Spatiotemporal variability of the relation between African Easterly Waves and West African Squall Lines in 1998 and 1999

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D11, 4332, doi: /2002jd002816, 2003 Spatiotemporal variability of the relation between African Easterly Waves and West African Squall Lines in 1998 and 1999 Andreas H. Fink and Andreas Reiner Institute of Geophysics and Meteorology, University of Cologne, Cologne, Germany Received 13 July 2002; revised 22 January 2003; accepted 5 March 2003; published 5 June [1] This study investigates the spatiotemporal variability of the relationship between African Easterly Waves (AEWs) and the lifecycle and characteristics of Squall Lines (SLs) over West Africa for the two six-month periods May October 1998 and In all, 81 AEWs have been tracked using analyses from the European Centre for Medium-Range Weather Forecasts (ECMWF) and 344 SLs have been identified by localizing their leading edges mainly from passive microwave rain rate retrievals. It is found that the area west of the AEW trough is a favorable location for SL generation over the entire tropical West Africa. In the Sahel, a secondary peak around the region of maximum AEW-related southerlies is observed. In these wave phases, 42% of all 344 SLs were identified and defined as AEW-forced. According to this definition, the contribution of AEWs to SL generation increases from 20% around 15 E to 68% at the West African coast (15 W), and is larger for the Sahel than for the Guinea Coast/Soudanian region. Furthermore, the impact of AEWs on SL genesis is greater at the height of the Sahelian rainy season (July September) than in the remaining early and late monsoon months. Few SLs form after midnight and in the morning hours, but if so, they mostly belong to the sample of AEWforced SLs. Since AEW-forced SLs exhibit no extraordinary characteristics (lifetime, propagation speed, size, and rain rate) compared to the remaining SLs, it is suggested that the impact of AEWs is largely restricted to SL initiation and organization processes. Finally, some potential physical mechanisms responsible for the AEW/SL genesis relationship are discussed. INDEX TERMS: 3314 Meteorology and Atmospheric Dynamics: Convective processes; 3354 Meteorology and Atmospheric Dynamics: Precipitation (1854); 3360 Meteorology and Atmospheric Dynamics: Remote sensing; 3364 Meteorology and Atmospheric Dynamics: Synoptic-scale meteorology; 3374 Meteorology and Atmospheric Dynamics: Tropical meteorology Citation: Fink, A. H., and A. Reiner, Spatiotemporal variability of the relation between African Easterly Waves and West African Squall Lines in 1998 and 1999, J. Geophys. Res., 108(D11), 4332, doi: /2002jd002816, Introduction [2] Since the beginning of the 1970s, all climatic zones in tropical West Africa, from the arid Sahelian to the humid Guinea Coast climate, have experienced a decade-long period of below-normal annual rainfall amounts [Wagner and da Silva, 1994; Figure 3 of Nicholson et al., 2000]. In this context, Le Barbé and Lebel [1997] noted a decrease in the number of rainy events over the central Sahelian country of Niger in the two dry decades from 1970 to Using hourly rainfall data from Niamey airport station ( N; E; see map of West Africa in Figure 1 to localize geographic place names mentioned in the text), Shinoda et al. [1999] pointed out that the frequency of heavy rainfall events were especially reduced in years with below-normal rainfall. Squall line systems belong to the most important rain-bearing weather systems in tropical West Africa. Based on characteristic sudden changes in observed meteorological parameters at surface weather stations, several studies found that the contribution of squall line systems to the Copyright 2003 by the American Geophysical Union /03/2002JD ACL 5-1 annual rainfall amounts increases from between 16 and 32% at the Guinea Coast [Acheampong, 1982; Omotosho, 1985] to around 50% in the Soudanian zone [Eldridge, 1957; Omotosho, 1985] and to about 80% in the Sahel [Dhonneur, 1981]. As a consequence, interannual rainfall variations are likely to be associated with the interannual variability in the occurrence and intensity of squall line systems. [3] Squall line systems are intense mesoscale convective systems that can generally be divided into two parts: (1) The squall line or convective region that is characterized by a leading line of several convective cells, typically a few hundred kilometers in length, which may, in extreme cases, exceed 1000 km [Houze et al., 1989; Roux et al., 1984]. This line of thunderstorms is associated with heavy precipitation of short duration preceded by strong gusts at the gust or squall front [Chong and Hauser, 1989; Peters et al., 1989]. In a plan view, the convective part of the squall line system has a linear or convex form and is often referred to as the leading edge of the system [Rowell and Milford, 1993]. (2) The trailing anvil or stratiform region with continuous light rains that may last up to several hours. [4] In the present paper, the term squall line (SL) refers to both parts of the squall line system, although the

2 ACL 5-2 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES Figure 1. Map of West Africa showing orography, political outlines, and some important geographic locations mentioned in the text. The study region is indicated by the bold rectangle. distinctive leading edge of the system will be used to identify SLs in microwave and infrared satellite imagery. Typical rain rates in the convective region of West African SLs are in excess of 30 mm h 1 for a period of about 30 minutes with peak values ranging from 60 to 110 mm h 1. The intense rains turn into continuous light rains in the stratiform part with a characteristic intensity of 4 mm h 1 and a duration of 2 3 h [Chong et al., 1987; Roux, 1988]. It has been shown by Chong and Hauser [1989] that 55 65% of the rainfall associated with SLs occurs during the passage of the leading edge and the remaining 35 45% due to stratiform precipitation. Further investigations revealed that long-lived (henceforth used for SLs with lifetimes of more than 6 h) West African SLs have a mean lifetime between 10 h [Payne and McGarry, 1977; Rowell and Milford, 1993] and 13 h [Aspliden et al., 1976]; occasionally durations of two days and more have been reported [Desbois et al., 1988; Peters and Tetzlaff, 1988]. It appears, however, that such extremely long life spans might be the result of multiple SL genesis or regeneration. The