Model Simulations of Complex Dust Emissions over the Sahara during the West African Monsoon Onset

The existing limitations in ground-based observations in remote areas in West Africa determine the dependence on numerical models to represent the atmospheric mechanisms that contribute to dust outbreaks at different space-time scales. In this work, the ability of the Weather Research and Forecasting model coupled with the Chemistry (WRF-Chem) model using the GOCART dust scheme is evaluated. The period comprises the West African Monsoon onset phase (the 7th to 12th of June, 2006) coinciding with the AMMA Special Observing Period (SOP). Different features in the horizontal and vertical dynamical structure of the Saharan atmosphere are analyzed with a combination of satellite and ground-based observations and model experimentation at 10 and 30 km model resolution. The main features of key Saharan dust processes during summer are identifiable, and WRF-CHEM replicates these adequately. Observations and model analyses have shown that cold pools (haboobs) contributed a substantial proportion of total dust during the study period. The comparative analysis between observations and WRF-Chem simulations demonstrates the model efficiency to simulate the spatial and 3D structure of dust transport over the Sahara and Sahel. There is, therefore, a strong basis for accurate forecasting of dust events associated with synoptic scale events when model dust emission parameterization is suitably calibrated.


Introduction
Mineral dust is the dominant natural atmospheric aerosol by mass [2], and although the extent of its contribution is still uncertain, it plays an important role in the Earth climate system [3,4]. The Sahara Desert in North Africa is the world most important dust source with ∼50% of the total dust global mass [5]. Following field campaigns and new satellite data information, aerosol and climate scientists have increased our understanding of the Saharan dust cycle, particularly (i) preferential dust sources and especially those associated with topographic and geomorphological features including the Bodélé Depression [6,7] the flanks of the Hoggar and Air mountains, amongst others [8], (ii) the complex meteorological processes leading to dust lifting in these sources (for a schematic of these mechanisms, refer to Cuesta et al., [9], Figure 1), (iii) the dynamical mechanisms controlling the structure of the Saharan atmosphere and the role of dust, and (iv) the radiative impact of dust aerosols.
Dust emissions over North Africa are closely coupled to the large-scale atmospheric circulation in the region and the pronounced seasonal and diurnal cycles [10][11][12]. In essence, the main seasonal climatic feature in North and West Africa is the annual cycle of dry and rainy monsoon seasons resulting from the north-south migration of the Intertropical Convergence Zone (ITCZ). The Intertropical Discontinuity (ITD) lies at the interface between the low-level moist southwesterly monsoon flow and the dry northeasterly "Harmattan"fl o w [13]. Harmattan winds are driven by the north-south pressure gradient across the zonal extension of the North African subtropical anticyclone (the Libyan high) and the monsoon trough, predominantly during winter months (October-March) over the entire Sahara-Sahel, as the ITD lies at ∼5 • N, close to the Gulf of Guinea coast.
