The impact of deep convection on the West African summer monsoon climate: a regional climate model sensitivity study


  • M. B. Sylla,

    Corresponding author
    1. Earth System Physics (ESP) Section, International Centre for Theoretical Physics (ICTP), Trieste, Italy
    • International Centre for Theoretical Physics (ICTP), Earth System physics Section (ESP), Strada Costiera 11, PO Box 586, I-34151 Trieste, Italy.
    Search for more papers by this author
  • F. Giorgi,

    1. Earth System Physics (ESP) Section, International Centre for Theoretical Physics (ICTP), Trieste, Italy
    Search for more papers by this author
  • P. M. Ruti,

    1. UTMEA-CLIM, Italian National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA), Casaccia Center, Rome, Italy
    Search for more papers by this author
  • S. Calmanti,

    1. UTMEA-CLIM, Italian National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA), Casaccia Center, Rome, Italy
    Search for more papers by this author
  • A. Dell'Aquila

    1. UTMEA-CLIM, Italian National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA), Casaccia Center, Rome, Italy
    Search for more papers by this author


The role of the representation of deep convection on key elements of the West African summer monsoon climate is addressed using the Regional Climate Model RegCM3. Two simulations in which a scheme of deep convection is activated and then turned off are performed and intercompared. Results show that the presence of deep convective heating along the intertropical convergence zone sustains increased lower-level baroclinicity favoring intensification of the jet core and leading to a more realistic African easterly jet. In addition, although the isentropic potential vorticity (IPV) is lower when the convection scheme is switched off, African easterly waves (AEWs) are still generated and propagate westwards but they dissipate around the west coast. Substantial differences between the two simulations occur mainly at the 6- to 9-day time-scale over land, when much weaker activity is simulated in the absence of convection. This indicates that orographic friction and low-level large-scale moisture convergence, generating high values of latent heat and IPV, may play the dominant role in the genesis and growth of AEWs and that deep convection acts to strengthen the overall wave activity and to favor their west coast development. Copyright © 2011 Royal Meteorological Society

1. Introduction

The main structure of the West African monsoon (WAM) circulation is critically affected by atmospheric deep convection, from the scale of a single convective cell up to the scale of synoptic systems (Redelsperger et al., 2002). One of the most prominent features during the boreal summer season is the mid-tropospheric (600–700 mb) African easterly jet (AEJ), a region of strong zonal winds (up to ∼10 m s−1) extending from East to West Africa and crossing the Atlantic Ocean. The disturbances around this zonal circulation pattern, so-called African easterly waves (AEWs), have been identified as the main mechanism in organizing rainfall patterns (Diedhiou et al., 1999) and therefore as key drivers of climate variability in this region. It is difficult to isolate the role of deep convection in maintaining the jet from its role in the propagation of AEWs (Leroux and Hall, 2009). However, to gain insight into the uncertainties inherent in the modeling of climate variability over this region (Ruti and Dell'Aquila, 2010), we test different settings of the deep convection scheme embedded in a realistic regional climate model and seek to provide useful information for future modeling efforts.

Burpee (1972) first established a relation between the observed AEJ, the existence of a surface baroclinic zone and the reversal of the temperature gradient in the mid troposphere. Later, Thorncroft and Blackburn (1999) identified the jet as the atmospheric response to diabatic heating associated with dry convection over the Sahara and moist convection over the intertropical convergence zone (ITCZ). Cook (1999) pointed out the importance of baroclinicity induced by the low-level meridional temperature and soil moisture gradients between Equatorial Africa and the Sahara during the summer season for the formation of the jet. The key role played by surface gradients has also been highlighted by Steiner etal. (2009).

The fluctuations of the position and strength of the AEJ induce rainfall variability at the intraseasonal and interannual time-scales in regions affected by the WAM. For example, a more equatorward position of the AEJ corresponds to dry conditions over the Sahel (Jenkins et al., 2005; Nicholson, 2008). More recently, Sylla et al. (2010a) demonstrated that the north–south displacement of the AEJ core during the monsoon season strongly affects the intraseasonal fluctuations in the position of the monsoon rainbelt. This indicates a strong interaction between the jet and the mesoscale convective systems embedded in the AEWs responsible for most of the rain over West Africa (e.g. Gaye et al., 2005; Mohr and Thorncroft, 2006).

