In summertime, the West African heat low (WAHL) is generally located over the Sahara, west of the Hoggar. During this period, the WAHL is considered to be a major dynamic element of the West African monsoon (WAM) system, and the main driver of precipitation over the Sahel (see Lavaysse et al., 2009, and references therein). In the low troposphere, the cyclonic circulation generated by the depression tends to organize and reinforce the southwesterly monsoon flow along its eastern flank and the northeasterly harmattan flow along its western flank. Parker et al. (2005) suggested that such a circulation is likely to enhance the probability of moist convection and cloud cover poleward of the main rainy zone to the east of the WAHL. In the midtroposphere, the anticyclonic circulation associated with the WAHL and the meridional gradient of temperature between the rain belt and the WAHL help to maintain the African Easterly Jet (AEJ) and modulate its intensity (Thorncroft and Blackburn, 1999). An intensification of the AEJ is indeed observed during strong phases of the WAHL (Lavaysse et al., 2010b). On an intraseasonal scale, the 3–10 day pulsation of the WAHL is associated with moist and cool advection in the lower troposphere linked to the southerly sector of African Easterly Waves (AEWs), which increases convective activity over the Sahel (Lavaysse et al., 2010).
The West Saharan source is one of the world's largest sources of mineral dust, with an annual peak of emission during summer (Engelstaedter and Washington, 2007). Dust emitted from desert regions and transported in the atmosphere is recognized as an important component of the Earth's climate system, scattering incoming solar radiation and absorbing and emitting infrared radiation. In consequence, mineral dust plays an important role in the dynamics of the region via its radiative properties. Over the source regions, the dust effect usually results in atmospheric warming located in the planetary boundary layer (PBL), where the dust is concentrated. In the case of elevated plumes, the dust effect can lead to a cooling of the near-surface layers (Tulet et al., 2008; Lemaitre et al., 2010; ). Furthermore, dust-related radiative forcing over the Sahara has remote impacts on convection over the Sahel. For example, a substantial weakening of the AEJ in the forecasts of the European Centre for Medium-Range Weather Forecasts (ECMWF) model is partly attributed to an overestimated direct radiative forcing by dust aerosols (Tompkins et al., 2005). At the two-day range, the use of a prognostic dust aerosol scheme in the regional model Meso-NH (Lafore et al., 1998), rather than a climatology, captures the observed convective activity over the Sahel better (Chaboureau et al., 2007). At the seasonal scale, the impact of dust leads to a reduction of precipitation over the Sahel, except over a latitudinal band over the northern Sahel and the southern Sahara (Solmon et al., 2008). In contrast, Lau et al. (2009) have analyzed the increase of seasonal mean precipitation over the whole of the Sahel associated with an intensification of the West Africa/Caribbean region circulation and resulting from an elevated heat-pump effect owing to dust forcing. At the diurnal scale, dust tends to stabilize the atmosphere in the daytime and reduce the stability during the night-time. Furthermore, Zhao et al. (2011) have shown that precipitation is reduced in the late afternoon, whereas it is increased during the night-time and in the early morning over the WAM region.
Summertime interactions between the WAHL and dust loads over the Sahara have not been examined so far. Dust is likely to influence the WAHL through direct radiative forcing and indirect effects via the West African circulation. Here, the impact of dust on the strength of the WAHL is investigated using the Meso-NH model during a six-day period between 14 and 20 July 2006. The episode was selected because of the abrupt variation in the West African circulation as well as the high values of dust aerosol optical depth (AOD) over the Sahara. Three numerical experiments were designed, differing only in their representation of dust. This simulation framework allowed us to study the effects of dust on the pulsation of the WAHL. Note that this study focuses on the direct and semi-direct aerosol radiative effect of dust only. The indirect aerosol effect is not considered here.
The article is organized as follows. Section 2 contains an overview of the intraseasonal variation of the WAHL during summer 2006. It also presents the dust and convective activities during the six-day episode. Section 3 describes the model and the numerical experiments for which the dust representation, WAHL intensity and associated circulations are assessed against satellite observations and ECMWF analyses. Section 4 discusses the effects of dust on the West African circulation with a focus on WAHL thickness, precipitation, the AEJ, AEW activity and moisture flux. Section 5 concludes the article.
