The 6 9 day wave and rainfall modulation in northern Africa during summer 1981

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D17, 4535, doi: /2002jd003215, 2003 The 6 9 day wave and rainfall modulation in northern Africa during summer 1981 David Monkam Département de Physique, Faculté des Sciences, Université de Douala, Douala, Cameroon Received 25 November 2002; revised 13 May 2003; accepted 2 June 2003; published 4 September [1] Zonal and meridional wind components and geopotential height from European Centre for Medium-Range Weather Forecasts model analyses and daily rainfall data from the Institut Français de Recherche Scientifique pour le Développement en Coopération (ORSTOM) are used to study westward propagating 6 9 day waves and rainfall modulation in northern Africa during summer The 6 9 day wave structure is determined using a composite method. In this structure, there are two vortices of opposite circulation on either side of the latitude 12.5 N. The rainfall maxima are associated with cyclonic vortices and the rainfall minima with anticyclonic vortices, coinciding with the minima and the maxima of geopotential height anomalies, respectively. The composite variability shows that the 6 9 day wave is associated with positive rainfall anomalies in West Africa in the band of latitude N, in the western part of the area around Senegal and Guinea and in the center toward Lake Chad. The rainfall anomalies are linked to the zonal wind anomalies, and the increase in rainfall is associated with large modulation of the African Easterly Jet zonal wind component, mainly in the cyclonic circulation. The main zones of decreasing rainfall appear north of N, toward Sudan, and south of 8 N, near Ivory Coast. INDEX TERMS: 3309 Meteorology and Atmospheric Dynamics: Climatology (1620); 3364 Meteorology and Atmospheric Dynamics: Synoptic-scale meteorology; 3374 Meteorology and Atmospheric Dynamics: Tropical meteorology; 3384 Meteorology and Atmospheric Dynamics: Waves and tides; KEYWORDS: 6 9 day wave, rainfall, data anomalies, vortices, high (H), low (L), cyclonic, anticyclonic Citation: Monkam, D., The 6 9 day wave and rainfall modulation in northern Africa during summer 1981, J. Geophys. Res., 108(D17), 4535, doi: /2002jd003215, Introduction [2] Most studies concerning undulatory phenomena in northern Africa and the tropical Atlantic troposphere have concentrated on easterly waves, commonly called African waves. Reed et al. [1977] have shown the existence of perturbations traveling from east to west in West Africa and the tropical Atlantic and have established relations between them and rainy/thundery phenomena. Burpee [1972, 1974, 1975], Reed et al. [1977], and Duvel [1990] found that the wavelength of African waves is km, with a phase velocity of 8 m s 1. African waves are visible mainly at 700 and 850 hpa levels. [3] However, African waves are not the only undulatory phenomena in this region; Viltard and de Felice [1979, 1982], using GARP Atlantic Tropical Experiment (GATE) data from land stations, found a large peak in the 6 9 day band period in the power spectra of surface pressure, averaged over 25 stations in northern Africa. De Felice et al. [1990], using European Centre for Medium-Range Weather Forecasts (ECMWF) analyses, have shown that lower-troposphere and midtroposphere waves of 6 9 day Copyright 2003 by the American Geophysical Union /03/2002JD periods moved westward during summer in northern Africa. At 700 hpa the waves had a westward velocity of 7 longitude d 1 and a wavelength of km. Diedhiou et al. [1999, 2001] analyzed the characteristics of African and 6 9 day waves and their interactions with rainfall and convection over West Africa using ECMWF and National Center for Environmental Prediction/National Center for Atmospheric Research reanalyses from 1979 to The mean wavelengths and phase velocities found were similar to those given above. Rainfall and convection were significantly modulated by the African waves, with convection in the Intertropical Convergence Zone (ITCZ) being enhanced in the trough. In comparison to the 3 5 day waves, the authors mentioned the intermittency of the 6 9 day waves, which also significantly affect rainfall but in a different way, inducing large zonal convective bands in the ITCZ. However, these results were obtained from mean composite wave patterns over a period of 17 years and would not be identical each year. For example, since they are intermittent, differences in rainfall modulation may occur between years when these 6 9 day waves are very active and other years. Viltard et al. [1998], using a composite analysis, established a link between these 6 9 day wave patterns and rainfall during summer 1989 within the area N, 17.5 W 20 E, in West Africa. Rain- ACL 5-1

2 ACL 5-2 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA fall was strongly modulated in the most northern (17.5 N) and most southern (7.5 N) bands. However, these results did not show the longitudinal modulations of rainfall, and Viltard et al. [1998] did not extend their study to other years. Composites of rain amount were based on data for summer 1989 only because, as the authors state, this was the year that the 6 9 day oscillations were clearly visible and that the rain amounts were disposable. [4] The present paper uses data from 6 years, : zonal and meridional wind components and geopotential height from ECMWF model analyses and daily rainfall data from the Institut Français de Recherche Scientifique pour le Développement en Coopération (ORSTOM). We examined the three parameters of the ECMWF analyses for the six summers (June September) in some detail, and it was noted that these waves were very active during the summer of 1981 but were also found during the other 5 years. So we propose to study the rainfall modulation by the 6 9 day waves during summer 1981 and examine the spatial variability of these waves in connection with rainfall in West Africa. We also try to make a contribution to how their activities that can be related particularly to the number of amplified waves can influence or affect rainfall in West Africa. In section 2 we describe the data used. In section 3 we proceed to the spectral and composite analyses to show evidence of the waves and determine their patterns in summer In section 4 we study composite rainfall in connection with the wave patterns and present the main results of the modulation. In section 5 we investigate the variability of 6 9 day waves and rainfall in West Africa. We study this variability on the intraseasonal timescale and compare it to the case of the African wave. The interannual variability of 6 9 day waves in connection with rainfall is examined throughout the six summers. 