The Harmattan is a dry dust-laden continental wind, and in the boreal winter Harmattan dust plumes affects many West African countries, including Ghana. When the Harmattan is strongest the southern part of Ghana is affected by the Inter Tropical Discontinuity (ITD). In this study, we investigate if the ITD functions as a barrier, preventing long transported Harmattan dust to settle south of, and below, it. This is done by analyzing a Harmattan dust outbreak, mapped using Earth observation (EO) data from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard the Meteosat Second Generation (MSG) platform, coupled with data from West African AERONET stations, and comparing these observations with wind data from NOAA's Air Resources Laboratory (ARL) program and the mineral suite of samples from seasonal dust deposits in north and south Ghana. In northern Ghana traces of minerals indicate a weak influence of particles from an arid environment, which is found consistent with the mapped dust plumes and NE wind directions. In southern Ghana the mineral composition show no sediments of an arid origin, the mapped dust plumes is less intense, and the surface wind directions and wind mass trajectories are more varying with lower wind speeds. Based on the results of this study it is concluded that dust deposited, or measured near ground, in the Harmattan period under the ITD, and south of it, does not contain material from the Chad Basin due to the local winds conditions.
 Each year dust plumes transported by the dry Harmattan wind covers the West African countries along the Gulf of Guinea [Dubief, 1979; Kalu, 1979; McTainsh, 1980; McTainsh and Walker, 1982; Cox et al., 1982; D'Almeida, 1986; Pye, 1987; Tiessen et al., 1991; Afeti and Resch, 2000; Washington et al., 2006; Sunnu et al., 2008]. One of the countries affected is Ghana, located approximately 2000–3000 km SW from the Bodélé Depression in the Chad Basin, which is consider to be the main dust source for the winter Harmattan [Engelstaedter et al., 2006; Washington et al., 2006]. The Harmattan wind raises large amounts of dust into the atmosphere as a result of a combination of special wind conditions and the topography surrounding the Bodélé Depression area [Washington et al., 2006]. However, other areas in the Sahara-Sahel region are also considered as a dust source to the countries bordering the Gulf of Guinea [Moreno et al., 2006; Rajot et al., 2008]. Furthermore, the sediment composition of a Harmattan dust plume is dynamic as an exchange of dust in suspension takes place along the transport path [Pye, 1987]. In Ghana, the Harmattan dust plumes can be experienced from November to March, when the NE Harmattan replaces the dominant SW maritime monsoon wind.
 In northern Ghana, studies of He et al.  and Lyngsie et al.  indicate that a significant amount of dust deposited in the Harmattan period is of local origin. Nevertheless, both studies find traces of sediment which may have its origin in an arid environment. However, in the southern part of Ghana the climatic conditions are affected by the Inter Tropical Discontinuity (ITD) when the Harmattan is strongest [Engelstaedter et al., 2006; Schwanghart and Schutt, 2008; Klose et al., 2010]. The ITD occurs where the warm and dry Saharan Air Layer (SAL) is undercut by the denser marine air and is located over Ghana between November to March and reaches its most southern location (approximately 5°N) in start January [Sunnu et al., 2008]. However, it is evident that during some years the ITD migrates to lower latitudes and that the position of ITD varies from year to year. The ITD is defined as the build up where the dry Harmattan wind converges with the wet Monsoon. The vertical structure of ITD is not perpendicular to the surface but is tilted upward toward the south all year round [Hastenrath, 1985]. Engelstaedter et al.  suggest that these converging winds favors dust emission by surface turbulences from dry convection, a phenomenon that results in vertical air currents. This disturbance will cause higher frequencies of events where the wind speed exceeds the threshold value required to bring a grain in suspension thereby leading to higher dust emissions under dry convection. These turbulent vertical wind currents will not influence the estimated mean horizontal wind speed [Engelstaedter et al., 2006]; hence, the area influenced by the ITD will appear to be characterized by calm or near calm near surface wind condition.