2. Description of the case study
2.1. The West African Heat Low in summer 2006
The mean location of the WAHL in summer 2006 is examined first. The location is calculated using the low-level atmospheric thickness between 925 and 700 hPa taken from the 0600 UTC ECMWF operational analysis (Lavaysse et al., 2009). As in Lavaysse et al. (2009), the WAHL is defined as the area associated with 10% of the largest thickness over West Africa. Between 1 June and 30 September 2006, the largest occurrence probability is located over northern Mali, just west of the Hoggar Mountains (Figure 1(a)). Over the Sahara, the circulation at 925 hPa is cyclonic and almost centred on the maximum occurrence of the WAHL. This result shows the consistency between the WAHL location and the 925 hPa wind field. Note that the position of the WAHL in 2006 is very close to its climatological position in summer shown in Lavaysse et al. (2009). Between 14 and 20 July 2006, the WAHL is centred on 25°N, moving from 0°E on 14 July to 5°W on 20 July, as typically observed in the summertime.
The WAHL activity between 14 and 20 July is shown in the context of summer 2006 (Figure 1(b)). In addition, the dust load in the WAHL region was examined using the ozone monitoring instrument (OMI) aerosol index (Torres et al., 1998) averaged over the area covered by the WAHL. The OMI aerosol index is positive for aerosols absorbing at ultraviolet wavelengths, such as dust. At the beginning of July, the WAHL displayed a pronounced increase in thickness. It then collapsed after 10 July due to the occurrence of rainfall over the Sahel, which was associated with widespread convection (Janicot et al., 2008). On 14 July, the WAHL thickness was again enhanced by a few metres for two days. On 16 July, it collapsed by 4 m in four days and exhibited a minimum on 20 July. This six-day episode was selected because of the relatively short period in the WAHL pulsation.
Over the same six-day period, the OMI aerosol index (AI) followed the same fluctuation as the WAHL thickness. This suggests a relationship between the intensity of the WAHL and the dust load in the WAHL region. When a longer period was examined, from 1 July–15 September, the aerosol index was only partly in phase with the WAHL thickness. As a result, a correlation coefficient of 0.47 was obtained. Note that a smaller correlation coefficient (0.27) was found between the WAHL thickness and Moderate Resolution Imaging Spectroradiometer (MODIS) AOD retrievals. The poor correlation during the first part of the period could be due to the uncertainties in dust detection in the low layers. Since dust absorbs radiation in the ultraviolet wavelength range, and because absorption by these aerosols in this part of the spectrum increases with increasing altitude, elevated dust plumes are generally detected better than aerosol layers confined at low levels (Torres et al., 1998; Hsu et al., 1999).The OMI aerosol index is more sensitive to the dust altitude than the MODIS AOD (Flamant et al., 2009). This might explain the larger correlation coefficient obtained with the OMI aerosol index than with the AOD. During a period of deep WAHL, the associated increase of turbulence can favour the increased advection of dust in the middle troposphere and thus increase dust detection using the OMI AI.
2.2. Convective activity
The convective activity and dust loads during the six-day episode were described using the Meteosat Second Generation based Spinning Enhanced Visible and Infra-Red Imager (SEVIRI) images produced from a combination of three infrared channels, namely channel 10 (12 µm), channel 9 (10.8 µm) and channel 7 (8.7 µm) (Figure 2). On 14 and 15 July, mesoscale convective systems (MCSs) (in red and brown) developed over the whole of West Africa around 10°N. On the two following days, the convective activity was less ubiquitous between 15°W and 15°E. For example, MCSs on 17 July were found west of 15°W and east of 0°E only. On 18 July, a large area of convective activity was triggered over southern Chad (15°E, 10°N). The convective cells then propagated westward through the Sahel and, on 19 July, two major MCSs were located over Mauritania and western Niger.
On 14 and 15 July, dust plumes (appearing pink or magenta in Figure 2) were located mostly over the Atlantic Ocean off the African coasts at 30°N and covering most of the Western Sahara. A few dust storms were visible northwestward of the convective systems, at 5°W, 25°N on 15 July, at 5°W, 20°N on 18 July and at 10°W, 20°N on 19 July, for example. This suggests that dust was mobilized there at the leading-edge cold pools of MCSs as observed over the Sahel (Flamant et al., 2007). Dust also appeared over the Sahara far away from any MCSs. On 16 and 17 July, dust appeared at the borders between Libya and Niger. On 18 July, dust was visible over southern Senegal at 10°W, 18°N and was organized in a circular pattern. Due to the presence of both convective activity and a dust plume over the Sahara, this period is used to investigate the interaction between the dust and the dynamics of the monsoon.