propagation speed of SLs ranges from 6 m s 1 to more than 33 m s 1, while the average speed lies between 14 and 17 m s 1 [Aspliden et al., 1976]. [5] Rowell and Milford [1993] summarized the necessary climatological background conditions for SL generation as strong vertical wind shear in the lower troposphere, conditional instability and an atmospheric layering characterized by warm, dry air at middle levels overlying colder, humid air at lower levels. Such favorable conditions prevail over tropical West Africa between May and October. The lowlevel vertical wind shear is related to the African Easterly Jet (AEJ) at 600 hpa, which owes its existence to the lowlevel meridional temperature gradient between the dry, hot desert air and the humid, cooler air originating from the Gulf of Guinea. The large convective available potential energy (CAPE) of near-surface air parcels and the mid-level dryness is associated with the shallow, humid monsoon layer capped by the dry Saharan air layer. [6] Important factors for the generation of SLs are daytime surface heating, topographic features, supply of moisture by the shallow southwesterly monsoon flow and largescale convergence [Rowell and Milford, 1993]. Another potential triggering factor are mesoscale surface gust fronts from convective storm outflow, especially if the flow interacts with a range of hills or mountains [Wilson and Schreiber, 1986]. The importance of solar heating for SL generation is evident from a peak of SL initiations during the mid-to-late afternoon hours. For example, Aspliden et al. [1976] have shown that 55% of the observed long-lived SLs originated between UTC. Additionally, preferred areas of SL origin are the western sides of the Aïr and Adrar des Iforas mountains and over the Jos Plateau [Omotosho, 1984; Rowell and Milford, 1993]. Therefore, forced ascent of the southwesterly monsoonal flow and/or daytime heating over elevated terrain also act as important SL triggers. [7] The role of African Easterly Waves (AEWs) for SL generation, however, is still somewhat controversial. AEWs have a period of 3 5 days [Carlson, 1969; Burpee, 1972] and their associated cyclonic vortices or wave troughs usually propagate along two tracks over West Africa: one is situated north, the other south of the AEJ. While the southerly waves originate from barotropic instability at the equatorward flank of the AEJ, the northerly waves grow in association with dry baroclinic energy conversions at the southern margin of the Sahara [Chang, 1993; Pytharoulis and Thorncroft, 1999; Diedhiou et al., 1999]. Nitta and Takayabu [1985] and Pytharoulis and Thorncroft [1999] found that the northerly and southerly vortices were coherent features. Rowell and Milford [1993] examined the life cycle of 186 long- and short-lived SLs for a region bounded by 2.5 W 14 E and 9 12 N using METEOSAT infrared (IR) imagery for August 1985 and pointed out that there is no significant relationship between AEWs and SL generation and decay. They speculated that AEWs influence SL generation and decay only west of their study region, where the wave related low-level meridional wind fluctuations are more intense [Albignat and Reed, 1980; Duvel, 1990]. Bolton [1981, 1984] was unable to find any correlation between AEWs and SL passage at Minna, Nigeria ( N; E). In contrast to these studies, many other authors revealed an impact of AEWs on SL lifecycles or on the modulation of rainfall. Reed et al. [1977] investigated the structure and properties of eight AEWs over West Africa and the adjacent Atlantic Ocean during Phase III (August 23 to September 19, 1974) of GATE (GARP Atlantic Tropical Experiment) and observed maximum vertical motion, cloud cover and precipitation west of the trough, between the region of maximum northerly winds and the trough axis. For the same three-week period, Payne and McGarry [1977] found a preferred region of SL generation west of the trough, immediately east of the region of strongest northerly winds, and a preferred region of decay just east of the ridge. More recent studies by Duvel [1990] and Diedhiou et al. [1999] using METEOSAT IR data and ground measurements of precipitation over longer periods seem to confirm the results of Reed et al. [1977] and Payne and McGarry [1977] for the land-based convection over

3 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES ACL 5-3 West Africa. However, regional differences in the phasing of maximum rainfall probability appear to exist. Burpee [1974] was first to note on the basis of synoptic weather reports that south of 12.5 N the rainfall maxima occur just west of the AEW trough, while north of this latitude maximum rainfall intensities are observed west of the ridge in the region of strongest southerly winds. Duvel [1990] and Mathon et al. [2002] corroborated Burpee s findings and further noticed that convection is suppressed west of the trough in the dry Saharo-Sahelian region between 15 and 20 N. However, most of the 49 SLs identified by Duvel [1990] formed just west of the maximum northerly winds. From Duvel s study and a companion paper by Machado et al. [1993], it appears that, while the AEW trough region and the region of maximum southerlies promote the enhancement of rainfall in the Sahel, the origin locations of SLs may exhibit a different phase relationship. The above mentioned results, though, are somewhat disagreeing with the findings of Peters and Tetzlaff [1988] and Peters et al. [1989] who carried out a composite study with 17 and 32 Sahelian SLs, respectively, for the period June to August They found a distinct peak in the occurrence of squall front positions west of the AEW trough in the region of maximum northerlies with a concomitant 50% increase in rainfall intensity compared to SLs occurring east of the trough. This contradiction might be consistent with the results of Druyan et al. [1996] who described a shift of the rainfall maximum from the region west of the AEW trough in the dry year 1987 to the region of maximum southerlies in the wet year [8] In conclusion, a clear-cut picture of the influence of AEWs on SL generation and decay cannot be inferred from the present literature. As a basic consensus one might summarize that different phase relationships hold for the dry Sahel and that the strength of the coupling increases toward the West African coast. A shortcoming of most of the above mentioned studies, except those of Duvel [1990], Machado et al. [1993] and Diedhiou et al. [1999], is that they are based on investigation periods