2. Data [5] We use ECMWF analyses for six summers: After computing the power spectra of certain meteorological parameters (wind components and geopotential height) we found that on the whole the peaks in the 6 9 day period were most prominent during summer So the details of some of our results were obtained from summer 1981, with daily (1200 UTC) analyzed wind and geopotential height fields at 850, 700, and 500 hpa levels on a grid, between the equator and latitude 30 N and the longitudes 20 W and 30 E. The study of the interannual variability of 6 9 day waves in connection with rainfall is made from summer 1981 to summer Reed et al. [1988a, 1988b] assessed the performance of the ECMWF system in analyzing easterly wave disturbances over Africa and the tropical Atlantic, and Laurent et al. [1989] have proved the performance evaluation and local adaptation of the ECMWF system forecasts over northern Africa in summer; hence we thought that the ECMWF model analyses data would also be convenient for investigating 6 9 day waves. [6] A daily rainfall data set compiled by ORSTOM is used in the study. The compilation consists of 1000 stations located in the domain 5 22 N and 17.5 W 22.5 E. The daily values have been interpolated in space on the ECMWF grid by assigning each station daily value to the nearest grid point and averaging all the values related to each grid point. They have also been interpolated in time, related to ECMWF daily wind fields since daily rainfall amounts were measured between 0600 and 0600 UTC of the following day. We had daily rainfall data from 1 June to 30 September (122 values) for the six years of the study ( ). The study with rainfall data is limited to the domain N, 17.5 W 22.5 E, where these data are regular on the ORSTOM rainfall data files. In this domain the mean fields of rainfall for the six summers had the same overall tendency. So we briefly present the results of summer 1981 (Figure 1). The mean rainfall pattern during this period consists of a rainband of 6 8mmd 1 at N, up to more than 16 mm d 1 toward the Fouta-Djallon Mountains along the southwest coast of West Africa, and a rapid rainfall decrease northward, leading to a mean amount of 2 3 mm d 1 toward 15 N. South of Lake Chad the maximum value is 7mmd 1. The ITCZ is located at its most northern position over West Africa during the rainy season in the Sahel. Finally, on the whole the rainfall mean field during summer 1981 shows that in this area, in particular, the deep convection in the ITCZ is at N. 3. Spectral and Composite Analyses: The 6 9 Day Wave Pattern 3.1. Evidence of the 6 9 day Wave in Summer 1981 [7] Evidence of large peaks in the 6 9 day band period is shown by means of spectral analysis using two different methods: The maximum entropy method and fast Fourier transform (FFT) are used to display the energy distribution over frequency. We applied these methods to compute the power spectra of zonal and meridional wind components at different grid points for the six summers, and we found that, in general, the peak of the 6 9 day period was very large on the spectra of the zonal wind component. The peak also appeared on the spectra of the meridional wind, but it was smaller than in the zonal wind spectra. Figure 2 is an example obtained for the zonal wind component at 700 hpa level at the grid point 15 N, 12.5 W in summer In Figure 2 the largest peak is at 7.2 days, while the peak of the 3 5 day period (African waves) is weak. [8] FFT also gives the phase spectrum (not shown), which is the phase distribution over frequency. Using the results of the phase distribution, we studied the wave propagation. For this purpose we selected the phase corresponding to the frequency 0.14 d 1 or (7.2 days) at each grid point at the 700 hpa level. Then we drew the Figure Mean field of rainfall (in mm d 1 ) for summer

3 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA ACL 5-3 Figure 2. Zonal wind power spectrum at 700 hpa for summer 1981 at the grid point 15 N, 12.5 W. isophases, from which we estimated the wave propagation in the studied area and its kinematic characteristics. The isolines obtained were approximately parallel to the meridian and increased from east to west of the studied area, showing westward propagation. From the isophases we computed the wavelength, and the value was km, which is around twice the wavelength of African waves. Timelongitude diagrams were used to study wave propagation at different latitudes. Time-longitude diagrams (not shown) obtained with the filtered zonal wind component, with a band-pass filter centered at 7.2 days at 700 hpa from the latitudes 5 20 N (seven latitudes), show that the waves are westward propagating. The mean velocity computed is 7 of longitude d 1. The corresponding mean wavelength is km. The band-pass filter centered on the frequency 0.14 d 1 [BMDP Statistical Software, 1985] removes periods longer than 12 days and shorter than 5 days. [9] A composite method is used to determine the wave pattern. This method has been described in some detail by Monkam [1990] and de Felice et al. [1990]. Composite methods had been used by Reed and Recker [1971] and by Burpee [1975]. Our method is very close to that of these authors. We choose the parameter, the level, and the latitude where the 6 9 day wave had large or rather very large amplitudes as the reference parameter, level, and latitude. We apply the band-pass filter centered around 7.2 days to the zonal wind component at each grid point of the studied area. In fact, the waves did not have large amplitudes during the whole summer. However, between 5 July and 12 September, there are 8 10 regular and rather large waves in the band of latitudes 5 20 N. An example illustrating this is visible in Figures 3a 3c, displaying the filtered and unfiltered zonal wind components at 700 hpa level for three grid points: 15 N, 2.5 W; 12.5 N, 7.5 W; and 10 N, 2.5 W in Figures 3a, 3b, and 3c, respectively. In Figures 3a 3c the 6 9 day waves appear clearly on the unfiltered zonal wind component (bottom curve). The 6 9 day waves (bold curve) have large or rather very large amplitudes in the Figure 3. Zonal wind component (positive eastward) at 700 hpa level at the grid points (a) 15 N, 2.5 W; (b) 12.5 N, 7.5 W; and (c) 10 N, 2.5 W during summer Shown are unfiltered data from the European Centre for Medium- Range Weather Forecasts model (bottom curve), data filtered by a band-pass filter at 7.2 days from day 12 (12 June) to day 111 (19 September) (bold curve), and data filtered by a band-pass filter at 3 5 days (thin curve).