3. Model design and validation
3.1. Meso-NH simulations
The numerical simulations were performed with the non-hydrostatic mesoscale model Meso-NH (Lafore et al., 1998) version 4.7. The case was simulated with a horizontal grid of 384 × 216 points with a grid spacing of 24 km over the domain shown in Figure 2. The vertical grid had 70 levels with a level spacing of 60 m close to the surface and 600 m at high altitude. The lateral boundary conditions were given by the ECMWF operational analyses. Initial conditions were provided by a four-day Meso-NH run nudged towards the ECMWF analyses with a six-hour relaxation time. This allows a spin-up for dust and cloud in the model while keeping the meteorological fields (wind, temperature and water vapour) close to the ECMWF analyses at the initial time. The simulations were initialized on 14 July 2006 at 0000 UTC and were integrated for 6 days. Three simulations were designed, namely REFE without any radiative effect of dust, CLIM using the monthly dust climatology of Tegen et al. (1997) and DUST using the prognostic dust scheme described below. Table I summarizes the differences characterizing the Meso-NH experiments.
Table I. Characteristics of the Meso-NH experiments.
No dust effect
Deep and shallow convective transport and precipitation were parametrized following Bechtold et al. (2001) based on the work of Kain and Fritsch (1993). The microphysical scheme included the three water phases with five species of precipitating and non-precipitating liquid and solid water (Pinty and Jabouille, 1998) and a modified ice-to-snow autoconversion parametrization following Chaboureau and Pinty (2006). Subgrid cloud cover and condensate content were parametrized as a function of the normalized saturation deficit by taking both turbulent and convective contributions into account (Chaboureau and Bechtold, 2002, 2005). The turbulence parametrization was based on a 1.5-order closure (Cuxart et al., 2000). The radiative scheme was the one used at ECMWF (Gregory et al., 2000) including the Rapid Radiative Transfer Model (RRTM) parametrization (Mlawer et al., 1997). The surface scheme was the Interactions between Soil, Biosphere and Atmosphere (ISBA) soil scheme of Noilhan and Planton (1989).
The dust prognostic scheme is described in Grini et al. (2006). In this parametrization, the three log-normal modes were generated and transported by the log-normal aerosol scheme of the ORganic and Inorganic Log-normal Aerosols Model (ORILAM: Tulet et al., 2005). These modes were described by their zeroth, third and sixth moments, with the last kept constant. Dust fluxes were calculated from wind friction speeds using the Dust Entrainment and Deposition (DEAD) model (Zender et al., 2003). The initial dust size distribution contained three modes with median radii of 0.32, 1.73 and 4.33 µm and standard deviations of 1.7, 1.6 and 1.5, respectively, as defined by Alfaro and Gomes (2001). Dust loss occurred through sedimentation and rain-out from convective clouds.
The radiative properties of dust were obtained, in the short-wave spectrum, from the photometers deployed during the African Monsoon Multidisciplinary Analyses (AMMA) programme while, in the long-wave, the standard formulation of absorption and re-emission for aerosols from the ECMWF model was used (see Tulet et al., 2008 for more details).
3.2. Evaluation of the simulated dust distribution
The distribution of dust in the Meso-NH experiments was evaluated against observations. First, the Meso-NH AOD valid for the spectral band of 440–690 nm was compared against the AOD at 550 nm from MODIS on board the Aqua satellite. The MODIS AOD was obtained from Collection 005 over ocean and vegetated surfaces and Deep Blue Collection 051 over bright, highly reflective surfaces such as deserts.
The MODIS AOD is shown for around 1200 UTC every day over land only (Figure 3). Over the western Sahara (west of 10°E), values of AOD larger than 1 were often observed during the six-day period. For example, large AODs were seen at 5°W, 25°N on 15 July, at 5°W, 20°N on 18 July and at 10°W, 20°N on 19 July, in agreement with the dust detected in the SEVIRI images. Dust was also observed over the southern tip of the Arabian Peninsula, as also seen in SEVIRI images. Because of the difference in time and sensing methods, other areas with large dust loads not detected by SEVIRI were revealed by MODIS. Large AODs were present almost every day around 15°W, 15°N. It is likely that dust there originated from the Bodélé depression, the most active source in the world, located in northern Chad. Another important area of dust was centred over northern Mauritania in the WAHL area. The related AOD maximum moved southwestward with time, from 0°E, 27°N on 14 July to 15°W, 20°N on 19 July.
In the DUST simulation, the AOD showed four areas with enhanced dust loads: the Arabian Peninsula, the Bodélé depression, Niger and the western Sahara around northern Mauritania (Figure 4). These locations agree rather well with those retrieved from MODIS observations. Correlation coefficient values between DUST and MODIS AODs range between 0.35 and 0.53 and biases between −0.11 and 0.04, the highest score being obtained for the last 4 days. In particular, the southeastward migration from 25°N to 20°N of the AOD maximum is correctly reproduced by the DUST simulation. However, the spatial extent of the dusty areas is overestimated in DUST compared with MODIS, possibly due to limitations in the AOD retrievals with the Deep Blue algorithm. Furthermore, the model overestimates AOD over, e.g., Mauritania and the Arabian Peninsula while dust sources in Algeria and Tunisia are missed. This might have led to an overestimation of the dust radiative effects in DUST.