AEWs are maintained by combined baroclinic and barotropic conversions around the mid-tropospheric AEJ (Hsieh and Cook, 2007). The early consensus for the generation and maintenance of AEWs focused mainly on the necessary conditions for instability introduced by Charney and Stern (1962), whereby a zonal flow is unstable in regions of reversed meridional gradient of the isentropic potential vorticity (Schubert et al., 1991; Thorncroft and Hoskins, 1994a). The interpretation of AEWs as the result of spontaneous growth of random fluctuations on an unstable zonal basic state has been recently challenged by Thorncroft et al. (2008) and by Leroux and Hall (2009), who propose that AEWs are triggered by localized finite-amplitude perturbations associated with latent heating upstream of the region of growth. Hence the key role of deep convection over the Darfur and Ethiopian highlands in initiating the AEWs and favoring its subsequent development is a consolidated result (Berry and Thorncroft, 2005; Mekonnen et al., 2006). In this framework, it is sufficient to consider the region of reversed sign of the meridional isentropic potential vorticity gradients as favorable for the propagation of AEWs through barotropic and baroclinic instabilities (Thorncroft and Hoskins, 1994a, 1994b; Hsieh and Cook 2008; Diedhiou et al., 2010).

However, the relative role of convection and AEJ in sustaining the AEWs still remains an open issue. Hsieh and Cook (2005) suggested that the AEJ is of lesser importance than convection associated with the ITCZ as a cause of wave activity. Conversely, Leroux and Hall (2009) showed that among the necessary conditions for the development of AEWs is a strong AEJ. These highlight the difficulty of establishing a clear relationship between convection, the AEJ and the characteristics of AEWs.

In this framework, we investigate the relationship between deep convection, AEJ and AEW activity using the International Centre for Theoretical Physics (ICTP) Regional Climate Model (RegCM3; Pal et al., 2007). Reg CM3 has different options to describe deep convection and it has shown a generally good performance over West Africa in simulating mean climatologies, intraseasonal and interannual variability of rainfall and related synoptic circulation features, including the AEJ, AEWs, and their interactions (Sylla et al., 2009, 2010a, 2010b, 2010c). It is thus a suitable tool to investigate the role of deep convection in the establishment of the dynamical features triggering and maintaining the WAM precipitation.

We here carry out and intercompare two model simulations in which the representation of deep convection is present and then switched off, respectively. We focus our analysis, in particular, on the dynamical response of the model to the representation of deep convection and on how this is reflected in the simulation of the AEJ, AEW activity and monsoon precipitation. Model and experiment design are first described in section 2, while the results are analyzed in section 3 and final considerations are given in section 4.

2. Model description and experiment design

We use the version of the Regional Climate Model Reg CM3 described by Giorgi et al. (1993a, 1993b) and Pal et al. (2007). For a detailed description of the model's characteristics and performance over West Africa, the reader is refereed to Sylla et al. (2010a).

Of particular interest in this study is the representation of deep convection. For this purpose, RegCM3 includes three options: the Kuo-type scheme of Anthes et al. (1977) in the simplified implementation described by Grell etal. (1994), the scheme by Emanuel (1991) and the scheme by Grell et al. (1994). Previous experiments have shown that the Emanuel scheme together with the BATS land surface scheme tends to overestimate monsoon precipitation (Steiner et al., 2009). Similarly, preliminary experiments showed that the Kuo-type scheme produces an unrealistically low amount of precipitation (not shown here for brevity). As a result, for this study we employed the scheme of Grell et al. (1994) using the closure assumption based on Fritsch and Chapell (1980). This choice was based on an in-depth analysis of the model performance in multidecadal simulations of present day climate (Sylla et al., 2010a, 2010b).

In the Grell scheme, convection is activated by deep conditional instability due to diabatic heating and moisture convergence between cloud base and cloud top (deep moist convection). Updrafts and downdrafts originate at the levels of maximum and minimum moist static energy, respectively. Mixing between clouds and environment exists only at the top and the bottom of the cloud layer, and the Fritsch and Chapell (1980) closure assumption assumes that convection removes the available buoyant energy over a time interval T set to 30 min. A critical parameter in the Grell scheme is the convective heating H generated in the layer comprised between the levels of maximum and minimum moist static energy (once convection is triggered). In this scheme, convection occurs when the convective heating is within a selected range: Hmin< H < Hmax and convective rainfall is scaled from the updraft mass flux. The Grell convection scheme can thus be switched off in the model by simply adjusting the convective heating thresholds to values not reachable in the model.

Large-scale (grid-resolvable scale) precipitation processes are treated using the sub-grid explicit moisture scheme (SUBEX) of Pal et al. (2000). This scheme assumes that rain-producing clouds are generated after a grid point relative humidity threshold is reached (in our study this threshold is 80%); therefore SUBEX is capable of describing fractional cloud cover formation. Rain is produced by autoconversion and accretion processes and raindrops can evaporate when falling below the cloud base. It should be emphasized that the convection and resolved precipitation schemes do interact with each other in the sense that excess moisture that is not removed by one scheme needs to be eventually removed by the other. The main difference between the two schemes is that while the Grell convection scheme redistributes heat and moisture through parameterized updraft and downdraft fluxes, the SUBEX scheme does so only via large-scale motions.