4 ACL 5-4 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA middle and at the end of summer in Figure 3a, in the middle of summer in Figure 3b, and during the whole summer in Figure 3c. The maximum amplitude is 4 5 m s 1.In Figures 3a 3c the amplitude of the African wave (thin curve) is not very large (2.0 m s 1 ). [10] Like in Figure 3b, at many other grid points of the latitude 12.5 N the waves were regular and had a large amplitude (2 m s 1 ) in July and August and at the beginning of September. So we selected the interval from 5 July to 12 September as the reference time interval for the compositing method. For these 6 9 day waves the reference parameter is the filtered zonal wind component at the reference level 700 hpa and the reference latitude 12.5 N. The compositing is accomplished by dividing each wave into eight categories determined from the filtered zonal wind component series at 700 hpa level and 12.5 N latitude in the reference time interval defined above. Category 1 dates are associated with maximum westerly wind and category 5 dates with maximum easterly wind. The dates of categories 2, 3, and 4 are at one quarter, one half, and three quarters of the interval between category 1 and the following category 5, respectively. The dates of categories 6, 7, and 8 are at one quarter, one half, and three quarters of the interval between category 5 and category 1 of the following wave, respectively. After estimating the value of each parameter for each category, for all the waves of the time interval from 5 July to 12 September, the mean values are computed for each grid point over all the waves of this time interval. So, as there are 10 waves of a 6 9 day period in this time interval, at each grid point a mean timecomposite wave is obtained, where the value at each category is a mean of 10 values. Then a mean composite wave is computed at each latitude over the grid points of the latitude. With 10 waves in the time interval and 17 grid points at each latitude between 17.5 W and 22.5 E, each composite value is a mean of = 170 values. [11] We applied this composite method to the rainfall data by considering dates of categories determined from the filtered zonal wind component at the 700 hpa level and latitude 12.5 N. The rain data series were obtained as described in section Wave Patterns [12] The wind component anomalies at each grid point were computed by subtracting the mean of summer. The composite method described in section 3.1 was applied to these data to obtain a mean 6 9 day wave pattern on the wind field. The results are displayed in Figure 4 for the 700 hpa level. The maximum wind perturbation caused by the 6 9 day wave is 4.0 m s 1 and appears toward the band of latitudes N. Figure 4 shows that the zonal wind component was strongly affected by the 6 9 day wave, while the meridional wind component was very weak, in particular around the reference latitude 12.5 N (v 0). Toward lower latitudes in the studied area (near the equator) the meridional wind component was strongly affected by the 6 9 day wave. This must be linked to the 6 9 day wave contribution to water fluxes across the equator, as mentioned by Cadet and Houston [1984] and Cadet and Nnoli [1987] and as confirmed by Monkam [1990]. The wave pattern is characterized by two systems of vortices south and north of 12.5 N. In the north, there is a cyclonic vortex Figure 4. The 6 9 day composite wave for wind vectors and geopotential height anomalies at 700 hpa level during summer The maximum wind vector is 4.0 m s 1. The isolines of geopotential height anomalies are in m, and the maximum (minimum) value is around +3 m ( 6 m). H stands for high, and L stands for low. from categories 8 to 2 and an anticyclonic vortex from categories 4 to 6. The two vortices have their zonal axis near the latitude 17.5 N. In the south, there is an anticyclonic vortex from categories 8 to 2 and a cyclonic vortex from categories 4 to 6. The cyclonic and the anticyclonic vortices coincide with the low (L) and the high (H) of the composite geopotential height anomalies, respectively (Figure 4). At 500 hpa and 850 hpa (not shown) the wind vector patterns are almost identical to those at 700 hpa, with large zonal wind components at 12.5 N and large meridional components south of 7.5 N and north of 17.5 N. [13] The pattern of the 6 9 day wave is very different from the pattern of the African wave. The same method was applied to determine the African wave structure. The reference parameter was the filtered meridional wind component (using a band-pass filter for 3 5 days) at 700 hpa and 12.5 N. Category 1 dates are associated with the maximum northerly filtered meridional wind component, and category 5 dates are associated with the maximum southerly filtered meridional wind component. The other category dates are determined following the procedure described for the 6 9 day wave. In the mean composite African wave (not shown), there is only one system of two vortices of opposite circulation, with a common zonal axis at 12.5 N. [14] The perturbed wind vectors composited for the 6 9 day waves (Figure 4) were obtained from the wind component anomalies. The filtered zonal wind data were used only to determine the dates of categories. We think that the 6 9 day waves exist. By compositing 170 values, an amplitude of 3.5 m s 1 is obtained on the zonal wind component when the wind mean velocity is 8 ms 1 at N and at 700 hpa. This result suggests that the 6 9 day wave-like disturbance pattern is rather robust. 4. Composite Rainfall Pattern on the 6 9 day Wave 4.1. Composite Rainfall [15] The composite method was applied to rain data series at different grid points. We recall that the dates of categories used to compute the rain values on the composite 6 9 day waves were determined from the filtered zonal wind at 700 hpa and 12.5 N. Briefly, for each grid point the rain