of two months or less and/or solely examine the region of wave origin and initial growth. Consequently, little is known about the intraseasonal variability of the coupling between AEWs and SLs. Furthermore, many studies do not directly deal with the examination of SLs, but with cloud clusters, cloudiness or rainfall. Among them are the above-cited three studies which investigated multiyear periods. [9] By investigating the relationship between 81 manually tracked AEWs and 344 SLs over tropical West Africa for the two six-month periods May October 1998 and 1999, the present study aims at advancing the understanding of the spatiotemporal variability of the relation between AEWs and the lifecycle and characteristics of SLs. Emphasis is put on the spatial variations of this relationship from the region of wave origin and initial growth in central West Africa to the Atlantic coast, as well as on the differences between the Guinea Coast/Soudanian regions and the Sahel, defined here as the region north of 12.5 N. Furthermore, temporal differences between the peak of the monsoon between July and September (JAS) and early (May and June) and late (October) monsoon months will be discussed. Additionally, some aspects of the diurnal cycle of the coupling between AEWs and SLs will be considered. [10] The following sections 2 and 3 deal with the data and methods used for the identification of AEWs and SLs. In section 4, the spatiotemporal variability of the relation between the two systems will be shown and analyzed. A summary and discussion is given in section Data [11] Reed et al. [1988a, 1988b] have shown that the analyses of the European Centre for Medium-Range Weather Forecasts (ECMWF) can successfully be used to track AEWs. The six-hourly (00, 06, 12, and 18 UTC) ECMWF pressure level analyses of the meridional and zonal winds and the relative vorticity at the 850 and 700 hpa levels were transformed from spherical harmonic space onto a 1 1 latitude/longitude grid for the periods May October 1998 and 1999.The investigated area between 20 W 20 E and 0 20 N is depicted by the bold rectangle in Figure 1 and comprises mainly tropical West Africa, but parts of the northern Gulf of Guinea and the eastern-most Atlantic Ocean are also included. The use of ECMWF gridded data for AEW tracking is described in section 3.1. The tracking of SLs has been accomplished by exploiting passive microwave rainfall retrieval techniques. For this purpose, microwave radiances from the TRMM Microwave Imager (TMI) on board of the Tropical Rainfall Measuring Mission (TRMM) satellite and from the Special Sensor Microwave Imager (SSM/I) borne on the Defense Meteorological Satellite Program (DMSP) satellites F11, F13, and F14 were used on average from eight overflights each day. Additionally, three-hourly METEOSAT IR images have been utilized to ensure a continuous SL tracking. These METEOSAT images were provided by the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT). The identification method is described in section 3.2. The examinations of monthly and seasonal rainfall anomalies (section 4.1) were carried out using monthly rainfall data from the U.S. National Oceanic and Atmospheric Administration (NOAA), from AGRHYMET (Centre Régional de Formation et d Application en Agrométéorologie et Hydrologie Opérationnelle, Niger), and from IRD (Institut de Recherche pour le Développement, France). 3. Methods 3.1. AEW Tracking [12] AEWs have been identified and located using a new three-step subjective tracking method. First, two sets of longitude-time (Hovmoeller) diagrams of temporal meridional wind anomalies were produced: one at 700 hpa averaged between 7 13 N, the other at 850 hpa averaged between N. The two types of Hovmoeller diagrams take the existence of northerly and southerly AEWs at different altitudes (see below) into account [Pytharoulis and Thorncroft, 1999; Nitta and Takayabu, 1985]. From these diagrams, the approximate longitudinal position of AEWs were easily identified by subjectively interpolating sign changes in the meridional wind with a straight line. An example for an AEW occurring in July 1999 is shown in Figure 2a. In a second step, six-hourly streamline maps of 2 6-day bandpass-filtered (bpf) wind at 850 hpa have been

4 ACL 5-4 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES Figure 2. Tracking of AEWs. (a) Longitude-time plot of the temporal anomalies of the meridional wind component at 700 hpa along 30 W 20 E between 00 UTC July 01 and 18 UTC July 15, The contours show the latitude average between 7 13 N in intervals of 2 m s 1. The bold lines denote zero speed anomaly, negative contours are dashed. The shading indicates negative meridional velocities at 850 hpa for the N latitudinal average. (b) Streamline map of 2 6-day bandpass-filtered wind at 850 hpa for 12 UTC July 09, 1999 (for further details see text). plotted. In most cases, a northerly and southerly center of cyclonic inflow at or close to the previously identified longitudes could unequivocally be localized in the streamline maps (Figure 2b). In case of absent closed circulation centers or ambiguity, maps of 2 6-day bpf relative vorticity at 700 hpa were investigated in a third step to diagnose the latitude/longitude coordinates of the cyclonic center of the northerly and southerly AEWs (not shown). The selection

5 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES ACL 5-5 of both the 850 and 700 hpa levels for AEW identification turned out to be the best choice for the following reasons: (a) the northerly waves are strongest at the lowest pressure levels [Pytharoulis and Thorncroft, 1999] and (b) the southerly waves very often show strong cyclonic inflow at 850 hpa, but strongest cyclonic shear at AEJ core heights ( hpa) [Reed et al., 1977]. From our tracking method, we obtained a total of 81 AEWs having a mean propagation speed of 7.3 degrees longitude per day (9.1 m s 1 ) and wavelengths between 2200 and 2750 km with a few extreme wavelengths around 3000 km. These values are consistent with previous synoptic studies that also derived AEW properties from sequences of weather maps [Burpee, 1975; Reed et al., 1977]. [13] In former studies, various AEW identification and tracking methods have been applied. Reed et al. [1988b] used horizontal distributions of streamlines of unfiltered wind and vorticity, whereas Reed et al. [1977] solely used streamline analyses of 2 6-day bpf wind. These techniques are similar to our method, except that no Hovmoeller diagrams were utilized. Druyan