Two 5-year simulations are carried out over the domain shown in Figure 1, which covers the entire African continent north of about 10°S with a horizontal grid spacing of 50 km: in the control simulation RegCM-GC (Grell convection) a tested setup of the Grell scheme is adopted, as used in the study of Sylla etal. (2010a), where the convective heating thresholds are set to Hmin = −250 K hPa−1 s−1 and Hmax = 500 K hPa−1 s−1. In the second simulation (Reg CM-NC, no convection) the Grell scheme is switched off by setting Hmin and Hmax to very low values so that deep convection is never triggered. Note that by switching off the convection scheme the model accumulates moist static energy in the lowest model level with no dissipation in the vertical. The main focus of the analysis is thus to investigate the influence of the representation of convection on the genesis and development of the AEJ and AEWs and the resulting effect on monsoon precipitation.

Figure 1.

Model domain and orography.

3. Results

Before examining the background climatology in which the AEJ and the AEWs evolve, it is useful to investigate the effect of the suppression of convection on key elements of the West African thermodynamics and rainfall.

3.1. Thermodynamics and rainfall

Figure 2(a)–(f) compares the summer mean rainfall from GPCP and TRMM observations with the RegCM-GC (total, convective and large-scale) and RegCM-NC simulations, respectively. Observations (both GPCP and TRMM) place the ITCZ in a zonal and tilted band between 8°N and 13°N, with rainfall decreasing south and north of it. Precipitation maxima are located in orographic regions such as the Guinea highlands, Jos plateau, Cameroun mountains and East Africa highlands. RegCM-GC (Figure 2(c)) reproduces well this spatial distribution, in particular the ITCZ position, the northern and southern gradients of rainfall and the location of maxima over orographic regions. The main model error compared to the observed datasets is an overestimation of precipitation over the Ethiopia highlands and an underestimate over the Congo Basin and the Nigeria/Cameroon border, both features being consistent with the results of Sylla etal. (2010a).

Figure 2.

Seasonal (JJA) mean rainfall (mm per day, 2001–2005) of (a) Total rainfall from GPCP, (b) total rainfall from TRMM, (c) total rainfall from RegCM-GC, (d) convective rainfall from RegCM-GC (e) large-scale rainfall from RegCM-GC and (f) total rainfall from RegCM-NC.

Note that in RegCM-GC the total rainfall is the sum of convective and large-scale precipitation, which are shown in Figure 2(d) and (e) respectively. Convective type precipitation dominates over most of the West Africa monsoon region, while non-convective precipitation plays a primary role particularly over the Ethiopian Highland precipitation maximum, and to a lesser extent the other topographic complexes. This indicates that in the control simulation most of the rainfall over West Africa originates from deep moist convection associated with deep conditional instability, while large-scale precipitation is mostly associated with topographic forcing.

In the RegCM-NC simulation, rainfall is only from large-scale, non-convective clouds. This model configuration exhibits a narrower ITCZ and monsoon rainfall band than observed (and than in RegCM-GC), with a substantially drier Guinea coast region. Also, the RegCM-NC configuration produces much stronger orographic maxima than observed. These are primarily due to the occurrence of very intense precipitation episodes, or ‘numerical point storms’. Previous studies (e.g. Giorgi, 1991) showed that, for warm climates in the absence of convection, the heat released by condensation of water vapor generates strong updrafts at the grid point level, which in turn enhance moisture convergence, condensation and rain in a positive feedback loop that produces unrealistically intense grid point precipitation events. This effect was particularly pronounced over topographic peaks, as also found in our RegCM-NC simulation. Conversely, when the convective scheme is activated the vertical redistribution of heat and moisture tends to remove buoyant energy and suppress convection, thereby acting as a limiting factor for intense numerical rain events.

Figure 3(a) and (b) shows the vertical cross-section of the mean vertical pressure velocity averaged between 10°E and 20°W for the RegCM-GC and RegCM-NC, respectively. As in Nicholson (2008) and Sylla et al. (2010a), RegCM-GC exhibits a low-level vertical motion core at around 5°N which is related to friction occurring at the interface between the ocean and land surface and it is connected with the mid-tropospheric ITCZ strong ascent centered at 10°N and the Saharan Thermal Low uplift centered at 19°N. In the RegCM-NC simulation, the center of motion associated with coastal friction is separated from the ITCZ ascent, which is displaced northward to 12°N and it is merged with a Saharan Thermal Low shifted southward around 15°N. This structure causes the disappearance of the descending branch of the low-level cell at 12°N. Also, the ascents are stronger in RegCM-NC than in RegCM-GC because of the condensation heating feedback process described above. Overall, consistently with Figure 2, the area of deep ascent is narrower in RegCM-NC than in RegCM-GC.

Figure 3.

Seasonal (JJA) mean omega cross-section (10–3 hPa s−1, 2001–2005) from (a) RegCM-GC and (b) RegCM-NC.