5 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA ACL 5-5 value for category 1 (5) corresponds to the day of maximum westerly (easterly) filtered zonal wind component. Likewise, rainfall values for categories 2 (6), 3 (7), and 4 (8) correspond to the day for dates at one quarter, one half, and three quarters of the interval between category 1 (5) and the following category 5 (1) in the reference time interval, respectively. A mean rain composite value is obtained at each grid point, for each category, over all the waves in the reference time interval. Finally, at each latitude, for each category, the mean composite value is computed over the 17 longitudes of the studied area. [16] We computed the composite daily rainfall for the 6 9 day wave in the studied area (not shown). South of 12.5 N the rainfall is maximum (9.3 mm d 1 )at N on categories 4 6 and minimum (4 5 mm d 1 ) on categories 8 2; north of 12.5 N, there is an opposition of phase comparatively to the south, with a minimum rainfall (0.5 mm d 1 ) at N on categories 4 6 and a maximum (1.5 mm d 1 ) on categories 8 2. [17] The tendency is systematically reproduced on the composite rainfall anomalies (not shown), with a maximum anomaly (2.70 mm d 1 )at N on categories 4 6 and a minimum ( 3.0 mm d 1 ) on categories 8 2. In the north, there is a maximum anomaly (0.4 mm d 1 )at N on categories 8 2 and a minimum ( 0.3 mm d 1 ) on categories 4 6. [18] The rainfall is strongly modulated at N and N, though the values are weak at N. This weakness of the composite values of rainfall anomalies is due to the fact that the rainfall is very erratic and the mean field in the studied area, as was described in section 2 (Figure 1), has strong values in the south (8 16 mm d 1 at N) and weak values in the north (1 2 mm d 1 at N). [19] Therefore, to illustrate the rainfall modulation, we computed the rainfall percentage at each grid point. The method of computation is close to the one used by Viltard et al. [1998]. The daily rainfall percentage is computed at each grid point by dividing the daily rainfall by the mean rainfall computed for the whole summer for the given grid point; the result of this operation is multiplied by 100 to obtain a percentage. The value of the rainfall percentage is relative to each given grid point. For example, 50% at a grid point near 7.5 N cannot correspond to the same rainfall amount as 50% for a grid point near 17.5 N. [20] The rainfall modulation is easily visible on the mean composite of these percentages (Figure 5). It is clear that the latitudes N and N are the two regions where the rain is strongly modulated by the 6 9 day waves. At 17.5 N, there is a rainfall percentage maximum of %, corresponding to 2.5 mm d 1 on categories 8 2 (8, 1, and 2), and a minimum of 60 80%, corresponding to 1.2 mm d 1 on categories 4 6 (4, 5, and 6). At 7.5 N the maximum is 140%, corresponding to 9 mm d 1 for categories 4 6, and the minimum is 40 60%, corresponding to 3 mm d 1 for categories Connection Between the 6 9 day Wave and Rainfall Patterns [21] Figure 4 displays the 6 9 day wave pattern on the perturbed wind vectors and the isolines of geopotential height anomalies, with L for low and H for high. In Figure 5. The 6 9 day composite wave on the local percentage of rainfall data during summer Figure 4, north of 12.5 N, there is a cyclonic circulation between categories 7 and 3, with the center near the low (L) at N, and an anticyclonic circulation between categories 3 and 7, with the center near the high (H) at N. The southern part of the latitude 12.5 N isin opposition of phase with the northern part. South of 12.5 N, there is an anticyclonic circulation between categories 7 and 3, with the center near the high (H) at N, and a cyclonic circulation between categories 3 and 7, with the center near the low (L) at N. [22] North of 12.5 N, at N, the centers of the cyclonic and anticyclonic circulations correspond to the maximum and minimum of rainfall, respectively (Figure 5). On the southern side of 12.5 N the centers of the cyclonic and anticyclonic circulations are toward N and correspond to the maximum and minimum rainfall, respectively (Figure 5). Hence in summer 1981 the rain was strongly modulated by the 6 9 day waves, mainly near the centers of the vortices and the extrema of the geopotential height anomalies. 5. Spatial Pattern of the 6 9 day Wave in Connection With Rainfall [23] The mean composite rainfall pattern displayed in Figure 5 shows that the rainfall is strongly modulated toward N and N. Figure 5 does not show the spatial variation in the meridional direction. It is a mean meridional composite wave at each latitude which, therefore, cannot display the longitudinal dependency. Furthermore, at some of the grid points of the reference latitude, in the reference time interval, there are some waves with weak amplitudes <2.0 m s 1 (Figure 3); at some other grid points of the reference latitude, there are also waves of large amplitude outside this time interval. So to see the effects of these waves on rainfall, it is convenient to consider the totality of strong waves over the whole summer. To study the variability of the 6 9 day waves and then examine their effects on the spatial variability of rainfall, we used another composite method close to the previous one but showing both latitudinal and longitudinal mean variations of parameters. [24] For this purpose, (1) we filtered the daily June September 700 hpa zonal wind component time series of each grid point with a band-pass filter between 6 and 9 days; (2) at each grid point we selected the days (dates or waves) when the filtered 700 hpa zonal wind component is maximum, with an amplitude 2.0 m s 1 ; for example, in

