become sucked into the SHL as described by Ashpole and Washington (2013) or form the ridge of an AEW (Jones et al., 2003). In both cases, dust may eventually be pushed towards the Atlantic and released as a dust plume once the SHL is in its western phase (Chauvin, 2010) or an AEW enters the Atlantic (Jones et al., 2003). Besides that, dust-loaded air masses are transported towards the Atlantic by the Harmattan flow. In that case, the dust export flux is rather continuous compared to the plume-like structure of dust export associated with the intermittent nature 30 of the SHL pulsation or the passage of AEWs. In summary, SHL and AEW can be understood as a modulation on the North African dust export within the Saharan air layer.

Data sets providing information on the atmospheric dust life-cycle
To elaborate various concepts of controlling mechanism on the atmospheric dust life-cycle, in particular dust emission and transport over the North African continent, this study combines two different data sets: (1) the DSAF (Dust Source Activation Frequency) data set providing information on the spatio-temporal distribution of active dust sources inferred from satellite observations, and (2) fields from numerical simulations of the atmospheric dust-life cycle using the atmosphere-dust model 5 system COMSO-MUSCAT.

Dust source identification
Satellite observations as these obtained from the geostationary MSG (Meteosat Second Generation) SEVIRI (Spinning-Enhanced Visible and Infra-Red Imager) instrument at 15-minute temporal resolution provide convenient data for identifying active dust source. As described by Schepanski et al. (2007), the infrared dust index, which is a red-green-blue (RGB) composite image 10 of different combinations of the brightness temperature of the three SEVIRI channels centred at 8.7µm, 10.8µm and 12.0µm wavelengths, is used to identify active dust source in terms of dust entrainment into the atmospheric boundary layer. Active dust sources are recorded on a 1 • × 1 • grid covering North Africa north of 10 • N building up the DSAF data set.
The dust index provides qualitative information on the presence of airborne dust and has previously been used to locate active dust source within the viewing field of SEVIRI over North Africa (Schepanski et al., , 2009b and southern Africa 15 (Vickery et al., 2013). As based on measurements at infrared wavelength, the index is sensitive to atmospheric humidity, which affects the identification of dust layer in regions with high atmospheric humidity such as cold pool outflows (Brindley et al., 2012). However, due to the visual identification of dust plumes the impact of this caveat is minimized compared to automatic retrieval algorithm. The sensitivity against colour shades and propagating plumes is much higher and provides reliable identification. As for most satellite-based observations, no information on the presence of dust is available underneath clouds.

20
Consequently, dust uplift in the vicinity of deep moist convection may remain unrecorded and may result into an underestimation of dust source activations driven by deep moist convection.
Besides the geographical location of active dust sources, the time of day when the dust source became active was recorded and binned at hourly resolution. In case of several dust source activation per 1 • × 1 • grid cell, the earliest one determined the recorded time. A comparison of dust source data sets based on 15-minute resolution satellite observation and daily observations 25 has pointed towards a shift in dust source regions identified in downwind direction due to the temporal off-set between onset of dust source activation and satellite overpass .

