Precipitation is a major socioeconomic factor in the Guineo-Soudanian zone of tropical West Africa with its distinct summer rainy season from May to October. Albeit rare, precipitation during the dry season can have substantial impacts on the local hydrology and human activities reaching from the rotting of harvests to improved grazing conditions. This study provides an observationally based synoptic and dynamical analysis of an abundant rainfall event during the dry season of 2003/04 that affected the countries of Nigeria, Benin, Togo, and Ghana. The results point to a forcing of the rainfalls from the extratropics in the following ways: 1) Upper-level clouds and moisture to the east of a weak, quasi-stationary extratropical disturbance enhance the greenhouse effect over the Sahel and the adjacent Sahara, and thereby cause a net-column warm anomaly and falling surface pressure. 2) One day before the precipitation event, negative pressure tendencies are further enhanced through warm advection and subsidence associated with the penetration of a more intense upper-trough into Algeria. 3) The resulting northward shift and intensification of the weak wintertime heat low allows low-level moist southerlies from the Gulf of Guinea to penetrate into the Soudanian zone. 4) Finally, daytime heating of the land surface and convective dynamics initiate heavy rainfalls. Operational forecasts of this event were promising, pointing to a strong control by the comparatively well-predicted extratropical upper-level circulation.
Tropical West Africa is characterized by a monsoon climate with the largest part of the annual precipitation falling during the boreal summer, when moist southwesterlies from the Gulf of Guinea penetrate far into the continent in response to the intense thermal low over the Sahara (Hamilton and Archbold 1945; Griffiths 1972a; Leroux 1983a, b, 2001; Buckle 1996). The duration and intensity of the rainy season is a major factor for agricultural production potential and thus socioeconomic development in this region (see, e.g., Carter 1997; Stahr 2000). During the boreal winter months in contrast, dry and often dusty northeasterly harmattan winds from the Sahara prevail and rainfalls are practically absent to the north of a narrow coastal belt.
On 20 January 2004, however, infrared (IR) satellite imagery shows deep moist convection spreading from the Guinea coast into the Soudanian zone (9°–12°N; Fig. 1). Many stations in Ghana, Togo, Benin, and Nigeria received more than 40 mm of rain with extreme values of more than 100 mm recorded in central and southeastern Benin (see section 3a). Arriving completely unexpectedly, the intense rainfalls had substantial impacts on the local residents. Drenched harvests stored in the open air such as cotton rotted, while the greening of pastures in the open savannah, a truly unusual sight at this time of year, was beneficial to cattle nomads. The large impact on the local hydrology is illustrated by soil moisture measurements from a site in central Benin (Fig. 2). Before the event soil moisture steadily increased from values ∼7 percent by volume near the surface to ∼27 percent by volume in the 120–140-cm layer. Shortly after the precipitation set in, moisture in the uppermost layer jumped from 7 to 30 percent by volume in only half an hour followed by smaller jumps in the layers down to 1 m. While the surface layer returned to normal values by mid-February, the deeper layers stored water for far more than a month, underlining the substantial disturbance of the hydrology brought about by this single event.
Naturally, studies on precipitation in tropical West Africa have focused on the rainy season from June to September, while work on rainfall during the boreal cool and transition seasons have concentrated on events in subtropical North Africa. The latter are often related to equatorward-penetrating extratropical disturbances, termed diagonal troughs due to their orientation from southwest to northeast (Flohn 1975; Nicholson 1981). The accompanying elongated bands of mid- and high-level clouds connecting the tropics and subtropics (as in Fig. 1) have been named tropical plumes (TPs; McGuirk et al. 1988) and are a frequent feature over the eastern Atlantic and adjacent Africa (Iskenderian 1995). Usually, TPs form along the equatorward flank of subtropical jet (STJ) streaks (Knippertz 2005). Diagonal troughs have caused extreme precipitation along the western (Buckle 1996, p. 214; Leroux 2001, 167–173; Knippertz and Martin 2005), northwestern (Fink and Knippertz 2003), and northern fringes of North Africa (Winstanley 1970; Flohn 1975). Dubief and Queney (1935) attribute two winter- and springtime rainfall events in the central Sahara to “Saharan disturbances,” surface lows that move westward in tropical West Africa and then cross the Sahara toward the Mediterranean Sea on an anticyclonic trajectory. Dubief and Queney (1935) mention that these situations are associated with rainfall in the Soudano-Sahelian zone of West Africa. Nicholson (1981) calls these systems Soudano-Sahelian depressions (see also Buckle 1996, p. 177) and suggested an interaction of diagonal troughs with low-level African easterly waves (AEWs). To the best of our knowledge, the exact dynamical link between diagonal troughs and heavy rainfalls in the Soudanian zone during boreal winter has not yet been documented.
The present study will address this issue on the basis of a detailed case study of the unusual January 2004 event that brought substantial precipitation to both the subtropical hyper-arid Algerian Sahara (accumulated amounts given in Fig. 1) and to tropical West Africa (see section 3a). In section 2, we provide information on the observational data used in this study. Section 3 contains a detailed analysis of the precipitation in the West African tropics and the accompanying evolution of the large-scale circulation. The dynamical analysis in section 4 is based upon a special form of the pressure tendency equation that allows us to distinguish between the contributions from different dynamic and thermodynamic processes. Section 5 contains a summary of the identified mechanism and a discussion of the implications of our findings for the predictability of dry-season rainfall events and the potential for the mitigation of their impacts.