 The hypothesis of this study is that the ITD functions as a barrier that prevents long transported Harmattan dust originating from the Sahara region to settle south and below this barrier. This hypothesis is in line with the general assumptions by Afeti and Resch , Breuning-Madsen and Awadzi , Awadzi and Breuning-Madsen , and Sunnu et al. . However, all these studies assume that dust deposited or in suspension in southern part of Ghana, but north of the ITD, generally are long transported and do not consider how the ITD affect the surface wind conditions in this zone or south of it. The aim of this study is therefore to improve our understanding of the dynamics and deposits of the Harmattan phenomenon. This is done by analyzing a single Harmattan dust outbreak over a case area in West Africa, using a high-temporal and high-spatial resolution time series of satellite remote sensing data coupled with ground based point measurements from West African AERONET stations. This provides observations of a dust plume covering a period of 5 days, from 6–10 January 2007. The observations are compared to wind data and the mineral suite of samples from seasonal dust deposits from two locations in Ghana, West Africa.
2. Study Areas
 The two dust deposit sampling locations in this study are Kade and Tamale (Figure 1). The locations were selected due to their different locations respective to ITD. Tamale is located north of the ITD southern turning point whereas Kade is located near the southern ITD turning point [Sunnu et al., 2008] and is therefore affected by the ITD climate conditions around January when the Harmattan is strongest in Ghana. The sampling sites are located away from public roads and urban areas, minimizing the possibility of contamination from these sources.
 In Kade the sampling site (6.158°N, −0.921°E) is located in a forested area on University of Ghana's Agricultural Research Station. The dust sampler is placed in the middle of an open grass field, with 50 m to the nearest tree line. The mean monthly temperature is between 25°C and 30°C. There are two rain seasons, one between April and July and the other from September to November, the annual precipitation is between 1400 and 1600 mm. The Kade sampling site is located in the vegetation zone “moist semideciduous forest” [Atta-Quayson, 2007]. The dominant wind direction is southwesterly. The geological base rock is precambric and from the Lower Birrimian formation. The Birrimian formation covers the western and southwestern part of Ghana. This formation consists of metamorphosed, sedimentary and plutonic rocks. Almost half of the formation consists of alkaline granites [Boher et al., 1992].
 In Tamale the sampling site is located at Savannah Agricultural Research Institute, a few kilometers west of the town of Tamale (9.407°N, −0.989°E) in northern Ghana (Figure 1). The mean temperature is the same as above. There is one rainy season from July to October and the annual precipitation is 1000–1200 mm [Dickson and Benneth, 1985]. Tamale is located in the vegetation zone Guinea savannah which covers 45% of Ghana and consists mainly of grasses, shrubs and a few trees [Dickson and Benneth, 1985]. Between December and March the dominant wind is northeasterly and for the rest of the year it is southwesterly. The soil parent material was formed during the Palaeozoic and belongs to the Voltaian formation [Affaton, 2008]. The Voltaian formation covers 40% of Ghana and is primarily comprised of sandstone, schist, and limestone.
3. Methods and Materials
3.1. AERONET Data
 The Aerosol Robotic Network (AERONET) is a global network of sun photometers measuring the atmospheric aerosol content [Holben et al., 1998]. This includes estimates of aerosol optical depth (AOD) in several wavelengths. As no stations are present in Ghana, the January 2007 AOD values in the 670 nm spectrum are used from the nearest three West African stations: Ilorin (Nigeria), Djougou (Benin), and Ouagadougou (Burkina Faso).
3.2. MSG SEVIRI Data
 The Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard the Meteosat Second Generation (MSG) geostationary platform provides observations every 15 min from its location at 0° longitude over the equator. The observations produced are at a 3 km sampling resolution at sub satellite point in 11 spectral channels, as well as a high-resolution broadband visible channel with a 1 km spatial sampling distance [Schmetz et al., 2002]. MSG SEVIRI data has previously been used to estimate aerosol optical depth [Popp et al., 2007]. In this study the georeferenced raw data values of channels 1 and 10 are used. Channel 1 records in the visible (VIS) spectrum (λmin = 0.56, λcen = 0.635, λmax = 0.71 µm), which is sensitive to the atmospheric aerosol content. Channel 10 records in the thermal infrared (TIR) spectrum (λmin = 11.00, λcen = 12.00, λmax = 13.00 µm), which is used for masking out clouds.
