A key feature of the WAM are African easterly waves (AEWs), which have periods of 3–5 days (Carlson, 1969a; Burpee, 1972; Reed et al., 1977; Diedhiou etal., 1999), with wavelengths near 3000 km and mean speed of approximately 8 m s−1. Cyclonic vortices (wave troughs) propagate westward along two tracks, one south and one north of the African easterly jet (AEJ), displaying a mixed barotropic–baroclinic growth mechanism. Hsieh and Cook (2005) showed that AEW generation may be correlated with convection within the ITCZ. More recently, Thorncroft et al. (2008) suggested that upstream MCSs may provide the initial perturbation into the jet entrance required for AEW generation.
Favourable large-scale dynamical forcing for the generation of rainfall through the organization of MCSs may be provided by AEWs. For example, Fink and Reiner (2003) indicated that 40% of SLs in West Africa were forced by AEWs; and Taleb and Druyan (2003) reported that 30–40% of total rainfall recorded at stations in West Africa from 1953 to 1978 was associated with AEWs. A number of previous studies, based upon satellite and reanalysis data, have reported the effects of AEWs on the modulation of precipitation (Burpee, 1974; Carlson, 1969a; Duvel, 1990; Diedhiou et al., 1999; Mathon etal., 2002; Fink and Reiner, 2003; Gu et al., 2004; Mekonnen et al., 2006), with latitudinal phasing dependence. Debate remains as to the phasing of deep convection in West Africa with AEWs, with deep convection occurring ahead of, within, or behind the AEW trough (Diedhiou et al., 1999; Fink and Reiner, 2003; Kiladis et al., 2006; Laing et al., 2008; Payne and McGarry, 1977). These studies suggest synoptic-scale characteristic modulation, as opposed to mesoscale modulation (i.e. MCS characteristics) discussed in this study.
African MCS lifetimes have been reported as between 2 and 3 hours (Chong et al., 1987) and greater than 2 days (Fink et al., 2006), though it is quite possible that extremely long lifetimes could be a function of convective regeneration. Fortune (1980) and Peters and Tetzlaff (1988) observed that Sahelian SLs can move faster than AEWs with propagation speeds that average between 14 and 17 m s−1 (Aspliden et al., 1976). Fink et al. (2006) found median values largely between 3 and 19 m s−1, dependent upon intensity metrics used. These observed MCS speeds indicate that propagation into or through multiple phases of an AEW synoptic environment is possible, which suggests complex spatial-scale interactions between AEWs and MCSs during the WAM season.
The objective of this study is twofold. First, we compare MCS statistics including precipitation and vertical structure for three distinct locations in West Africa. This study benefits from the African Monsoon Multidisciplinary Analysis (AMMA; Redelsperger et al., 2006; Lebel et al., 2010) and the NASA-AMMA (NAMMA; Zipser et al., 2009) experiments, which allowed access to a great deal of observational data in what is normally a sparsely observed area of the globe. The temporal extent of the study was determined by the common operating period (19 August–16 September 2006, Table I) of the three ground-based radars. Secondly, differences in convective characteristics between MCSs associated with AEWs and those that are not are explored for each radar location.
Table I. Sampling characteristics and locations of TOGA, NPOL, and MIT radars.
TOGA (maritime) Praia, Cape Verde
NPOL (coastal) Kawsara, Senegal
MIT (continental) Niamey, Niger
3 dB beamwidth (°)
Pulse repetition frequency (Hz)
Repeat cycle (min)
Range gate size (m)
Unambiguous range (km)
Pulse width (µs)
Period of operation, 2006
15 Aug–16 Sep
19 Aug–30 Sep
5 Jul–27 Sep
2. Data and methodology
2.1. Radar data and precipitation feature analysis
Data and methodologies largely follow Cifelli et al. (2010), with additions elaborated hereafter. Radar data were the primary datasets for the analysis of convective characteristics. Figure 1 shows the location of each radar site and the approximate maximum unambiguous range about each radar location. The continental (Massachusetts Institute of Technology, MIT) and maritime (Tropical Ocean and Global Atmosphere, TOGA) radar systems were both C-band, single polarimetric, while the coastal radar system (NASA Polarimetric Radar; NPOL) operated at S-band with dual polarimetric capabilities. The NPOL dual polarimetric information was not used in this study. Sampling characteristics, along with the location of each radar system, are listed in Table I. Quality control was performed on each dataset to remove spurious echoes. In addition to internal calibrations performed in the field, calibration comparisons of radar reflectivity against the TRMM precipitation radar (TRMM PR) were performed following methodologies described in Anagnostou et al. (2001) and Lang etal. (2009). Adjustment values are shown in Table II. The attenuation correction algorithm of Patterson et al. (1979) was applied to both C-band radar systems.