Day-to-day synoptic scale variability in the Harmattan winds is associated primarily with variability in the mid-latitude circulation, especially ridging of the Libyan high [14], coastal "Sharav" cyclones [15],andcoldsurges [16]. Accordingly, during winter, dust events are generally associated with anomalously strong Harmattan wind events at the synoptic scale [17]. Within this, there is a pronounced diurnal cycle of dust emission, which peaks in the midmorning. This is associated with the diurnal cycle in low level and surface winds driven by surface heating and cooling through the inertial oscillation mechanism [18]. This results in an outof-phase coupling between nocturnal low-level jet maximum and midmorning surface wind maximum as momentum from the LLJ is mixed down to the surface by heating in the hours after sunrise [17,19]. In boreal summer, the West African Monsoon (WAM) dominates the circulation over the Sahara-Sahel region [20] as the ITD lies at its northernmost location ∼15-20 • N, separating the relatively moist, cool monsoon southwesterlies and the hotter, dry northeasterlies. A key driver of the circulation over the western Sahara during the summer is the Saharan heat low (SHL), which in June is centred close to 0 • E, 20 • N where surface temperatures peak [21,22]. Data limitation in the central Sahara has restricted analysis of the SHL, but studies based on satellite observations and model experiments highlight a contribution of the SHL to the enhancement of meridional monsoon circulation [9,23,24]. In summer midtropospheric flow is dominated at ∼600 hPa by the African Easterly Jet (AEJ), centred at ∼10 • N, driven to the first order by the surface temperature gradient between the cool equatorial Atlantic and the Sahara [25]. Instabilities in the AEJ lead to the formation of synoptic scale African Easterly Waves (AEWs) [26]. The AEWs typically form west of 20 • E, propagate westward, and play an important role in the organization and propagation of Mesoscale Convective Systems (MCSs) over the Sahel and occasionally the Sahara [27]. MCS activity can be organized along a squall line at leading edge of the upper level AEW [28]. Surface troughs embedded in the AEW are sometimes referred to as Sudano-Sahelian depressions (SSDs) and can propagate towards the northwest into the Sahara where they can become cut-off lows [29].
During the WAM onset in May-June an extended portion of the western Sahara and Sahel exhibit high dust loadings with a maximum approximately collated with the SHL extending over southern Algeria, northern Mali, and northeast Mauritania [11,12]. It is now believed that this aerosol maximum is due to a variety of emission and transport processes. On the synoptic scale, surface troughs frequently within AEWs [30] create anomalously strong winds and often dust emission in both the dry northeasterly flow leading the trough [31] and the moist southwesterly flow behind (e.g., the monsoon surges documented by Flamant et al. [32] and Bou Karam et al. [33]). Once again, surface wind velocity is closely coupled to the diurnal cycle of the nocturnal LLJ in both the Harmattan and monsoon flow [23].
At the mesoscale, gravity currents within MCSs generated by subsidence of air cooled by evaporation in dry air [34,35] can cause strong surface winds. These cool, moist outflows ("cold pools") from deep convective systems cause the characteristic dust fronts called Haboobs [36][37][38][39], which can propagate at speeds on the order of 10 ms −1 for more than 24 hours, over many hundreds of kilometers, sometimes far away from the parent convective clouds [40]. In addition, small-scale turbulence from dry convection [21]h a sb e e n implicated in dust emission [12], although the importance of this mechanism has been a subject of debate [41]. Once dust is uplifted by these various mechanisms during summer, it is quickly mixed vertically by dry convection throughout the depth of the planetary boundary layer, which can exceed 5 km over the central Sahara, and by upgliding over the more dense monsoon flow [42]. Subsequently, dust is typically advected westward over the Atlantic within the elevated Saharan Air Layer (SAL) [43].
Model representation of the complex features affecting the dynamics of atmospheric circulation, which determine dust mobilization, is still a matter of intensive study. Adequate model simulation requires that models represent the large-scale, synoptic and mesoscale features of the Saharan atmosphere. Based on these premises, the ability of the Weather Research and Forecasting (WRF) model with a Chemistry module (WRF-Chem) to represent these key multiscale processes associated with dust mobilisation in the Sahara during the WAM onset is evaluated. The analyzed period (7th-12th of June 2006) corresponds to a period which encompasses the complexity of processes that operate during summer. This work is organized as follows. Section 2 describes model configuration and employed data for validation. Section 3 provides the results of model experiments including synoptic scale conditions and dust emissions and transport associated to the atmospheric mechanisms described previously. Section 4 summarizes the findings made in this work and conclusions.