et al. [1996] identified AEWs in time series of 700 hpa meridional winds from Niamey radiosonde data and from cyclonic circulations on latitude-longitude analyses at 700 hpa. Duvel [1990] and Diedhiou et al. [1999] selected AEW periods from threshold amplitudes of the bpf low-level meridional winds at reference latitudes. An alternative approach presented in Thorncroft and Hodges [2001] is based on an automatic tracking of vorticity maxima. Their algorithm yields good results over the Atlantic Ocean and the Caribbean, but the relation between vorticity maxima and AEWs is not as clear over West Africa. One advantage of the method employed in the present study is that the use of Hovmoeller diagrams of unfiltered meridional wind anomalies and the meticulous subjective analysis of the described set of six-hourly maps avoids problems with locating the waves and ensures that only AEWs will be identified. Disadvantages include the facts that the manual identification method is subjective and extremely time consuming, therefore inhibiting the investigation of multiyear periods SL Tracking [14] The distinction between SLs and cloud clusters is not always clear-cut, because cloud clusters may contain or begin/end as SLs. In this study, any system with a sharp leading western edge is defined as a SL, since a line of organized convective cells is the essential distinguishing feature between SLs and cloud clusters (cf. section 1). Due to the sparsity of conventional observations, many authors used satellite data in order to identify SLs. Often only IR imagery is employed [Aspliden et al., 1976; Payne and McGarry, 1977; Rowell and Milford, 1993], although the extensive cirrus shields blown out of or sheared away from cumulonimbus anvils by the upper-level Tropical Easterly Jet sometimes prevent the unequivocal identification of the leading edge, which negatively impacts on the reliability of the identification method. Furthermore, the determination of SL sizes from IR data is somewhat uncertain due to the fact that the observed cirrus shields are generally more extensive than the area affected by ground precipitation. In particular, IR data are unsuitable for a quantitative computation of SL rainfall intensities (cf. section 4.2), since no clear relation exists between cloud top temperature and precipitation at the surface. [15] In this study, a new method of SL tracking has been applied. Passive microwave measurements of the SSM/I on board of the polar-orbiting DMSP satellites F11, F13, and F14, and of the TMI on the TRMM satellite, were converted into rain rates. For this purpose, the algorithm of Ferraro and Marks [1995] was used, which is based on the attenuation of radiances at 85.5 GHz due to scattering by ice particles in rain clouds. We have adapted the original Ferraro-Marks algorithm for West Africa, both with ground truth data and the precipitation radar aboard the TRMM satellite. The size of the 85.5 GHz ellipse (pixel) at the surface (footprint) is 7 by 5 km for the TMI and 13 by 15 km for the SSM/I. Since the respective swath widths of the SSM/I and TMI radiometers are 1394 and 759 km and since approximately eight overflights of the four satellites were available each day for the investigated region, not only SL identification, but also SL tracking is feasible. [16] Figures 3a and 3b illustrate that microwave retrieval of instantaneous rain rates allow an easy identification of leading edges. Figure 3a displays the swath of the SSM/I radiometer and the derived rain rates for an overflight of the satellite F13, while Figure 3b shows the corresponding features for a F11 overflight that occurred 90 minutes later. These images reveal a westward moving SL consisting of a well developed leading edge (rain rates > 35 mm h 1 ) and an area of stratiform precipitation at the trailing region. The position of the SL is defined as the position of the leading edge. The dashed convex curves on the map of West Africa in Figure 3c demonstrate how the leading edges visible in Figures 3a and 3b have been approximated. The heavy solid curves mark the positions of the leading edges inferred from antecedent and subsequent satellite overflights. The analyzed SL track is indicated in Figure 3c by a heavy horizontal line connecting the center points of the subjectively analyzed leading edges. Finally, the area affected by the SL is defined by the dashed envelope connecting the outer ends of the leading edges. This kind of tracking has been carried out for the entire investigation period. [17] Since the overflights of the DMSP satellites over West Africa were clustered between 16:30 22:00 UTC and 4:45 10:00 UTC and TRMM overflights occurred two or three times in between (around midnight and early afternoon), many short-lived SLs might have been missed. In fact, only 11 SLs with lifetimes of less than 6 h were detected and the majority of identified SLs have lifetimes greater than 9 h. In order to secure that the same SL system in subsequent overflights of the DMSP and TRMM satellites was traced, METEOSAT IR imagery with a temporal resolution of 3 hours have been visually inspected. Moreover, if overflights of TRMM or DMSP satellites were of limited use (e.g., if the leading edge was only partly located in the swath) or unavailable for SL tracking over a longer period, the SL position, but not its rainfall area and intensity, has been subjectively determined using IR images. The time when the SL was first (last) detected in DMSP/TRMM overflights or METEOSAT IR imagery was taken as the time of generation (decay). SLs entering the investigation domain from the east were not considered. Although this tracking method is again very time consuming, it allows a reliable identification of the leading edge and a computation of SL size and intensity in