Table II. Data and methodology characteristics for TOGA, NPOL, and MIT radars.
Radar reflectivity adjustment via TRMM comparison (dB)
Partial features (%)
Z = 230R1.25
Z = 368R1.24
Z = 364R1.36
Radar polar coordinate data were interpolated to a 2 km vertical and horizontal resolution Cartesian grid using the National Center for Atmospheric Research REORDER software (Mohr et al., 1986). The grid extended 130 km in the x and y directions from the radar location listed in Table I. The spacing chosen was a direct result of different sampling characteristics (Table I), so data interpolated to the Cartesian grid did not exceed maximum spatial resolution of any dataset at the maximum unambiguous range.
Radar reflectivity (Z) volume scans were partitioned into convective and stratiform components using the Steiner et al. (1995) algorithm. This technique uses a convective threshold value to identify convective cores, along with a convective peakedness criterion which evaluates surrounding pixels for convective classification. Rainfall (R) estimates were made from power-based Z (in mm6 m−3)–R (in mm h−1) relationships (Table II). The relations chosen were based upon radar–rain gauge comparisons for the GARP Atlantic Tropical Experiment (GATE) region (Hudlow, 1979) for TOGA; and disdrometer data in Dakar (Nzeukou et al., 2004) for NPOL and Niamey (Sauvageot and Lacaux, 1995) for MIT. The TOGA relationship produced a higher rainfall rate for the same Z value when compared to the corresponding Z–Rs used for NPOL and MIT, consistent with previous Z–R relationships for oceanic regimes. A single Z–R relation was used for both convective and stratiform portions at each location. As with any precipitation estimator, there is inherent error associated with the use of Z–R relations.
Divergence profiles were calculated using the method described by Mapes and Lin (2005), which is a refined derivative of the Browning and Wexler (1968) velocity azimuth display (VAD) method. Briefly, polar-coordinate radar data were processed via a space–time binning algorithm (CYLBIN) to retain range-dependent characteristics. Data were pooled into 50 hPa vertical levels to account for sparse data at upper levels, while 8 km annuli were used in horizontal processing over ranges of 4–92 km. Methodology for attaining mean divergence profiles follows Hopper and Schumacher (2009), in which 40 km annuli (five-range pooling) centered about 28, 44, 60, and 76 km were used.
Ground-based radar observations in this study do not allow examination of the evolution and structural differences of the largest MCSs, due to both geographic position and limited observational domains. The spatial extent of MCSs may be several times larger than the radar scan domain and Hodges and Thorncroft (1997) showed that MCSs are preferentially generated >5° from the most eastward radar location. Despite these limitations, ground-based observations allow detailed analysis of a smaller subset of convective systems. In order to analyze convective characteristics of radar data, precipitation features (PFs) were identified using an objective algorithm described in Cifelli et al. (2007), related to an approach developed for TRMM satellite observations (Nesbitt et al., 2000, 2006). A contiguous echo region that meets a minimum threshold reflectivity (10 dBZ) and size criterion (8 km or 4 pixels in this case) was identified in the lowest grid level (1 km AGL). The algorithm then broke these into three categories: MCS (≥1000 km2, with at least one convective grid point), sub-MCS (<1000 km2, with at least one convective grid point), and no convective (NC; features that did not display required convective criteria regardless of spatial scale). Statistics of associated precipitation, reflectivity, vertical structure, and number of elements were recorded for each feature. In addition, representative thermodynamic variables (discussed later) were retained.
While ground-based radar data provide high spatial and temporal resolution, it is important to note the limitation of the radar scan area. Owing to the high frequency of large convective systems extending beyond the scan range of the radar, it is impossible to completely sample the largest convective systems within the view area of a single radar system (confirmed through infrared (IR) satellite animations; not shown). Buarque and Sauvageot (1997), using radar and rain gauge data from Niamey, suggested that rainfall estimates may scale radar estimates, dependent upon the mode of convection (i.e. SL, convective line, stratiform region) for an area calculation technique. Nesbitt et al. (2006) showed that feature area is comparable for continental and ocean systems, though the continental site did display overall larger system size in West Africa (confirmed in the present study). In this study, the continental site exhibited the largest fraction of partial features (features that occur at the edge of the scan region and were not fully sampled; Table II). Further analysis showed that partial features contributed similar fractions of rain volume and feature area at all sites. Because the emphasis of this study is on understanding the relative trends in the statistics, the occurrence of partial features should not adversely affect the analysis.