COSMO-MUSCAT dust simulations
Fields describing the state of the atmosphere and the dust aerosol distribution are taken from the meso-scale atmosphere- Aerosol Transport Model), which is a non-hydrostatic atmosphere model (COSMO version 5.0) that is coupled on-line to the 3D chemistry tracer transport model MUSCAT. Schemes on dust emission (Tegen et al., 2002) and deposition (Berge, 1997;Jakobson et al., 1997;Zhang et al., 2001) are implemented in MUSCAT and on-line driven by atmospheric and hydrological fields calculated by COSMO (Heinold et al., 2011). Thus, the full atmospheric dust cycle consisting of dust mobilization and emission, transport, and dry/wet deposition is represented by the COSMO-MUSCAT simulations. Five size bins ranging from 0.1 to 48µm are defined to resolve the size dependency of dust transport and deposition. Depending of the state of the atmosphere, in particular the near-surface wind distribution, dust emission fluxes are calculated for regions identified as active 5 dust source using the dust source activation frequency (DSAF) mask inferred from MSG SEVIRI observations as described by Schepanski et al. (2007). By applying this DSAF mask, all grid cells, where within the time period 2006-2012 dust source activation was observed at least twice, were treated as potential dust sources. The dust flux then is calculated using the actual wind fields and atmospheric/hydrological conditions calculated by COSMO and considering soil texture and soil size distribution as proposed by Tegen et al. (2002). Additionally, the aerodynamic roughness length z 0 is implemented for adjusting the dust Active dust sources were inferred from MSG SEVIRI IR dust index images for the time period 1 June to 31 July 2013 as outlined in section 3. The two-month period covers the SALTRACE field campaign, which took place from 10 June to 14 July 2013. Whereas the dust measurements obtained over Barbados characterise the last part of the atmospheric dust cycle and thus the final stage of the life-cycle of dust in the atmosphere, an encompassing analysis includes the discussion of dust sources and dust emission processes. E.g., Groß et al. (2015) identified different phases of atmospheric dustiness over the Caribbean 30 and Barbados, which were characterized by differing optical properties and atmospheric aerosol layer structures. These phases were related to active dust source regions in order to investigate the relation between dust source characteristics, dust transport route and arrival over the Caribbean. Here we focus on locally occurring wind speeds that are sufficiently high to foster dust entrainment. The diurnal wind speed distribution is determined by the large-scale atmospheric circulation modulated by smaller-scale processes such as regional wind regimes and local boundary layer processes. Prominent examples for large-scale features are the Harmattan or WAM circulation; regional wind regimes are for example the mountain and valley breeze, but also the formation of LLJs. As topography 5 and atmospheric circulation, which determine the local wind speed distribution and thus the frequency of fulfilling the major atmospheric precondition for dust emission, are not equally distributed over North Africa, spatial and temporal variabilities in dust source activation are expected.
The analysis of the spatial distribution of frequently active dust sources as shown in Fig. 1a illustrates that frequently active dust sources that also dominate the June-July 2013 period are located within the mountain regions over North Africa besides 10 the dust sources located in the Bodélé Depression. In particular, dust source embedded in the mountain foothills of the Hoggar-Adrar-Air Massif become frequently active during this period. Dust sources with pronounced activity are also observed over the Al Jabel Akhadra Massif in Libya and the southern fringe of the Atlas Mountains. Also, dust sources over southern Mauritania are found to be active. The predominance of dust source embedded in mountain regions agrees with earlier findings based on the DSAF data base (i.e., Schepanski et al., 2009bSchepanski et al., , 2012. 15 Comparing the June-July 2013 period to the four-year June-July 2006-2009 period (Fig. 1b), the patterns of the spatial distribution of DSAF are matching. However, from a broader perspective, the level of occurrence frequency is generally higher for  The temporal distribution of dust source activation, in particular its diurnal cycle, provides implicitly information on the meteorological driver fostering dust source activation and subsequent dust entrainment into the atmosphere (Schepanski et al., 2009b). Dust source activations during the first half of the day are predominantly linked to the break-down of the nocturnal LLJ, whereas dust source activations occurring during the second half of the day are predominantly related to (moist) convection (Schepanski et al., 2009b). The climatology of the occurrence frequency of the LLJ shows in particular high frequencies 25 where the Harmattan is the dominating wind regime, and over region of maritime air inflow such as the Atlantic and Mediterranean ventilation areas, and along the northern front of the WAM (cf. Schepanski et al., 2009b;Fiedler et al., 2013). Moist convection and related MCSs develop more frequently towards the south, where the arid Sahara transition into the semi-arid Sahel zone. With northward propagation of the monsoon front, MCSs frequently migrate into the desert, where downdrafts generate dust storms, which are frequently associated with the formation of impressive dust fronts -the so-called Haboobs.