The precipitation data used in this study come from both rain gauge observations and satellite estimates. The first category includes standard synoptic station measurements, usually with a temporal resolution of 3–6 h. In addition data from a dense network of rain gauges in the upper Ouémé Valley (UOV) in central Benin with a temporal resolution of up to 1 min are used (Schrage and Fink 2007). To get a spatial distribution of the precipitation, we used the IMPETUS1 MultiSensor Algorithm developed by M. Diederich of the University of Bonn (Speth and Diekkrüger 2006, 82–89). This algorithm uses passive microwave sensors from five satellites [Special Sensor Microwave Imagers (SSM/Is) aboard the three Defense Meteorological Satellite Program (DMSP) satellites F13–F15, the Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) aboard Aqua, and the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) aboard the TRMM satellite]. The microwave information is merged with Meteosat-7 IR data using morphing and probability/histogram techniques. The precipitation estimates have a temporal resolution of 30 min and a spatial resolution of 5–6 km. The product was validated with rain gauge observations from almost 100 stations in Benin for the period 2001–06. In addition, we used 3-hourly TRMM 3B42 version 6 (V6) rainfall estimates from the National Aeronautics and Space Administration (NASA) to assess rainfall associated with a cloud band across the Sahara.
b. Other data
The large-scale atmospheric circulation and dynamics are investigated using operational analysis fields from the European Centre for Medium-Range Weather Forecasts (ECMWF) at 1° × 1° latitude–longitude horizontal resolution (Persson and Grazzini 2007). It would have been desirable to use direct satellite measurements of top of the atmosphere (TOA) solar and thermal radiation, but unfortunately data from the Geostationary Earth Radiation Budget Experiment (GERB) are not available for the period under study and radiometers aboard polar orbiters yield a too poor temporal sampling to calculate daily radiation balances. Therefore, we took TOA estimates from 24-h operational ECMWF forecasts. Climatological means of TOA radiation and dynamical quantities are based upon the satellite period of the 40-yr ECMWF Re-Analysis data (ERA-40; Uppala et al. 2005) from 1979 to 2002, which is usually regarded as the more reliable part of the dataset for climatological analysis (Kållberg et al. 2005). We also used standard surface observations from synoptic stations and 10-min averages from an automated station at Mont de Gaulle in central Benin (9.1°N, 1.9°E, 333 m) to delineate the changes in surface parameters before and during the rainfall event. The soil moisture measurements presented in Fig. 2 are taken from a time domain reflectometry (TDR) sensor buried in a cornfield close to the Mont de Gaulle site near the village of Dogué. For the analysis of clouds, we use standard Meteosat-7 images in the IR channel (10.8 μm).
3. Synoptic evolution
The following two subsections provide an analysis of the synoptic evolution during the period 15–21 January 2004. The first part concentrates on the extreme precipitation event in tropical West Africa and investigates the involved convective dynamics. The following part places this event into the context of the large-scale circulation and the local pressure and humidity distribution over West Africa. Section 4 then provides a detailed analysis of the involved dynamics based upon the pressure tendency equation.
Climatologically, the month of January is the heart of the dry season in tropical West Africa, and long-term average precipitation ranges between 0 and 7 mm to the north of about 8°N (WMO 1996). The 48-h accumulated precipitation sums on 20 and 21 January 2004, as estimated from rain gauges and from satellites, show substantial deviations from this climatology over a wide region including parts of Ghana, Togo, Benin, and Nigeria (Fig. 3). Maximum rainfall was observed in southern and central Benin and neighboring Nigeria with several stations recording sums of more than 100 mm [e.g., Ketou (7.35°N, 2.60°E) and Ouando (6.55°N, 2.62°E), both operated by the Direction de la Météorologie Nationale (DMN) of Benin]. A maximum of 122.2 mm was observed at Adiangdia-Ouest [9.43°N, 1.98°E; run by the French Institute de Recherche pour le Development (IRD)]. Even areas north of 10°N received substantial amounts of precipitation (e.g., 64.3 mm at Fo-Boure, 10.12°N, 2.40°E; DMN). The maxima of 103 mm over southern and of 75 mm over central Benin in the satellite data broadly agree with the rain gauge results (Fig. 3). The northernmost station with substantial precipitation is Dassari (10.78°N, 1.13°E; DMN) with 9.0 mm, while the satellite product indicates rainfall even farther north. Unfortunately, the large precipitation sums over eastern central Ghana in the satellite estimate cannot be verified due to a limited number of surface observations. The isolated rainfall area at the Ghana–Burkina Faso border near 10°N, 3°W is also remarkable. The rainfall bands in the satellite estimates have an orientation from southwest to northeast in accordance with the cloud structure seen in Fig. 1.