The climatological AOD valid for July and used in CLIM is shown in Figure 5. The largest AOD values (in excess of 1) are located at the southern tip of the Arabian Peninsula. Over the western Sahara, the AODs are below 0.5, i.e. much smaller than the AODs retrieved from MODIS observations and simulated in DUST (both show values larger than 1 in this area). This underestimation may be partly explained by the fact that July 2006 was particularly dry with respect to the climatological conditions as the monsoon onset (i.e. the occurrence of rainfall) over the Sahel occurred nearly 2 weeks after its climatological average (24 June ±8 days: Sultan and Janicot, 2003; Janicot et al., 2008). The vegetation cover and soil moisture were reduced in 2006 relative to the climatological mean, which was favourable for dust source activation. As a result, the underestimation of climatological AODs over the western Sahara should lead to contrasting results between CLIM and DUST for the WAHL dynamics.
The comparison between MODIS and Meso-NH AODs gave a quantitative assessment of the horizontal distribution of dust over land. Observations from lidar allowed a further quantitative comparison between observation and simulation of the vertical structure of dust. This is of importance, as the impact of the radiative heating is strongly related to the altitude of dust (Tulet et al., 2008; Lemaitre et al., 2010). The attenuated backscatter (ATB) signal at 532 nm from the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) is shown in Figure 6(a). The vertical cross-section of ATB was measured at 0300 UTC on 17 July between (0.6°E, 0°N) and (10.3°E, 40°N). As checked using SEVIRI images, observed ATB signals larger than 10−2 km−1 sr−1 were due to clouds: high clouds around 10°N, mid- to high-level clouds around 35°N and planetary boundary layer (PBL) clouds near 5°N. Dust characterized by ATB signals around 10−3 km−1 sr−1 occurred between 10°N and 35°N in the first 5 km of the atmosphere. These results are in accordance with the frequency distribution of dust-layer depth observed in southern Morocco during May and June 2006 during the SAharan Mineral dUst experiMent ((SAMUM), Tesche et al., 2009) and also the observations above the Sahara (Cuesta et al., 2010) and the Sahel (Flamant et al., 2007) made during AMMA. The authors found that the dust-layer top generally reached a height of 4–6 km above sea level.
In DUST, the ATB signal was calculated from the atmospheric profiles using the lidar simulator developed by Chaboureau et al. (2011). The Meso-NH DUST simulation reproduced the observed features of the vertical structure (Figure 6(b)). High clouds associated with tropical deep convection at 10°N were simulated at about the right latitude. Between 35°N and 38°N, high-level clouds were simulated at the right altitude but mid-level clouds were missed. The clouds capping the PBL over the Hoggar Mountains covered too large an area compared with the isolated cell observed at 21°N. In the southern part of the domain, clouds at the PBL top extended to 15°N. In between the clouds over the Sahara, the ATB signal associated with the dust spanned the right latitude range and signal intensity (with some discrepancies locally, e.g. around 30°N where ATB signals above 1.5 reach higher altitudes in DUST than in CALIOP). Similar agreement in the vertical structure of dust was seen for all the CALIOP profiles obtained during the six-day period. These results suggest that dust was correctly distributed in the vertical direction in the DUST simulation.
3.3. Comparison with analyses
The location and intensity of the WAHL and AEJ in the Meso-NH simulation REFE is evaluated against the ECMWF operational analyses. ECMWF analyses account for dust radiative effects though data assimilation. However, radiosondes over Africa are sparse (Parker et al., 2008), while only upper-level information from satellite observations is assimilated over land (typically above 5 km where dust is almost absent). As a result, analysis over Africa mainly relies on the model first guess. Therefore, we first compare the REFE simulation with the ECMWF analyses in order to verify the accurate representation of the key WAM features in the REFE simulation. The impact of the dust is then analyzed by comparing the reference simulation REFE with the CLIM and DUST simulations. The WAHL thickness, 925 hPa wind and precipitation averaged between 15°W and 10°E are shown for ECMWF analyses and REFE in Figure 7. This longitude band is used to define the Sahel area, where the monsoon circulation displays a strong zonal circulation. In Figure 7(a), the thickness and wind fields are taken from ECMWF analyses while precipitation is taken from the EPSAT-SG product (Bergès et al., 2010).
