1. Introduction
African easterly waves (AEWs) are westward-propagating synoptic-scale weather systems observed over northern tropical Africa and the tropical Atlantic during the boreal summer (e.g., Reed et al. 1977; Thompson et al. 1979). AEWs are an important component of the tropical North African and Atlantic climate. They modulate convection and rainfall on daily time scales (e.g., Carlson 1969a, b) and are associated with the formation of most tropical cyclones in the Atlantic (e.g., Avila and Pasch 1992). Past studies that considered AEW–convection relationships using composite analyses indicate that convection, including mesoscale convective systems (MCSs), tend to occur in preferred regions in the AEWs: at or ahead of the AEW trough in the vicinity of the intertropical convergence zone (ITCZ) over land (e.g., Carlson 1969a, b; Reed et al. 1977; Duvel 1990; Diedhiou et al. 1999; Payne and McGarry 1977; Fink and Reiner 2003; Kiladis et al. 2006), and collocated with the trough over the ocean (Thompson et al. 1979). In the region north of the ITCZ over land, however, convection is enhanced ahead of the AEW ridge (e.g., Burpee 1974).
Although it is clear from the composite analyses and case studies cited above that a coherent relationship exists between AEWs and convection, we lack a detailed knowledge of the significance of the synoptic modulation of convection, including how this varies in space and time. Our understanding of the interannual variability of AEWs is limited and consists of composites of AEWs during wet and dry years (e.g., Grist 2002; Grist and Nicholson 2001) or is just based on comparison of short periods (e.g., 3 yr as in Duvel 1989). Thorncroft and Hodges (2001) demonstrated the presence of interannual variability of coherent vorticity structures associated with AEWs but included no analyses of how this relates to convection.
Most of our knowledge of AEWs, including their relationship with convection, is based on the analysis of AEWs in West Africa (e.g., Burpee 1972; Albignat and Reed 1980, hereafter AR80) and the tropical Atlantic (e.g., Thompson et al. 1979). These are regions where the AEWs have peak amplitudes and are most coherent. We still know little about the nature of AEWs and their relationship with convection in the region east of 10°E. This is most likely due to a combination of factors including data sparsity and also because AEW structures are in the early stages of development. Indeed, there is still no consensus regarding the source region of AEWs and the mechanisms for their genesis. AR80 concluded that the source region of AEWs was west of 10°E. Using 10 000 ft (nearly 700 hPa) streamline analysis and cloud images, Carlson (1969b) found that nearly one-half of the waves he identified originated east of 18°E and suggested that the possible generating mechanism for wave disturbances may involve the interaction between convective processes and highlands over Sudan and Ethiopia. Recently, using both kinematic and convective analyses, Berry and Thorncroft (2005) hypothesized that AEWs form when the basic state is perturbed by convection over the Darfur mountains (Fig. 1). They argued that AEWs form as a dynamic response to large convective outbursts consisting of several MCSs leading to downstream development along the African easterly jet (AEJ). The idea that AEWs are forced by finite-amplitude convective precursors is also consistent with a recent numerical modeling results by Hall et al. (2006), who suggested that the AEJ is stable to small perturbations. Most of the observational studies quoted above were based on short time periods or individual cases. More analysis using time periods longer than a season or few seasons is clearly required to shed light on the eastern extent of AEW activity and its relationship with convection, including how this varies from year to year.
Motivated by the issues highlighted above, the two key aims of the present study are (i) to provide quantitative analysis of convection and AEWs and how this varies in space and time, and (ii) to consider how synoptic-scale convective activity relates to dynamical measures of AEW activity, and the initiation of AEWs in the east. Specifically, the role of periodic convection over eastern Africa in triggering AEW initiation in West Africa will be explored. Using brightness temperature data, Yang and Slingo (2001) showed that the intradiurnal variance (variance ≤1 day) contributes about 40% to the total variance in deep convecting Tropics. Our study extends the work of Yang and Slingo (2001) and describes the relative contribution of the synoptic time-scale variability to the total observed variability of convection. The phase relationships between AEWs and convection that have been highlighted in previous composite studies (e.g., Reed et al. 1977) will not be considered here. Instead, we will focus on the more general issue of AEW activity and how this relates to convection.
The paper is structured as follows. In section 2, a brief description of the datasets and methods used are provided. Section 3 presents spectral analysis and analysis of variance based on satellite brightness temperature data, including a quantitative assessment of the significance of synoptic time scale on the convection. In section 4, we investigate the relationship between the convection at synoptic time scale and dynamical measures of AEW activity, and its year-to-year variability. We also discuss the role of convective outbreaks over central and eastern African regions in initiating AEWs westward. Finally, in section 5, major findings of this study and directions for future work are summarized.
2. Data and methods
In this study, we will use brightness temperature (TB) from the Cloud Archive User Service (CLAUS; Hodges et al. 2000; Yang and Slingo 2001) of the European Union to investigate the characteristics of convection, and the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) to study dynamic measures of AEW activity.
Compiled from multiple satellite observations and a high-resolution dataset (3-hourly, global 0.5° by 0.5° grid), the CLAUS data provide a unique opportunity for a detailed study of interannual and seasonal variability of convection, including regional variations and comparison of synoptic-scale variance with intradiurnal variance. The CLAUS data were obtained for the period 1 January 1984–30 September 2001. This dataset has been successfully applied to the study of the diurnal cycle by Yang and Slingo (2001).
The dynamical fields are derived from the twice-daily ERA-40 (2.5° grid) dataset. Although observational data are sparse over Africa, especially over central and eastern parts, we are confident that this dataset is adequate for the analysis of the scale we are interested in for this work. It should be noted that the performance of the earlier ECMWF analyses for the synoptic-scale study of the African easterly waves was tested by Reed et al. (1988) and was found to be satisfactory. Tompkins et al. (2005) have also suggested that the ECMWF analyses in this region are of good quality, despite sparse observations.