WRF-Chem Model
Description. WRF is a next-generation mesoscale forecast model and assimilation system to advance both the understanding and prediction of weather. WRF is applied in a variety of research and operational contexts, from storm scale to regional climate scale [44]. Meteorology   3 The WRF-Chem incorporates a chemistry module that simultaneously simulates the emission, turbulent mixing, transport and chemical transformation, of trace gases and aerosols [45]. In this work, WRF-Chem implements the Goddard Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) dust scheme, which includes emission, advection, and deposition and is fully described in Ginoux et al. [46]. In essence, the GOCART dust scheme simulates dust emission as a function of surface wind speed and surface erodibility and surface wetness [47]. Emission flux Fp for aerosol particle size group p is expressed as

Advances in
where C is a dimensional, scheme-specific constant of proportionality equal to 1 mg s 2 m −5 and S is the erodibility source function, both values originally prescribed as in Ginoux et al. [46]. sp is the fraction of each size class within the soil, u 10 m is the wind speed at 10 m, and u t is the threshold velocity of wind erosion, determined by particle size and surface wetness. Due to the uncertainties in the available soil texture data, a simple particle size distribution following Tegen and Fung [48] is implemented. The fraction of clay is based on the assumption that erodible clay represents 1/10 of the total mass of emitted silt and that of each silt subclass is assumed to be the same. The sp values are thus 0.1 for the class 0.1-1 mm and 0.3 for the classes 1-1.8 mm, 1.8-3 mm, and 3-6 mm, respectively. The source function S is defined from an erodibility map representing the fractional grid cell area of alluvium available for wind erosion as follows: where S is the probability of accumulated sediments in the grid cell i of altitude zi and z max and z min are the maximum and minimum elevations in the surrounding 10 × 10 topography, respectively. Only land surface with bare soil is considered as possible dust sources. The bare soil surface map is obtained from the 1 × 1 vegetation data set derived from the Advanced Very High-Resolution-Radiometer (AVHRR) data [49]. In general, source regions are characterized as erodible fraction areas higher than 30% particularly with an elevated proportion of sand. For this research, a new version from this erodible fractional map derived from AVHRR at a higher resolution 0.25 • × 0.25 • is used ( Figure 1). This new map provides a more detailed representation of the source regions within the Bodélé Depression, Chad, and identifies smaller local dust sources elsewhere. Like most dust emission schemes in models it is necessary to "tune" emission to account for many unconstrained parameters the magnitude of some depending on model horizontal resolution. As described above, constant C in (1)i sa n empirical proportionality constant that was originally estimated based on regional datasets. In this analysis, we tuned this constant C as in Zhao et al. [50]. These authors tuned C values of 0.65 in contrast to the 1 mg s 2 m −5 from Ginoux et al. [46] to replicate consistence with AERONET datasets. Following a similar approach, we did fix this to value 0.5 mg s 2 m −5 to produce consistency between model   Model configurations selected are based on previous sensitivity experimentation of the different available cumulus, PBL and microphysics parameterization by the authors not included in this work and the findings of several sensitivity analyses documented by the modeling community (e.g., [51][52][53]). The cumulus parameterization scheme employed is the G-3D, which is an improved version of the Grell-Devenyi (GD) ensemble cumulus scheme [54]. The GD scheme has been recommended for smaller model grid scales (e.g., 10-30 km) as it tends to allow a balance between the resolved scale rainfall and the convective rainfall. Multiple cumulus ensembles are averaged within each grid box in an attempt to represent different cloud processes, such as detrainment and precipitation as well as explicit representation of updraughts and downdraughts. The G-3D version also allows for subsidence in neighboring columns. Sensitivity experiments in the WRF using the GD scheme have shown a good performance to detect convection development [55]. The Planetary Boundary Layer (PBL) scheme utilized is the Mellor-Yamada-Janjic TKE scheme as it has proved to represent more realistically in terms of temperature and humidity and PBL height [53].