2.2. Radiosonde data
Radiosonde data collected near each radar location (Praia (TOGA), Kawsara/Dakar (NPOL), and Niamey (MIT)) have undergone extensive quality control and corrections (Nuret et al., 2008; Agustí-Panareda et al., 2009). Sounding launch intervals were approximately 4 hours (0000, 0400, 0800, 1200, 1600, 2000 UTC) at Praia, roughly 6 hours (0000, 0600, 1200, 1800 UTC) at Niamey and Kawsara, and twice daily (0000, 1200 UTC) at Dakar. Missing data at Kawsara reduced the number of usable soundings. Inspection of time series and variable distributions showed that Kawsara and Dakar soundings were nearly the same, therefore, given their close proximity (approximately 40 km); these two datasets were combined to improve temporal resolution. A number of thermodynamic characteristics were calculated for each sounding to characterize local environments for convective generation conditions, including convective available potential energy (CAPE), convective inhibition (CIN), and low-level shear (surface to low-level maximum zonal wind). Pseudo-adiabatic parcel ascent from mixed layer (bottom 50 hPa of sounding) was used in CAPE and CIN calculations. The methodology of Lucas et al. (2000) was used for the shear calculations. Time series correlations were tested using rank and product–moment correlations. The Wilcoxon–Mann–Whitney hypothesis test was applied at the 95% significance level for non-normal data distributions (i.e. CAPE), which is a nonparametric rank method that tests whether two samples are from the same or different populations.
2.3. Reanalysis data and easterly wave analysis
The NASA Goddard Space Flight Center (GSFC) Global Modeling and Assimilation Office (GMAO) Modern Era Retrospective analysis for Research and Applications (MERRA) product (Bosilovich et al., 2006), based upon the Goddard Earth Observing System Version 5 (GEOS-5) general circulation model (Rienecker et al., 2008), was used for identifying AEWs during the 2006 season.
Wave identification was performed using 700 hPa winds and the methodology discussed by Berry et al. (2007), in which the westward advection of curvature vorticity is employed for trough tracking. This algorithm is one of the few that explicitly attempts to reduce noise associated with the vorticity field from individual MCSs through the elimination of divergent flow. Results using the native resolution (0.5° latitudinal × 0.67° longitudinal) resulted in a discontinuous vorticity field between time steps. By degrading the reanalysis product to a 1° × 1° grid, the algorithm produced a cleaner vorticity field in which trough tracking was easily attained. Wave identification results using MERRA were found to be comparable to those produced by the European Centre for medium-range weather forecasts interim reanalysis (not shown). An example of an objectively identified AEJ axis (dashed line) and AEW troughs (solid lines) are shown in Figure 1.
Owing to previous association of precipitation and AEW troughs, it was of interest to analyze convective characteristics while in the presence of an AEW (wave regime) and while no AEW was present (no-wave regime). To accomplish this, systems occurring within 500 km of an AEW trough identified in the MERRA dataset were assumed to be associated with the wave regime (approximately a mean AEW wavelength). Previous research has shown that triggering and maintenance of convection occurs at this scale (Berry, 2009), including the top 10th percentile of intense convective events (Nicholls and Mohr, 2010).
3.1. Study area and environmental characteristics
The 2006 July–September (JAS) season was found to have small precipitation anomalies (less than 5%) from climatological values at each location using TRMM 3B42 gridded precipitation data. A time–longitude diagram using data from 2006 averaged over 12–17°N is shown in Figure 2. Data below 12°N (southern humid region in continental Africa) may skew regional analyses via the introduction of an area with less variability (Mohr et al., 2009), and was therefore excluded. The depiction of precipitation fraction data allows the cycle of convective decay and regeneration to be observed as a function of system propagation. Streaks of variable precipitation were observed, corresponding to westward propagating PFs—similar to outgoing longwave radiation diagrams (Laing et al., 2008; Cifelli et al., 2010). More than 10 advecting (or propagating) modes are evident in Figure 2, with varying life cycles in terms of precipitation intensities, duration, and phase speed. Objectively identified AEW trough tracks are overlaid (solid black lines). Ten AEW troughs were associated with the continental site, eight with the coastal site and 12 with the maritime site. Propagating modes averaged a speed of 14.9 m s−1, while mean AEW speed was 8.5 m s−1. In some cases, westward propagating precipitation events were evident along AEW trough tracks, while other trough regimes were void of precipitation. It is possible that variations in thermodynamic conditions and topography could have driven precipitation irregularity observed in the propagating modes (Laing et al., 2008).