3.3. Wind Data
 Wind data was obtained from National Oceanic and Atmospheric Administrations (NOAA), Air Resources Laboratory (ARL). ARL's archive program produces a global 1° longitude/latitude data set (based on the surface pressure in 10 m) every 3 h. These data sets are gathered in a weekly file which is uploaded to ARL's public available server. Data were collected as eight data sets a day covering the 5 day period where the dust plume was observed (6–10 January 2007) and throughout the 3 month Harmattan season (1 December 2006 to 28 February 2007). The percentage of wind in each direction class and the percentage of the wind in each wind class from the two periods were summed and weighted.
 Also from NOAA ARL, the online web version of the Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) [Draxler and Hess, 1997; Draxler and Rolph, 2013; Rolph, 2013] has been used to compute the trajectories of air masses during the dust event occurring between the 6 January 2007 and 10 January 2007. This has been done calculating 3 day backward trajectories and 3 day forward trajectories for 8 January at 09:00 h for Kade and Tamale, which fit the time period studied using SEVIRI data, when satellite images show high concentrations of atmospheric dust over Tamale. The HYSPLIT model calculations are based on archive data from the National Centers for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS) meteorological database in 1° latitude/longitude grid.
3.4. Sampling of Dust Data
 To ensure that resuspension did not occur, a sampler retaining the dust in water was chosen. Dust was collected in three individual plastic bowls, each approximately 40 cm deep and with an inlet area of 0.26 m2. The bowls were filled with distilled water to 10 cm below the rim, and the water level was topped up every week. A thin mesh grid with a mesh width of approximately 1 cm was used to cover the inlet area of the bowls to prevent animals from polluting the water. The samplers were installed on a frame 1 m above the ground. For a more detailed description of sampling technique and a discussion of sample height see Breuning-Madsen and Awadzi . When harvested, most of the water in the bowls was decanted on site and the remaining water containing the sediment was stored in 1.5 L plastic bottles. After drying in the laboratory (University of Ghana) all three samples were mixed before being shipped to the University of Copenhagen for further analysis. The inspection of the sample bowls at the end of the sampling period showed that algal bloom had occurred in some samples as the water in some of the bowls was greenish. The Harmattan dust sampling took place between December 2006 and March 2007. Topsoil (down to 10–15 cm) was collected in July 2008 as a bulk sample mixed by three jacks within a radius of 10 m from the dust samplers.
3.5. Dust Analyses
 The dust and soil samples were treated with 35% H2O2 to remove organic matter [Moore and Reynolds, 1997]. The samples were treated twice a day until they showed no further reaction with H2O2. Grain size distribution was measured using a Malvern Mastersizer 2000. Approximately 0.5 g of each sample was put in 0.01M Na4P2O7 solution and treated twice with ultrasound for 2 min. The pattern of diffraction was converted to particle size by Mie diffraction theory [Operators Guide, 1998]. The mineral suites of all samples were studied with powder X-ray diffraction on bulk, silt and clay fractions. For bulk and silt analyses randomly oriented specimens were made. For clay mineral suite the samples were prepared as described in Moore and Reynolds . The clay fraction was flocculated with a divalent ion (Ca2+) to stabilize the swelling clays (these were expected based on the study of He et al. ). The clay fraction was split into five subsamples. Each subsample was saturated with either (a) Mg2+, (b) Mg2+ and glycerol, (c) K+ and heated to 300°C, or (d) K+ and heated to 500°C, and if the sample did not collapse at 500°C, another sample was heated to 550°C. Oriented samples of the clay fraction were prepared on glass slides using the suspension method [Moore and Reynolds, 1997]. Bulk, silt and clay mineral suites were analyzed at room temperature with a Bruker D8 Advance, Cu-Kα source powered by 40 kV and 40 mA, equipped with a Ge monochromator, a fix slit, a step size 0.02, a step speed 1 s step−1, and a linear Position Sensitive Detector (PSD) detector (Lynx-eye) with an opening angel of 3.3° (bulk and silt) or 1° (clay).
3.6. Data Processing: Mineral Composition
 All the minerals were identified with Bruker Evaluation (EVA). The quantitative analyses on the bulk samples were made by Rietveld refinement on Bruker Total Pattern Analysis Solutions (TOPAS) 4, refinement was accepted when the weighted profile R-factor <12 and goodness-of-fit <3. The semiquantitative clay mineral analysis was conducted by an estimate of the peak area relative to the other phases.