Figure 4(a) and (b) shows the horizontal wind divergence and superimposed low-level horizontal wind vectors at 850 hPa for RegCM-GC and RegCM-NC, respectively. Although RegCM-GC displays a stronger monsoon flow entering West Africa from the Gulf of Guinea, the wind vectors are weaker compared to RegCM-NC between 8°N and 15°N over land. As a result, low-level mass convergence is higher in the case of RegCM-NC over the northern part of the Guinea Coast, along the West African ITCZ and over western Sudan. Therefore, in general, the presence of deep convection weakens the low-level convergence over continental West Africa. This is consistent with the stronger ascent found in the case of RegCM-NC. In fact, due to the absence of deep convection, the vertical motion associated with the West African ITCZ is only driven by the low-level mass convergence in an intense and concentrated uplift region.

Figure 4.

Seasonal (JJA) mean low-level (850 hPa, 2001–2005) divergence and superimposed wind vectors (m s−1, 2001–2005) from (a) RegCM-GC and (b) RegCM-NC.

Figure 5(a)–(c) compares the RegCM-GC and RegCM-NC outgoing long-wave radiation (OLR) with corresponding NOAA observations (Liebmann and Smith, 1996). The OLR follows the distribution of cloudiness, with lower values in the presence of high cloudiness, and thus its pattern essentially follows that of precipitation, with minima along the ITCZ and over the mountainous area of Guinea, Jos, Cameroun, Central and East Africa. South and north of the ITCZ the OLR increases and reaches values in excess of 260 W m−2 around 3°N and 15°N. RegCM-GC captures most of these features. Both simulations generally reproduce the observed patterns; however, they tend to overestimate the OLR over some regions of Central Africa and to underestimate it over the Ethiopian Highland maximum. This latter feature is evidently related to the precipitation overestimate there. The main differences between the two model experiments occur over West Africa, where RegCM-GC produces lower OLR associated with greater convection and cloudiness. Over East Africa the two simulations show similar results, indicating that non-convective clouds play the dominant role there.

Figure 5.

Seasonal (JJA) mean of top of the atmosphere outgoing long-wave radiation (W m−2, 2001–2005) from (a) NOAA, (b) RegCM-GC and (c) RegCM-NC. The lower OLR corresponds to more cloud cover. OLR values above 260 W m−2 are not shaded.

3.2. The African easterly jet

In this section we focus on the impact of suppressed convection on the mean position and strength of the AEJ through the simulation of baroclinicity, meridional temperature gradients and soil moisture. Since convection affects the redistribution of heat, it is likely to affect the structure of the AEJ.

Figure 6(a) and (b) shows the latitude–height crosssections of meridional temperature gradient and superimposed potential temperature averaged between 10°E and 20°W for RegCM-GC and RegCM-NC, respectively. Both simulations show a similar increase of potential temperature with height and a maximum near the surface at 20–25°N. Note that the level of the AEJ (650 hPa) occurs in correspondence of the isentrope 315 K. The isentropes exhibit a large deflection toward the ground from 10°N, suggesting that baroclinicity increases north of that latitude, consistently with the findings of Hsieh and Cook (2005).

Figure 6.

Seasonal (JJA) mean meridional temperature gradient (in shaded,°C km−1, 2001–2005) and superimposed isentropes (in contours, K) cross-sections averaged between 10°E and 10°W from (a) RegCM-GC and (b) RegCM-NC.

The existence of the higher low-level baroclinicity is also noticed in the vertical profile of the meridional temperature gradient. RegCM-GC exhibits a positive meridional temperature gradient from the surface to the mid troposphere (650 hPa) centered at 17°N and a negative gradient over the Sahara extending from the surface to the upper tropospheric levels. RegCM-NC shows the same profile of the spatial distribution of the temperature gradient; however, the positive low-level meridional gradient centered at 17°N is weaker than in RegCM-GC. In addition, a reversed sign of the gradient occurs in RegCM-GC just above the positive surface maximum around 17°N and 550 hPa, which is much weaker in RegCM-NC. Another variable affecting the AEJ is the soil moisture distribution. The seasonal mean soil moisture content of the root zone layer (top 1–2 m) is shown in Figure 7(a) and (b) for RegCM-GC and RegCM-NC, respectively. In RegCM-GC, following the rainfall spatial distribution, the highest soil moisture content is found in the ITCZ and over complex terrain features, while the lowest amounts are located north of 15°N. The RegCM-NC case shows a similar spatial distribution of soil moisture, but lower values. This is related to the sharper definition of the ITCZ along with the larger local amounts which favor moisture loss due to surface runoff. Correspondingly, surface evaporation is greater in RegCM-GC than in RegCM-NC over the monsoon rain regions of West and East Africa (not shown for brevity). It is thus evident that the RegCM-GC case exhibits greater temperature and soil moisture gradients. These differences in the meridional temperature and soil moisture gradients should strongly impact the simulation of the AEJ (Cook, 1999; Hsieh and Cook, 2005; Steiner et al., 2009).