Focusing on the radar locations, time series of radar reflectivity-estimated precipitation, CAPE, and CIN are shown for the continental (Figure 3), coastal (Figure 4), and maritime (Figure 5) locations, with AEW trough passages superimposed (hatched shading). Precipitation events of long duration and large spatial coverage were generally represented by unconditional rain rates (mean over entire scan domain) greater than 0.5 mm h−1. Continental convective systems were linear in organization and faster moving than those found over the east Atlantic (Laing and Fritsch, 1993; Hodges and Thorncroft, 1997) resulting in narrower peaks inland.
Time series of maritime CAPE in Figure 5(b) showed more variability than one might expect, likely because soundings were launched from an island large enough for nocturnal surface cooling to help establish a low-level inversion prior to daytime heating. Small values of CIN were prevalent at the coast, with greater variability at the continental and maritime sites. The largest values of CIN were observed with more frequency at the continental site. No significant correlations or anti-correlations (including lag correlation) were found between time series at each radar location. Spectral analysis of the time series showed no common precipitation or environmental periodicities between variables plotted in Figures 3–5, suggesting little dependence upon wave-driven dynamics. However, AEW and no-wave regime environmental variable populations were shown to be significantly different (discussed later).
More frequent occurrence of MCSs ahead of AEW troughs at the continental and maritime sites (60%) and an even distribution ahead and behind the trough at the coastal location were observed, in agreement with earlier studies (Carlson, 1969b; Payne and McGarry, 1977; Duvel, 1990; Machado et al., 1993; Diedhiou et al., 1999; Kiladis et al., 2006). During MCS events at each radar site, the AEJ was predominantly located north of the radar, which agrees with observations from Mohr and Thorncroft (2006), who found that the most intense convective systems occurred south of the jet axis in September.
Interpretation of PF results was dependent upon understanding the environment within which convection occurs. Mohr and Thorncroft (2006) showed that environments of high shear and high CAPE can result in a high probability of the most intense convective systems (SLs) in West Africa, in agreement with simulations (Weisman and Klemp, 1982). Vertical wind shear is an essential component to linearly organized convective systems (Bluestein and Jain, 1985; Weisman et al., 1988; Coniglio et al., 2006). Nicholls and Mohr (2010) found that the top 10th percentile West African convective systems exhibited significantly stronger low-level shear. Though MCSs exist in environments with a wide range of shear, organization and system strength tend to increase with increasing shear (also true in this study despite low correlation values). Figure 6 depicts the relative frequency of CAPE (top row), CIN (middle row), and low-level shear magnitudes (bottom row) at each site. The distribution of CAPE at the maritime location (Figure 6(a); 1090 J kg−1 median value) was skewed toward lower values, while the coastal location displayed a tendency toward larger CAPE values (Figure 6(d); 1842 J kg−1 median value). The continental site (Figure 6(f); 1044 J kg−1 median value) was centered about more moderate CAPE values; though extremely large quantities up to 6000 J kg−1 were observed (not shown), but confined to less than 1% of cases. Extreme CAPE values at the continental site were mostly unrealized, occurring in unfavourable conditions for convection (e.g. lack of synoptic-scale convergence, very little vertical shear, and large CIN). The continental domain exhibited a 50% larger CAPE value during AEW regimes (found to be significant to the 95% confidence level), while the coastal and maritime locations remained nearly unchanged between wave and no-wave regimes. Median values and distributions of CAPE are generally consistent with Fink et al. (2006) and Nicholls and Mohr (2010).
Distributions skewed toward small CIN values were observed at each location. Maritime and continental values (Figure 6(b) and (h)) showed occasional large CIN, with tails extending beyond 400 J kg−1, while the relative occurrence of small CIN was most frequent at the coastal site (Figure 6(e)). Occurrence fraction of sub-MCS features (fraction of time when sub-MCS convection was present) at the coastal site was 23%, while it was only 16% at the continental site, indicating that smaller CIN at the coastal site may have allowed for a higher relative occurrence of sub-MCS systems. Convective storms able to overcome the larger convective cap (shown by larger CIN values) inland resulted in more ‘intense’ convection in terms of vertical growth and reflectivity statistics (shown later) and is consistent with Nicholls and Mohr (2010), who found both larger CAPE and CIN values were present during intense events when compared to less-intense occurrences.