6 ACL 5-6 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES Figure 3. Tracking of the SL that was first detected on 14:54 UTC September 18 and dissipated after 06 UTC September 19, The intensity of the shading of the filled circles in the upper two panels indicates pixel rain rates in units of mm h 1 (see label bar) obtained from the adapted microwave-based Ferraro-Marks algorithm. The swaths of the SSM/I radiometers on board of the DMSP satellites (a) F13 in and (b) F11 in are shown by dashed straight lines. Times in the top right-hand corner correspond to the flight times of the respective satellites from the equator to 20 N. (c) SL track obtained by connecting the center points of the subjectively analyzed leading edges. The heavy dashed lines correspond to the leading edges in (a) and (b). The area affected by the SL is defined by the dashed envelope connecting the outer ends of the leading edges. terms of rainfall. In the period under investigation, 344 SLs have been identified with a mean lifetime of 12 h and a mean westward propagation speed of 15 m s 1 (corresponding to 54 km h 1 ), agreeing well with the results of Aspliden et al. [1976]. The maximum life time of the SLs observed was 32.1 h, therefore, confirming that reported durations of several days [Desbois et al., 1988] might be associated with multiple SL (re-)generation in long-lasting cloud clusters Assignment of SLs to AEW Phases [18] In order to examine the phase relation between AEWs and SLs, each wave was subdivided into eight wave phases or categories following the classical approach of Reed and Recker [1971]. Phase 2 is centered on the region of maximum northerly wind, phase 4 on the trough axis, phase 6 on the region of maximum southerly wind, and phase 8 on the ridge axis. Figure 4 shows an AEW occurring over central West Africa on June 20, 1998, and illustrates how the partitioning into wave categories was subjectively carried out from 850 hpa streamline maps of 2 6-day bpf wind. It is worth noting that only AEW phases relevant for determining the phasing of the SL were analyzed (see Figure 4). In general, trough (ridge) axes were identified from cyclonic (anticyclonic) vortices and flow curvature (Figure 4), while the lines of maximum northerly and southerly wind were analyzed on isotach maps of 2 6-day bpf wind at the same pressure level (not shown). Phases 1, 3, (see heavy dashed lines in Figure 4), 5 and 7 were determined as being located midway between the other phases. [19] In a second step, the leading edge of a SL embedded within an AEW was assigned to a wave phase interval. For this purpose, rain rates within the swaths of overflying satellites were superimposed on both the streamline (see example in Figure 4) and isotach maps (not shown). In general, the overflight times of the available satellites did not correspond to the analyses times of the ECMWF (00, 06, 12, and 18 UTC), so that each overflight was compared to streamline/isotach maps of the nearest analysis time. The temporal discrepancy was in most cases far less than 3 hours and, given the fact that AEWs propagate westward less than 1 longitude within 3 hours, the resulting errors are negligible. In Figure 4, the crossing time of the TRMM satellite between 23:22 and 23:31 UTC corresponds well to the 00 UTC analysis time of the streamline map. It is obvious from Figure 4 that the leading edge of the SL has to be assigned to the AEW phase interval 8 1 at this time of its lifecycle. Figure 4. Assignment of SLs to AEW phases. The swath of the TMI radiometer borne on the TRMM satellite is shown by thin dashed lines. The corresponding rain rates are shaded in units of mm h 1 (cf. Figure 3). Times in the top right-hand corner correspond to the flight time from the equator to 20 N on June 20, The streamlines were obtained from 2 6-day bandpass-filtered winds at 850 hpa on 00 UTC June 21, AEW phases are subjectively analyzed on the streamline map and are defined in the text. The letters R, N, and T are abbreviations for Ridge, Northerlies, and Trough.

7 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES ACL 5-7 This SL appeared between AEW phases 1 and 2 on the METEOSAT IR image of 21 UTC and was last observed on a F14 SSM/I image at 07:43 between phases 8 and 1, thus having an analyzed lifetime of 10 hours and 43 minutes. 4. Results 4.1. General Results [20] In order to place the results of the present study into the climatological background, monthly and seasonal (May October) rainfall anomalies relative to the base period were computed from the raingauge data mentioned in section 2. In the period May October 1998, dry conditions were observed at the Guinea Coast, whereas rainfall was above average for most parts of the Sahel due to a wet August and September. The rainy season of 1999 started dry in May and June, but abundant and widespread rainfalls in the following four months lead to positive rainfall anomalies almost everywhere across tropical West Africa. Thus, the AEWs and SLs investigated in this study occurred during two wet Sahelian years when compared to the base period It should be stressed, however, that our 33-year base period fell within anomalously dry decades in the Sahel so that the two years investigated have to be categorized as normal and slightly above average when compared to the longer 20th century records (cf. Figure 9 in Parker and Alexander [2001]). [21] The paths of all 81 AEWs obtained from our tracking method described in section 3.1 for the two six-month periods in 1998 and 1999 are shown in Figure 5. Figure 5 reveals the existence of two tracks reflecting the fact that an AEW usually consists of two troughs or cyclonic centers that travel westward at either side of the AEJ. The numbers of northerly and southerly AEWs are displayed in the top and bottom left-hand corner of each panel, respectively. Northerly (southerly) wave troughs without a complementary southerly (northerly) vortex occurred in May, July, and September 1998 and in May, June and October In the latter month, an equal number of single northerly and southerly troughs occurred. For the entire investigation period, 12 (15%) out of 81 AEWs were only identifiable at one flank of the AEJ. This percentage is smaller than the 68% given in Diedhiou et al. [1999], but supports the notion of Pytharoulis and Thorncroft [1999] and Nitta and Takayabu [1985] that the northerly and southerly waves are usually coherent features. [22] In the two years considered, the comparison of the AEW activity in terms of their number and maximum values in bpf 700 hpa relative vorticity with the monthly rainfall anomalies (not shown) suggest a much weaker relationship at the beginning/end (May, June, and October) than at the height of the Sahelian rainy season (July, August, and September). For example, the dry months October 1998 and June 1999, but also the wet month October 1999 coincide with strong AEW activity. On the other hand, dry conditions in July 1998 are associated with weak AEW activity, and wet conditions in July 1999, and August and September 1998 and 1999 concur with strong wave activity. [23] Figure 6 shows the result of the SL tracking (section 3.2) for the rainy seasons 1998 and 1999, during which a total of 344 SLs were