Various methods of analysis for the period 1984–2001 have been employed. The analyses that involve Lanczos filtering of the TB field are based on 1984–2000. We apply the Lanczos filter (Duchon 1979; Lanczos 1956) to separate the time scales of interest. A bandpass filter in 2–6 days is used to isolate the synoptic time scale and a high-pass filter at 1.125-day cutoff is applied to isolate the intradiurnal variance. A low-pass filter with at 2-day cutoff is carried out to identify variabilities associated with periods of more than 2 days. To sharpen the response of the filter, 30-day data points were used on each side of the time series to be filtered. Also, as will be discussed in section 4, the time filtering revealed westward- and eastward-propagating convective structures. To distinguish these two types of convective structures, space–time filtering has been carried out following the method of Wheeler and Kiladis (1999; hereafter WK99).
3. Objective analysis of brightness temperature
a. Spectral analysis
As a first step toward assessing the association between AEWs and convection, spectral analysis is performed on the 3-hourly brightness temperature data for the period 1 July to 30 September (JAS). To investigate the regional variations in the spectra, we consider two regions: the tropical Atlantic (5°–10°N, 40°–20°W) and tropical Africa (10°–15°N, 15°W–40°E) in the vicinity of peak convection (Fig. 2). These regions are further subdivided into a number of 5° by 5° latitude–longitude boxes (4 in the Atlantic and 11 over land; 121 grid points in each box). The latitudinal belt of the locations (10°–15°N over land and 5°–10°N over the ocean) are chosen to coincide with regions of peak variance (cf. Fig. 4a). Spectral analysis was carried out on the JAS time series of each box, and the mean of the spectrum (MOS) was obtained by averaging over all 18 yr. The MOSs are compared against mean red noise1 spectra to examine the significance of peak periods (cf. Slingo et al. 1992). Here, significant power is defined as a peak in the spectra greater than the corresponding red noise value.
Figure 3 shows the spectral returns. The shading highlights significant power, and vertical dashed lines are intended to highlight the power for the 2–6-day periods, typical of AEW time scales. In the eastern tropical Atlantic, significant peaks are present between 4 and 5 days, but the diurnal signal is absent, consistent with past studies in this region (e.g., Yang and Slingo 2001). In the tropical North African region, significant power at both the diurnal time scale and in the 2–6-day range is seen. Also evident is a shift toward shorter synoptic periods over land. Perhaps the most striking result from Fig. 3 is the west-to-east change in the shape of the spectra. Over West Africa, west of 15°E, significant peaks are found near 3 days. In the region between 20° and 25°E, significant peaks in 2–6 days are weak or nonexistent, a region where dry air is advected from the Sahara (Thorncroft and Haile 1995). This is also a region where weak 2.5–5-day variance has been seen previously in dynamical fields (AR80, their Fig. 3). Arguably, the most pronounced spectral peaks in convection, at the synoptic time scales, are seen between 25° and 35°E, immediately west of Ethiopia. Past studies (e.g., AR80) have concluded that this region is not well known for synoptic disturbances or AEWs. It remains unclear why there should be such a pronounced peak at these time scales, and this needs to be examined further. The question as to whether this region is important for AEWs in this region or downstream will be examined in section 4. As expected, no significant peak in the 2–6-day range is seen in the regions east of 35°E, over the Ethiopian highlands.
The presence of significant power in the 2–6-day period over most of tropical North Africa and the tropical Atlantic motivates us to explore, in more detail, the spatial and temporal variability of convection at this time scale.
b. Analysis of variance
The JAS mean of TB (Fig. 2) highlights a narrow zone of peak convection over the tropical Atlantic in the vicinity of the ITCZ and a much broader region over the continent, consistent with the African monsoon and a summer rainfall climate over tropical North Africa. Also prominent are convective peaks over the elevated terrain that act to complicate the pattern over the continent as compared to the ocean. Over the Atlantic, the maximum variance (Fig. 4a) is located between 5° and 10°N in the vicinity of peak convection (Fig. 2). Over the land, the maximum variance is located slightly north of the axis of coldest brightness temperatures. As indicated by Salby et al. (1991), this is related to the variance alternating between high surface temperature and low cloud temperature. As shown in Fig. 3, the total variance has contributions from two key time scales: 2–6 days and diurnal. To highlight the contribution of these time scales to the total variance and to determine where each of the time scales is most important, we isolate the two time scales.
Figure 4b shows the variance of the 2–6-day-filtered TB. Over the Atlantic, the peak variances are confined to a narrow region between 5° and 10°N within the ITCZ. Over the land, the maximum variances are located between 10°–15°N and 15°W–10°E in West Africa, and slightly equatorward of this between about 7°–12°N and 20°–35°E in eastern Africa. This is consistent with the spectral analysis, which highlights the significant power in 2–6-day band in two regions. The regions of maximum variances in West Africa are regions where we expect to see peak easterly wave activity (e.g., Reed et al. 1977; AR80). Although similar magnitudes of variance are seen in both regions, past studies have shown that synoptic-scale easterly wave activity is either extremely weak or nonexistent east of 15°E (e.g., Burpee 1972; AR80). We revisit this in section 4 using dynamical measures of AEWs. Another prominent variance maximum is located in equatorial Africa between about 4°S and 5°N. Preliminary analysis of wind structures and propagation characteristics (not shown) indicate that this variance is not generally associated with AEW activity, and thus, further analysis of this region will be left for future work.
The significance of the 2–6-day variance on convection is now considered through the analysis of the percentage ratio of this time scale to the total variance (Fig. 4c). Over the land, percentage ratios of 20%–35% characterize most regions of summer time convection (cf. Fig. 2). Over the eastern Atlantic, a ratio of 30%–40% characterizes the ITCZ area. These results are consistent with the calculations of Dickinson and Molinari (2000), who found that 2–6-day-filtered variance of outgoing longwave radiation (OLR; for the period 1986–95) accounts for 25%–35% of the total variance over Africa. There are notable minima in the contribution from deep convecting regions over high terrain in Ethiopia, Cameroon, Nigeria (Jos mountains), and Guinea highlands, and also over the Sahara where we expect the diurnal cycle in the surface temperature to be more prominent (cf. Fig. 1).