Atmospheric dust distribution: Validation of dust aerosol optical depth
The atmospheric dust load is often used to assess the different elements of the atmospheric dust life-cycle consisting of (1) dust emission, (2) dust transport, and (3) dust deposition. Dust emission is determined by source characteristics (i.e. sediment 10 supply and availability) and wind strength, and thus dust emission fluxes are a representative for the connection of these factors.
Dust transport results from prevailing wind regimes, but also particle buoyancy, which is influenced by atmospheric stability  As the COSMO-MUSCAT AODs consider dust aerosols only in these simulations analysed here, sun-photometer measurements in contrast are affected by all types of aerosol particles (i.e., mineral dust, soot, sea spray aerosol), coarse-mode AODs (O'Neill et al., 2003) predominantly represent the dust fraction. This is a common approach in order to allow for the best Atmos. Chem. Phys. Discuss., doi: 10.5194/acp-2017-309, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 15 May 2017 c Author(s) 2017. CC-BY 3.0 License. comparability with model simulations where mineral dust is the only simulated type of aerosol Schepanski et al., 2015b. The comparison of COSMO-MUSCAT dust AODs with AERONET coarse mode AODs as shown in Fig. 3 depicts the model's ability to simulate the temporal variability of the atmospheric dust load over the particular sites. Generally, the model represents the range of AOD values and the temporal evolution of the coarse mode AOD derived from sun-photometer measurements at these sites well, although some minor difference on single event basis are evident. This sug-5 gests that COSMO-MUSCAT sufficiently well captures the atmospheric dust life-cycle including its relevant and determining atmospheric processes. In particular, matching dust AODs suggest that COSMO-MUSCAT is able to capture the meteorology correctly and, furthermore, to balance dust emission dust removal fluxes sufficiently good. The model results therefore are trustworthy for deriving the relations between the dust cycle and the drivers of atmospheric circulation in the study period.
6 Atmospheric circulation modulating the atmospheric dust burden over North Africa 10 In the following subsections, the role of the Harmattan, the SHL and the WAM will be examined exemplarily for summer (June-August) 2013 with regard to their impact on the atmospheric dust load and dust transport towards the north Atlantic.

Harmattan
The strength of the Harmattan, which names the trade winds over North Africa, is determined by the pressure gradient between the subtropics and the tropics. Due to the geographic situation, the gradient is strongest over the North African continent. The in the Harmattan flow. The index is similar to the North Atlantic Oscillation (NAO) index, but represents the pressure gradient over the North African continent, which is of direct relevance for the strength of the Harmattan winds. The NAFDI is the normalized difference of the anomaly of the geopotential Φ at 700hPa between two 3 • × 3 • boxes located north and south of 20 the Sahara desert. For this study, the NAFDI index F is calculated as follows: with Φ n denoting geopotential averaged over a box located north of the Sahara ( Similar to the distribution of the geopotential reflecting the atmospheric circulation at low level, composites are calculated for 15 the atmospheric dust loading expressed as dust AOD (Figs. 5a and b) and the meridional wind components (Fig. 6). The AOD composites for low and high NAFDI index values address the question for a link between atmospheric dust load and NAFDI phase. To illustrate the difference in AOD during particular low respectively high NAFDI phases to the JJA average, the relative differences are shown in Figs. 5c and d. As the NAFDI is based on the pressure gradient and so is the wind determining dust entrainment and transport, composites of the meridional wind speed complement the relation between NAFDI and dust AOD.

20
In order to distinguish between north-easterly Harmattan and south-westerly Monsoon inflow, the meridional wind components are considered separately for northerly (representing Harmattan conditions) and southerly (representing Monsoon conditions) winds. The composite shows the average wind speed for the corresponding wind direction over the selected time period. In agreement with a weaker Harmattan for periods characterised by low NAFDI index values (Fig. 6), the dust plume over the Sahara extends further towards and into the Mediterranean basin as southerly winds prevail (Fig. 5). As the pressure gradients 25 are weaker, less dust is suspended in the atmosphere resulting in lower dust AOD values. During periods characterised by positive NAFDI index values, the atmospheric dust loading over the central Sahara at about 20-25 • N is increased and so is the dust AOD over the dust outflow region off the West African coast. This situation is complemented by the dust export flux (Fig. 7), which shows that the core of the dust flux at 20 • W is shifted to the south for days characterised by a particular low (negative) NAFDI index. Furthermore, the re-circulation of dust resulting into positive (eastward) dust fluxes at 20 • W 30 is enhanced compared to the JJA 2013 mean. The dust export (westward dust flux) is enhanced for days characterised by a positive NAFDI index representing strong Harmattan winds.