identified. The meaning of the Figure 5. AEW tracks for the 12 months from May October 1998 and The shaded areas are displayed for orientation. The digits in the top (bottom) left-hand corners denote the numbers of northerly (southerly) AEWs for the respective months. The northerly (southerly) AEW tracks are dashed (solid). A total of 81 AEWs were identified within the two six-month periods. different coloring of SL tracks in Figure 6 will be explained in the next section. The number of SLs in the top left-hand corner of each image reveals that most SLs occur in August and September. Furthermore, Figure 6 shows that SL tracks migrate northward from May to July/August and progressively retreat southward in September/October. This seasonal shift occurs in association with the seasonal migration of the AEJ, and, hence, with the region of strongest vertical shear (section 1). It is consistent with the double peak in SL frequency in May/June and September/October south of 12 N and the single peak in July/August in the Sahel [Omotosho, 1984]. Figure 6 also indicates that the preferred propagation direction is west-southwest, in agreement with previous findings [cf. Aspliden et al., 1976]. [24] Figure 7 displays the regions of SL generation (Figure 7a), decay (Figure 7b), and occurrence (Figure 7c). The intensity of the grid box shading in Figures 7a and 7b increases with the number of SL origins and decays per latitude/longitude grid box. The grid box counts of SL frequency in Figure 7c were similarly obtained, except that the crossing of leading edges over 1 1

8 ACL 5-8 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES Figure 6. SL tracks for the 12 months May October 1998 and The blue tracks are SLs appearing west of the AEW trough, the red tracks indicate SL that formed within the AEW phase interval 5 and 7 north of 12.5 N. The shaded areas are displayed for orientation. The digits in the top left-hand corner denote the number of SLs for the respective months. A total of 344 SLs were identified within the two six-month periods. latitude/longitude boxes at any time of the SL life cycle was counted. In other words, a crossing was counted for each 1 1 grid box lying partly or entirely within the area affected by the SL, whose determination is exemplified by the light dashed envelope in Figure 3c. According to Figure 7a, three preferred regions of SL generation are discernible. The first is situated over the Nigerian Jos Plateau (7 10 E) where surface heating over elevated terrain and forced ascent of the southwesterly monsoon flow help to generate SLs from afternoon convection. The Jos Plateau has been identified as a major source region of SLs using both ground [Omotosho, 1984] and satellite IR data [Aspliden et al., 1976; Rowell and Milford, 1993]. Consistent with these studies, a second maximum of SL generation in the vicinity of the Greenwich Meridian (GM) is perceivable from Figure 7a. A third maximum not mentioned in the studies cited above is observed around 10 W. As is evident from the lack of a 500 m altitude contour in Figure 7, a major range of hills or mountains is clearly absent around the GM. In contrast, the presence of the flat Guinea rise and its southern extension, the Fouta Djalon mountains (see also Figure 1), that together form the watershed between the Niger and the Senegal/Gambia river catchments, does not exclude a major role of orography around 10 W. It is interesting to note that Martin and Schreiner [1981] found that during GATE a maximum of cloud cluster genesis, of which 10% were SLs, existed in roughly the same area. In order to further clarify this point, the role of AEWs for forcing SLs in these regions will be examined in section 4.2. [25] The regions of maximum SL dissipation (Figure 7b) reflect to a large extent the mean lifetime and propagation characteristics of the SLs observed. Taking into account their mean direction of propagation west-southwest, their mean lifetime of 12 h and their mean westward propagation speed of 15 m s 1 (section 3.2), the locations of maximum SL dissipation are expected to be approximately six degrees longitude (or a little more than two grid box distances in Figures 7a and 7b) to the west-southwest of the maxima of SL generation. Therefore, the broad SL lysis maximum around the GM is related to SLs that formed over and to the west of the Jos Plateau, and the coastal maximum is connected to the genesis maximum at 10 W, although increasing low-level stability over coastal boxes due to colder sea surface temperatures might also have contributed to the latter. Finally, the highest overall frequencies of SL occurrences are found west of the Jos Plateau, while a second extensive region of enhanced SL occurrence is situated over the Guinea rise and the headwaters of river Senegal (Figure 7c). In contrast, a third maximum west of the GM is not well pronounced, because this region also exhibit a large number of SL decays (Figure 7b). The fact that the eastern- and western-most genesis maxima are collocated with regions of relatively few SL dissipations further supports the role of orography in these areas. [26] The diurnal cycle of SL genesis and lysis, represented by the number of SL generation and dissipation events, within eight three-hourly intervals between 0 3 UTC and UTC, is shown in Figure 8. The times in UTC are very close to the local solar time in this part of the tropics. Most SLs (52%) are triggered between 15 and 21 UTC (Figure 8a) and decay (49%) between 3 and 9 UTC (Figure 8b) in agreement with the mean lifetime of 12 hours. It should be mentioned that the majority of the SLs were generated between 15 and 19 UTC and that only few SLs were triggered between 19 and 21 UTC. A considerable portion (35%) of SLs develop between 21 UTC and 12 UTC, and are, thus, not the direct result of surface heating. The potential role of AEWs for the organization of these nighttime and early morning SLs will be examined in section 4.3. [27] Figure 9 depicts the relation between AEW phases and the genesis and lysis of the 344 long-lived SLs detected Figure 7. (opposite) Grid counts of (a) SL generation and (b) decay for the periods May October 1998 and Counts for each latitude/longitude grid box were incremented by one, if parts of the leading edge of a SL were located within the box at the time of its (a) first and (b) last detection. (c) Grid counts per 1 1 latitude/longitude grid box of SL passages at any time of its life cycle. The light solid contour depicts the 500 m altitude isohypse.