In the ECMWF analyses (Figure 7(a)), the WAHL centred between 20°N and 25°N is associated with the southeasterly monsoon flow between 5°N and 15°N and the northwesterly Harmattan winds between 25°N and 35°N. The WAHL and circulations at 925 hPa present a diurnal cycle in opposite phase, the WAHL being maximum at 1800 UTC and minimum at 0600 UTC. As previously seen in Figure 1(b), the WAHL thickness increases from 15 July, after the passage of an MCS. It then collapses on 19 July just after a large-scale convective event reaches northern Mali. The occurrence of the precipitation estimated by EPSAT-SG agrees well with the occurrence of clouds in SEVIRI images (Figure 2), as the EPSAT-SG product relies on a rainfall probability from SEVIRI.
In the REFE simulation (Figure 7(b)), the location of the WAHL is consistent with the analysis, centred between 20°N and 25°N, but its thickness is generally between 5 and 10 m shallower than in the ECMWF analysis. This reduces not only the maximum thickness, especially during the night, but also the latitudinal extension of the WAHL. The mean wind circulation is reproduced in the REFE simulation, but with higher wind speed at 10°N and 25°N. Over this area, the wind speed in the REFE simulation exceeds 10 m s−1, 6–8 m s−1 higher than the analysis. Similar faster monsoon flow is obtained, together with a systematically shallower boundary layer, for the Meso-NH forecasts made in summer 2006 using a similar model configuration (Söhne et al., 2008). This drawback can be attributed to insufficient vertical mixing, which also explains the too large cloud cover at the PBL top seen previously in the lidar profile. Despite these discrepancies, the REFE simulation shows an increase in the WAHL intensity followed by a decrease similar to the evolution shown in the ECMWF analysis. Because of the absence of dust effects in REFE, this result suggests that large-scale forcing primarily drives the variation of the WAHL intensity.
In both the ECMWF analyses and the Meso-NH REFE simulation, the northerly wind increases when the WAHL strengthens (around 16 July). On 19 July, a monsoon burst occurs as depicted by a large increase in the southerly winds north of 15°N. This evolution resembles the intraseasonal WAHL variability studied in Lavaysse et al. (2010b), where the influence of the temperature and water-vapour advection on the WAHL thickness are strong. These advections are associated with wind anomalies and an enhanced convective activity over the Sahel.
The enhanced convective activity observed in SEVIRI (Figures 2 and 7(a)) on 14 and 19 July fits the scenario well. As shown in Figure 7(b), two strong precipitation events are generated at 15°N on 14 and 19 July in the REFE simulation. In both cases, a minimum in WAHL thickness is detected just after each strong precipitation event. These rain events favour a decrease in temperature at low levels, thereby explaining the decrease in WAHL thickness. In contrast, the WAHL is observed to be deep (from 17–19 July) when the rainfall is reduced.
The AEJ is another important synoptic component of the West African Monsoon system, which is shown with the wind speed at 600 hPa averaged between 15°W and 10°E (Figure 8). The 600 hPa level is the level at which the AEJ is strongest in the Meso-NH simulations (Chaboureau et al., 2007). In the ECMWF analyses, the AEJ is zonal until 16 July. It is centred at 12°N with a maximum wind speed increasing from 12 m s−1 on 14 July to more than 18 m s−1 on 16 July. On 17 July, a broadening of the AEJ toward the south is observed and its maximum intensity south of 10°N reduces to 10 m s−1. On 19 July, the AEJ reintensifies with winds up to 17 m s−1 around 15°N, further north than a few days before. In the REFE simulation, the maximum wind speed is stronger, over 14 m s−1, throughout the period. A maximum of 16 m s−1 (less than in the ECMWF analyses) is reached on 16 July. Until 16 July, the REFE simulation shows two branches in the AEJ more distinctly than can be seen in the ECMWF analyses. However, they both show a southward displacement of the AEJ core toward 10°N on 17 July and the presence of the jet around 15°N on 19 July.
4. Impact of dust on the West African circulation
As seen in the previous section, a reasonable agreement with observations and analysis was achieved for the Meso-NH simulations, be it in terms of dust distribution, location of rainfall or synoptic circulation. The impact of the dust in the CLIM and DUST simulations is now examined, firstly in the form of temperature and wind anomalies with respect to REFE. Then the dust impact on the AEJ and the AEW activity is discussed. It is then shown that the enhanced circulation when dust effects are added leads to an increase in the moist monsoon flow.