The relative significance of 2–6-day variance is also examined by comparing the ratio of this variance to the intradiurnal variance. This will allow us to identify regions where forecasters need to give increased importance to synoptic time-scale signals relative to diurnal and higher-frequency signals. Figures 5a and 5b show the intradiurnal variance, and the percentage ratio of the 2–6-day time scale to the intradiurnal variances, respectively. The highest intradiurnal variances (Fig. 5a) are largely confined to the region between 10° and 15°N over land, consistent with Yang and Slingo (2001). Large values are located at the West African coast and over the highlands of Guinea, Cameroon, Jos, Darfur, and Ethiopia (cf. Fig. 1). Through comparison of Figs. 4b and 5a, we note that the 2–6-day signal is disrupted in the vicinity of the elevated topography, suggesting that local processes associated with orography are more important for convection there than the processes associated with propagating synoptic weather systems. As seen in Fig. 5b, the percentage ratio of the 2–6-day convective variance to the intradiurnal variance is more than 60% between 5° and 20°N over West Africa, and between 5°S and 10°N in central and eastern Africa, except over the highlands where this ratio is less than 30%. Around Senegal and in the region 5°–10°N, 5°–10°W (southern West African coast), the ratio increases to more than 80%, indicating a nearly equal importance of synoptic and diurnal time scales there. These are known regions where AEWs attain maximum amplitudes (e.g., Burpee 1972; AR80). Ratios of more than 200% in the Atlantic are consistent with a weak contribution from the intradiurnal time scales there (masked for clarity).
The intradiurnal variance contributes 40%–50% to the total variance in tropical Africa (not shown). Another way of measuring the significance of 2–6-day variances is to examine the ratio of this variance to the variance where intradiurnal and higher-frequency variances are removed (i.e., variance computed on a low-pass-filtered TB at 2-day cutoff). As seen in Fig. 5c, the contribution from 2–6-day variance is more than 50% within the eastern Atlantic ITCZ, and more than 70% over much of the deep convecting regions over tropical Africa. With the high-frequency variances removed from the total, therefore, the synoptic time scale stands out as the most distinctive part of the variability, particularly over land.
Using high spatial and temporal resolution brightness temperature (TB) data, we have shown the presence of significant 2–6-day TB variance over tropical Africa and the Atlantic. To highlight how this relates to AEW activity, an important second step is to compare these results with the dynamical measures of AEWs. This is explored in the next section.
4. Comparison with dynamical measures of AEWs
The result that there is indeed a strong and significant 2–6-day TB variance in West Africa and the tropical Atlantic is consistent with previous studies that have shown significant AEW activity in these regions (Carlson 1969a, b; Burpee 1972; AR80; Reed et al. 1977). The present study has also highlighted the presence of significant 2–6-day convective variance in eastern Africa (east of 10°E). Past studies have concluded that AEW activity east of 10°E is either very weak or nonexistent (e.g., Burpee 1972; AR80); it is unclear, therefore, how this is consistent with our analysis of the 2–6-day TB variance. This section highlights the AEW activity using dynamical measures and will explore how this relates to the analysis of 2–6-day TB variance.
a. Dynamical measures of AEW activity
As in many previous studies, time-filtered meridional wind (υ) at 700 and 850 hPa is used as dynamical measures of the AEW activity (e.g., AR80; Pytharoulis and Thorncroft 1999). As seen in Fig. 6a, over land, the maximum AEW activity at 700 hPa is located along 10°N, south of the AEJ axis and in the vicinity of peak convection (cf. Fig. 2). The maximum variance near 20°N over the ocean is somewhat surprising as it is located far to the north of the mean peak convective activity. This activity may be related to the midlatitude synoptic systems rather than equatorial ones. At 850 hPa, the waves attain maximum amplitudes on the northern flank of the jet axis, near 20°N over the Sahara within the baroclinic zone, and on the southern flank, along 10°N near the West African coast in the region of maximum convection (Fig. 6b). This is broadly consistent with AR80. At both levels, amplitudes in central and eastern Africa are weaker than in the west, again, consistent with AR80.
To further highlight the west–east differences, we briefly consider the low-level heat fluxes and midtropospheric potential vorticity fluxes. Since AEWs grow through interactions between Rossby waves on the positive low-level meridional potential temperature (θ) and the jet-level negative meridional potential vorticity (P) gradients (e.g., Pytharoulis and Thorncroft 1999), we expect baroclinic growth to be associated with significant equatorward (negative) heat fluxes at low levels, and poleward (positive) P fluxes at the jet level and in the vicinity of the jet core. We also expect AEWs to be associated with equatorward P fluxes south of the jet consistent with barotropic growth. Figure 7 shows the meridional potential temperature flux (
While there exists a significant difference in dynamical measures of AEW activity between the eastern and western tropical North Africa, the convective measure of synoptic activity shows very little difference and is, in fact, slightly larger in the east (Fig. 4b and Figs. 6 –7). We suggest that the simplest interpretation of this analysis is that AEWs are initiated by convection east of 10°E, consistent with the case studies of Berry and Thorncroft (2005), Lin et al. (2005), Carlson (1969b), and also with the composite study of Kiladis et al. (2006), who showed that the first indication of an AEW event with strongest amplitudes over West Africa was a synoptic-scale region of convection around 30°E. To explore this idea further, we consider the nature of the relationship between convective and dynamical measures of synoptic activity through the analysis of year-to-year variability and two extreme years.
b. Interannual variability of AEWs and convection
1) The 2–6-day-filtered convection
The interannual variability of the 2–6-day-filtered TB variance is presented to highlight variations over the tropical Atlantic and tropical North Africa. Regions chosen for this are eastern Atlantic (7°–12°N, 40°–20°W), western Africa (10°–15°N, 10°W–10°E), and central and eastern Africa (7°–12°N, 10°–30°E). The choice of these regions is based on the regional maximum centers in 2–6-day variance statistics (Fig. 4b). Figure 8 shows the anomalous variabilities of 2–6-day-filtered TB variance for the regions. These anomalies are calculated as deviations from the 17-yr (1984–2000) mean and serve as a simple measure of the interannual variability.