6 ACL 5-6 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA Table 1. Number of Maxima Having an Amplitude 2.0 m s 1 on the 700 hpa Filtered Zonal Wind Component in the 6 9 Day Band for Summers 1981 and 1982 a Longitude, deg Latitude, N 20W 17.5W 15W 12.5W 10W 7.5W 5W 2.5W 0 2.5E 5E Summer Summer a Two maxima define one wave. The number of maxima (or waves) are counted at each grid point from 12 June to 19 September. summer 1981 (Table 1), nine cases have been retained at the grid point 10 N, 17.5 W; (3) the latitude having the greatest number of cases is considered as the reference latitude. From the different dates or cases retained at each grid point (reference point) of the reference latitude we computed for each grid point of the studied area the mean anomaly values of a given parameter (zonal and meridional wind components, geopotential height, rainfall) over the number of cases of that reference point. Finally, after doing this for the different reference points of the reference latitude we computed the mean over all the reference points of the reference latitude. We think that by selecting a threshold of 2.0 m s 1 on different points at a reference latitude, we have a large number of cases of regular and well-amplified waves from which we can obtain stable or very stable mean composite fields of anomaly values of different parameters in the studied area. [25] To choose the area where the 6 9 day waves were active with large amplitudes, de Felice et al. [1990] computed the percentage of variance of these waves on the zonal wind component at each grid point of the domain 5 30 N, 17.5 W 40 E. They found large percentages in the band of latitudes N, west of longitude 2.5 W [see de Felice et al., 1990, Figure 2a]. They chose the latitude 12.5 N as their reference latitude. With our method the reference was 10 N, close to 12.5 N and belonging to the band N. Therefore the results are practically the same with the two methods Variability in Summer 1981 [26] Table 1 displays the numbers of days or dates when the filtered zonal wind component is 2.0 m s 1. At each grid point the number of days or dates contributes to defining the number of waves, as was done for the previous composite method, where 8 10 waves were selected in the time interval 5 July to 12 September. In Table 1 we presented only the results for grid points between the latitudes 5 N and 17.5 N, which, according to Reed et al. [1977], is inside the area where the African waves develop. In 1981 the latitude 10 N had the greatest number of dates. The reference latitude on which we considered 11 reference points from 20 W to5 E is10 N because to the east of 5 E the number of cases is small (<4). At some of the reference points, such as 10 N, 10 W and 10 N, 7.5 W, there are 11 dates selected, corresponding to 10 waves of a 6 9 day period, which is a very big number for series of 100 values. So these waves were very active during summer 1981, in particular to the west of the Greenwich meridian. At the reference latitude of 10 N, there are 99 cases for the 11 points. For a given parameter at a grid point a composite value can be a mean of 99 anomaly values. The numbers of dates selected with the filtered meridional wind component (not shown) were small compared to those obtained using a filtered zonal wind component. This confirms the choice of the filtered zonal wind component as the reference parameter and the fact that the 6 9 day waves are visible on the meridional wind component. The other numbers presented in Table 1 concern the summer of We shall deal with these quantities in sections 5.2, 6, and 7. [27] Figure 6 displays the spatial composite variability of wind vector anomalies with rainfall anomalies (Figure 6a) and with geopotential height anomalies (Figure 6b) for summer These perturbed wind vector fields were very similar to the filtered signal (not shown). The maximum wind magnitude anomaly is 4.5 m s 1. There is a large modulation of the zonal wind component, in particular to the west of the Greenwich meridian, with circulations of opposite signs on either side of the African Easterly Jet (AEJ) location (10 15 N, shown in Figure 11). Figure 6a also displays the rainfall anomalies. Negative values appear mainly between 5 N and 10 N and between 10 W and 7.5 E, with a mean core of about 3 mmd 1. Positive rainfall anomalies are observed mainly at N, W with a core of 5 mmd 1, in the west part of the studied area toward the zone of maximum rainfall observed on the mean field near the Fouta-Djallon Mountains (Figure 1), and at N, E with a core of 3.2 mm d 1, close to Lake Chad, where we observed a maximum rainfall on the summer mean field (Figure 1). So the 6 9 day waves can contribute to increased rainfall in West Africa, particularly toward southern Senegal and Guinea-Bissau (the mean rainfall is mm d 1 ) and near Lake Chad (the mean rainfall is 6 8 mm d 1 ). There are good or very good coincidences between the zones of maximum positive rainfall anomalies and the low (L) and between the core of negative rainfall anomalies and the high (H) of the composite geopotential height anomalies displayed in Figure 6b. [28] To examine the variability in the intraseasonal timescale in summer 1981, it was necessary to establish mean composite fields for every summer month, but the 6 9 day waves are not very amplified during every month of summer. The results obtained in Figure 6 correspond mainly to the months of July and August. For example, at the grid point 10 N, 10 W, there are 11 dates, 19, 26, 42, 48, 55, 62, 69, 76, 83, 91, and 98, when the filtered zonal wind component is 2.0 m s 1. Apart from dates 26 and 42 the difference between two successive dates is equal to 6, 7, or 8 days, belonging to the 6 9 day band. So two dates corresponding to one wave appear in June (19 and 26), and eight dates (seven waves) appear in July and August (42,

7 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA ACL 5-7 Figure 6. The mean composite variability of (a) wind vectors and rainfall anomalies at 700 hpa for the 6 9 day waves and (b) wind vectors and geopotential height anomalies, with H for high and L for low, in summer The composites were computed by selecting the dates when the filtered 700 hpa zonal wind is maximum and 2.0 m s 1 at 11 reference points of the reference latitude 10 N from 20 W to 5 E. The mean of wind and rainfall anomalies is computed at each grid point over the dates selected for each reference point. Finally, the mean is computed over all the reference points. Maximum wind magnitude anomaly V max =4.5ms 1. 