The arrival of the precipitation in central Benin during the evening hours of 20 January 2004 is documented in surface observations from the station at Mont de Gaulle (Fig. 4). During the afternoon, winds are fairly light and mainly from the northerly directions, while temperatures reach a maximum of about 34°C indicating strong daytime heating. Shortly before 1800 UTC, winds suddenly accelerate to 12 kt and turn to predominantly southerly directions. Winds remain strong and gusty over the following hours, while the temperature begins to drop. More than 1 h after the gust front, precipitation begins and quickly reaches an intensity of more than 50 mm h−1. During this time temperatures fall dramatically and the relative humidity increases to more than 90%. Two weaker precipitation periods follow around midnight and in the early morning hours of 21 January, accumulating to 57.8 mm. Due to the diurnal and semidiurnal atmospheric tides in the tropical belt (e.g., Dai and Wang 1999), we express synoptic pressure tendencies in terms of 24-h changes, which reach values of +2 to +5 hPa for several hours after the onset of the rains. Wind and temperature observations suggest an initiation of the precipitation by an outflow boundary of a convective system farther south. To test this, we compiled an isochrone chart from the dense rain gauge network in the UOV (Fig. 5). Isochrones are lines connecting points with the same onset of precipitation recorded by the automatic rain gauges, as in Schrage and Fink (2007). The analysis indicates a northward propagation over the western part of the UOV with rainfall setting in around 1900 UTC in the south and around midnight in the north. A coherent propagation is more difficult to see farther to the east, where the station density is smaller.
The presented observations suggest a penetration of the rainfalls from the Guinea Coast into the Guineo-Soudanian zone during the course of 20 January 2004. Satellite images and half-hourly rainfall estimates from the IMPETUS MultiSensor-Algorithm (not shown) corroborate the general northward or northeastward progression of the convection. It should be noted that throughout most of the year convective clusters in the investigated area are steered southwestward by the ambient flow (e.g., Fink et al. 2006), such that the propagation direction in the present case is quite unusual. The beginning of the rainfalls in the afternoon suggests that daytime heating is important to sufficiently destabilize the atmosphere. The triggering of such intense precipitation is only possible in the presence of sufficiently moist air at low levels from the Gulf of Guinea. Consequently, the next section will address changes in the regional circulation that cause this moistening.
b. Local circulation
The strength of the heat low and the associated monsoon circulation are known to experience a distinct out-of-phase diurnal cycle. While the former peaks in the afternoon hours (i.e., at around 1800 UTC), when surface winds are weak and erratic due to the turbulent boundary layer, the latter is strongest in the morning (Parker et al. 2005). We have opted to construct surface maps from synoptic station observations of mean sea level pressure (MSLP), 2-m dewpoint, and 10-m wind at 1200 UTC, since data coverage is generally largest at this time and the surface circulation is not too strongly disturbed. The wind observations are not assimilated in ECMWF analyses and are therefore an important additional source of information.
On 17 January 2004, 3 days before the precipitation begins, weak MSLP gradients and light winds prevail across tropical West Africa (Fig. 6a). Easterlies and low dewpoints suggest a Saharan origin of the air over the northern part of the domain. These dry and often dusty harmattan winds are the result of the pressure gradient between the subtropical anticyclone centered around 30°N and the weak continental heat low, usually located just to the north of the Guinea coast at 7°–8°N (Hamilton and Archbold 1945; Leroux 2001). The sharp boundary between the Saharan air and the moister air from the Gulf of Guinea is called the intertropical front2 (ITF) and is depicted by the 14°C-dewpoint isopleth in Fig. 6, following the definition of Buckle (1996, p. 140). On 17 January, the ITF is displaced north over the western part of the domain and close to its climatological position over the region that is affected by the rainfalls on 20 January (Fig. 3). To the south of the ITF, the winds are predominantly from the west or southwest.
Two days later, on 19 January, the MSLP distribution has changed considerably with a now clearly defined minimum of 1007.5 hPa near the Ghana–Burkina Faso border (Fig. 6b). In this region the pressure has fallen by about 2 hPa over the previous 48 h. With respect to the climatological position, this pressure minimum is shifted northward by about 4°–5° latitude. Wind observations reflect the cyclonic circulation around the low center with a southerly component reaching up to 12°N over Benin, Togo, Ghana, and Burkina Faso. This flow brings moist air from the Gulf of Guinea much farther into the continent than is usual, as demonstrated by the change in dewpoint and in the position of the ITF, respectively. From 17 to 19 January precipitable water over central Benin increases from 30 to 45 mm (not shown). During the following night the southerly flow continues as revealed by the observations from Mont de Gaulle (Fig. 4, middle panel), leading to a further increase in precipitable water.
Surface observations from synoptic stations in the Sahelian part of Mali and Niger, that is, to the north of the area shown in Fig. 6, reveal falling MSLP over most of the period from 15 to 19 January 2004 (Fig. 7). At some of these stations this decrease is not monotonic at some synoptic hours (e.g., at Tombouctou at 0600 UTC), which is most likely related to local effects such as precipitation (see section 3c). While the pressure fall is slow, but continuous during 15–18 January, a quite pronounced drop is observed on 19 January followed by an equally distinct rise over the course of 20 January. This evolution is best seen at 1800 UTC, the time with the strongest heat low effect. The magnitude of the total pressure fall between 1800 UTC on 15 and on 19 January decreases from north to south with almost 5 hPa at Tessalit (Fig. 7a) and 2–3 hPa at Tilabery, where the MSLP even rises slightly between 15 and 16 January (Fig. 7d). This pattern results in a reduction of the north–south pressure gradient over this region and a relatively weak harmattan wind (Fig. 6b). The presented observations show that the intensification and the northwestward shift of the weak wintertime heat low has allowed an unusual northward penetration of the monsoon layer, ultimately the prerequisite for the strong rainfalls on 20 January. In the following, we are going to place this development into the context of the large-scale circulation over northern Africa during the previous days.