As seen in Fig. 8, there is large year-to-year variability in the 2–6-day TB variance (dark shading). Notable extreme years include 1988 when all the regions were characterized by anomalously high variance and, in contrast, 1992 and 1997 when all the regions were characterized by anomalously low variance. Included in Fig. 8 are anomalies of mean TB (a measure of mean convection) to study the association with the filtered variance. Anomalously high 2–6-day TB variance over the tropical Atlantic and West Africa tend to be associated with anomalously active mean convection (Figs. 8a,b). A high correlation over eastern tropical Atlantic (r = −0.73; significant at 95%2), and relatively stronger correlation over West Africa (r = −0.32; significant at 90%) is found between the 2–6-day TB variance and mean convection. The same is very weak over central and eastern Africa (r = 0.15). As seen in Fig. 8c, anomalously high 2–6-day variance over central and eastern Africa were present in 1984, 1987, and 1991, but the mean convection was anomalously low. Also, central and eastern Africa experienced weak 2–6-day variances in 1992, 1998, and 1999, but the mean convection was strong. This suggests a potentially more important role of other time scales (e.g., diurnal cycle) in the mean in some years.
2) The 2–6-day-filtered meridional wind
Figure 9 shows the year-to-year variability of 2–6-day-filtered meridional wind variance at 700 hPa for the three regions in tropical Africa and the Atlantic (light shading). Over the eastern Atlantic, this ranges from anomalously high variance in 1999 and 2000 to anomalously low variance in 1993 and 1997. Over West Africa, anomalously high variance occurred in 1995 and 1996, while anomalously low variance occurred in 1984, 1990, and 1997. Over central and eastern Africa, anomalously high variance occurred in 1986 and 1996, while low variances occurred in 1993, 1994, and 1997. The wave activities (as measured by these dynamical variables) over eastern and western Africa are strongly correlated (r = 0.8; significant at 95%), consistent with AEWs propagating between these two regions. In contrast, the relationship between AEW activity over West Africa and eastern tropical Atlantic is weak (r = 0.19). As they leave the land and move over the ocean, AEWs experience a different environment, suggesting that processes local to the tropical Atlantic are more important for determining AEW activity there than AEW activity upstream.
Reproduced in Fig. 9 are the 2–6-day-filtered TB variances to explore their association with the dynamical measures. A correlation analysis between the convective and dynamical measures indicates a strong relationship over West Africa (r = 0.41, significant at 90%). The correlations over the Atlantic (r = 0.19) and over central and East Africa (r = 0.1) are weak. The strong association between these measures over West Africa is consistent with earlier studies that show a coherent relationship between AEWs and convection over the region (e.g., Reed et al. 1977; Carlson 1969a, b; Kiladis et al. 2006). The weaker correlation between dynamical and convective measures over the Atlantic latitudinal belt (7°–12°N) may be somewhat unexpected, given the known strong relationship between AEWs and convection (e.g., Thompson et al. 1979). One reason for this may be the fact that the dynamical measures of AEW activity and convective activity peak in different locations (cf. Figs. 4b and 6a). Indeed, the correlation increases to 0.34 (significant at 90%) if we consider the region between 10° and 20°N (for the same longitudes). Lack of a significant correlation over eastern Africa is consistent with the result that AEWs have their main development downstream of convection over central and eastern Africa.
3) Two extreme years
The 2–6-day TB variances above reveal anomalously high and low variabilities in 1988 and 1990, respectively, over West Africa. We now present an analysis of AEWs and convection for these two extreme years. For illustration, August is chosen because it is the month of peak convective activity over land. Consistent with the JAS anomalies (Fig. 8), Augusts of 1988 and 1990 were characterized by anomalously high and low 2–6-day TB variance over West Africa, respectively (not shown). The contrast between these two years is clearly seen in the 2–6-day-filtered TB and 700-hPa υ Hovmöller diagrams (Fig. 10; averaged over latitudinal band 10°–15°N). The choice of averaging over the 10°–15°N band is based on the results of 2–6-day-filtered TB variances (Fig. 4b).
August 1988 shows coherent westward-propagating convective structures across the region, with a clear link between east and west (Fig. 10a). The convective structures are characterized by an approximate wavelength in the range 2500–3000 km and phase speed in the range 10–13 m s−1, consistent with previous analyses of AEWs (e.g., Burpee 1972; Carlson 1969a, b; Reed et al. 1977). These coherent convective features can be tracked back as far east as 35°E over western Ethiopian highlands (especially during the second half of the month) and at times even east of it (e.g., see 1 August). This indicates a potentially important role of the eastern African region for initiating synoptic-scale convection that can propagate over large distances and last for several days. As seen in Fig. 10a, convective events over land tend to be located in the northerlies, in agreement with composite studies cited earlier. Also, the dynamical signatures exhibit coherent westward-propagating structures starting from eastern Africa around 30°E (e.g., see third and fourth week). Most of the wave amplitudes increase westward over land. However, the AEWs are weaker and move more slowly when they are in the eastern Atlantic, consistent with the poor relationship between AEW activity over West Africa and eastern Atlantic discussed in the previous section. An important feature of this figure is that both convection and dynamical structures are characterized by similar wavelengths and phase speeds, suggesting a dynamic coupling between them.