3.7. Data Processing: Satellite Aerosol Product
 The period of January 2007 was examined using data from the SEVIRI instrument. Cloud and cloud shadow masking were performed by implementing multiple thresholds on a combination of both VIS and TIR data. Cloud shadows were found to be most effectively removed by using a threshold on spatial standard deviations in the TIR spectrum. Hourly estimates (09:00–16:00) of AOD in the red (635 nm) spectrum were produced by separating top of atmosphere reflectance into surface and atmospheric components using hourly minimum reflectance composites. The hourly minimum reflectance composites were generated on a monthly basis, justified by the slow change in surface reflectance properties during the dry season. Atmospheric reflectance component was regressed to AOD values using spatiotemporal concurrent measurements from the three AERONET stations. A comparison of the AOD values produced can be seen in Figure 2. The AERONET stations within the study area also supplies Ängstrom exponents (between 440 and 870 nm) that can be used to differentiate between mineral dust aerosols and aerosols from biomass burning over the stations [Dubovik et al., 2002], confirming the plumes as dust. A detailed description of the empirically based SEVIRI aerosol product, and the advantages of geostationary remote sensing of atmospheric aerosols, will be presented in a forthcoming publication (in preparation).
4.1. Remote Sensing Observations
 The monitoring of Harmattan dust plumes using geostationary observations from the SEVIRI instrument provides a subdiurnal temporal resolution available for tracking atmospheric aerosols during the Harmattan period. The Harmattan dust plume observed (Figure 1) using the 09:00 and 16:00 SEVIRI estimates, crosses the study area in 5 days (6–10 January 2007) from north-east to south-west with an average speed of 13–14 km/h. The coherent area of increased AOD value, here defined as the area where AOD > 1.75, was calculated for the 7 January at 16.00 scene and estimated to be ∼350.000 km2. This dust plume is one out of four clearly observed plumes in January 2007. All four plumes followed the approximate same trajectory and with similar speed, but size of the coherent center of increased AOD values varies from ∼100.000 to 500.000 km2.
4.2. Wind Data
 In Tamale a distinct northeasterly wind direction is seen in both investigating periods (Figures 3a and 3c). For the 5 day period the wind is from the northeasterly direction for 77% of the time, NNE being the dominating wind direction (35%) (Figure 3a). In the 3 month Harmattan season the wind is from a northeasterly direction 66% of the time, NNE (26%) direction being slightly dominating over NE (22%) (Figure 3c). For 94% of the investigated 5 day period the wind speed is 1–3 m s−1 and the last 4% of the 5 day period have a wind speed between 4 and 6 m s−1. In the 3 month period calm wind condition (less than 1 m s−1) is recorded for 11% of the time, wind speed of 1–3 m s−1 for 86% of the time and 4–6 m s−1 for 3% of the 3 month period.
 In Kade the wind rose is characterized by more variability during the 5 day period and no prevailing wind direction is seen, however wind from the northeasterly direction is slightly dominating (27%) (Figure 3b). For the 3 month sampling period the dominating wind direction is southwesterly (40%) but similar to the 5 day period the wind rose show more variability than in the corresponding period in Tamale (Figure 3d). The dominating wind speed in Kade is 1–3 m s−1 (approx. 60% of the time) and calm wind conditions are recorded approx. 40% of the time. This is the case for both the 5 day and the 3 months periods (Table 1).
Table 1. Wind Speed, Deposition Rate, and Mean Grain Size
 HYSPLIT computations of both backward and forward air mass trajectories are shown in Figure 4 and the corresponding vertical paths in Figure 5. The trajectories have been calculated at ground level (10 m), 500 and 2000 m height for both Kade and Tamale from 8 January 2007 at 09:00. The distance between points is the distance the air mass traverse in 1 h, and the lowest speeds are observed at ground level, while the air masses passing Tamale at 500 and 2000 magl (meters above ground level), as well as the 2000 magl at Kade, move at higher speeds. All trajectories crossing Tamale arrives from east-north-east and proceeds toward the Gulf of Guinea, at Kade only the 2000 magl trajectory do likewise. The trajectory of the air mass passing Tamale at 500 magl indicates it passing through the Bodélè depression 70–55 h earlier approx. 600 magl, Figure 5. 10–30 h after passing Tamale, the 500 magl trajectory moves to higher altitudes at ∼2000 masl (meters above sea level). An uplift to 1500 magl is also seen for the 10 magl trajectory 20 h after passing Tamale.The 10 m and 500 m trajectories from Kade show a different pattern, coming in off the coast at very low speeds and turning toward the Gulf again at 7.5°N (10 m trajectory) and 6°N (500 m trajectory). Both the 10 and 500 magl forward trajectories are lifted to 2000 masl ∼26 and 2 h respectively after passing Kade. Kade's 2000 magl trajectory starts near ground but is lifted up when passing the ridge between Togo and Ghana where it remains.