To test the sensitivity to model horizontal resolution two WRF-CHEM experiments are performed, with model grid sizes of 30 and 10 km, EXP30 and EXP10, respectively. In the vertical plane the model has 50 vertical levels heavily biased towards the boundary layer (22 layers in the bottom 1 km). The two model domains encompass 5 • to 30 • N-20 • W to 40 • Ea n d8 • N-29 • N, 12 • W-22 • E for the 30 km and 10 km grid resolutions experiments, respectively. The period considered in this study is part of the Special Operation Period Phase 1 (SOP1, 2006) of the African Monsoon Multidisciplinary Analysis (AMMA) during the monsoon onset (May-June). This campaign consisted of a multiscale and multiprocess analysis of the west African monsoon [56]. Refer to Redelsperger et al. [57]f o rac o m p r e h e n s i v ed e scription of the AMMA field campaign. The initial boundary conditions are generated from NCEP Final Analysis (FNL from GFS) 1 × 1 degree resolution, every 6 hours. Model simulation EXP30 is performed from 5th to 15th June 2006, whilst EXP10 focuses on a shorter period from 7th to 10th of June, when mesoscale convection was particularly active over the southern Sahara.

Observational Datasets.
The spatial distribution of modelled dust is compared with a comprehensive set of satellite datasets providing information on atmospheric dust aerosols. These consist of (i) the EUMETSAT dust product color-composite from the SEVIRI sensor available at a 15minute temporal and 5 km nadir spatial resolution; the dust Product is generated from the SEVIRI Brightness Temperatures in three channels located in the infrared (8.7, 10.8, and 12.0 µm) [58] and is a qualitative dataset proven useful to identify active dust sources (e.g., [8]); (ii) AOT from MODIS-Aqua Collection 5.1 Level 2 (10 km spatial resolution, daily at local noon) using the latest version of the Deep Blue algorithm [1], providing AOT even over highly reflective desert surfaces; (iii) several transects from the CALIOP Lidar Version 3 Level 2 extinction coefficient at 532 nm and the Vertical Feature Mask data product (VFM) with 30 m vertical resolution that are used to evaluate dust vertical structure [59,60]. Although the L2 extinction coefficient retrieval from CALIOP is still a beta-quality product, it is useful to evaluate model dust profile structure in combination with a more validated VFM product. Meteorological data within Sahara itself is sparse and we use observations enhanced by the AMMA project from within the domain of EXP10, specifically surface SYNOPS reports from Agoufou and Bamba ( Figure 1). June, 2006. The meteorological situation during the period under analysis has been partly described in Flamant et al. [64]. These authors focused their analysis on dust emission and transport from the Bodélé Depression and Sudan regions from the 9th to the 14th of June, 2006. In the present research, analysis is extended to include the whole extent of the Sahara-Sahel. The large-scale circulation simulated by WRF ( Figure 2) at 06 UTC each day (the approximate time of maximum nocturnal low-level monsoon winds [23]) is consistent with reanalysis data (not shown) and shows a number of key features: (i) a positive sea level pressure (SLP) anomaly over the central/eastern Mediterranean and Libyan sector driving enhanced northeasterly Harmattan winds over the eastern sector of the Sahara (Chad-Sudan); although this SLP anomaly declines gradually over the period, northeasterly Harmattan winds are strong throughout; (ii) a series of westward propagating tropical troughs characteristic of SSD events (features SSD1 and SSD2 in Figure 2), which curve northwestward as they approach the western extremity of the Sahara. The propagation of the SSD troughs is highlighted in the SLP time-longitude analysis ( Figure 3)at15 • N and suggests a displacement speed of about 880 km day −1 consistent with AEW activity at 15 • N in the study of Kiladis et al. [65]. Figure 2 highlights pronounced meridional ITD perturbations associated with the passage of SSD troughs. Moist (dry) air is advected northward (southward) within the southwesterly (northeasterly) flow behind (ahead of) the center of the SSD trough. For example, monsoon "surges" extend as far north as ∼21 • Na t∼0-5 • E on the 7th-8th of June, associated with SSD1 (Figure 2(b))and20 • E on the 8th-9th of June, associated with Feature SSD2 (Figure 2(d)), and play an important role in preconditioning moist convection over the Sahara as described in later sections.

Mesoscale Moist Convection.