Figure 7.

Seasonal (JJA) mean soil moisture (mm per day, 2001–2005) from (a) RegCM-GC and (b) RegCM-NC.

To characterize this impact, a latitude–height cross-section of zonal wind during June–July–August (JJA) is shown in Figure 8(a) and (b) for both RegCM-GC and RegCM-NC. As documented by Sylla et al. (2010a), the simulation displays (1) the monsoon flow and harmattan winds below 800 hPa at 5–20°N and 25–30°N respectively; (2) the AEJ in the mid-levels centered at 15°N; and (3) the tropical easterly jet (TEJ) in the upper tropospheric levels at 200 hPa centered at 10°N. In RegCM-NC, the monsoon flow is stronger (as already mentioned above) and the TEJ slightly weaker than in RegCM-GC, probably as a result of lower surface temperature gradients. Of particular interest is the core of the AEJ in the mid-tropospheric levels, which is greatly weakened (by up to 4 m s−1) in the absence of convection. This is more apparent in Figure 9(a) and (b) showing the JJA averaged zonal wind at 650 hPa for both the RegCM-GC and RegCM-NC cases. RegCM-GC places a well-defined AEJ core (wind speeds greater than 10 m s−1) at 15°N over West Africa, with the easterlies stretching towards the east up to the Ethiopia highlands. In RegCM-NC the eastward stretching of the easterlies is less pronounced, the core of the AEJ over West Africa is weaker and less defined and two maxima of easterlies cover only a small area in the coastal regions around 15°N and off the Guinea coast.

Figure 8.

Seasonal (JJA) mean zonal wind (m s−1, 2001–2005) cross-section averaged between 10°E and 10°W from (a) RegCM-GC and (b) RegCM-NC.

Figure 9.

Seasonal (JJA) mean zonal wind at 650 hPa (m s−1, 2001–2005) from (a) RegCM-GC and (b) RegCM-NC.

Therefore, in summary, RegCM-GC shows stronger low-level baroclinicity and temperature and soil moisture gradients and therefore a much better-defined and more realistic AEJ than RegCM-NC (Cook, 1999; Thorncroft and Blackburn, 1999; Steiner et al., 2009), indicating that the absence of convective heating leads to an unrealistically weak AEJ core.

3.3. The African easterly waves

Since convection induces instabilities and fosters the genesis of AEWs, in this section we turn our attention to the impact of suppressed convection on the propagation of AEWs. A useful indicator of the conditions most favorable for supporting the propagation of the disturbances is the isentropic potential vorticity (IPV).

The Ertel IPV is computed as given in Molinari et al. (1997) and Hsieh and Cook (2005):

equation image(1)

where g is the acceleration due to gravity, θ is the potential temperature, ζθ is the vertical component of the relative vorticity computed on isentropic surfaces, p is the pressure, f = 2 Ωsinφ is the Coriolis parameter and equation image is the inverse of the mass density. Equation (1) indicates that the IPV anomaly is associated with both an absolute vorticity anomaly (related to barotropic instability) and with a mass density anomaly (related to baroclinic instability).

Figure 10 shows the IPV on the 315 K isosurface of the RegCM-GC and RegCM-NC experiments (Figure 10(a) and (b), respectively), along with their corresponding meridional gradients (Figure 10(c) and (d)). In RegCM-GC, a zonal strip of high IPV values extends from East to West Africa and into the Atlantic Ocean between 8°N and 17°N, with maxima located on the western side of the Ethiopia highlands and along the West African ITCZ. This strip is similar to the one found by Berry and Thorncroft (2005). A similar IPV strip is found in RegCM-NC. The maximum over the East Africa highlands has similar magnitudes in both cases but values over the West Africa ITCZ are generally lower in RegCM-NC. In addition, the locations of these IPV maxima closely follow those of low OLR, indicating a tight connection with latent heating. Note that, although the deep convection scheme is switched off in RegCM-NC, the IPV in the ITCZ is still present but with lower values compared to RegCM-GC. This suggests that the IPV over the Central and East African highlands results from orographic friction and low-level convergence, while the maxima along the ITCZ are driven by low-level convergence only. In RegCM-GC, these processes are further strengthened by deep convective heating, especially along the West African ITCZ.

Figure 10.

Seasonal mean (JJA) isentropic potential vorticity (in PVU, 2001–2005) on the 315 K isosurface from (a) RegCM-GC and (b) RegCM-NC, and seasonal (JJA) mean meridional IPV gradient (PVU km−1) on the isosurface 315 K from (c) RegCM-GC and (d) RegCM-NC.