9 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES ACL 5-9

10 ACL 5-10 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES Figure 8. Histograms showing the diurnal cycle of (a) SL generation and (b) decay for the periods May October 1998 and The diurnal cycle is represented by the numbers of SL generations/decays within eight three-hourly intervals between 0 3 UTC and UTC. The sample size is 344 (333) for genesis (lysis). Eleven SLs moved out of the investigation domain and dissipated west of 20 W. area of strongest southerly winds (phase interval 6 7, Figure 9a) that will be discussed in some more detail in section 4.2. SLs tend to decay in the vicinity of the ridge (40% of 225 SLs) between wave phase intervals 7 8 and 8 1 (Figure 9b). Consistently, Payne and McGarry [1977] found that SLs often form west of the wave trough axis and terminate just to the east of the preceding ridge. From the phasing of the lysis peak, it is tempting to conclude that SLs move into the ridge area due to a higher propagation speed than the AEWs and dissipate therein because of the unfavorable environmental conditions. Reed et al. [1977], for example, found weak synoptic descent at 850 hpa within the ridge phase interval 7 1, with the strongest descent at 15 N. This sinking motion is also present in the wave composites presented in Duvel [1990], who showed waverelated temperature perturbations at 850 hpa that would tend to increase the dry static stability of the lower troposphere mainly to the west, but also within the ridge area. [29] Some insight into this problem can be gained by looking into the origin statistics of SLs that dissipate in the ridge. In our sample, only 25% of the SLs emerging west of the trough (between phases 2 and 4) decay in the ridge (phase interval 7 1). Assuming a typical wavelength of 2500 km and taking into account the mean AEW (SL) speed of 9.1 m s 1 (15 m s 1 ), 14.7 h are needed for a SL to traverse the 1/8 wavelength distance from phase 2 to 1 and, thus, to arrive at the ridge. Since the mean SL life time is only 12 h, the above mentioned percentage of 25% can only be explained by variations in wavelength and propagation speed of SLs and/or AEWs. From the 90 SLs dissipating in the ridge, 26 (equivalent to the aforementioned 25%) originated between phases 2 4, 21 emerged between wave phases 1 2, 20 evolved east of an AEW, moved into the ridge area of a preceding AEW and dissipated therein, and 18 SLs appeared and disappeared in the ridge phase intervals 7 8 and 8 1. The genesis points for the remaining 5 SLs were assigned to AEW phase intervals 4 5 and 6 7. While the two peaks in the phase distribution of SL genesis in Figure 9a point to an influence of the waves on SL generation, the corresponding peak in the lysis statistics for the ridge phases (Figure 9b) alone do not allow such a over West Africa between May and October 1998 and The assignment of SLs to AEWs phases was described in section 3.3. A total of 227 (225) SLs evolved (dissipated) in association with the presence of an AEW. Thus, 34% of the SLs investigated are unrelated to AEWs or could not unambiguously be assigned to a wave phase. The fact that more SL generations were assigned to AEW phases than decays is mainly due to SLs which decayed west of the region examined, i.e., west of 20 W. This number is partly compensated by SLs that were detected in an area of absent AEW activity within the investigation domain, caught up the precursory AEW ridge due to the higher propagation speed and disintegrated in the ridge area. [28] From the percentages of SL generation in the eight wave phases (Figure 9a), it is evident that, in accordance with the results of Payne and McGarry [1977], a preferred area of SL initiation exists west of the trough between wave phases 2 and 4. In these two wave phases, 48% of the 227 AEW-related SLs formed. A secondary peak occurs in the Figure 9. Histograms showing the relation between AEW phases and (a) SL generation and (b) decay for West Africa and for the periods May October 1998 and The dashed line represents the AEW trough axis. The total number of SL generations/decays in association with an accompanying AEW is shown in the top right-hand corner (for details see text).

11 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES ACL 5-11 conclusion; it remains unclear if the lysis peak in the ridge is the simple result of SLs arriving here at the end of their life time or a consequence of environmental conditions less conducive to SL maintenance Spatial Variability [30] In this section, we will describe to what extent the impact of AEWs on SL generation and decay varies from east to west and from south to north. Earlier studies by Burpee [1974] and Duvel [1990] have found a phase opposition in the AEW-rainfall/cloudiness relationship between areas north and south of 12.5 N (section 1). Therefore, the statistics of the AEW-SL genesis/lysis relationship were partitioned in Figure 10 into areas north (henceforth alternatively termed the Sahel) and south (also referred to as the Guinea Coast/Soudanian regions) of 12.5 N. The differences between the absolute number of SL appearances and disappearances in the respective region (top right-hand corner of Figure 10) are to a large extent related to the most frequent west-southwestward propagation direction of SLs: i.e., some SLs developed in the Sahel, but dissipated south of 12.5 N. Moreover, some SLs have decayed west of the region examined (cp. section 4.1). [31] As a first result, we note that 61% (73%) of the 196 (148) SLs that formed south (north) of 12.5 N were associated with AEWs. The larger percentage in the Sahel probably reflects that here the maximum of SL occurrences coincides with the period of strongest and most sustained AEW activity between July and September. Moreover, Figure 10 reveals that south of 12.5 N, SLs are mainly organized west of the trough between categories 2 and 4 (65% of the 119 SLs), whereas in the Sahel two peaks are observed: one west of the trough (42% of the 108 SLs), the other in the area of maximum southerly wind between phases 5 and 7 (32% of the 108 SLs). The latter peak suggests that, in the arid Sahel, the increase in the conditional instability of near-surface parcels by the AEWinduced moist low-level southerlies around phase 6 can overcompensate the adverse wave-related factors of lowlevel weak subsidence and cooling west of the AEW ridge. However, the AEW will never directly trigger a SL, but instead will create a region of higher CAPE between phases 5 7 in the Sahel, thereby increasing the likelihood of SL initiation. Similarly, the AEW-induced vertical velocities west of the trough are far to small to overcome the convective inhibition (CIN). In observations, that include the effects of latent heat release on vertical motion, ascent rates are on the order of 1.4 cm s 1 [cf. Reed et al., 1977]. More likely, the AEW will modulate the thermodynamic and low-level wind profiles that make SL triggering and organization more likely west of the trough. A more indepth investigation of this aspect is, however, beyond the scope of the present paper. [32] In view of these arguments and the excess of SLgenesis within the aforementioned AEW-phases, we will consider SLs forming in AEW phase intervals 2 3 