Large-scale organized convection was active over the Sahel south of ∼15 • N throughout the study period. Of particular interest were MCS events further north than normal (Figure 4) within the monsoon surges around the SSD troughs, specifically, (i) deep convection on the nights of the 8/9th of June at ∼0 • Eo v e r Mali/Algeria (Feature DC1 in Figure 4(c)) and (ii) on the evening and night of the 9/10th of June over the Hoggar mountains of Algeria (Feature DC2 in Figure 4(e))a n d further east at ∼15-20 • E over Niger/Chad on the early morning of the 10th of June (Feature DC3 in Figure 4(e)). These MCS events produced pronounced cold pool outflow winds, evident from satellite as "haboob" dust emission events (Features H1, H2, and H3, resp., in Figure 4).

Dust Emission and Transport Model Simulations and
Observational Data. The following sections describe the synoptic and mesoscale low-level wind features combined to drive a complex set of dust emission events over the Sahara during the study period. Dust source regions and plume transport can be most unambiguously identified from the high-temporal-resolution SEVIRI imagery ( Figure 4).
Quantitative estimates of AOT from MODIS are shown in Figure 5 whereas WRF-CHEM EXP30 AOT estimates for comparison are shown in Figure 6.T h ec o n t r i b u t i o n of processes identified for dust emissions and transport in the model and from observations over the study period is analysed below.

Dust Emission Associated with Enhanced Harmattan
Northeasterlies. Dust emission is apparent from a number of source regions in Sudan and Chad in the pink product derived from SEVIRI imagery, including the BodéléDepression every day from the 8th to the 11th of June (Figure 4 features B1 to B4, also apparent as high AOT values in Figures  5 and 6 from MODIS and model estimations). Over the Bodélé dust emission appears most active on the 9th of June ( Figure 6). The same day peak wind speeds at 10 m height Advances in Meteorology exceed the 18 m s −1 as the synoptic scale flow is channeled between the Tibesti and the Ennedi. Dust emission shows the diurnal cycle with the characteristic late morning peak in dust activity, which has been documented by Washington et al. [6] and Todd et al. [19]. This is the result from the out-of-phase relationship of LLJ, and surface wind speed is well represented by WRF-CHEM EXP30 (Figure 8). The dayto-day variability in wind speeds over the Bodélém a t c h e s that in approximate dust plume magnitude with peak winds and plume size on the 9th of June (Figure 6(a))wh e n1 0m wind speeds in excess of 14 m/s are simulated (Figure 8(a)).
High wind speeds are maintained the following two days (the 10th and 11th of June). This coincides with high dust emissions from this region as shown in the SEVIRI images ( Figure 4 features B1 and B2). WRF-Chem EXP30 represents both the Bodéléplumeswell( Figure 6) (Figure 7). To the west, in southern Algeria weaker dust emission is evident on the 7th and 8th of June (Features A1 and A2 in Figure 4) where a topographic influence is apparent. Discrete dust plumes emerge during the morning from the lacustrine deposits west of the town of In Salah driven by high wind speeds up to 12 m s −1 , accelerated around the northern flank of the Hoggar mountains. These dust plumes are well represented in WRF-CHEM (Figure 6), and the model indicates a diurnal cycle of LLJ and surface winds similar to but weaker than that over the Bodélé (Figure 10(b)), which drive the characteristic diurnal cycle in dust emission creating plumes A1 and A2.
Other active dust source areas influenced by low-level northeasterly winds are located in North Sudan (over an area approximately bounded by 15-20 • N, 20-25 • E). The Sudanese sources have been reported to be particularly active in May-July by several authors [64,66]. Clear examples are regions near the Nubia Desert in northern Sudan where the topography is mainly characterized with dense drainage channels like wadis likely to provide fluvial sediments. A particularly active source is located especially close to 16 • N, 25 • E. Generally, WRF-Chem captures these features reasonably well although it produces a rather generalized emission without representation of the major localized sources. Overall, WRF-CHEM captures well the space/time distribution of dust plumes from a number of preferential source regions in the Sahara, activated by synoptic scale winds constrained by topographic influences, although the fine detail of plume source location is lost in some cases. (Haboobs). The most prominent "haboob"e v e n t sa r eo b s e r v e d in SEVIRI imagery and are labelled H1-H3 in Figure 4.