48, 55, 62, 69, 76, 83, and 91). There is only one wave in September. The other reference points of the reference latitude have approximately similar repartitions. So, since the 6 9 day waves seem to increase rainfall in some zones in West Africa and as they are very active in July and August, they may contribute to the greatest rainfall values generally observed in summer or explain why rainfall is greater during these 2 months than in June and September. The contribution of the 6 9 day wave is evaluated for the 2 months of June and September At each reference point of the reference latitude we choose the dates when the filtered zonal wind component is maximum without condition on the amplitude (the maximum amplitude must be positive, so every wave is selected). For the 11 reference points, 166 dates were obtained. Figure 7 displays the composite wind vectors with rainfall anomalies (Figure 7a) and with geopotential height anomalies (Figure 7b). The maximum wind magnitude anomaly is 3.1 m s 1 (<4.5 m s 1 ). The tendencies for Figures 7a and 7b are practically the same as for Figures 6a and 6b, respectively, but their maxima are weak. So when the 6 9 day waves are not very active, their influence on rainfall is weak. Therefore we think that on the intraseasonal timescale the 6 9 day wave effects on rainfall are weak in June and September and strong in July and August. [29] Now we try to see the effects of 6 9 day waves on rainfall compared to the African waves of a 3 5 day period. For this purpose we chose the filtered meridional wind component in the 6 9 day and 3 5 day bands of period as the reference parameters for two regimes. We chose the latitude 5 N as the reference latitude because it was at that latitude that we observed a large number of waves for the two regimes. We selected the dates (or waves) for which the filtered meridional wind component was 1.5 m s 1 for the African waves and 1 m s 1 for the 6 9 day waves. The results of the composite wind vectors with rainfall anomalies and with geopotential height anomalies are displayed in Figures 8a and 8b for the 6 9 day waves and in Figures 8c and 8d for the African waves. There are 97 cases for the 6 9 day waves, with a wind vector maximum equal to 3.4 m s 1, and 139 cases for African waves, with a wind vector maximum equal to 2.6 m s 1. Figures 8a and 8b have good similarities to Figures 6 and 7, obtained with the filtered zonal wind component as a reference parameter. The circulation is of opposite signs on either side of the AEJ core in the band of latitudes N, inside which a strong rainfall Figure 7. (a and b) Same as in Figure 6, but the composites were computed by selecting the dates when the filtered 700 hpa zonal wind component is maximum; V max =3.1ms 1.

8 ACL 5-8 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA Figure 8. (a and b) Same as in Figure 6, but the composites were computed by selecting the dates when the filtered 700 hpa meridional wind in the 6 9 day band period is maximum and 1.0 m s 1 at 11 reference points of the reference latitude 5 N from 20 W to5 E; V max =3.4ms 1. (c and d) Same as in Figure 6, but for African waves. Dates were selected for filtered meridional wind (in the 3 5 day band period) maximum 1.5 m s 1 ; V max =2.6ms 1. increase is observed from east to west of the studied area. A maximum rainfall anomaly equal to 7.7 mm d 1 is observed at 10 N, 15 W. This maximum is probably associated with the Fouta-Djallon Mountain in Guinea. There is a strong anticyclone north of 15 N with a large meridional extension. For the African waves, there is a large meridional wind modulation with a counterclockwise circulation and with a very large meridional extension from the equator to more than 20 N, with a core around the AEJ location west of the Greenwich meridian. This cyclonic vortex, contrary to the 6 9 day waves, is associated with negative rainfall anomalies. The rainfall increase is observed mainly east of the Greenwich meridian, at E, near the Lake Chad location. Finally, the 6 9 day waves increase rainfall between 7.5 N and N, from east to west of the studied area, while the easterly waves increase rainfall mainly east of the Greenwich meridian. Similar results were obtained when the reference latitude was 10 N Interannual Variability [30] Now we examine interannual variability through the six summers For this purpose, since the 6 9 day waves were not very active during every summer (see, for example, Table 1 for summer 1982), at each grid point (reference point) of the reference latitude 10 N we selected the dates when the filtered zonal wind component is 1 ms 1. The total number of cases selected for the 11 grid points of the reference latitude are 122, 113, 121, 123, 96, and 106 for , respectively. Figures 9a 9f display the results of composite wind vectors and rainfall anomalies. For these summers the maxima of wind magnitude anomalies are 3.4, 2.2, 2.5, 2.3, 3.0, and 1.9 m s 1 for the summers of , respectively. In fact, we found 99, 44, 42, 40, 80, and 38 cases where the filtered zonal wind component is 2.0 m s 1 for , respectively. So if the waves selected at the reference latitude do not have an amplitude maximum, the amplitude of the composite wave will be too small. Waves were very amplified for summers 1981 and 1985 (V max =3.4ms 1 and V max =3.0ms 1, respectively). Viltard et al. [1997] had also noted the existence of active 6 9 day waves in summer These authors mentioned that the 6 9 day phenomena do not appear clearly each summer. The results of 1985 are similar to those described above for Waves were not very amplified during the summers of 1982, 1983, 1984, and For example, in 1984, among the 123 cases selected, there are only 40 dates when the filtered zonal wind component is 2.0 m s 1. Therefore the composite wave has a maximum wind magnitude anomaly of 2.3 m s 1. In these 6 9 day waves, rainfall anomalies are linked to the zonal wind anomalies. Figure 9c for summer 1983 displays a maximum of 2 3 mm d 1 at 7.5 N, E. In 1984 (Figure 9d) a rainfall maximum of 1 2 mm d 1 is observed at N, W. These maxima of rainfall anomalies in 1983 and