The summer of 1990 represents a dry year where AEW activity and 2–6-day TB variance over West Africa were anomalously low (Fig. 9b). In contrast to 1988, the convective structures are weaker and less coherent and they are not clearly linked to the region east of 35°E (Fig. 10b). In general, tracks of convective structures over West Africa are shorter, and propagation speeds are larger in 1990. The direction of convective phase propagation is also more complicated. While most of the structures in August 1988 are westward moving, some structures that form between 0° and 20°E in 1990 are seen to propagate eastward (e.g., during 2, 5, and 20 August). Convective events over central and East Africa are much weaker than in 1988 and do not appear to have coherence with the dynamical wave signals. The AEW activity in 1990 was also weaker. It appears that the convective and AEW phase propagation speeds as well as their wavelengths are not in the same range, implying weaker coupling between the dynamics and convection. In this month, in contrast to August 1988, two coherent AEW structures are seen over West Africa during the second and third weeks and yet these did not have coherent convective signatures.
The contrast between August 1988 and 1990 is striking. A brief examination of the basic states for these years indicate the expected relationship between the AEJ and the mean rainfall (not shown), with a stronger and more equatorward AEJ in 1990, consistent with 1990 being much drier (Newell and Kidson 1984). A cursory look at the basic states did not reveal any obvious differences that could explain marked contrasts in the synoptic behavior. The simplest explanation for this difference is that 1988 was a much wetter year over tropical North Africa, and associated with this, we would expect greater numbers of finite-amplitude precursors for triggering AEWs (cf. Hall et al. 2006). More detailed analysis and modeling studies are required to explain the causes of the observed variability.
c. Role of eastern African convection on initiation of AEWs
In August 1988 both the synoptic convection and dynamical AEW signatures were seen in the vicinity of East Africa. In general, convection appeared to precede wave signals in the east, consistent with the hypothesis that AEWs are initiated by convection. In contrast, 1990 was characterized by a lack of AEW structures and a complicated pattern of convection that moved both eastward and westward (Fig. 10). This motivates us to isolate westward- and eastward-propagating structures based on the method of space–time filtering (WK99).
The AEWs are the major components of westward-moving structures in tropical Africa and the Atlantic. As in Kiladis et al. (2006), westward-propagating structures are isolated within periods of 2–7.5 days and wavenumbers 6–20, the tropical disturbance (TD)-type range shown in the frequency–wavenumber spectrum by WK99. Synoptic-scale eastward-propagating convective structures may be associated with Kelvin waves. However, we do not know a priori that they are the major components of the eastward-moving structures. Therefore, eastward-propagating structures are isolated using a wider window with periods of 2–10 days and wavenumbers 1–18. Note that, unlike WK99, the filtering scheme is carried out on the total TB, without first separating the spectrum into symmetric and antisymmetric components of the TB field. The basis for using the total field is because the variances are not symmetrical about the equator.
Figures 11a–d show distinct westward and eastward convective structures for the two JAS seasons of extreme years, 1990 and 1988. Overlaid are solid lines (defined as phase lines) on the propagating structures intended to highlight coherent propagation of convection. The phase lines are objectively identified using the following criteria: (i) convective anomaly should be less than or equal to −5 K (i.e., greater than one standard deviation of the filtered field) between the source and terminal regions; (ii) convective propagation must last for at least 2 days; and (iii) for westward phase lines, there must be continuous propagation for at least 2500 km, consistent with AEW wavelength scale. For eastward phase lines, the distance threshold is relaxed to 2000 km, which is close to the minimum resolved wavelength scale in the filtering parameters. These objective criteria were chosen to highlight the most coherent and long-lived systems and reduce potentially spurious signals from being counted. Comparing these objectively defined “tracks” with a manual analysis, based on the Hovmöllers, indicates very good agreement in most of the cases suggesting that the criteria are well chosen. Occasionally the objective tracks break where a manual analysis would suggest continuity (see 21 July and 3 August in Fig. 11a). However, since this occurs in less than 5% of the cases these events do not significantly affect the statistics or conclusions reached in this study.
Figures 11a–d clearly illustrate distinguishable behavior of convective structures relevant to AEWs and convective outbreaks at the source region. Consistent with Fig. 10, in 1988, nearly all of the westward-moving structures originated from east of 35°E (shown by vertical line, Fig. 11a), while in 1990, only a few originated over these regions (Fig. 11b). This leads us to consider an objective analysis of the total number and the year-to-year variability of propagating convective systems from source regions in the east to assess the extent to which these regions are important for initiating AEWs downstream.
As a first step toward assessing the origin of the structures, we consider the starting points of each of the objectively identified phase lines in the region between 10° and 15°N and between 15°W and 70°E. Recall that this is where 2–6-day TB variance is maximized over land (Fig. 4b). Then, the counts are binned in 5°-wide boxes and are shown in Fig. 12. It is evident that there are many more westward coherent convective structures during JAS 1984–2001 that originate from 20°–30°E, around Darfur, than from other regions. A secondary peak is seen in the east between 35° and 45°E, the Ethiopian highlands, suggesting a relatively more important role of the Darfur region for triggering AEWs.
Although the histograms in Fig. 12 show structures that propagate more than 2500 km, it is important, for forecasting purposes, to estimate how many of these actually reach a region of interest in West Africa, say, Niamey (13°N, 2°E). The structures that reach Niamey are shown by dashed lines in Fig. 12. Almost all of the convective structures formed over Darfur reach the Niamey region (0°–5°E), while some from the Ethiopian highlands do not. This would suggest that forecasters concerned with predicting the passage of AEWs over West Africa should, in general, be concerned with convective events breaking out or enhanced in the Darfur region (cf. Berry and Thorncroft 2005). This result updates the findings of AR80, who concluded that wave activity in the region east of 10°E is insignificant. In their work, AR80 did not include a measure of convection, and their findings were based on time-filtered meridional wind at 700 and 850 hPa for the short period available at that time.
Another important aspect that is not revealed in Fig. 12 is the year-to-year variability of westward convective structures. To examine this, two boxes, 20°–25°E and 35°–40°E, that show the highest number over the Darfur and Ethiopian regions are objectively chosen. As seen in Fig. 13, in 1988, five westward convective systems start over the box in Darfur region, when there was only one in 1992. In 1988, about five convective structures propagated westward from the box in Ethiopia, and in 1992 there were none. Thus, in some years, when the vicinity of Ethiopian highlands is “better connected” to the west, forecasters should also be concerned with convective events starting farther east than Darfur thereby potentially providing an extra day of warning of AEW passage over West Africa.