The meridional gradients of IPV for RegCM-GC and RegCM-NC are shown in Figure 10(c) and (d). In both cases, the most prominent feature is the existence of a large zonal band of negative IPV gradients from 12°N to 20°N along the axis of the AEJ from East to West Africa. This gradient is somehow similar and narrower in RegCM-NC than in RegCM-GC over East Africa. These negative gradients appear after a reversal in the sign of the meridional IPV gradient at 10°N. This reversal, occurring along the axis of strongest easterlies (AEJ) and the region of high potential temperature gradient, suggests that both the Charney–Stern and Fjortoff criteria for instability are satisfied. Therefore, the region roughly lying between 10°N and 20°N is barotropically and baroclinically unstable in both cases and can support AEW growth.

Note that although the meridional IPV gradient is similar in both experiments (because of the high resolution of the regional model), the mean IPV values are larger (indicating more instability) in RegCM-GC than in RegCM-NC. In a linear-instability interpretation of the propagation of AEWs, the weaker IPV in RegCM-NC would imply an environment that is less favorable for the growth of perturbations. We would then expect RegCM-NC to produce AEWs with different characteristics from RegCM-GC in terms of their overall structure, amplitude, frequency and intermittency.

In order to test this hypothesis we analyze the propagation of wave disturbances by adopting two different bandpass filters to separate the 3- to 5-day and the 6- to 9-day wave regimes as observed by Diedhiou et al. (1998, 1999) and Hsieh and Cook (2005).

The ERA-Interim reanalysis (which recall forces the regional model at the lateral boundaries) shows relatively high 3- to 5-day wave activity over West Africa slightly north of the ITCZ and along the west coast between 5°N and 25°N, although some weak activity is also present around the Gulf of Guinea and southern Sudan (Figure 11(a)). The NCEP reanalysis displays approximately the same distribution but the wave activity is weaker over the west coast and substantially stronger over Central–East Africa (Figure 11(b)). Theses differences between reanalysis products in terms of the representation of AEW activity have been identified and analyzed in detail by Ruti and Dell'Aquila (2010).

Figure 11.

Seasonal (JJA) mean variance of the 3- to 5-day bandpass-filtered meridional wind from (a) ERA-Interim reanalysis, (b) NCEP reanalysis, (c) RegCM-GC and (d) RegCM-NC and at 650 hPa during 2001–2005.

In both RegCM simulations (Figure 11(c) and (d)), strong 3- to 5-day wave activity is found between 10°N and 20°N in the barotropic and baroclinic unstable zone. This wave activity is slightly shifted northward in Central–East Africa compared to the reanalyses. The regional model overestimates the maxima along the ITCZ and somewhat underestimates them near the Gulf of Guinea. The wave activity distribution follows closely the reversal of 315 K IPV gradients discussed above and is similar to the climatology shown by Mekonnen et al. (2006) and Ruti and Dell'Aquila (2010), both based on the ERA40 reanalysis. The larger values found in the model compared to the ERA-Interim reanalysis (Sylla et al., 2010a) may also be due to the higher model resolution. In fact, although some precursors of AEWs may enter the domain through the boundary forcing, this model overestimation of frequency and strength of waves compared to the ERA-interim field indicates that they mainly originate internally within the regional model simulations. Comparing the two simulations, both models display similar AEW activity maxima in central–eastern Sudan and along the West African ITCZ. This implies that low-level moisture convergence matters more than deep convection for sustaining the 3- to 5-day AEWs. Note that the core of the AEJ was stronger in RegCM-GC. Therefore, in the absence of convection, the AEWs might have grown at the expense of the AEJ through combined baroclinic and barotropic conversions (Diedhiou etal., 1999; Thorncroft and Blackburn, 1999). RegCM-NC shows markedly lower wave activity into the Atlantic Ocean, whereas RegCM-GC displays some strong activity centers around 20°N (as also found in the reanalyses fields), suggesting that deep convection is critical for the high-frequency variability over the west coast (Berry and Thorncroft, 2005; Mekonnen et al., 2006).

Analogous to Figure 11(a)–(d), in Figure 12(a)–(d) we report the variance of the 6- to 9-day bandpass-filtered meridional wind at 650 hPa from respectively ERA-Interim, NCEP and both RegCM simulations. The reanalyses (Figure 12(a) and (b)) show the location of higher activities off the coast of regions from Senegal to northern Africa extended inland up to the Sudanese complex terrains. Compared to the 3- to 5-day variance, these wave activities are located farther north, in good agreement with Diedhiou et al. (1998) and Hsieh and Cook (2005). Although the pattern is similar in both reanalyses, the magnitudes are substantially different. In fact, NCEP reanalysis shows larger variance in the land, especially along the ITCZ and Sudanese highlands. RegCM-GC (Figure 12(c)) captures the location of the maxima but overestimates markedly the one over the Sudanese highlands, where the 3- to 5-day easterly waves are generated. In addition, it significantly underestimates the activity in the ocean probably because of the proximity of the western boundary of the domain. Nevertheless, the 6- to 9-day wind perturbations are weak south of 5°N and stronger between 10°N and 15°N, consistent with the findings of Hsieh and Cook (2005). It is thus evident that RegCM3 performs well in reproducing the overall pattern of the wave activity but has difficulty in capturing the proper magnitudes probably because of the intermittent character of the 6- to 9-day wave regime.