Dust Emission Associated with Cold Pool Features
Time development of H1 case can also be followed in Figure 9. Note that the "haboob" dust front may only activate emission for a short period of perhaps a few hours. Subsequently, the dust plume is transported and modified. The major events are (i) over northern Mali/southern Algeria on the 8th/9th of June (Feature H1 in Figures 4(c)-4(d)), (ii) over central-southern Algeria on the 10th of June (Feature H2 in Figure 4(f)), and (iii) over eastern Niger on the 9th/10th of June (Feature H3 in Figure 4(f)). In the case of (i) extensive deep convective systems (Feature DC1 in Figure 4(c)) developed in a SW-NE oriented squall line extending across Niger in the evening of the 8th of June, propagating westward towards the Mali/Niger border on the night of the 8th-9th of June, within the monsoon extension around the eastern flank of SSD1 (Figure 2(b)). Th e s ec o n v ecti v ec ell sg e n e ra t edc o l dpoo lo u tfl o w sevi d e n t from the associated dramatic haboob dust front extending at least 1000 km in length from central eastern Mali to the Hoggar uplands of southern Algeria at 00 UTC on the 9th of June. This dust front propagated towards the NW to result in a dust plume extending from central Mali to southern Algeria by 12 UTC on the 9th of June, and then transported north to central-western Algeria on the 10th of June, in the northward intrusion of the monsoon flow. This extensive dust emission is not accurately simulated by WRF-Chem EXP30 at 30 km resolution (Figure 6(c)). Although a weak dust plume oriented SW-NE across Mali-Algeria is evident, the emission occurred in the midmorning hours of the 9th of June associated with mixing of the nocturnal monsoon LLJ to the surface after sunrise, reminiscent of the mechanism identified by Bou Karam et al. [33]. In contrast, the model simulation EXP10 at 10 km resolution resolves well the timing and extent of convective systems within the monsoon surge and, crucially, the associated westward propagating cold pool outflows ( Figure 12). In the model wind speeds at 10 m height in excess of 18 m s −1 are simulated along these moist gust fronts over western Niger and eastern Mali over the period 18 UTC on the 8th to 06z on the 9th of June (Figure 9), which generate dust emission subsequently transported towards the northwest with elevated AOT values. These simulated gust/dust front features map well in terms of timing and location with the observed dust front (Feature H1 in Figure 4). Unfortunately, it is not possible to verify quantitatively the model AOT with MODIS AOT product set due to low data quality caused by cloud screening in this particular area during this period.
The time series of 10 m wind speed ( Figure 10) in EXP10 compare well with the observed wind speeds at stations Bamba and Agoufou. The timing of the passage of the cold pool events on the early hours of the 9th of June is coincident with observations although the magnitude of peak winds is underestimated. The timing and magnitude of peak wind speeds are far better resolved in the 10 km model run EXP10 than in the 30 km run EXP30.