c. Large-scale circulation
Figure 8 shows the large-scale evolution of cloudiness, the 300-hPa geopotential height, and the MSLP during 15–20 January 2004. On 15 January, the MSLP distribution shows the typical zonal structure with a relatively strong gradient between high pressure over northwestern Africa and the monsoon pressure trough with the ITF just to the north of the Guinea coast (Fig. 8b). Note that the MSLP values in Figs. 6 and 8 cannot be compared to each other directly due to the strong diurnal cycle in the tropics. There is a weak upper-level trough over the west coast of West Africa with a pronounced TP on its eastern side that reaches from the tropical Atlantic Ocean to central Algeria (Figs. 8a and 8b). Below the TP, light rainfalls were observed on 14 and 15 January that accumulated to 3 mm at Linguere in Senegal (15.38°N, 15.11°W) and 12 mm at Tidjika in Mauritania (18.34°N, 11.26°W). Such rare winter events are locally known as heug (Griffiths1972b; Buckle 1996, p. 214) or mango rains (Leroux 2001, p. 173).
On 16 January, the upper trough over the west coast intensified, and the TP reached deeper into the tropics and adopted a more meridional orientation (Figs. 8c and 8d). Tessalit in northern Mali (20.2°N, 0.98°E) reported some rain on this day. Below the TP, the MSLP fell when compared to the previous day (Figs. 8b and 8d). The northward excursion of the ITF in Fig. 6a is most likely the result of this pressure fall and the subsequent deeper penetration of southerlies into the continent. In section 4 we will present arguments that this pressure fall is related to the heating of the atmospheric column caused by the locally enhanced greenhouse effect due to the high clouds and a higher water vapor content in the region of the TP (for documentation of the latter, see Waugh 2005). Along these lines of argument, the pressure rise over Ivory Coast and Ghana, where weak precipitation of up to 2 mm is observed, could be related to a cooling caused by the increase in albedo due to thick clouds. On 17 January, the upper trough and the associated clouds continue to move eastward across West Africa (Figs. 8e and 8f), bringing traces of precipitation to parts of Mali (e.g., Mopti, Tessalit, and Tombouctou). The MSLP fell over large parts of the Sahel, consistent with the station observations presented in Fig. 7, the locations of which are marked in Fig. 8.
Over the next day the trough remained almost stationary, and the TP became shorter and more compact (Figs. 8g and 8h). Again, traces of precipitation were observed in Mali and southern Algeria with a maximum of 2 mm at the elevated station Tamanrasset in the Ahaggar Mountains (22.47°N, 5.31°E). TRMM rainfall estimates suggest that more showers occurred in southern Algeria and adjacent Mali on this day (not shown). The MSLP rose slightly in this region, while the pressure continued to fall over parts of the Sahel and the Soudanian zone, in agreement with Figs. 6 and 7. A slight northward shift of the ITF can also be observed. A pronounced upper trough begins to penetrate into Algeria from the western Mediterranean Sea (Fig. 8h). It is associated with low MSLP in a band from western Algeria to Sardinia. On 19 January, this trough merges with the one over West Africa forming a large system with a strongly tilted horizontal axis (Fig. 8j). The TP moved eastward to Niger and Libya, while new clouds formed along the southern and eastern sides of the upper trough (Fig. 8i). Over western Algeria a surface low forms in the lee of the Atlas Mountains and eliminates the north–south pressure gradient usually found in this area (Fig. 8j). The MSLP falls over large parts of the Sahel and the Soudanian zone leading to the northward shift of the heat low and the ITF (cf. Fig. 6). In contrast to the previous days, no precipitation is observed over the Sahara or the Sahel during this period.
Finally, on 20 January the upper-level trough has cutoff over Algeria with a minimum geopotential height of about 890 gpdm, while the surface depression has deepened to 1004 hPa and moved to eastern Algeria (Fig. 8l). A trailing pressure trough has formed over the Sahara with an axis from western Niger to southwestern Libya. Three distinct cloud features are evident in the IR satellite image (Fig. 8k): A compact convective region just below the cutoff produces strong precipitation over the Algerian Sahara (see Fig. 1). Weaker precipitation over western Libya is associated with the upper cold frontal cloud band farther to the east. Finally, the remnants of the TP have taken a more zonal orientation and stretch from West Africa to the Red Sea. To the southwest of the cutoff, the MSLP has begun to increase, which is consistent with Fig. 7. Despite the pressure rise over Burkina Faso, the moisture that has penetrated into the continent over the previous days allows for the generation of deep convection starting soon after the time shown in Fig. 8k (see Fig. 1 and section 3a).