Figure 12.

Seasonal (JJA) mean variance of the 6- to 9-day bandpass-filtered meridional wind from (a) ERA-Interim reanalysis, (b) NCEP reanalysis, (c) RegCM-GC and (d) RegCM-NC and at 650 hPa during 2001–2005.

The maximum over the complex terrains of Sudan is also found in RegCM-NC but it is much weaker, while the one along the ITCZ is missing (Figure 12(d)). In addition, the 6- to 9-day variance does not extend further west and into the Atlantic Ocean in RegCM-NC. It is worth highlighting that these discrepancies between the two simulations arise just in correspondence of the different distribution of IPV reported in Figure 10(a) and (b), indicating the key role of deep convection in favoring the genesis and propagation of 6- to 9-day synoptic disturbances in the RegCM simulations. Therefore, although the 3- to 5-day wave activity is similar in the two simulations over land, the presence of deep convection along the ITCZ further strengthens the 6- to 9-day wave development, leading to stronger overall activity and enhanced propagation of the waves across the West African coast.

The synoptic properties of the AEWs generated in the two experiments at the two distinct time-scales are illustrated adopting the technique introduced by Diedhiou et al. (1998). First, we consider the 3- to 5-day filtered meridional wind speed averaged over a 1° × 1° box that includes the relative maximum of variance over West Africa shown in Figure 12(c) and (d) (i.e. latitude = 15°N, longitude = 5°W for RegCM-GC; latitude = 12°N, longitude = 2°W for RegCM-NC) and then we display its time series, which can be considered as a reliable index of wave activity (Diedhiou et al., 1998). The two simulations exhibit very similar behavior in terms of frequency and intermittency of AEWs (Figure 13(a) and (c), and the Kolmogorov–Smirnov test performed on the probability density function (PDF) distributions of the selected indicator proves that the two time series have the same continuous distribution at 99% level of statistical significance (not shown).

Figure 13.

Time series of 3- to 5-day (left panels) and 6- to 9-day (right panels) bandpass-filtered 650 hPa meridional wind speed (m/s) averaged over a 1° × 1° box around the relative maximum of variance during 2001–2005 for respectively RegCM-GC (blue line, upper panels) and RegCM-NC (red line, lower panels).

After having compared the 3- to 5-day wave activity index between the two simulations, we now turn our attention to the 6- to 9-day wave regime. Diedhiou etal. (1998) have discussed the northern flank of the 6- to 9-day AEWs over the ocean. However, we find this activity center to be less active in our simulations and probably too close to the model's boundary. Therefore we focus on the southern lobe of activity over Central Africa and consider as reference point the maximum over the Sudanese highlands (i.e. latitude = 12°N, longitude = 25°E for RegCM-GC; latitude = 12°N, longitude = 20°E for RegCM-NC).

As for the 3- to 5-day wave regime, in Figure 13(b) and (d) we show the time series of the filtered fields averaged around its maximum, respectively for RegCM-GC and RegCM-NC. Although the frequency and intermittency are close in the two simulations, the RegCM-GC index displays larger amplitudes. This is confirmed by the PDF distribution where the two time series have the same continuous distribution at the 99% statistical significance level, although RegCM-GC exhibits higher density (consistent with the larger amplitudes) than RegCM-NC for the intense events (not shown).

The spatial pattern of AEWs is best portrayed by showing the correlation maps of the previously introduced AEW index with the filtered 650 hPa meridional wind field. The correlation map (Figure 14(a) and (b)) for both experiments shows a wave-like pattern of alternating negative and positive centers, originating downstream of the Ethiopian highlands and getting amplified over West Africa while moving westward over the Atlantic Ocean along the AEJ track. This result confirms that the maxima in the variance shown in Figure 12(c) and (d) correspond to AEW-like waves. The spatial patterns of the disturbances, as well as their tracks and spatial wavelengths, are comparable for the two simulations, even if in RegCM-NC the genesis of these disturbances seems to be just near the Ethiopian highlands (10°N–30°E), while in RegCM-GC it is more localized over Central Africa (10°N–10°E). It is also interesting to compare the main pathway of the simulated AEWs to the implied precipitation patterns (Figure 2). A visual comparison between RegCM-GC and RegCM-NC shows that, over the continent, the area of intense summer rainfall follows very closely the path of AEWs in both simulations. This tight connection between AEWs and rainfall can also be noticed using a Hovmöller diagram (not shown). However, by taking into account the decomposition between large-scale and convective precipitation reported in Figure 2, we highlight that atmospheric convection spreads rainfall over a wider latitudinal band, whereas in the absence of convection the narrower band of more intense rainfall is concentrated around the core of synoptic disturbances. Furthermore, in the RegCM-NC simulation, the AEW genesis near the Ethiopian highlands is accompanied by more intense rainfall (seasonal average up to 20 mm per day). As discussed previously, these unrealistic intense precipitations events (or ‘numerical point storms’) occurring in the absence of convection are due to the heat released by condensation (e.g. Giorgi, 1991), which may also act as a local heating perturbation triggering and enhancing AEWs (Thorncroft et al., 2008).