Similar median low-level wind shear values were observed for the maritime (easterly 3.7 × 10−3 s−1), coastal (easterly 4.1 × 10−3 s−1), and continental (easterly 4.1 × 10−3 s−1) locations; however, distributions differ for each location. Mean vertical wind profiles in Figure 7 show the presence of the AEJ near 650 hPa, and a westerly low-level jet (LLJ) near the surface for both the continental and coastal sites. This configuration is consistent with the higher frequency of larger easterly shear values at these locations. Southwesterly flow at the surface gives way to easterly flow aloft inland. At the coast, mean southwesterlies were overlaid by northeasterlies up to the AEJ level. The largest difference between the AEW and no-wave regime wind profiles occurs at the coast, where a difference of approximately 3 m s−1 existed throughout the profile. In addition, the westerly LLJ was more pronounced during the no-wave regime. Calculations of shear from the surface to the westerly LLJ (not shown) revealed that the coastal site exhibited larger mixing potential at low levels during wave passage. Despite prominent changes in environmental wind profiles between AEW and no-wave regimes, the coastal location exhibited the smallest inter-regime changes in precipitation and convective characteristics in the study (shown later). Along with the relative homogeneity of CAPE mentioned earlier, this suggests that when favourable large-scale dynamics are absent, MCSs at the coastal location draw upon buoyancy to maintain their intensity, despite less environmental shear.
Mean VAD divergence profiles (Figure 8) may be used to assess the effect of MCSs on the large-scale environment. Convective cells are characterized by convergence at the surface and divergence aloft, while stratiform regions display divergence at the surface, midlevel convergence and divergence aloft (Gamache and Houze, 1982; Mapes and Houze, 1993a). Standard deviation associated with the profiles was too large to yield significant differences between the AEW and no-wave regimes. The maritime profile (Figure 8(a)) exhibited the same structure as the intermediary case (a system during the conversion process from being convective to stratiform in nature) discussed in Mapes and Houze (1993b), also for an oceanic profile. The coastal profile (Figure 8(b)) showed divergence near the surface, mid-level convergence, and divergence aloft. The continental site exhibited the same general pattern (Figure 8(c)), with decreased divergence at the surface and peak convergence occurring lower in the atmosphere. This suggests distinct heating profiles for each location. It should be noted that these profiles could be driven by time-of-arrival of propagating MCSs that were often in a similar stage of development (see section 3.3).
3.2. Precipitation characteristics
Table III lists statistics derived from PF analysis for the study time period, along with the statistics for both AEW and no-wave regimes. Less than 4% of total scans over the continent and even less over the coastal and maritime locations contained MCS events. Even though MCSs where infrequent, MCS rain volume fractions (of total observed precipitation) were large, in line with previous studies using IR (80–90%; Mathon and Laurent, 2001) and TRMM microwave satellite data (60–80%; Mohr et al., 1999; Nesbitt et al., 2006), with a mean Sahelian value near 80% of annual precipitation.
Table III. Convective system characteristics derived from precipitation feature analysis for all study times and within AEW and no-wave regimes.
MCS occurrence fraction (%)
MCS rain volume fraction (%)
MCS area fraction (%)
Convective (stratiform) rain volume fraction (%)
Convective (stratiform) area fraction (%)
Number of precipitation features identified
A marked decrease at successive westward locations is observed in MCS area fractions (echo area coverage contributed by MCS-scale features; Table III). The percentage of area covered by continental and coastal MCSs was larger than sub-MCSs, while maritime MCSs and sub-MCSs covered an equivalent percentage of area, which agrees with Liu et al. (2008), who showed that the population of large satellite-observed systems decreased from West Africa into the East Atlantic at this latitude.