The cyclone over Algeria intensified over the next 2 days and caused strong dust storms over Libya, Egypt, and parts of the Middle East (not shown). The associated winds are sometimes referred to as khamsin or ghibli (Buckle 1996, 83–84). Cyclones that form in the lee of the Atlas and then track to the eastern Mediterranean Sea have been termed sharav cyclones and occur most often in springtime (e.g., Alpert et al. 1990; Trigo et al. 1999). These systems depend on the lee effect (Egger et al. 1995) and a strong upper-level potential vorticity anomaly (Thorncroft and Flocas 1997), and can at times cause severe weather in the Mediterranean (Horvarth et al. 2006).
a. Pressure tendency equation
The analysis in section 3 revealed the importance of falling MSLP over the Sahel and the adjacent Sahara for the rainfalls farther south. Consequently, the dynamical analysis will concentrate on a physical understanding of this process. Over lowland areas the MSLP is closely related to surface pressure, psfc. The classic approach to this problem is expressing the psfc tendency in terms of mass divergence integrated from the surface to the TOA using the hydrostatic and the continuity equations (Panofsky 1956):
where g is the acceleration of gravity, ρ air density, and u the three-dimensional wind vector. Here, we take the alternative approach proposed by Kong (2006) and express the local density tendency in terms of total density tendency and advections, and then use the first law of thermodynamics and the ideal gas law to finally yield
where v is the horizontal wind vector in Cartesian–pressure coordinates, Tυ the virtual temperature, R the gas constant for dry air, ω the vertical velocity in pressure coordinates, cp the specific heat capacity at constant pressure, and Q̇ representing the diabatic heating rate (for details see the appendix).
The first term on the right-hand side of Eq. (2) expresses the contributions from the horizontal advection of the virtual temperature and will be denoted the temperature advection term hereafter (as in Kong 2006). Warm advection causes a psfc fall; cold advection a psfc rise. The second term is essentially the product of vertical motion with the difference between the dry adiabatic and the actual lapse rate. We will call this term the vertical motion term. Assuming a stable stratification, subsidence (i.e., positive ω) leads to adiabatic warming and thus falling psfc, while ascent causes cooling and rising psfc. For a dry-adiabatic atmosphere, the contribution of this term is always zero. The last term is denoted as the diabatic term. The main contributors to Q̇ are phase changes of water in the atmosphere, latent and sensible heat fluxes from the underlying surface, and radiation. For positive Q̇, the column expands and psfc falls, while a net cooling causes psfc to rise.
In the following we will discuss the contributions of the three terms to the development of three stereotypical weather systems. For baroclinic waves in the extratropics, an important part of the dynamics is essentially dry (e.g., Thorncroft et al. 1993). Assuming relatively small contributions from sensible heat fluxes and radiation, the pressure tendency equation will be dominated by the temperature advection and vertical motion terms. Overall, rising motions in regions of warm advection and descent in regions of cold advection lead to an extensive cancellation of the two terms. Surface cyclones form in regions where the warm advection dominates and causes psfc to fall. In some cases latent heating supports this process. In tropical cyclones temperature advection is usually weak. Here, the latent heat release in the eyewall and spiral bands causes psfc falls, which are not compensated by the opposing effect of ascending motions, as long as the vertical stability is sufficiently low (Kong 2006). Subsidence in the stably stratified eye also contributes to negative psfc tendencies. For a summertime heat low over a dry continent, the sensible heat fluxes from the ground are the dominating contribution to the negative pressure tendency during the day.
b. Climatological balance
In this section climatologies of the different contributions to Eq. (2) are discussed in order to qualitatively understand the prevailing balances that keep psfc constant over West Africa on longer timescales. In section 4c we will analyze the disturbances to this balance that caused the pressure fall during the January 2004 precipitation event. For the temperature advection and vertical motion terms, the 24 Januarys from 1979 to 2002 from the ERA-40 dataset were used. Vertical integrals were computed for the entire troposphere from the surface to 100 hPa, for the surface to 700 hPa, and for 700 to 100 hPa, but for the sake of brevity only the results for the whole troposphere are shown. The psfc tendencies were averaged over the four synoptic hours (0000, 0600, 1200, and 1800 UTC) and converted into units of hectopascals per day. The former is necessary to eliminate the strong diurnal cycle in, for example, psfc and temperature.
Unfortunately, a quantitative assessment of the diabatic term is impossible, because vertical profiles of the heating rates are not archived at the ECMWF (C. Maas, ECMWF, 2007, personal communication). We can therefore only provide a qualitative discussion of the contributions from single diabatic processes. Over the dried-out soils of the Sahara and the Sahelo-Soudanian zone during its dry season (see Fig. 2), latent heat fluxes from the surface and latent heat release into the atmosphere are small and will not contribute much to the diabatic term in a climatological sense. Significant contributions can be expected from radiation and sensible heat fluxes. In principle, one would have to explicitly calculate a vertical integral over the radiative and sensible heating rates as in Eq. (2). However, for our application the total energy that heats an atmospheric column is most important. In the absence of latent processes, this quantity is closely tied to the radiation budget at the TOA. It is not important whether the heating occurs due to a positive radiation balance in a particular atmospheric layer or through radiative heating of the surface and subsequent sensible heat fluxes into the atmosphere. One process that could disturb this relation is a net flux of sensible heat into the ground, which we assume to be of minor importance in a climatological sense.
Another problem is the temperature in the denominator of the diabatic term, which implies a larger impact on the density and thus the pressure of heating in cold regions. Consequently, radiative cooling in the upper troposphere affects psfc more than the equivalent solar heating of warmer lower layers. It is therefore conceivable to have a positive TOA radiation budget without there being an effect on psfc. Given the usual range of temperatures in the atmosphere, we do not expect this to be a first-order effect. We conclude that the TOA radiation budget is a good proxy for the diabatic term under the special circumstances of wintertime North African climate. In contrast to standard meteorological variables, the TOA radiation fields are not operationally analyzed and are therefore taken from 24-h ECMWF forecasts for the 24 Januarys of 1979–2002.