4.3. Deposition Rate and Grain Size
 In the 3 months of Harmattan period the average deposition rate in Tamale is 0.25 t km−2 d−1, Table 1. This is more than double of Kades deposition rate of 0.12 t km−2 d−1. The mean grain size of the deposited dust is the same (9 µm) at the two sampling sites.
4.4. Mineral Composition
 The deposited dust in Tamale is dominated by quartz (62%), followed by clay minerals (28%) and approximately 10% of feldspar (Table 2). The clay fraction is strongly dominated by kaolinite, but also contains smectite. Traces of pirssonite and analcime occur in the clay fraction. In Kade quartz is also the dominating mineral (57%) in the deposited dust. Clay minerals account for 26% and kaolinite is the dominating clay mineral, but the fraction also contains considerable amount of smectite. Fourteen percent of the dust deposited in Kade belongs to the feldspar group and 2% to the amphibole group, specifically pargasite. The topsoil in Kade shows a high degree of weathering by being highly dominated by quarts (72%) and contains of small amount of feldspar (1.5%). The topsoil contains 21% clay minerals which is dominated by kaolinite but also contains considerable amounts of smectite and traces of chlorite.
Table 2. Mineralogical Composition of Harmattan Dust in % From Tamale and Kade, Ghana, Determined by XRD and Quantitative Bulk Analysis by Rietveld Refinement
 In Tamale traces of pirssonite and analcime in the clay fraction indicate a deposition of material from an arid environment. These findings suggest that dust deposited in Tamale over a 3 months period is to some degree influenced by sediment from an arid environment. This is in agreement with the satellite based observations of atmospheric aerosols, estimated as AOD, as a 5 day dust plume is clearly observed to pass over Tamale. The climate conditions, as seen from wind rose and HYSPLIT backward trajectories during the 5 day dust plume period, are representative for the prevailing wind conditions of the 3 month Harmattan period which implies that deposition of material from the Chad Basin is possible throughout the season. The four observed dust plumes in January further confirm this. All this indicates that deposition of long transported dust at Tamale is possible and can be assumed to happen. Nevertheless, a study by Lyngsie et al.  finds that deposited dust sampled at Tamale outside the Harmattan season has similar elemental composition as dust sampled in the Harmattan season and further that dust sampled within the Harmattan period is missing the trace element signature found in Bodélé sediments, making it likely that a majority of deposited dust is of local origin, with long transported dust being a minor constituent.
 The mineral composition from the two locations resembles each other, but the dust sample from Kade has no traces of any minerals with a specific arid origin. The pargasite and smectite found in the deposited dust from Kade further indicate the contribution of minerals with pure local origin as these minerals is also found in the vicinity of Kade. Even though pargasite is not found in the topsoil it cannot be used as a tracer as pargasite could be expected as granite is common in the Biriman formation (cf. Study area).
 Compared with Tamale the AOD values are lower in Kade and the surface wind directions are more varying with lower wind speeds throughout the 3 month period. The observed speed of the dust plume is much higher than the surface wind speed (∼13 km/h as opposed to ∼3 km/h) indicating a majority of aerosols in suspension higher in the troposphere where wind speeds are higher. This corresponds well with the back and forward trajectories, where the surface trajectories (10 m) from both sites are much slower than the ones observed at higher altitudes at Tamale (500 and 2000 m).
 A study by Washington et al.  shows a significant variation of transport paths and altitude of air mass throughout an average year based on median annual cycle of 10 day trajectories for the Bodélé over a 29 year average. Compared to this long-term mean the altitudes of the trajectories found in this study shows some similarity of those found for the northern hemisphere winter months where an uplifting of the air masses from near ground to higher altitudes is seen around the same latitude (∼5°N).The paths of back trajectories during the dust event presented here are not similar to the long term mean direction of forward trajectories from Bodélé. This is consistent with dust events crossing Ghana being observed as individual events separated in time, and not a continuous phenomenon.