During the night of the 9th and daytime of the 10th of June further deep convection occurs over central southern Algeria (Feature DC2 in Figure 4(e)). Associated haboob events are evident over central-western Algeria (H2) and augment the existing dust plume. Moist convective processes Advances in Meteorology Figure 4: SEVIRI false color dust product imagery at 1200 UTC each day over the study period. Pink tones indicate dust and red-brownblack tones indicate clouds. Letters correspond to major dust and cloud features of interest (see text for details).    the Harmattan flow on the 8th of June lays over eastern Niger by 1200 UTC on the 9th of June (Feature B2, Figure 4(d)). On the 10th of June the cold pool outflow from MCS system DC3 (Figure 4(f)) is evident from SEVIRI data as it both generates dust and advects northward the existing dust plume B2 over Niger. The resulting "hybrid" plume constitutes the eastern axis of the inverted "V" shape, lying over western Niger/southern Algeria on the 10th-11th of June. This was probably augmented by further emission over centraleastern Mali on the 11th of June associated with further scattered convective systems, constituting the western axis of the inverted "V", although cloud cover renders interpretation of emission sources and processes problematic. It appears that the inverted "V-" shaped plume at 1200 UTC on the 11th of June (Feature H3 + B2, Figure 4  moist airmass region, a feature also noted by Knippertz and To d d [ 31]. By the 12th of June the "V" plume has dispersed and dust transported south where peak AOT lies at ∼14 • N. The WRF-Chem EXP30 at 30 km resolves reasonably well the inverted "V" plume. However, the eastern axis is rather weak because WRF-CHEM at this coarse resolution does not represent cold pools over Niger from DC3 and the associated haboobs. As such the simulated dust plume from Chad/Sudan (Section 3.3.1) in the Harmattan is effectively transported westward over Niger without pronounced modification by MCS outflow but rather is then advected northwards towards Algeria in the large scale southerly monsoon flow around the eastern flank of SSD2. Thus, the resulting EXP30 model simulation of the plume structure largely reflects emission Advances in Meteorology and transport within the largescale flow, culminating in the organisation of dust along the Harmattan/monsoon boundary around SSD2. However the model does not advect the dust far enough southwards over the monsoon flow, a feature highlighted in analysis of vertical dust structures in Section 3.3. The model produces a reasonable representation of large-scale dust emission and transport but without accurate representation of the mesoscale haboob phenomena that are substantially responsible. From an analysis of longterm satellite and reanalysis data Knippertz and Todd [31] suggest that the aerosol "hotpot" over the western Sahara in Mauritania-Mali-Algeria collocated with the SHL during summer is substantially a reflection of such emission and transport processes rather than local emission as previously inferred. As such, in this case WRF-CHEM EXP30 produces a broad scale dust plume with a realistic horizontal location but resulting from processes different from reality. The averaged surface dust emissions for the period from the 7th to the 12th of the June in the WRF-Chem model are equal to 200 g m −2 d −1 .  analyse this we examine a full diurnal cycle and comparison with CALIOP extinction coefficient profiles. Vertical profiles of dust, potential temperature, humidity, and horizontal winds from WRF-CHEM EXP30 along a latitude at 10 • E for over one characteristic diurnal cycle on the 9th of June are examined ( Figure 12). Lidar extinction coefficient observations from CALIOP orbits close to 10 • E at 12.41 UTC on the 9th of June and 01.06 UTC on the 10th are shown for comparison ( Figure 13). A number of key structures in the atmosphere can be identified. (i) The shallow moist and cool monsoon flow south of the ITD undercut the dry Saharan air to the north. The diurnal cycle in the monsoon flow is such that the ITD advances/retreats by ∼5 • latitude from about 12 • N at 18.00 UTC to about 17 • N at 06.00 UTC ( Figure 12). (ii) A well mixed, deep and dry convective boundary layer develops during the day up to height of about 6 km in the afternoon. This is the Saharan convective boundary layer (SCBL) and extends north from the ITD to ∼27 • N. (iii) At night, surface cooling decouples this layer resulting in an elevated residual Saharan convective boundary layer above the surface inversion. This residual layer then overrides the monsoon flow.