4.1. Change in temperature and WAHL thickness
The impact of the dust is first analyzed in the temperature fields by examining a vertical meridional section of the temperature anomaly between DUST and CLIM with respect to REFE averaged between 15°W and 10°E and between 14 July and 20 July (Figure 9). The impact of the dust differs depending on the altitude. In the low troposphere (925 hPa) the temperature is higher in REFE than CLIM and DUST over the WAHL area (between 20° and 30°N). This is due to the presence of the dust plume above 800 hPa, which reduces the incoming solar radiation, and other indirect effects such as the occurrence of clouds. These differences are greater in DUST. Between 800 and 400 hPa, the temperature difference is positive over the WAHL area, especially in the afternoons (not shown). Farther north, the DUST and CLIM experiments show a positive anomaly of temperature even close to the surface. These differences are related to the presence of the dust-source areas north of 20°N. In the low layers, these results are different from Lau et al. (2009), where the authors found on average from June–August a positive anomaly in 2 m temperature between 20 and 30°N; in this case this positive values appear northward of 30°N.
The time variation of the WAHL thickness, 925 hPa wind and precipitation anomalies with respect to REFE is shown over West Africa for CLIM and DUST (Figure 10). For both experiments, the WAHL has a diurnal cycle of similar amplitude to that of REFE. In CLIM, the anomalies of the WAHL thickness increase throughout the period of the simulation, and appear larger than REFE, especially in the northern part of the WAHL. This is in accordance with the strong increase in temperature visible in Figure 9(a). The wind-speed anomalies with respect to REFE show a slight decrease in the northerly wind at 20°N on 17 July. Over the Sahara, between 20°N and 25°N, where AOD was greater than 0.5 in CLIM, the easterly component of the Harmattan increases with respect to REFE. This increase also occurs, but faster, in DUST and is characterized there by larger AOD than in CLIM. Unlike the other experiments, the DUST simulation shows the WAHL collapsing twice: once around 17 July, associated with a large southerly anomaly of the wind speed, and again on 19 July during a period of enhanced convective activity. These anomalies are strongly linked with the negative anomalies of temperature at 925 and 850 hPa in DUST with respect to the REFE (not shown).
The convective activity is also modified as shown by precipitation anomalies (Figure 10). Anomalies for CLIM are small except on 19 July when the southern Sahel (around 15°N) exhibits a positive anomaly in precipitation and in southerly monsoon flow. In DUST, positive anomalies in rainfall are more frequent, right from the beginning of the simulations on 14 July. The two largest episodes occur on 17 and 19 July, in phase with the two WAHL collapses discussed above. These increases in precipitation could be due to a modification of the circulation or a modification of the atmospheric stability due to the dust radiative heating. On 14 July, the wind-field anomaly in CLIM or DUST with respect to REFE does not change significantly. Moreover, this area undergoes low-level warming (not shown). This suggests that the origin of the enhancement of the rainfall is linked more with modification of the atmospheric stability.
Because of the configuration of the simulations, these changes in the WAHL, wind and rainfall between the Meso-NH experiments are due to the representation of dust. The direct effect of dust is on the heating budget through radiative processes. This point is examined with temperature tendencies calculated over the western Sahara (9°W–3°E; 20°N–30°N) (Figure 11). During the daytime, heating in REFE occurs over the whole troposphere, with the maximum heating larger than −5 K day−1 in the PBL (below 5 km). During the night, cooling is experienced down to −5 K day−1 in the first 3 km. There is an additional effect due to dust, larger than 0.5 K day−1, during daytime especially below 5 km altitude. This increase is accompanied by PBL cooling that can be explained by a decrease in the incoming short-wave radiative flux. Compared with DUST, CLIM generally shows more heating in the low layers, which is consistent with the larger WAHL thickness in CLIM than in DUST. In DUST, the additional heating is largest on 18 July when the difference in the WAHL thickness between DUST and REFE is maximum. Thus, the radiative impact of dust leads to an increase in the mean temperature in DUST. Following the hypsometric equation, the thickness of the layer is closely linked to its mean temperature. This means that the low troposphere is, on average, warmer in DUST than in REFE. Using the decomposition of the temperature difference at each level between CLIM or DUST with respect to REFE (not shown), a negative temperature anomaly is detected over the WAHL area at 925 hPa. Over the other levels (from 850–600 hPa), the temperature is lowest in the REFE simulation, especially at 700 hPa and north of 25°N. This increase in temperature over the Sahel leads to a larger temperature gradient between the Gulf of Guinea and the Sahara. This temperature gradient is one of the main mechanisms that controls the West African monsoon system.