 The more than double deposition rate in Tamale compared to Kade is in line with the findings of Klose et al. . They find that the zone of maximum dust event activity is located north of ITD as based on synoptic station data from 1983 to 2008. Furthermore they find that this zone is closely linked to ITD spatiotemporal evolution and that this is especially evident for the months between November and March when Ghana is influenced by the ITD. The remote sensing based observations of the 5 day Harmattan dust plume supports these findings, as estimated AOD values for the column of atmosphere above Tamale are higher than above Kade. However, the lower deposition rate in Kade could also simply be a consequence of the more dense vegetation cover in southern Ghana which, ceteris paribus, reduces the overall dust suspension compared to northern Ghana where vegetation is sparser.
 Considering the low deposition rate, the low wind speed, the spread wind rose in the 5 day plume period, the prevailing wind direction in the 3 months sampling period, the 10 and 500 magl back trajectories and finally the AOD values, it is most likely that none or very little long transported Harmattan dust have been deposited on ground in Kade. This study therefore indicates that atmospheric disturbances near the surface are responsible for the dust transportation in Kade and not the NE Harmattan wind. It could be hypothesized that when the Harmattan dust cloud converges with the (non perpendicular) ITD it is undercut by the moist Monsoon air and raised to higher levels in the atmosphere, as reported in by Kalu  in Nigeria. This is in agreement with observation by Stuut et al.  who, offshore of the Ghanaian coast, finds weak surface winds from S to SE, but aeolian material transported from NE at higher altitudes (700–3400 masl), which they attribute to the tilted vertical structure of the ITD. This tilted structure is also confirmed here by the 10 and 500 m Kade wind mass trajectories, as the 10 m trajectory moves further north than the 500 m before turning toward a NE direction (Figure 4) and by the gradually uplift of both the 500 and 10 magl forward trajectories in Kade and Tamale. It could also be argued that the uplifting of the dust plume is necessary to get the dust affecting the countries around Gulf of Guinea transported to the Amazon forest as reported by, e.g., Koren et al. , Ben-Ami et al. , and Huang et al. .
 All material presented in this study indicates that long transported Harmattan dust do not deposit in southern Ghana when the ITD is positioned in its yearly average southernmost position around 5°N, contrary to what is concluded by, e.g., Aboh et al.  and Dionisio et al. . Additionally the current results imply that the interpretation of the higher dust mass concentrations in the Harmattan season as reported by Afeti and Resch  and Sunnu et al.  may be caused by dry convection rather than being long transported Harmattan dust.
 The results of this study can only serve as an indication of the dynamics of dust deposits in West Africa and the following methodological improvements are suggested to future studies: In order to determine provenance of the deposited dust additional isotope analysis on the dust samples is recommended. Furthermore in situ climatic observations at the sampling site would provide a more detailed view of the local climate conditions.
 This study presents a multidisciplinary approach to test the hypothesis that ITD functions as a barrier for long transported Harmattan dust in West Africa. The wide range of data presented here all confirms this hypothesis: In northern Ghana traces of pirssonite and analcime indicate a weak influence of particles from an arid environment. This is consistent with the passing of dust plumes as observed using remote sensing estimates of AOD, the wind mass trajectories and northeasterly wind directions observed using data from NOAA's ARL program. In southern Ghana the mineral composition does not show any sediments of an arid origin. Furthermore the observed AOD values are lower, and the surface wind directions are more varying with lower wind speeds. The wind mass trajectories at higher altitudes and the 5 day dust plume shows much higher speeds than the surface wind speed, indicating a majority of aerosols in suspension higher in the troposphere. Based on this study it is therefore concluded that long transported Harmattan dust, while clearly observable over Southern Ghana, is not deposited because of the ITD both acting as a barrier and lifting the dust plume to higher altitudes due to its tilted structure.
 We would like to thank the AERONET for making the data freely available, especially the principal investigators and site managers of the three stations used in this study: Ilorin station: Rachel T. Pinker, Clement Akoshile. Djougou station: Philippe Goloub, Armand Mariscal, Luc Blarel. Ougadougou: Didier Tanri. The authors will also like to thank the staff at Ecological Laboratory, University of Ghana, for technical support in laboratories and practical help with the fieldwork.