Contrary to results from previous studies, the stratiform precipitation fraction increased from west to east. Stratiform precipitation fractions for the maritime (36%) and coastal (37%) regions generally agree with Schumacher and Houze (2006), while the continental site (49%) was larger by nearly 10%. This difference may be explained by the fact that this study used only 1 month of data compared to 5 years in (Schumacher and Houze, 2006) and that spaceborne precipitation estimates do not account for the evaporation of precipitation in the boundary layer. Stratiform area accounts for 90% of MCS area, which may lead to underestimation from ground-based observations, which view a much smaller domain than space-borne observations and may not sample the entire MCS area. Strong, easterly low-level shear in this region produced leading convective line, trailing stratiform MCSs that greatly affected boundary layer properties. Boundary layer relative humidity (not shown) increases an average of 5% (>20% in some cases) with the passage of the convective line of these MCSs (denoted by arrow heads along the bottom abscissa in Figure 3), thereby retarding evaporation of the following stratiform precipitation and increasing observed stratiform precipitation fraction. Generally, upper-level humidity increases via transport by strong convective updrafts were observed during periods of high precipitation.
To further investigate precipitation in terms of vertical structure, precipitation contributions as a function of two characteristic reflectivity levels were calculated at each vertical level. First, 20 dBZ (Figure 9(a)–(c)) echo top heights were chosen to closely match the minimum threshold of the TRMM PR and minimize contamination from spurious echo missed in the radar QC process. Second, 30 dBZ (Figure 9(d)–(f)) echo top heights were chosen to identify intense convective cells with significant mixed-phase processes (DeMott and Rutledge, 1998). Data for all occurrences are shown; exclusion of sub-MCSs did not affect the distributions.
A strong bimodal distribution at the maritime location (9 and 15–17 km peaks), a weak bimodal distribution at the coast (9 and 15 km peaks), and a broad, unimodal distribution (15 km peak) at the continental site were seen in the 20 dBZ distributions. The maritime and coastal distributions suggest distinct modes, while the continental site appears to be influenced by a deeper spectrum of vertical development. Convective precipitation controls the contribution from the deep mode at all sites, while the stratiform precipitation occurs at a lower height. The stratiform contribution generally exhibits a more narrow vertical distribution, with the exception of the broad stratiform distribution at the continental site.
The 30 dBZ distributions indicate that the continental and coastal locations had deeper, more intense convective modes than the maritime site. The continental and coastal distributions fall off less rapidly from the 7–9 km peak, with a secondary peak in the coastal distribution near 13 km. More vertically developed storms display a greater propensity for mixed-phase processes (DeMott and Rutledge, 1998; Nesbitt et al., 2006), enhancing the stratiform region and leading to larger precipitation contribution over the course of the study from deep convection observed over the continent (Figure 9(e)–(f)). DeLonge et al. (2010) showed that MCSs transitioning from land to ocean exhibit signs of disorganization, resulting in less intense convective characteristics over the ocean. Figure 4 indicates that storms at the coast experienced a higher likelihood to enter a region with higher CAPE, which would theoretically produce stronger updrafts and significant lofting of precipitation-sized particles. Greater low-level shear over the land could act to enhance linear organization, resulting in two distinct modes (ocean and land) of vertical development present during the study period.
It is well established that the diurnal cycle of precipitation in West Africa is largely controlled by propagating MCSs (Reed, 1978; Shinoda et al., 1999; Mohr, 2004; Fink et al., 2006; Laing et al., 2008; McGarry and Rickenbach et al., 2009) and is a function of distance from genesis and redevelopment regions (e.g. high terrain; Hodges and Thorncroft, 1997; Mohr, 2004). This pattern was confirmed in this study in conjunction with Meteosat imagery (not shown), showing peak precipitation occurring near 0800 LT at the continental site, 0200 LT at the coastal and maritime sites.
3.3. Convective characteristics and easterly waves
Analysis of longer time period radar-estimated precipitation at the continental site showed a peak precipitation interval every 3–4 days (Nieto Ferreira et al., 2009), suggesting a correlation to AEW trough passage at this longitude. While precipitation events did occur near trough passages during the time frame of this study, many events also occurred when no wave was identified (see Table II). As a result, no significant correlation between AEWs and precipitation was noted at the continental site. Given the current debate concerning the impact of AEWs on precipitation, it was of interest to compare convective system characteristics during periods of AEW passage and periods of no AEW forcing. The intent was to take advantage of radar data from the three sites to further elucidate effects of AEW forcing on convective characteristics (i.e. vertical and horizontal structure) and to understand possible feedbacks of these mesoscale features onto the larger scale (i.e. MCS latent heat release).