Figure 14.

Seasonal (JJA) correlation map for the 3- to 5-day bandpass-filtered meridional wind from (a) RegCM-GC and (b) RegCM-NC and for the 6- to 9-day bandpass-filtered meridional wind from (c) RegCM-GC and (d) RegCM-NC at 650 hPa during 2001–2005.

For the 6- to 9-day wave regime, the correlation maps (Figure 14(c) and (d)) clearly show wave-like patterns in both simulations but with some relevant discrepancies. Specifically, the wave patterns in RegCM-GC are characterized by a broader meridional width, while in RegCM-NC the AEW signal is confined to a narrower band superimposed onto the rainfall region as described above.

In summary, our results suggest that the existence of significant wave activity in RegCM-NC indicates that deep convection is of minor importance for the generation of the 3- to 5-day AEWs; however, it is critical at the 6- to 9-day time-scale. The uplift of air resulting from large-scale low-level moisture convergence and orographic forcing, leading to high values of latent heat and IPV gradients, may be sufficient to trigger AEWs in the genesis region (Thorncroft et al., 2008). The low-level moisture convergence along the ITCZ maintains the 3- to 5-day waves. Deep convection acts to increase the 6- to 9-day wave activity and to favor the overall west coast wave development.

4. Summary and conclusions

In this paper the background state in which the AEJ and AEWs evolve is studied using regional climate model simulations aimed at investigating the importance of the representation of deep convection for the development of the WAM circulation. This is accomplished by comparing two simulations: one with (RegCM-GC) and one without (RegCM-NC) a parameterization for deep convection.

The results suggest that the clouds and rainfall over the Central and East African highlands are primarily driven by orographic uplifting and low-level convergence, while precipitation is mainly maintained by low-level mass convergence over the ITCZ. In the RegCM-GC simulation, these processes are further strengthened by deep convection.

We also found that the lack of convective heating causes weaker meridional soil moisture and temperature gradients and hence a weaker baroclinic zone. As a result, the core of the AEJ is weak (and in fact almost non-existent) in RegCM-NC, although some easterlies stretching from East to West Africa are still present. Consistent with previous studies, when deep convection is activated the core of the AEJ intensifies and it is well set around 15°N at 650 hPa. This indicates that the AEJ core owes its existence to the presence of deep convection leading to a more pronounced baroclinic surface associated with stronger low-level temperature gradients.

An analysis of IPV fields and IPV gradients on the 315 K isosurface shows that the region between 10°N and 20°N can support the propagation of disturbances maintained by barotropic and baroclinic conversion, independent of the presence of deep convection. Despite the differences in IPV fields, the 3- to 5-day easterly wave regime in the two simulations exhibits quite similar activity, frequency and amplitudes over land, while RegCM-NC shows less variance over the ocean. For the 6- to 9-day wave regime, both wave activity and amplitude are stronger in RegCM-GC than in RegCM-NC over the genesis region (Sudanese highlands), along the ITCZ and over the Atlantic Ocean.

We conclude that the representation of deep convection is not necessary for the genesis and growth of the 3- to 5-day wave regime but is critical for their west coast development and for the propagation of the 6- to 9-day easterly wave regime. Orographic friction and low-level moisture convergence handled by the resolvable scale components of the model may be sufficient for the wave genesis and development. The magnitude of the activity of the 6- to 9-day wave regime is strengthened in the presence of deep convective heating, implying that their interactions with convection favors wave development.

Finally, our work confirms that small-scale processes, such as the interaction between complex orography and atmospheric deep convection, are important drivers of intraseasonal atmospheric variability over the tropical African continent. In particular, our simulations describe how deep convection affects the representation of monsoon location and strength, including the simulation of the associated precipitation band. This implies that care needs to be taken to select, test and optimize deep convection schemes over the region for a better representation of the West African monsoon climate towards the use of regional climate model projections for impact assessment studies.


We would like to gratefully thank the associate editor and the two anonymous reviewers for their thoughtful comments and suggestions, which helped to improve this paper.