4.2. Change in AEJ and AEWs
The AEJ centred between 10°N and 15°N changes with the representation of dust in the simulations. This is shown on the basis of the 600 hPa wind-speed anomalies with respect to REFE (Figure 12). In CLIM, the intensity and location of the AEJ are quite similar to the REFE situation despite the thickening of the WAHL. The two experiments differ at the end of the episode. The AEJ increases at the southward edge in CLIM, with a positive anomaly of the wind speed south of 5°N. Around 10°N, the AEJ decreases over the area where rainfall occurs (Figure 10). In contrast, DUST shows a larger departure from REFE. The southern flank of the AEJ, south of 10°N, is increased, especially before 18 July. The maximum of the AEJ wind speed occurs between 16 and 17 July, associated with the maximum difference of the temperature anomaly (not shown), whereas the maximum occurs 2 days later in the other experiments. Nevertheless, the intensities of the maximum wind are close (around 18 m s−1). These differences of the AEJ cannot be explained by the WAHL activity, where the intensity is lower in the DUST than the REFE simulation. The origin of this strengthening could be associated with the location of the dust plume in the DUST simulation (Figure 4), in which the southern edge is located around 12°N on 16 July then moves southward to reach 10°N on 19 July. This increases the latitudinal gradient of temperature, especially between 700 and 600 hPa (Figure 9(b)), and underlines the possible large impact of the dust on the AEJ intensity and location on an intraseasonal time-scale.
These differences in the AEJ should affect the energy available for AEWs. For example, Leroux and Hall (2009) show that a necessary condition for the development of strong waves is a strong AEJ. A common method for identifying AEWs is based on the 3–5 day filtered meridional wind. This allows both the location and intensity of the AEWs to be estimated (Diedhiou et al., 2002; Lavaysse et al., 2006). In general, the 700 hPa level is used to detect the southern branch of the AEWs while the 850 hPa level is better to identify the northern branch. In the ECMWF analyses (not shown) during the period of interest, the meridional filtered wind has a spatial and temporal structure reminiscent of the characteristics of AEWs, such as propagation speed (around 10 m s−1) and wavelength (3500 km with a 3–5 day period). Over the Sahel, the wave activity is larger in the middle of the episode as also observed with other methods (see wave #3 in figure 19 of Janicot et al., 2008). The Meso-NH simulations are, however, too short for the filtering method to be used. Instead, the meridional wind with respect to its meridional average is shown for the ECMWF analyses and the Meso-NH simulations (Figure 13). Correlation coefficient values between ECMWF analyses and Meso-NH simulations are slightly larger for CLIM and DUST (0.83) than for REFE (0.80).
Overall, the 850 hPa meridional winds in the Meso-NH experiments present a pattern similar to the ECMWF winds (Figure 13). They all show an AEW on 17 July as seen with a strong northerly wind at 0°E embedded between two southerly winds. The propagation and wavelength of this AEW are also quite similar between the analysis and the simulations during the period. In CLIM, the intensity of the AEWs remains close to REFE. In DUST, the intensity of both the southerly and northerly wind increases by 1 or 2 m s−1, which represents a relative increase of more than 25%. This is in accordance with the strengthening of the AEJ previously shown. Remarkably, the difference in AEW intensity occurs on 17 July when the AEJ is increased in DUST.
4.3. Change in the moisture flux
To understand better the impact of each experiment on the West African monsoon system, the flux of the water-vapour mixing ratio is calculated over the WAHL area (10°W–7°E; 20°N–32°N) for the Meso-NH experiments (Figure 14). The flux is taken as positive in case of humidity inflow. Most of the time, the variation in the moisture flux is associated with a similar variation in the moisture advection over the WAHL area. It has been shown that these mechanisms have a strong impact on the intraseasonal variability of the WAHL activity (Lavaysse et al., 2010b).
In REFE and CLIM, the periods with low WAHL intensity (i.e. before 15 July and after 19 July) are characterized by large moisture fluxes below 700 hPa (Figure 14(a) and (b)). These fluxes tend to contribute to the collapse of the WAHL due to increased cloud occurrence, rainfall events or because they are generally associated with relatively cold air masses. From 15–19 July the fluxes are much weaker in magnitude, with two short periods with negative values on 16 and 17 July during daytime. A strong negative moisture flux appears on 17 July due to incoming warm, dry air (not shown). This feature was much weaker in REFE and CLIM. This change in moisture flux is consistent with the temporal evolution of the WAHL intensity in the experiments. These results also suggest a possible impact of the increase in AEW activity shown previously. In comparison with REFE and CLIM, the moisture flux in DUST is largest on 19 July (Figure 14(c)).