The conclusion here that the region in the vicinity of Darfur is generally more important than the region around Ethiopia or to the east of it for producing convective precursors that trigger AEWs downstream is consistent with recent numerical modeling results of Hall et al. (2006; N. Hall 2006, personal communication). They indicated that the entrance region to the AEJ (close to Darfur) is the preferred region to initiate downstream AEW growth by localized heating anomalies. Our statistical analysis combined with this modeling work and also the case studies by Berry and Thorncroft (2005) and Lin et al. (2005) provide a growing body of evidence to support the hypothesis that AEWs are triggered by finite-amplitude convective precursors in the east.
It should be made clear that even in the years when convective precursors near Ethiopia do not appear to have a role in initiating AEWs (as in 1984, 1990, and 1992), this region can still be characterized by convection. Instead of moving westward in a coherent manner (as in 1988), convection in these years generally decays before reaching Darfur, consistent with the unfavorable dry environment between the two regions (cf. Thorncroft and Haile 1995). The region between Darfur and the Ethiopian highlands (around 30°–35°E) is associated with a relatively lower number of propagating structures (Fig. 12). During these “unconnected” times, the convection in this region is still characterized by marked 2–6-day periodicity as shown in Fig. 3. An explanation for this periodicity is still lacking and will be the subject of future work.
It is perhaps a surprising result of this analysis that on average during JAS, more convective systems move eastward toward Ethiopia from Darfur than move westward away from Ethiopia (see Figs. 12 and 14). The number of systems moving eastward toward Ethiopia varied from about 11 in 1987 to about 5 in 1992. In their 850-mb wind analysis, AR80 also found propagating disturbances moving from around 20°E toward the east. More analysis of the nature of these systems and their variability is required since they are likely to have an impact on the weather and climate of the countries in the Horn of Africa, and when active, represent periods when East Africa is not triggering AEWs. Possible aspects to consider will be the role of equatorward penetration of midlatitude troughs and/or equatorial Kelvin waves (G. Kiladis 2006, personal communication).
5. Summary and final comments
The spectral analysis of an 18-yr JAS time series using satellite brightness temperature from the CLAUS dataset shows a pronounced 4–5-day periodicity over the Atlantic and a 3–4-day periodicity over western and eastern Africa, except for the region of 20°–25°E, which presents weak or no peak in the spectrum. The 2–6-day time-scale TB variance is of the same order of magnitude in eastern and western Africa and accounts for 25%–35% of the total variance in the deep convective regions. The contribution from this variance increases to more than 70% when compared with the variances computed from 2-day low-pass filter, particularly over land. Also, the 2–6-day variance is nearly as important as the intradiurnal variance, particularly over West Africa, except over the highlands and the west coast where intradiurnal variance is much higher.
While convective variance amplitudes are similar across the whole of tropical North Africa, dynamic measures of wave activity show marked east–west differences. The dynamic measures of AEW activity are strongest west of 10°E and are much weaker east of this. We suggest that these observations are consistent with AEWs that are initiated by convective precursors in the east and grow as they move westward, supported by combined baroclinic and barotropic growth along AEJ (Burpee 1972). This hypothesis is also supported by recent case study work of Berry and Thorncroft (2005) and Lin et al. (2005), and by idealized modeling work by Hall et al. (2006). We explored this hypothesis further by considering the year-to-year variability in AEW activity and its association with convection.
Our analysis highlighted the presence of large interannual variability in 2–6-day-filtered convection across the whole tropical North Africa and Atlantic region. Over West Africa and the Atlantic this varies coherently with the mean convection, while over central and East Africa there is no clear relationship with the mean. The weaker relationship in the east suggests a more important role of other time scales in the mean convection (e.g., diurnal cycle). Also, stronger correlations between convection and dynamical measures of AEW activity are found over West Africa, consistent with the peak amplitudes known to characterize the region. In contrast, the correlation between the 2–6-day-filtered convection and AEW activity over central and eastern Africa is very weak.
The analysis here showed distinct variability of the synoptic time-scale convection that originates from central and eastern Africa (Figs. 12 –14). In most years studied, however, synoptic convection originated over the region around Darfur is more consistent and coherent than those signals initiated from farther east. This suggests that the role of the Darfur region is generally more important for producing convective precursors that trigger AEWs and have stronger coherence with wave activity downstream. However, in some years such as 1988, the Ethiopian highlands have an important role in wave initiation downstream. Future work will closely examine the causes of this year-to-year and spatial variability in convection and AEW activity and the relative roles of the basic state and precursors.
Also, the results presented here have highlighted the need for further detailed analysis of the convective systems in central and eastern Africa, both because of their potentially important role in determining variability of AEW activity in the west but also because of how these systems impact the region directly. Further analysis is required to determine the causes of the marked periodicity in convection that characterizes the East African region and what determines eastward movement of convection and their intermittency from year to year and within a season.
Acknowledgments
This work is supported by a grant from the National Science Foundation (PTAEO: 10239111-24796). CLAUS data are available at the British Atmospheric Data Centre (BADC; http://badc.nerc.ac.uk/cgi-bin/data_browser/data_browser/badc/claus/). We thank Dr. Kevin Hodges of the Reading University, Reading, United Kingdom, for his help and useful suggestions about the CLAUS data. ERA-40 data were obtained from the ECMWF data server (available online at www.ecmwf.int/). Spectral analysis was carried out using an FFT package developed at the Department of Atmospheric Science, University of California, Los Angeles (www.atmos.ucla.edu/tcd/ssa/). We extend our gratitude to Dr. George Kiladis of the NOAA/Aeronomy Laboratory, Boulder, Colorado, and Dr. Doug Parker of the University of Leeds, Leeds, United Kingdom, for useful discussions and comments on this paper. We also thank the three anonymous reviewers for their useful comments, which greatly improved the paper.