and 3 4 and north of 12.5 N, additionally between phases 5 and 7, as being AEW-forced SLs. In other words, it is presumed that SLs owe their existence to the passage of an AEW. In order to test this assumption statistically, a Monte Carlo simulation was performed, that consisted of experiments in which the total of 227 SL events and the 108 Figure 10. As in Figure 9, but SL generation/decay statistics for the regions (a) north and (b) south of 12.5 N are displayed separately. northern SL events, respectively, were randomly assigned to eight wave categories to obtain a frequency distribution complying with the null hypothesis that the waves have no impact on SL generation. For the entire West African sample, the Monte Carlo tests reveal that the null hypothesis can be rejected with 99% significance for phases 2 3 and 3 4 (cf. Figure 9a). This statistical confidence was also obtained in the reduced Sahelian sample for wave category 6 7, but not for 5 6 (Figure 10a, left panel). In this phase interval, a substantial number of Sahelian SLs originated in 1998, but not in The statistics for 1998 and the physical arguments given above were the motivation to include the concerning 11 SLs in the sample of 145 AEWforced SLs. [33] For West Africa as a whole, it was found that 42% of all 344 SLs detected are AEW-forced SLs. In Figure 6, the tracks of these AEW-forced SLs are colored in blue for SLs that formed west of the trough and in red for SLs that appeared in the Sahel in the AEW phase interval 5 7. If West Africa is divided into the Sahel and Guinea Coast/ Soudanian regions, the percentages of AEW-forced SLs are 54% and 33%, respectively. It is critical to compare the latter two numbers from the statistical point of view, since twice as many wave phases were considered in the Sahel. Nevertheless, the fact that two different mechanisms, that favor SL genesis and generation, act in two wave phase intervals in the Sahel led us to conclude that AEWs have a larger impact on SL generation here than farther south. [34] The positions of the origin points of AEW-forced SLs (defined as the center points of the leading edges) relative to the simultaneous locations of the associated southerly and northerly AEW vortices are plotted in the longitude/latitude scatter diagram shown in Figure 11. As expected, SLs, that originate west of the AEW trough axis (between phases 2 and 4), mostly developed to the north-

12 ACL 5-12 FINK AND REINER: AFRICAN EASTERLY WAVES AND WEST AFRICAN SQUALL LINES Figure 11. Scatterplots of the positions of origin points of SLs relative to the simultaneous locations of the accompanying (a) northerly and (b) southerly AEW vortices. The abscissa (ordinate) denotes the longitudinal (latitudinal) distance with negative values indicating that the SL origin is west (south) of the respective vortices. The open circles represent SLs that formed west of the AEW trough in phase intervals 2 3 and 3 4, while the stars denote SLs assigned to AEW phase intervals 5 6 and 6 7 north of 12.5 N. The mean positions of either class of SLs are indicated by the large bold circles and stars, respectively. Both diagrams cover the periods May October 1998 and west of the southerly AEW and to the southwest of the northerly AEW (open circles in Figures 11a and 11b, resp.). However, the many open circles in the lower-right quarter of Figure 11a reflect the fact that various SLs originated east of the longitude of the associated northerly AEW vortex. For all this cases, the southerly AEW vortex must be located east of its northerly counterpart (otherwise the SL would not have formed west of the trough axes). As is evident from the number of open circles in the top-right quarter of Figure 11b, the inverse relationship holds for a fewer number of cases, and, thus, the mean relative position of SLs originating west of the trough is only 1.6 degrees longitude west of the northerly vortex, while the corresponding distance is 3.8 degrees longitude west of the southerly vortex (bold circles in Figures 11a and 11b, resp.). The corresponding individual and mean relative positions of the AEW-forced SLs that occur in the Sahel between phases 5 and 7 are indicated by asterisks in Figures 11a and 11b. The ensemble mean relative position of these SLs is located 10.9 (9, 5) degrees longitude east of the southerly (northerly) AEWs, thus, indicating that the northerly AEW vortex is more often located east of the longitude of the accompanying southerly vortex, contrary to what is observed for SLs appearing west of the trough. The mean latitudinal position for the SLs forced west of the AEW trough is 12 N, while for SLs forced in the area of maximum southerlies it is 15 N. The corresponding mean southerly (northerly) AEW vortex positions are around 9 N (17 N). This means that the SLs mostly formed in the AEJ region with strong low-level vertical wind shear, in between the two vortices. [35] The influence of AEWs on SL decay is not as obvious for the region south of 12.5 N as for the Sahel (right panels in Figure 10), where SLs frequently (47%) decay in the two ridge phase intervals 7 8 and 8 1. It is again tempting to speculate that the larger number of ridge dissipations in the Sahel is a result of the stronger adverse effects on SL maintenance prevailing in this region due to a more intense cold air advection and subsidence in the ridge (cf. Duvel, 1990). However, a close inspection reveals that 11 SLs moved from an AEW-free area to the east into the precursory AEW ridge and far more than 50% of the 45 ridge dissipations could also be explained by the SL genesis points, the mean SL lifetime, and the propagation speed of AEWs and SLs. These arguments have already been explained in more detail at the end of section 4.1. For the region south of 12.5 N, the secondary peak in the lysis frequency west of the trough axis (Figure 10b) reflects the fact that SLs were often generated west of the trough near the west coast, but decayed in the same wave phase as they propagated to the Atlantic, where cold sea surface temperatures inhibit the maintenance of squall clusters (cp. section 4.1). [36] In order to gain further insights into the spatial variability of the influence of AEWs on the life cycle of SLs, we investigated the frequencies of SL disintegrations (occurring within the ridge phase intervals 7 8 and 8 1) and AEW-forced SL initiations for 5 degrees longitude strips between E and W. As can be seen from Figure 12a, the absolute number of AEW-forced SL origins (solid line), as well as SL disappearances in the ridge (dashed line) increases from 20 E to the west coast. Figure 12b reveals that only 20% of the SL initiations were AEW-forced according to our definition at the longitudes of Lake Chad ( E), while this number increases to 68% at the West African coast. The importance of AEWs for SL generation west of the GM is consistent with the observed growth of the wave amplitude in terms of peak values of bpf vorticity toward the coast (not shown). In the region west of the GM, more than 50% of the SLs are likely forced by AEWs and this percentage exhibits the strongest overall increase from 34% to 54% at the GM (Figure 12b). Hence, a dominant role of synoptic wave activity for the SL genesis maxima around the GM and 10 W (see Figure 7a) is