The geographical origin of the moisture flux is determined by splitting the flux up into the four canonical directions. This decomposition is shown for DUST only (Figure 15). Very similar results were obtained for the other experiments with the notable exception of the southern part. The moisture flux from the north and west sides leads to a weak drying of the WAHL, enhanced at 0600 UTC each day. This drying is from the north until 16 July and then from the west. This change in direction from north to west is due to the veering of the harmattan wind with time. From the east, the moisture flow results in a moistening due to the AEJ, also associated with a diurnal cycle maximum at 0600 UTC. The magnitude of the moistening from the east is weak and is mostly counterbalanced by the northerly and westerly components.
In fact, most of the variation seen in the moisture flux is due to the southern component, which is characterized by moistening on 14 and 19 July and drying on 16 and 17 July. Compared with REFE and CLIM, the drying is greater in DUST, which explains the strengthening of the WAHL. Moreover, the moistening is slightly increased in DUST with respect to REFE and CLIM, especially at the end of the episode. So the larger variation in the WAHL seen in DUST is associated with an enhanced southern flux, leading to either moistening at the beginning and end of the episode or drying during the WAHL climax. These results confirm the strong impact of the increase in the AEW intensity during the DUST simulation, which tends to enhance the humidity and cold advection from the Gulf of Guinea to the Sahel.
A short pulsation of the WAHL in summertime has been studied between 14 and 20 July 2006. As seen with the ECMWF analysis, the WAHL increased during the first two days and then collapsed quickly from 16 July onwards. The collapse was associated with convective rainfall and dust bursts observed over the Sahel, while the strengthening period was characterized by the absence of convection over the central Sahel. Moreover, the change in the WAHL intensity was partly in phase with the variation of the OMI aerosol index in the area. Three Meso-NH simulations, REFE, CLIM and DUST, were performed in order to examine the potential impact of dust on the WAHL intensity over the Sahara and the convective activity over the Sahel.
These simulations were assessed against false-colour SEVIRI images and MODIS AOD (CLIM and DUST), CALIOP observations (DUST), WAHL intensity and wind fields from ECMWF analyses (REFE) and rainfall products from EPSAT-SG (REFE). The false-colour SEVIRI images and MODIS AOD revealed a strong variability in dust location that was captured well by the DUST simulation. Although the AOD was overestimated in DUST, the dust was distributed well along the vertical compared with lidar observations. In CLIM, the climatological AOD is too small over the Sahara. In comparison with the ECMWF analysis, the REFE simulation showed a shallower WAHL due to insufficient vertical mixing and some discrepancies in the AEJ pattern. However, both the pulsation of the WAHL and the variation in the AEJ intensity were reproduced, as were the two large convective events over the Sahel at the beginning and end of the episode. This indicates that large-scale forcing primarily drove the evolution of the WAHL intensity.
The impact of dust on the West African circulation was then analyzed with respect to REFE and is illustrated in Figure 16. Both CLIM and REFE simulations showed enhanced daytime heating and night-time cooling in the low troposphere over the Sahara. However, the dust forcing was stronger in DUST than in CLIM because of the larger AOD. In DUST, the anomaly of the temperature displayed a larger latitudinal gradient which was more restricted over the altitude of the AEJ, whereas in the CLIM experiment the larger latitudinal gradient covered a considerable part of the 925–400 hPa layer. In consequence, only a small contrast between REFE and CLIM was noticed at the end of the episode, while the largest impacts of dust on the AEJ were seen with the DUST simulation. Although not specific to this simulation, the AEW structure was, however, more intense in DUST, because its development and maintenance was favoured by a stronger AEJ. This intensification of the AEWs was closely linked with the modification of the moisture flux in the low troposphere and synoptic winds in the DUST simulation. In particular, the WAHL was enhanced in DUST on 16 and 19 July just after two strong northward advections of warm, dry air associated with the passage of an AEW. This enhancement in the AEW activity explained the increase in drying with southerlies and thus the variation in the WAHL pulsation.
In summary, dust affects the WAHL intensity in two ways. Firstly, there is a direct impact through dust-related radiative processes leading to an increase in the WAHL thickness. Secondly, an indirect effect is due to the increase in the southerly and northerly wind flows during an AEW episode. As West Africa is not always crossed by AEWs, a longer simulation over the whole monsoon season should provide a more robust analysis of the dust feedback on the WAHL in summertime.
This work was supported by the AMMA project. Based on a French initiative, AMMA was built by an international scientific group and is currently funded by a large number of agencies, especially from France, UK, US and Africa. It has been the beneficiary of a major financial contribution from the European Community's Sixth Framework Research Programme. Detailed information on scientific coordination and funding is available on the AMMA International web site http://www.amma-international.org. Computer resources were allocated by GENCI (project 0569). MSG observations were obtained from Climserv. We express our thanks to Marco Carrera (Environment Canada) and two anonymous reviewers for the careful review and the helpful suggestions.