In the following we will discuss balances between climatological values of the three terms in Eq. (2). The horizontal temperature advection term (Fig. 9a) shows a region of weak negative psfc tendencies over the West African coast, which is associated with warm advection in the right entrance region of the STJ (see the vertically averaged winds in Fig. 9a). Positive values of up to 5 hPa day−1 due to cold advection are analyzed over northern Africa, while the deep tropics reveal very small tendencies. By far the largest contributions come from the levels above 700 hPa (not shown). Mean subsidence and stable stratification over most areas to the north of 10°N lead to a strongly negative vertical motion term (Fig. 9b) and thus negative net dynamic psfc tendencies as low as −5 hPa day−1 (Fig. 9c). Again, most contributions come from upper levels, but subsidence below 700 hPa is also of importance (not shown).
The mean net TOA solar radiation has an almost zonal distribution over northern Africa with a maximum of more than 320 W m−2 near the Guinea coast (Fig. 10a), just south of the climatological position of the heat low. The zonal increase from the Sahara to the adjacent Atlantic reflects the high albedo over the desert, while the decrease toward the equator is caused by the cloudiness in the intertropical convergence zone (ITCZ). The longwave component also displays a zonal structure, but its highest absolute values are shifted northward (Fig. 10b), reflecting the suppression of thermal radiation loss through the higher moisture contents and cloudiness in the ITCZ. Over the mostly dry and cloudless Sahel and Sahara, energy loss through thermal radiation dominates solar energy input, leading to a net radiative cooling. The positive pressure tendencies associated with this diabatic effect appear to balance the dynamic effects (Figs. 9c and 10c), which supports the use of TOA radiation as a proxy for the diabatic term in this region.
Over the Gulf of Guinea ascent associated with convection in the ITCZ leads to strongly positive net psfc tendencies (Figs. 9b and 9c). In this environment, the net radiative heating (Fig. 10c) and latent heat release (not shown) contribute to balancing the dynamic effects. In a narrow strip near 12°N there is net radiative heating and weakly negative net dynamic pressure tendencies (Figs. 9c and 10c). This result may seem inconsistent at first sight but it is most likely associated with the simplifications discussed above, that is, the division by temperature in the diabatic term and the neglect of the latent heat fluxes from the surface. It should be stressed that an exact quantification of all terms is not necessary in order to understand the January 2004 case. The most important thing is to carefully estimate the disturbances to the climatological balance, as we will do in the next subsection.
c. Perturbations during the January 2004 event
Between 15 and 17 January, when falling surface pressure is observed over the Sahel (Figs. 7 and 8), the effects of warm advection and ascent on the eastern side of the first upper disturbance almost cancel each other out and therefore net dynamical psfc tendencies are low over West Africa to the south of 20°N and from 10°W to 10°E (not shown). The TOA net radiation, however, is clearly disturbed by the presence of the TP on these days (Figs. 8a, 8c and 8e), as demonstrated by the example of 16 January 2004 (Fig. 11). In the region with enhanced cloudiness (Fig. 8c), the 24-h accumulated net solar and net thermal radiation are both reduced in magnitude (Figs. 11a and 11b) as compared to climatology (Figs. 10a and 10b). Due to mostly thin and therefore transparent mid- and upper-level clouds and higher column-integrated water vapor contents, the impact on the thermal radiation is larger, leading to a locally enhanced greenhouse effect (Fig. 11c). Since there is no indication of a substantial disturbance in any other diabatic contribution to the pressure tendency equation, we conclude that the anomalous radiative warming is the main cause of the pressure fall during this period.
On 18 January when the pronounced extratropical trough from the Mediterranean begins to affect northern Africa (Fig. 8h), the dynamic pressure tendencies begin to increase (Fig. 12). Two distinct regions can be distinguished: (a) To the west of the trough axis there are positive contributions to the psfc tendency due to cold advection (Fig. 12a) and negative contributions due to subsidence (Fig. 12b) summing to a net dynamic tendency of about −4 hPa day−1 (Fig. 12c). Over the subsidence region, the thermal radiation loss (not shown) is enhanced compared to the climatology (see Fig. 10b), which will compensate part of the negative dynamic psfc tendencies. (b) To the southeast of the trough axis there are strongly positive contributions related to rising motions (Fig. 12b) and somewhat weaker negative values related to warm advection (Fig. 12a), resulting in net dynamic pressure tendencies of up to 8 hPa day−1 near 21°N and 6°E (Fig. 12c). As mentioned in section 3c, weak rainfall was observed in this region, which has presumably supported the ascent. The diabatic term in this region will have contributions from radiation, latent heating, and evaporational cooling. While the former two generate warming in the upper troposphere, the latter causes cooling at lower levels. Vertically integrated, the latter two terms will compensate each other to a fairly large degree, since only small amounts of precipitation actually reach the ground. The fact that rising pressure was analyzed over this region between 17 and 18 January (Figs. 8f and 8h) suggests that the diabatic warming by radiation and latent heat release is indeed not able to compensate the dynamical effects and the evaporational cooling. In all other parts of the Sahel and the adjacent Sahara, the pressure was still falling during this time.