REFERENCES
Albignat, J. P., and R. J. Reed, 1980: The origin of African wave disturbances during Phase III of GATE. Mon. Wea. Rev., 108 , 1827–1839.
Avila, L. A., and R. J. Pasch, 1992: Atlantic tropical systems of 1991. Mon. Wea. Rev., 120 , 2688–2696.
Berry, G. J., and C. D. Thorncroft, 2005: Case study of an intense African easterly wave. Mon. Wea. Rev., 133 , 752–766.
Burpee, R. W., 1972: The origin and structure of easterly waves in the lower troposphere of North Africa. J. Atmos. Sci., 29 , 77–90.
Burpee, R. W., 1974: Characteristics of North African easterly waves during the summer of 1968 and 1969. J. Atmos. Sci., 31 , 1556–1570.
Carlson, T. N., 1969a: Synoptic histories of three African disturbances that developed into Atlantic hurricanes. Mon. Wea. Rev., 97 , 256–275.
Carlson, T. N., 1969b: Some remarks on African disturbances and their progress over tropical Atlantic. Mon. Wea. Rev., 97 , 716–726.
Dickinson, M., and J. Molinari, 2000: Climatology of sign reversals of the meridional potential vorticity gradient over Africa and Australia. Mon. Wea. Rev., 128 , 3890–3899.
Diedhiou, A., S. Janicot, and A. Viltard, 1999: Easterly wave regimes and associated convection over West Africa and tropical Atlantic ocean: Results from NCEP/NCAR and ECMWF reanalysis. Climate Dyn., 11 , 795–822.
Duchon, C. E., 1979: Lanczos filtering in one and two dimensions. J. Appl. Meteor., 18 , 1016–1022.
Duvel, J. P., 1989: Convection over tropical Africa and the Atlantic Ocean during northern summer. Part I: Interannual and diurnal variations. Mon. Wea. Rev., 117 , 2782–2799.
Duvel, J. P., 1990: Convection over tropical Africa and the Atlantic Ocean during northern summer. Part II: Modulation by easterly waves. Mon. Wea. Rev., 118 , 1855–1868.
Fink, A. H., and A. Reiner, 2003: Spatio-temporal variability of the relation between African Easterly Waves and West African Squall Lines in 1998 and 1999. J. Geophys. Res., 108 .4332, doi:10.1029/2002JD002816.
Grist, J. P., 2002: Easterly waves over Africa. Part I: The seasonal cycle and contrasts between wet and dry years. Mon. Wea. Rev., 130 , 197–211.
Grist, J. P., and S. E. Nicholson, 2001: A study of the dynamic factors influencing the rainfall variability in the West African Sahel. J. Climate, 14 , 1337–1359.
Hall, N. M. J., G. Kiladis, and C. Thorncroft, 2006: Three-dimensional structure and dynamics of African easterly waves. Part II: Dynamical modes. J. Atmos. Sci., 63 , 2231–2245.
Hodges, K. I., D. W. Chappell, G. J. Robinson, and G. Yang, 2000: An improved algorithm for generating global window brightness temperatures from multiple satellite infrared imagery. J. Atmos. Oceanic Technol., 17 , 1296–1312.
Kiladis, G. N., C. D. Thorncroft, and N. G. Hall, 2006: Three-dimensional structure and dynamics of African easterly waves. Part I: Observations. J. Atmos. Sci., 63 , 2212–2230.
Lanczos, C., 1956: Applied Analysis. Prentice-Hall, 539 pp.
Lin, Y. L., K. E. Robertson, and C. M. Hill, 2005: Origin and propagation of a disturbance associated with an African easterly wave as a precursor of Hurricane Alberto (2000). Mon. Wea. Rev., 133 , 3276–3298.
Newell, R. E., and J. W. Kidson, 1984: African mean wind changes between Sahelian wet and dry periods. J. Climatol., 4 , 27–33.
Norquist, C. D., E. E. Recker, and R. J. Reed, 1977: The energetics of African wave disturbances as observed during Phase III of GATE. Mon. Wea. Rev., 105 , 334–342.
Payne, S. W., and M. M. McGarry, 1977: The relationship of satellite inferred convective activity to easterly waves over west Africa and the adjacent ocean during Phase III of GATE. Mon. Wea. Rev., 105 , 414–420.
Pytharoulis, I., and C. Thorncroft, 1999: The low-level structure of African easterly waves in 1995. Mon. Wea. Rev., 127 , 2266–2280.
Reed, R. J., D. C. Norquist, and E. E. Recker, 1977: The structure and properties of African wave disturbances as observed during Phase III of GATE. Mon. Wea. Rev., 105 , 317–333.
Reed, R. J., E. Klinker, and A. Hollingsworth, 1988: The structure and characteristics of African easterly wave disturbances determined from ECMWF operational analysis/forecast system. Meteor. Atmos. Phys., 38 , 22–33.
Salby, M. L., H. H. Hendon, K. Woodberry, and K. Tanaka, 1991: Analysis of global cloud imagery from multiple satellites. Bull. Amer. Meteor. Soc., 72 , 467–480.
Slingo, J. M., K. R. Sperber, J. J. Morcrette, and G. L. Potter, 1992: Analysis of temporal behavior of convection in the tropics of the European Centre for Medium-Range Weather Forecasts model. J. Geophys. Res., 97 , D16. 18119–18135.
Thompson Jr., R. M., S. W. Payne, E. E. Recker, and R. J. Reed, 1979: Structure and properties of synoptic-scale wave disturbances in the Intertropical Convergence Zone of the eastern Atlantic. J. Atmos. Sci., 36 , 53–72.
Thorncroft, C. D., and M. Haile, 1995: The mean dynamic and thermodynamic fields for July 1989 over tropical North Africa and their relationship to convective storm activity. Mon. Wea. Rev., 123 , 3016–3031.
Thorncroft, C. D., and K. Hodges, 2001: African easterly wave variability and its relationship to Atlantic Ocean tropical cyclone activity. J. Climate, 14 , 1166–1179.