Vertical
Dust emission and transport are strongly controlled by these features. During the early hours of the 9th of June a dust plume centred on 17 • N resides throughout the vertical extent of the residual boundary layer from above the surface temperature inversion to about 5.5 km height (Figures 9(a)-9(d)). This is the plume from the Bodélé Depression Advances in Meteorology   13   3N 6N 9N 12N 15N 18N 21N 24N 27N 30N 6N 9N 12N 15N 18N 21N 24N 27N 30N 6N 9N 12N 15N 18N 21N 24N 27N 30N  (B1 in Figure 4) emitted the previous day and advected east to ∼10 • E. The horizontal and vertical extent is verified from the CALIOP observations at 12.41 UTC on the 9th of June (Figure 13(a)). During the morning local emission occurs at 16-21 • N over Niger (Figures 9(d)-9(e)) as the nocturnal LLJ in the Harmattan flow is mixed to the surface, where 10 m winds exceed 12 m s −1 at 09.00 UTC. This emission is then rapidly mixed through the deep boundary layer to ∼6km height by 18 UTC (Figures 10(f) and 10(g)) representing an effective diurnal "dust pump" into the SCBL. The following night (not shown) the near surface temperature inversion decouples the surface and atmosphere above creating another elevated dust layer in the residual SBL.
This dusty residual layer appears to be advected along slanting isentropes over the monsoon flow leading to an elevated wedge of high aerosol south to about 10 • Nby12UTC. This elevated southward dust transport is again confirmed by comparison with the CALIOP extinction coefficient profiles (Figures 13(a) and 13(b)). Overall, both WRF-CHEM and CALIOP broadly represent the characteristic features of a deep dusty SCBL to about 6 km height north of the ITD and southward advection of dust over the monsoon layer. We cannot expect the model to replicate all aspects of convective layer (not shown). As a result entrainment of clear air from the free troposphere may be rather too strong resulting in a more variable dust profile.

Summary
The ability of the WRF-Chem model incorporating the GOCART dust scheme and the MOSAIC aerosol model to simulate processes of dust emissions and transport from over North Africa during summer was evaluated. The study focussed on the period from the 7th to 15th of June, 2006, during the AMMA Special Observing Period (SOP) in the West African Monsoon onset phase [57]. This period is particularly interesting for model evaluation as dust activity is pronounced and the associated processes of emission are manifold. Model simulations were confronted with a suite of satellite and surface observations. Evaluation implicate of synoptic and mesoscale atmosphere processes as drivers of dust plume events. Within the northeasterly Harmattan flow strong surface winds and dust outbreaks were caused by anomalies in the pressure gradient across north Africa associated with synoptic scale variability in both the midlatitudes and African easterly wave disturbances. Within the southwest monsoon circulation, monsoon surges around the AEW surface troughs facilitated moist convection well into the Saharan sector. These convective events often produce cold pool density currents driven by evaporation of precipitation in dry lower layers, which cause the haboob dust front events. WRF-Chem running at 30 km over most of North Africa represents well these synoptic scale features and in most cases the associated dust emission from preferential dust sources in the BodéléDepression,Libya,Sudan,andAlgeria, many of which are associated with erodible material surface drainage systems. The 4-dimensional space-time evolution of emission, transport, and organisation of dust plumes broadly matches observations although there are errors in the fine details.
However, the haboob dust features are not at all well simulated at 30 km model resolution due to inadequacies in the convective parameterisation to simulate downdraughts. During the study period haboobs contributed a substantial proportion of total dust. The dust fronts can be of the order of 1000 km in length and propagate distances of many hundreds of km. Limited analysis of WRF-CHEM at 10 km resolution indicated that the structure of cold pool flows associated the deep convection and the associated dust emission can be well simulated.
Following emission the dust plumes are transported and organised in the Saharan/monsoon circulation such that dust from many processes is mixed horizontally and vertically. The comparative analysis between observations from CALIOP and WRF-Chem simulations demonstrates the model efficacy to simulate the 3D structure of dust transport over the Sahara and Sahel.
Overall, the main features of the idealised summary of Saharan dust processes during summer of Cuesta et al. [9] are identifiable in this case and WRF-CHEM replicates these well. There is, therefore, a strong basis for accurate forecasting of dust events associated with synoptic scale events, when model dust emission parameterisation is suitably calibrated. Forecasting of the haboob events, an important driver of the dust cycle in summer, is more problematic. Further analysis of high-resolution model simulations is necessary to address this issue.