9 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA ACL 5-9 Figure 9. Mean composite of the 6 9 day waves for the summers of (a) 1981, (b) 1982, (c) 1983, (d) 1984, (e) 1985, and (f ) The composites were computed by selecting the dates when the filtered 700 hpa zonal wind component is maximum and 1.0 m s 1. Isolines of rainfall anomalies are in mm d 1. The wind magnitude is in m s are probably associated with the anomalies of the zonal wind component. In 1986, waves were not very active (V max =1.9ms 1 ) and their effects on rainfall were weak or seemed to be opposite to those of 1981 and We computed the mean fields of rainfall for the other five summers from 1982 to 1986 (not shown) as was done for summer 1981 (Figure 1). The mean rainfall values were weak in 1982, 1983, 1984, and 1986 compared to the results of summers 1981 and So we think that when the 6 9 day waves are active, they contribute, in general, to increased rainfall in West Africa in summer. 6. Discussion [31] In the composite method used for the wave pattern the category dates were determined from the filtered zonal wind component at 12.5 N and at 700 hpa level. This method implies that the wave features at other latitudes maintain a constant phase difference relative to those at 12.5 N. If it is not so, the amplitude of the composite wave is lower than what would be obtained by determining the category dates at each latitude. For example, the wind vectors are very weak south of 7.5 N and north of 20 N and are large between 7.5 N and 20 N in the African wave computed by Viltard et al. [1997] using the same method, with a reference latitude of 12.5 N. Duvel [1990] used a similar method to study the African waves and noted that this effect was very amplified when the latitude considered was far from the reference latitude and that the effect did not exist when the two latitudes were not too distant. Following Duvel [1990], since the waves of 6 9 day periods were also amplified on the meridional wind component, in particular

10 ACL 5-10 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA Figure 10. The 6 9 day composite wave for wind magnitude anomalies at 700 hpa level during summer The reference parameter and latitude are the filtered zonal wind component and 12.5 N for the latitudes N, respectively; in this band of latitude the maximum wind magnitude anomaly is 4.0 m s 1. The reference parameter and latitude are the filtered meridional wind component and the equator for the latitudes 7.5 S 5 N, respectively; in this band of latitudes the maximum wind magnitude anomaly is 2.9 m s 1. toward lower latitudes near the equator (the meridional wind component anomaly is about ±2.5 m s 1 at 5 N), we computed the composite wind vector anomalies for the 6 9 day waves between 5 N and7.5 S using the filtered meridional wind component and the equator as the reference parameter and the latitude, respectively. The result is shown in Figure 10, where the northern part ( N) corresponds to the part of Figure 4 obtained with the filtered zonal wind component with 12.5 N as the reference parameter. Between 5 N and7.5 S the maximum wind magnitude anomaly is 2.9 m s 1. So the 6 9 day wave is rather robust. The horizontal structure displayed by composite analysis (Figure 10) looks very much like the waves centered to the equator displayed by Matsuno [1966]. This likeness suggests that these 6 9 day waves correspond to a normal mode of the atmosphere. [32] African waves and 6 9 day waves are observed in the same season over the same area. Therefore it is important to examine the occurrence of the 6 9 day phenomena relative to the waves of a 3 5 day period. In Figures 3a 3c the African waves appear clearly on the filtered zonal wind component, though their amplitudes are not very large. The 6 9 day waves are not regular during the whole summer, and when they are weak, African waves are strong and vice versa. We counted the dates when the filtered zonal wind component is 2.0 m s 1, and we found 8, 9, and 11 cases for Figures 3a 3c, respectively. This counting was made at each grid point of the studied area from the filtered zonal and meridional wind components for the six summers. In Table 1, displaying the results for 1981 and 1982 in the band of latitudes 17.5 N and 5 N, the number of cases is weak in 1982, when the 6 9 day waves were not very active. The results for summers 1983, 1984, and 1986 (not shown) are very similar to those for 1982, while the results for 1985 (not shown) are closer to those for The 6 9 day waves are intermittent on the intraseasonal and interannual timescale. Kuo [1949] and Rennick [1976] invoked the meridional and vertical shears of the zonal wind to explain the African waves. Since the 6 9 day waves and African waves are observed in the same region, following these authors, the occurrence of the 6 9 day wave would depend on the zonal wind meridional profile. Therefore we computed the mean meridional profiles of the zonal wind component of the summers between 30 N and the equator for the longitudes 20 W to 30 E (21 grid points). The profiles were almost identical for the six summers. We present only two profiles in Figure 11, one for summer 1981 (Figure 11a), when waves were active, and the other one for summer 1982 (Figure 11b), when waves were weak. In Figure 11a the AEJ is around 8ms 1 between 15 N and 10 N and hpa. The Tropical Easterly Jet is about 14 m s 1 between 10 N and 2.5 N at 200 hpa. Figure 11b is practically the same as Figure 11a, though the AEJ seems to be slightly under the 500 hpa level or seems to have more latitudinal extension. We suggest that the two profiles are suitable to the wave development, though the 6 9 day waves were not very active in summer [33] The connection between the 6 9 day waves and rainfall, as was observed in the composite of percentage (Figure 5), is also visible in Figure 2d of Diedhiou et al. [2001], obtained from a mean of 12 years of August September rainfall data with the filtered meridional wind component as the reference parameter and at the reference latitude 17.5 N. In Figure 5 and in Figure 2d of Diedhiou et al. [2001], south of 15 N, around the ITCZ, positive rainfall anomalies appear behind the trough, with a zonal extension behind the ridge. The maximum rainfall is located at 7.5 N in Figure 5 and in Figure 2d of Diedhiou et al. [2001]. Figure 11 (with time sequence t =0)ofDiedhiou et al. [1999] for the mean of June September rainfall Figure 11. Mean meridional profile of the zonal wind component in summer (June September) (a) 1981 and (b) 1982 (in m s 1 ) from longitude 20 W to30 E (21 grid points).