Tompkins, A. M., A. Diongue, D. J. Parker, and C. D. Thorncroft, 2005: The African easterly jet in the ECMWF integrated forecast system: 4DVar analysis. Quart. J. Roy. Meteor. Soc., 131 , 2861–2885.
Wheeler, M., and G. N. Kiladis, 1999: Convectively coupled equatorial waves: Analysis of clouds and temperature in the wave-number-frequency domain. J. Atmos. Sci., 56 , 374–399.
Yang, G., and J. Slingo, 2001: The diurnal cycle in the Tropics. Mon. Wea. Rev., 129 , 784–801.
Topographic map of tropical Africa. Contour interval is 500 m, and shading shows areas of elevation greater than 500 m above mean sea level. Highlands are designated by letters: E = Ethiopian highlands, D = Darfur mountains, C = Cameroon highlands, G = Guinea highlands, and J = Jos mountains.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
Mean brightness temperature for JAS averaged over 1984–2001. Rectangular boxes over the tropical Atlantic and tropical North Africa indicate regions where spectral analysis was carried out based on 5° by 5° boxes.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
Power spectra of brightness temperature for regions over the Atlantic and tropical North Africa (see Fig. 2 for orientation). Significant power (power greater than the corresponding red noise values) is shaded. Vertical dashed lines delineate the power for the periods of 2–6 days. Power (in K2 units) on the ordinate is scaled to aid visualization. Frequencies on the horizontal axis are in cycles per day. Symbols denote the following: Eastern Atlantic Ocean (EAO), western Africa (WA), central Africa (CA), and eastern Africa (EA).
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
(a) Total variance of brightness temperature (TB; contour interval is 100 K2 and values greater than or equal to 400 K2 are shaded); (b) 2–6-day-filtered TB variance (contour interval is 20 K2 and values greater than or equal to 140 K2 are shaded); (c) percentage ratio of 2–6-day TB variance to the total TB variance (contour interval is 5% and values greater than or equal to 20% are shaded). All figures are for JAS and averaged over 1984–2000.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
(a) Intradiurnal variance of TB (contour interval is 30 K2 and values greater than or equal to 210 K2 are shaded); (b) percentage ratio of 2–6-day-filtered TB to intradiurnal variance (contour interval is 10% and values greater than or equal to 60% are shaded); (c) percentage ratio of 2–6-day TB variance to the TB variance computed with periods greater than 2 days (contour interval is 10% and values greater than or equal to 60% are shaded). In (b), values over the ocean are more than 200% and masked for clarity. All figures are for JAS and averaged over 1984–2000.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
The 2–6-day meridional wind variances for JAS (averaged over 1984–2001): (a) at 700 mb; (b) at 850 mb. Contour interval is 1 m2 s−2 and values greater than 5 m2 s−2 are shaded.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
The flux fields (2–6-day-filtered covariances) for JAS (averaged over 1984–2001): (a) meridional wind and potential temperature at 850 hPa (contour interval is 0.5 m s−1 K; values less than −0.5 m s−1 K are shaded); (b) meridional wind and potential vorticity on the 315-K potential temperature surface (contour interval is 1 × 10−2 m s−1 PVU, where PVU denotes potential vorticity units: 1 PVU = 10−6 m2 s−1 kg−1 K; values more than 1 × 10−2 m s−1 PVU are shaded). Zero contour is thick.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
Anomalies of the 2–6-day-filtered TB variance (dark shaded; in K2) and mean TB (light shaded; in K) for JAS: (a) eastern Atlantic (7°–12°N, 40°–20°W); (b) West Africa (10°–15°N, 10°W–10°E); and (c) central and eastern Africa (7°–12°N, 10°–30°E).
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
Same as in Fig. 8, but for anomalies of the 2–6-day-filtered meridional wind variance at 700 hPa (light shaded; in m2 s−2) and 2–6-day-filtered TB variance (dark shaded; in K2).
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
Hovmöller diagrams averaged between 10° and 15°N for the 2–6-day-filtered TB and meridional wind at 700 hPa: (a) August 1988 and (b) August 1990. To emphasize the stronger convective events, negative filtered TB (less than −5 K) are shaded. Meridional winds are contoured every 1 m s−1 (negatives are dashed; positives are solid). Vertical lines in the figures at 20°W and 35°E indicate the West African coast and western Ethiopian highlands, respectively.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
Space–time-filtered TB for westward-moving structures in (a) 1988 and (b) 1990, and for eastward-moving structures in (c) 1988 and (d) 1990. The averaging was carried out in the 10°–15°N latitude band. Filtered TB less than −5 K are shaded and coherent structures are identified by phase lines. Vertical lines at 35°E in (a) and (b) indicate western Ethiopian highlands, and at 20°E in (c) and (d) indicate western Darfur mountains.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
The histogram of convective structures that originate in a 5°-wide region (shaded, bar graph), and the total number of convective structures that reach Niamey (0°–5°E; dashed lines). The total number on the ordinate is for JAS 1984–2001. The abscissa shows center longitude of each box for the regions between 15°W and 65°E (the latitudinal band is 10°–15°N). Each of the structures propagate for more than 2 days and farther than 2500 km.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
The year-to-year variability of westward-propagating convective structures that originated from 20°–25°E in Darfur (solid lines) and from 35°–40°E in Ethiopia (dashed lines) and that propagate for more than 2 days and farther than 2500 km.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
The year-to-year variability of eastward-propagating convective structures that reach east of 35°E and that propagate for more than 2 days and farther than 2000 km.
Citation: Journal of Climate 19, 20; 10.1175/JCLI3920.1
For each box, red noise is computed using the autocorrelation coefficient and FFT results from each year’s time series and averaged over all years to form the mean.
The significance of the correlations is obtained by creating 10 000 pairs of synthetic time series through random ordering of the 18-yr anomalous indices, and the percentile levels are the ranked correlations of those pairs.