On 19 January, the diagonal trough penetrates even farther into northern Africa and the TP shifts to the east (Figs. 8i and 8j). Changes in psfc tendency over West Africa become dominated by the dynamic terms that indicate widespread warm advection and ascent on the eastern and southern sides of the trough, and strong subsidence and cold advection on its western side (Figs. 13a and 13b). The two terms do not cancel out entirely, but create a band of negative pressure tendencies from the West African coast across southern Algeria to the Libyan coast (Fig. 13c). These tendencies might be further enhanced by cloud effects in the eastern part of the band (Fig. 8i), which together explain the drop in MSLP on this day (Figs. 8h and 8j). On 20 January, precipitation begins over tropical West Africa and the dynamical pressure tendencies are largest close to the developing cyclone over Algeria (Fig. 8l). This analysis reveals that the diabatic and dynamic effects associated with the two upper-level troughs contribute to the falling psfc over West Africa that allows the inflow of moist air from the Gulf of Guinea and the development of heavy rainfall. Precipitation in the area of the Ahaggar Mountains on 18 January appears to locally counteract the overall decrease in MSLP.
5. Summary and conclusions
We investigated the synoptic evolution and the dynamics of an intense and widespread dry-season precipitation event in tropical West Africa in January 2004 and its relation to forcing from the extratropics. The abundant rainfall had large impacts on human activities in several West African countries. The event was accompanied by intense precipitation over the Algerian and Libyan Sahara and by dust storms over northeastern Africa. Besides analysis, satellite, and synoptic station data, high-resolution observations from a field site in central Benin were used. The analysis of the involved physical mechanisms was based on a special form of the pressure tendency equation that separates the effects of temperature advection, vertical motion, and diabatic processes.
The picture that emerges from this study is schematically depicted in Fig. 14 and has, to the best of our knowledge, not been discussed in the literature so far. The early stages of the event (“tropical plume phase,” 15–18 January) are characterized by the presence of a weak upper-level trough that remains quasi-stationary over the coast of West Africa (Fig. 14a). The enhanced cloudiness and moisture content along its eastern side, sometimes referred to as a tropical plume (TP), locally increases the greenhouse effect over the Sahel and the Sahara, which are in undisturbed conditions subject to net energy loss through radiation. The resulting column-integrated warm anomaly leads to slowly, but continuously, falling surface pressure in this region. The next stage is characterized by the penetration of a more substantial upper-level trough from the western Mediterranean Sea into Algeria that merges with the system over West Africa on 18–19 January (“dynamical phase”; Fig. 14b). Falling pressures are observed over large parts of tropical West Africa during this period. Subsidence to the southwest of the trough and warm advection on its southern and eastern sides lead to widespread negative dynamical pressure tendencies. The TP has moved eastward and the influences from an enhanced greenhouse effect are restricted to this area. Precipitation over the Ahaggar Mountains on 18 January locally counteracts the pressure trend through strong cooling in the ascent region.
At the end of the 4-day period with falling pressure over West Africa, the weak wintertime heat low has shifted northward and intensified. The increased pressure gradient toward the south leads to an inflow of moist air from the Gulf of Guinea, resulting in an anomalous northward excursion of the ITF, the border between the dry Saharan and the moist monsoon air (Fig. 14b). High low-level moisture contents and strong daytime heating over land on 20 January lead to the development of deep moist convection that spreads northward through lifting by convective outflow boundaries. Storm totals exceed 100 mm at some places, which is on the order of 10–20 times the long-term January average. Over northern Africa the upper-level disturbance produces intense precipitation over the Algerian and Libyan Sahara (Fig. 14b). It is one of the interesting outcomes of this work that the almost synchronous rainfalls to the north and south of the central Sahara are dynamically related. The upper disturbance also triggers a surface low that later causes dust storms over northeastern Africa and the Middle East.
There are factors of potential importance that have not been addressed in this analysis. One is the influence of coherent tropical modes (Wheeler and Kiladis 1999) on the development of the convection over West Africa. For instance, a multivariate index of the Madden–Julian oscillation (MJO) developed by Wheeler and Hendon (2004) shows an active MJO signal passing over the African tropics during the period under study (information online at http://www.bom.gov.au/bmrc/clfor/cfstaff/matw/maproom/RMM/). Even though the MJO signal is generally weaker over Africa than over other parts of the tropics, indications for an almost global influence have been found in the recent literature (e.g., Donald et al. 2006). In general, however, it is rather difficult to pin down the influence of an intraseasonal variability pattern like the MJO on a single localized meteorological event. Another factor not considered in this study pertains to the convective dynamics over West Africa. In contrast to summertime convection, the January event occurred in an environment of upper-level anticyclonic shear along the southern side of the STJ. The associated small absolute vorticity leads to low inertial stability near the tropopause, a situation that reduces energy loss due to subsidence and therefore favors the intensification of convection (Blanchard et al. 1998; Mecikalski and Tripoli 1998; Knippertz 2005). A thorough investigation of this effect would require information on convective momentum transports and vertical stability in the outflow region that cannot be derived from standard operational analysis products (Mecikalski and Tripoli 2003).