11 MONKAM: WAVE AND RAINFALL MODULATION IN NORTHERN AFRICA ACL 5-11 anomalies is very similar to Figures 6a and 8a for summer 1981 and Figure 9e for summer 1985 shown in this work, with positive rainfall anomalies around the ITCZ and with a large zonal extension from 15 E to west of the studied area. In contrast, Figure 11 of Diedhiou et al. [1999] is very different from Figures 9b 9d and 9f, corresponding to the summers of 1982, 1983, 1984, and 1986, respectively, when the 6 9 day waves were not active. The rainfall anomalies of Diedhiou et al. [1999, 2001] are weak compared to our results obtained for summers 1981 and 1985, and this may be due to three main reasons: (1) the authors selected waves of weak amplitude (0.5 m s 1 ); (2) they extended their region east of 20 E, where waves were weak; in general, these waves exist west of 20 E, which is obvious in Table 1 of Diedhiou et al. [2001], showing that east of the Greenwich meridian, less than 8% of 6 9 day waves had amplitudes >2 m s 1 at 17.5 N; (3) they took into account years when the 6 9 day waves were not active. [34] These reasons may also explain the differences observed in the wave pattern displayed in this work in Figure 4 and the stream lines in Figure 2a of Diedhiou et al. [2001]. The effects of the reference parameters and latitudes must also be invoked. For the 6 9 day waves the filtered zonal or meridional wind component can be used as a reference parameter. Only the filtered meridional wind component is frequently used for the 3 5 day waves. Hence for studies dealing with interactions between the two easterly waves regimes it is convenient to choose the filtered meridional winds as the reference parameters. This has been done by Diedhiou et al. [2001], who chose the latitude 17.5 N for the 6 9 day wave because they found that the latitude 17.5 N corresponds to the latitude of these waves tracks. For the same reason the reference latitude 5 N was chosen for the 3 5 day waves. In Figure 4, displaying wind vectors and geopotential height anomalies for the composite 6 9 day waves, the meridional wind component is maximum at 17.5 N and toward lower latitudes (<5 N; the maximum is at the equator), while the zonal wind component is maximum at the latitudes N. To obtain the spatial pattern of the 3 5 day waves (Figures 8c and 8d), we used the reference latitude 5 N because we found that at this latitude these waves had large amplitudes on the filtered meridional wind component. So there is a good coincidence between the latitudes of waves tracks found by Diedhiou et al. [2001] and the latitudes where waves had large amplitudes on the meridional wind component for the two regimes, as was seen in this work. Therefore, considering the effects of the choice of the reference latitude mentioned by Duvel [1990] and discussed in this section, the horizontal pattern of the 6 9 day waves in the band of latitudes 5 S 30 N in northern Africa could be improved by choosing the filtered meridional wind component and the equator, the filtered zonal wind component and 10 N (or 12.5 N), and the filtered meridional wind component and 17.5 N as the reference parameters and latitudes, respectively. We used a similar approach to obtain Figure 10, displaying the 6 9 day wave pattern between 5 S and 20 N. [35] Furthermore, from the study of variability of the composite wave during the summers of 1982, 1983, 1984, and 1986, when waves were not very active, there is a strong anticyclone south of 30 N. The cyclonic vortices are weak and exit in a narrow band of latitudes. To the south of the area, there is a large band of latitudes with westerly wind which may be linked to the African monsoon. So the occurrence or the activities of the 6 9 day waves could be influenced by the Azores and Libyan anticyclones and by the African monsoon. 7. Summary and Conclusion [36] In this work we investigated the 6 9 day waves and their connections or effects on rainfall in West Africa, mainly during summer We have shown that the 6 9 day waves exist with a large peak at 7.2 days on the power spectrum of the zonal wind component at 700 hpa level and that they are visible even in the unfiltered data. The pattern determined using a compositing method is characterized by the existence of two vortices of opposite circulation on either side of the latitude 12.5 N. The composite rainfall pattern displays a good coincidence between cyclonic (anticyclonic) vortices and positive (negative) rainfall anomalies, coinciding with negative and positive geopotential height anomalies, respectively. The rainfall modulation by the 6 9 day wave has been clearly established using rainfall percentages. The modulation is clearly visible in the percentage with strong values (>100%) associated with cyclonic vortices and on the weak percentage (<100%) with anticyclonic vortices. [37] The study of the variability of composite wind vectors, rainfall, and geopotential height anomalies has shown that the 6 9 day waves increase rainfall in the band of latitude N in the studied area, with a maximum to the west near the Fouta-Djallon Mountains in Guinea and in the center near Lake Chad, and decrease rainfall north of 15 N and south of 7.5 N. There are good coincidences between the zones of rainfall increasing (decreasing) and negative (positive) geopotential height anomalies. A comparison between the effects of these 6 9 day waves and African waves on rainfall shows that (1) the African wave is associated with meridionally extended positive rainfall anomalies between 5 N and 15 N, mainly at the east of Greenwich meridian, consistent with the meridional extension of the wave and that (2) the 6 9 day wave is linked to a zonally extended rainfall anomaly over West Africa, with positive rainfall inside the band of latitudes N and a negative value north of 15 N, toward Sudan, and south of 7.5 N, around Ivory Coast. In these 6 9 day waves the rainfall anomalies are linked to the zonal wind anomalies, and the increase in the rainfall anomalies is associated with the large modulation of the AEJ. [38] On the intraseasonal timescale the rainfall modulation is weak in June and September, when the 6 9 day waves are not very active. On the contrary, they are very active in July and August and may probably contribute to increased rainfall during these 2 main months of strong rainfall during the rainy season in the Sahel region. For the interannual variability the rainfall anomaly increasing or decreasing also depends on the 6 9 day wave activity during the summer of a given year. From the six summers considered in this work these waves were very active during the summers of 1981 and 1985, for which we observed strong positive rainfall anomalies. For these two summers