Table IV lists contributions of AEW regime PFs during the study period, with 37–57% (32–45%) of total rain volume (feature area) associated with the AEW regime at all sites; less than previous results. Only the continental site showed that greater than half of the total estimated precipitation occurred during the AEW regime. Laing et al. (2008) found that about 80% of the deep convective area (as identified by satellite cold cloud streaks) was associated with AEWs for a region from 10°W to 10°E. The discrepancy with the current study may be attributed to differences between the PF definition used here and the classification of convection based upon minimum IR brightness temperatures. Ground-based radar observations yield a more direct picture of the spectrum of precipitating features, while IR precipitation estimates are based upon persistent, high cloud shields associated with MCSs. Therefore the estimation and temporal evolution of precipitation may differ between these methodologies and result in partitioning differences.
Table IV. Contribution of AEW regime precipitation features as a function of study period totals.
Total feature fraction (%)
Total rain volume fraction (%)
Convective rain volume fraction (%)
Stratiform rain volume fraction (%)
To further consider differences during AEW and no-wave regimes, cumulative frequency distributions (CFDs) of the feature area (Figure 10(a)) and rain volume (Figure 10(b) and (c)) were created and the distribution differences (CFDAEW − CFDno−wave) analyzed. Regime populations were found to be significantly different from the 99% confidence level. The AEW regime was associated with broad increases in PF size at the continental and coastal locations, with the coastal peak increase offset to larger systems. Feature size decreased during the AEW regime at the maritime location, with a maximum decrease at the sub-MCS scale. Examination of convective and stratiform precipitation volume distributions revealed that continental convective precipitation (Figure 10(b)) was enhanced during the AEW regime, while stratiform precipitation (Figure 10(c)) decreased. Little change was observed at the coast for convective precipitation, with an increase in stratiform precipitation during the AEW regime. Convective precipitation decreased at the maritime site during the AEW, while stratiform precipitation showed little deviation between regimes. Inspection of PF distribution along with environmental variables may help clarify the differences shown in convective and stratiform precipitation.
The increase in system size and convective rain volume at the continental site (Figure 10(a)) is consistent with increased CAPE, decreased CIN, and weaker shear, resulting in a shift of sub-MCSs to MCS-scale that occurred during the AEW regime compared to the no-wave regime. Increased large system population at the coast may have been driven by increased CIN and vertical shear, which resulted in a greater thermodynamic triggering barrier and provided increased organization for larger systems at the expense of smaller systems. A reduction in CIN and weaker vertical shear in the lowest 3 km were observed during the wave regime at the maritime site, explaining the formation of weaker, smaller convection.
Differences in convective and stratiform contributions can be further elucidated in terms of mean vertical reflectivity profiles (Figure 11). Convective (stratiform) profiles for each site are similar, with near surface mean values of 36–40 (22–28) dBZ. The decrease in reflectivity with height is similar for all three locations. Continental and maritime AEW regime convective profiles were more intense and exhibited higher reflectivity values aloft compared to the no-wave regime, suggesting hydrometeor loading aloft due to strong updrafts. Note that the number of points was an order of magnitude less for the continental no-wave regime, also suggesting less vertical growth overall. The coastal site exhibited very little difference in convective reflectivity profiles for AEW and no-wave regimes, which agrees with convective precipitation differences noted earlier (Figure 10(b)). Land-to-ocean transitioning MCSs (coastal site) have been shown to diminish in strength (e.g. DeLonge et al., 2010) due to less favourable thermodynamic (e.g. lower specific humidity) and dynamic (e.g. reduced vertical wind shear) conditions. The changes associated with this transition may have mitigated enhanced synoptic-scale moisture flux convergence and potential vorticity during the AEW regime to mediate vertical reflectivity profiles.
A brightband signature, owing to the melting of aggregates common in organized MCSs (Houze et al., 1989), was observed near 3–5 km in the stratiform profile at the coastal and continental sites. Decreasing reflectivity below this level toward the surface is a signature of droplet evaporation below cloud base, consistent with mesoscale descent. The continental AEW regime stratiform profile decreased more rapidly with height when compared to the no-wave regime, consistent with the reduction in stratiform rain area (Figure 10(c)) noted earlier. The order of magnitude difference in the continental number of points profile might suggest the importance of large-scale dynamics during the AEW regime on MCSs for the maintenance of the stratiform shield at this site. The consequences of these profiles are that MCS heating profiles at the coastal and maritime locations are comparable for the wave and no-wave regimes, whereas at the continental site differences arise due to the modification of stratiform structure. Weaker upper-level stratiform signal during the AEW regime results in a reduction of heating aloft and a less top-heavy heating profile (Mapes and Houze, 1995).