For the warm season, AEWs and their interactions with the diagonal upper troughs in the westerlies can play an important role in the organization of precipitation (Flohn 1975; Nicholson 1981; Knippertz et al. 2003). However, for January 2004 we could not find indications of westward-propagating coherent disturbances in the lower-tropospheric streamlines, which is consistent with the absence of a distinct baroclinic zone over tropical West Africa during this time of the year. The surface pressure maps between 15 and 20 January 2004 (Fig. 8) are somewhat reminiscent of a typical trajectory of a Soudano-Sahelian depression moving from tropical West Africa on an anticyclonic path to the Mediterranean Sea (Dubief and Queney 1935). The presented results, however, suggest that the causes of cyclogenesis along this track may vary. Further research on the dynamics of Soudano-Sahelian depressions and their relation with diagonal troughs, TPs, and AEWs is needed to clarify this.
The present dynamical analysis unveiled a close relationship between tropical rainfall and synoptic-scale processes in the extratropics. The latter are usually much better forecasted by state-of-the-art numerical weather prediction models than is mesoscale tropical convection, suggesting a relatively high degree of predictability of the event under study. In fact the operational 78-h precipitation forecast from the ECMWF starting at 0000 UTC 19 January 2004 gives clear evidence of a northward excursion of the major rain zone into West Africa up to about 12°N (Fig. 15). However, absolute values in the Guineo-Soudanian zone reach only ∼10 mm, which is presumably due to deficiencies in the convective dynamics at a resolution of T511 (∼40 km). Nevertheless, the fact that the model does give indications of rainfall more than a day in advance provides the potential for taking actions to mitigate the impacts for the local societies. In this respect, our study contributes to the aims of the World Weather Research Programme’s The Observing System Research and Predictability Experiment (THORPEX; information online at http://www.wmo.ch/thorpex/).
Further research is needed to achieve a more solid understanding of the mechanism of extratropically triggered dry-season precipitation in West Africa. This shall include rainfall events in the pre- (April–mid-June) and postmonsoon (October–November) seasons. In principle, two different approaches appear promising. First, more case studies and ultimately a climatological investigation of cool- and transition-season precipitation events will help to identify the typical ingredients that can then be compared to the ones found in this paper. Special attention should be paid to the factors that modify the location and strength of the heat low and, thereby, affect low-level moisture fields over West Africa. Second, the use of numerical models would allow a control over single-influence factors for sensitivity and predictability studies. Output from such experiments or reruns with the ECMWF forecast model could be used to study the diabatic contributions, which were estimated from TOA radiation budgets alone in the present work, more quantitatively. In either approach, a use of the pressure tendency equation employed herein promises deeper insights into the dynamics of the heat low. Finally, one could also examine whether the proposed mechanism plays a role for other monsoon regions such as Australia, where TPs are frequently observed, too (e.g., Kuhnel 1989).
The authors acknowledge funding under the Emmy Noether program of the German Science Foundation (DFG; Grant KN 581/2–3) and under the IMPETUS Project (BMBF Grant 01LW06001A, North Rhine-Westphalia Grant 313-21200200). We are especially indebted to Simone Kotthaus, Volker Ermert, Susan Pohle, and Sonja Eikenberg for their help in data analysis and visualization. Simone Giertz provided the data from the Mont de Gaulle site, and Malte Diederich the IMPETUS MultiSensor Algorithm rainfall data. We thank the Ghana Meteorological Agency and its former Deputy Director General George Wilson; Francis Didé, director of the DMN of Benin; IRD; and Ernest Afiesimama from the Nigerian Meteorological Service for providing rainfall data. We are grateful to the helpful comments of Sharon Nicholson and one anonymous reviewer.
Derivation of the Pressure Tendency Equation
The local tendency of the air density ρ in Cartesian–pressure coordinates (x, y, p) is related to the same quantity using the geometric height, z, as a vertical coordinate as follows:
Substituting the hydrostatic equation ∂p/∂z = −gρ, where g denotes the gravitational acceleration, into term one on the rhs and expanding term two, we get
Now substituting the hydrostatic equation into term two and combining the terms yields
Dividing by −ρ/g and expressing the local density tendency in terms of total tendency and advection gives an expression for the vertical derivative of pressure tendency:
After further modification, this equation will finally be integrated vertically to give the tendency equation for surface pressure.
Using the ideal gas law for moist air ρ = p/RTυ, where R is the gas constant for dry air and Tυ the virtual temperature, the first two terms on the rhs of (A4) become
To modify the last term on the rhs of (A4), we use the entropy form of the first law of thermodynamics:
where cp is the specific heat capacity at constant pressure and Q̇ is the diabatic heating rate (W kg−1). Substituting the ideal gas law in for T yields
Rearranging terms and using cp − R = cυ, we get
Integrating this equation with respect to z from the surface to the TOA, where we assume pressure tendencies to vanish, yields the tendency equation for the surface pressure, psfc, used in section 4:
Corresponding author address: Peter Knippertz, Institut für Physik der Atmosphäre, Johannes Gutenberg-Universität Mainz, D-55099 Mainz, Germany. Email: firstname.lastname@example.org
IMPETUS is a German project that translates into “An integrated approach to the efficient management of scarce water resources in West Africa.”
Also termed the intertropical discontinuity or monsoon trough.