Saharan Heat Low
The low-level atmospheric thickness (LLAT) between 925hPa and 700hPa is used to identify the SHL as suggested by and significantly stronger phase occurred during 14 to 28 July 2013, the period which also includes the overall maximum LLAT depth for that summer. The LLAT maximum values were continuously above the 75th percentile, expect for one day (21 July).
The second, stronger SHL phase is framed by two periods of a significantly shallow SHL with LLAT maximum values below 20 the 25th percentile (2692m) during 9 to 11 July and 31 July to 13 August. The latter phase was briefly intermitted by three days of increased LLAT values, however, the generally shallow extend of the SHL dominated.
The dust AOD distribution varies among the different strengths of the SHL, which is illustrated in Fig. 9 for the 75th percentile (a) and 25th percentile (b) of the LLAT respectively. A particular deep and strong SHL (75th percentile) is accompanied by higher dust AODs over the central Sahara compared to a particular shallow and therefore weak SHL, which is accompanied by 25 a generally lower AOD levels. Regarding the dust export towards the east Atlantic, the dust outflow region is slightly shifted to the south for the stronger SHL condition, however, the dust AODs corresponding to the dust outflow region are at a comparable level. Comparing the dust AOD distribution for these two contrasting SHL conditions, significant differences are evident. The differences are strongest north of 35 • N, where the dust AOD level is significantly enhanced over the Atlas region due to the unusual northward extend of the Saharan dust plume during strong SHL conditions (Fig. 9a and c). Also, the level of dust AOD AOD levels are evident over coastal Libya, associated with the generally low atmospheric dust load over that region (Fig. 9b and d). This may be due to the more southerly position of the SHL during these periods.

West African Monsoon
The West African Monsoon circulation is driven by the Hadley-circulation and supported by the gradients in temperature and humidity resulting from the contrast between the hot and dry Sahara and the humid and cooler Gulf of Guinea (Parker et al., The objective of this study was set to investigate the atmospheric controls on the atmospheric dust life-cycle with particular focus on the dust export towards the Atlantic and ultimately towards the Caribbean Sea. Apparently, dust emission and transport pathways over the North African continent determine the dust export. The amount of dust crossing the 20 • W meridian 30 14 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-309, 2017 Manuscript under review for journal Atmos. Chem. Phys. is characterised by a function of latitude, vertical height and time describing the intermittent nature of the dust flux towards the eastern north Atlantic. Figure 11 presents its variability over time for June to August 2013. The figure panels ( Fig. 11a and Fig. 11b) are complementary: Figure 11a illustrates the predominant height of dust transport, but does not distinguish for different latitudes. Figure 11b pictures the variability of the vertically integrated dust transport fluxes over the different latitudes disregarding the variability in transport height.

5
The height of dust transport through 20 • W is crucial for the cross-Atlantic transport. Dust transported within the Saharan Air Layer (SAL), which is situated above the trade wind inversion at heights between 800 and 500hPa is often described as the transport-highway for dust across the Atlantic. Whereas dust transported within the marine boundary layer shows enhanced dust removal rates, and consequently high deposition fluxes, dust transport within the SAL is more efficient and higher dust concentrations are more likely to reach the Caribbean. Generally, dust export within the SAL is characteristic for the summer 10 season (e.g., Schepanski et al., 2009a).
Bringing together Figs. 11a and 11b, JJA 2013 dust transport through 20 • W occurred predominantly within the SAL and showed an intermittent character, which is typical for summertime dust export (Jones et al., 2003;Schepanski et al., 2009a).
The intermittent character of dust plumes or dust pulses leaving the North African continent and entering the Atlantic becomes also evident when considering the dust flux through 20 • W as a function of latitude and time (Fig. 11b).The major fraction of