A comparison of convective characteristics via ground-based radar statistics for three unique geographic locations (continental, coastal, and maritime) was presented for the peak of the 2006 WAM season. High spatial and temporal resolution ground-based radar observations were complemented by sounding locations near each radar system, allowing characterization of convective events and the thermodynamic environment in which they occurred. A precipitation feature algorithm was employed to analyze precipitation and area characteristics at each site. Partitioning data in terms of convective and stratiform precipitation modes and AEW (or no-wave) presence was used to analyze geographic and regime variability of convective system structure and characteristics.
The diurnal cycle of precipitation at each location was largely dependent upon the time-of-arrival of propagating MCSs. The distribution of environmental conditions was important in determining differences in convective characteristics at each site. Key environmental and dynamical characteristics of each location are listed in Table V. All locations were generally favourable for the formation and/or maintenance of deep convection, though the extent organization, and therefore characteristics, depended upon the environment associated with each location. For example, the continental and coastal sites supported more organized, linear convection, while the maritime site exhibited less organized convective systems (important in terms of vertical growth and heating profiles of MCSs).
Table V. Comparison of key characteristics for each location during 19 August–16 September 2006.
Ahead of AEW trough
Equal ahead and behind of AEW trough
Ahead of AEW trough
N of radar
N of radar
N of radar
Phasing of MCSs with AEW troughs differed on a regional basis and precipitation was uncorrelated to trough passage during this period. The mean speed of MCS systems was greater than AEWs, suggesting a somewhat complex interaction between mesoscale and synoptic disturbances. Table VI notes the tendency of observed environmental and convective characteristics for each location during the AEW regime in comparison to the no-wave regime. The continental AEW regime displayed the greatest total precipitation, with near equal contributions from stratiform and convective components and an increase in precipitation during the AEW regime. In addition, less total precipitation was observed at the coastal and maritime locations during the AEW regime compared to the no-wave regime. Increased occurrence of large MCSs at the coast resulted in increased stratiform fraction and therefore increased stratiform precipitation. Together the results suggest important differences exist longitudinally across West Africa and to some degree whether convective systems interact with an AEW. Regional differences were generally more striking than those found between AEW and no-wave regimes at each site and were largely driven by differences in environmental characteristics.
Table VI. AEW regime characteristics in relation to no-wave regime characteristics during 19 August–16 September 2006.
Results presented here are limited by the short analysis period. Future work should examine climatological PFs via satellite observations, where a large number of convective systems can be sampled to better characterize differences between features associated with AEW and no-wave regimes. Timing of convection in terms of AEW passage would also be of interest to investigate systematic changes of environment by preceding convection for large systems leading to favourable feedback mechanisms with the synoptic scale (and possible cyclogenesis precursor). In addition, comparisons to inferred microphysics of different synoptic regimes and convective and stratiform components within MCSs should be compared against model simulations. Detailed study of MCS kinematics at each region would help quantify structural variability associated with geographic location and AEW and no-wave phasing, leading to a better understanding of latent heating and momentum transfer in comparison to model simulations.
This research was supported by the CEAS fellowship grant NNX08AT77G and NASA Precipitation Measurement Mission under grant NNX10AG88G. Based on a French initiative, AMMA was built by an international scientific group and is currently funded by a large number of agencies, especially from France, UK, USA, and Africa. It has been the beneficiary of a major financial contribution from the European Community's Sixth Framework Research Programme. Detailed information on scientific coordination and funding is available on the AMMA International web site http://www.amma-international.org. The NAMMA program was made possible by Dr Ramesh Kakar (NAMMA mission director). The authors would like to thank the AMMA and NAMMA project communities for maintaining and distributing datasets. MIT radar data were provided by Earle Williams. NPOL radar data were provided by Paul Kucera. Sounding data at Niamey, Kawsara, Dakar, and Praia were provided by Doug Parker, Marcia DeLonge, Serge Janicot, and Francis Schmidlin, respectively. We would like to acknowledge the Global Modeling and Assimilation Office (GMAO) and the GES DISC for the dissemination of MERRA. Gareth Berry provided computer code for AEW analysis. We had important discussions with Chris Thorncroft, Ed Zipser, Arlene Laing, John Zawislak, and Tom Rickenbach. The authors would also like to thank Karen Mohr, who provided invaluable advice on this manuscript.