Interannual Variability of Rainfall over the West Africa Sahel

Interannual variability of the precipitation over West Africa Sahel is analyzed based on 32 years (1979-2010) from monthly and daily database of the Global Precipitation Climatology Project (GPCP). In this region, we found that there is a link between the West Africa Monsoon (WAM) and the daily means of the precipitation in the summer, unseasonal rains can occur in the transition seasons and even in the heart of the dry season. Rainfall is the most important element for agro-pastoral activities in this region. The 850-hPa wind and wind divergence structure show a maximum convection over Mountain region (Fouta-Djalon and Mont-Cameroon) which corresponds to the high precipitation and OLR observed in these regions. The trend and empirical orthogonal function (EOF) of the precipitation are presented, including the mid-July variability of the precipitation. The dominant EOF of GPCP precipitation accounts for around 25.3% of the variance with slightly large am-plitude in the north while relatively small in the equatorial band respectively. The second and third EOF which accounts for 20.5% and 14%, describes a longitudinal contrast with a zonal gradient.


Introduction
Sahel is an area of semi-arid expanse of grassland, shrubs and small thorny trees found in Southern Sahara Desert (Nicholson, 2013). This region extends about was removed (Kidson, 1977).
Northern Hemisphere circulation shifts and weakening have been linked to anomalously low Sahelian rainfall to weakening. Weakening in Sahel region appears as near disappearance of the 850-mb trough (Kidson, 1977), a weakened TEJ (Newell & Kidson, 1984;Fontaine et al., 1995), and a decrease in southwesterly monsoonal flow (Fontaine & Janicot, 1992). Some studies have examined drought during years 1967 and 1968 (Lamb, 1978) and years 1953-63 and 1968-75 (Fontaine & Janicot, 1992 and showed that it corresponded to a southward displacement of the ITCZ. On the other hand, some studies covering different time periods haven't detected a systematic southward movement of the ITCZ. These studies can be found in Newell and Kidson (1984) which analyzed data covering 1958-63 and 1969-73, or Nicholson (1981 which examined a dataset extending from 1900 to 1973. These authors found that during some wet (dry) years, the ITCZ extended northward and the Hadley cell's descending branch expanded (contracted). Fontaine et al. (1995) showed that rainfall anomalies are related to wind behaviour in the coast of Guinea and Sahel. They found that rainfall anomalies in these regions were of opposite sign most often. In addition, the Sahel zone drought-Guinea zone flood scenarios corresponded to an increased, southward, low-level, meridional temperature gradient and a stronger wind velocity at the AEJ's southern border. This study proposes to explore the interaction between rainfall and dynamics mechanisms events at annual time scales and evaluate their modification in the climate system over West Africa Sahel in way to add the literature. Datasets and methods used are briefly described in Section 2. Section 3 investigates the characteristic of precipitation and annual cycle associated with African Easterly wave's activity. Finally, Section 4 summarizes the main results and gives some prospects for future work.

Data
The in-situ data are obtained from CERAAS (Centre d'Etude Regional pour l'Amélioration de l'Adaptation à la Sécheresse) which is a national laboratory of ISRA (Institut Sénégalais de Recherches Agricoles) and a regional pole of excellence in the study of drought adaptation of species cultivated in dry areas. The knowledge acquired on the agro-physiological behaviour of plants in drought conditions has been enhanced by the development of tools and methodologies for the early estimation of agricultural production. These daily precipitation data  (Salack et al., 2016). Figure 1 shows the evolution of the standardized rainfall index over these synoptic stations where rainfall was regular at Louga, Linguere and Matam in contrary to Saint-Louis and Podor where the dry years are dominant. Rainfall data are obtained from the Global Precipitation Climatology Project (GPCP) one-degree daily (1DD) version 2.2. It is combined with the precipitation dataset from polar satellite-estimated (SSMI/I emission estimates) and GPCP geostationary satellite precipitation data centre (GPI and OPI estimate rain gauge). The dataset used covers the period from 1998 to 2008 (daily) and 1979 to 2010 (monthly). This product has been validated regionally and globally by McCollum et al. (2003); Dinku et al. (2007) and Ruiz et al. (2010). While the 1DD GPCP rainfall data is available over a rather short period, this product seems to be essential over Africa where topography and surface conditions force high rain spatial variability (Thomas et al., 2009;Vondou et al., 2010;Tchakoutio et al., 2012) and where OLR values may not be always highly correlated with rainfall quantities.
We also use 6-hourly ERA-Interim data for 36 years  on the latitude-longitude resolution of 0.75˚ × 0.75˚ (available at http://data.ecmwf.int/data/). We derived wind speed (U) and wind divergence from ERA-Interim. Vertical cross-sections extending from 1000 to 500 hPa (every 25 hPa from 1000 to 750 hPa and every 50 hPa from 750 to 500 hPa) were used to identify heat, moisture, and wind-speed gradients from 20˚W to 30˚E and from Equator to 30˚N (Dee et al., 2011).
Most often, outgoing longwave radiation (OLR) is used to investigate AEW (Wheeler & Kiladis, 1999;Roundy, 2008;Kiladis et al., 2009, Zebaze et al., 2017. The present study uses daily OLR dataset obtained from NOAA from 1974 to the present day with a time resolution of two measurements per day and per grid point (Liebmann & Smith, 1996). The local time of passage of the satellites varied during the entire period of observation. Hence, for maximum consistency in the data sets, we have retained only the years from 1980 to 2009 for which measurements were made at 7:30 and 14:30 LST. As has already been shown in previous studies (Janicot, 2011), this provides a realistic estimate of the daily average, especially for analysing the intra-seasonal time scale. Nguyen and Duvel (2008) showed that strong and reproducible 5 -6 days oscillation characterise dominant modes of boreal spring synoptic variability. Seasonal cycles were removed before calculating spectra for each month segment during 1980-2009. Spectra are calculated on each individual month and averaged over 11 years. a 3-point Daniell smother is applied to the spectral estimates (Daniell, 1946) to maximize degree of freedom. A red noise background spectrum is computed from the formula of Gilman et al. (1963). The 95% confidence limits about this red noise spectrum are determined using F-statistic. Wavelet analysis is a common tool for decomposing a time series into a time-frequency space and detecting time-frequency variations. Wavelet transform allows comparing a signal to a wavelet function called mother wavelet (here the Morlet wavelet is used; Torrence & Compo, 1998). After testing different mother wavelets, the results obtained look similar. Because the wavelet transform is a band pass filter with a known response function (the wavelet function), it is also a powerful filtering technique. In order to identify the dominant synoptic modes, the wavelet analysis method is, therefore, employed here over West Africa during 1980-2009 since in signal processing; wavelets are very useful for processing non-stationary signals.

Empirical Orthogonal Function
Also called Principal Component Analysis (PCA), the Empirical Orthogonal where the transpose operation is denoted by the superscript T and each of the M eigenvectors contains one element pertaining to each of the K variables, x k .

Trends
The We applied the MKTT to test the presence of the trend in precipitation and to analyze the local seasonal trends in this parameter. The magnitude of the trend of precipitation is estimated by linear regression, and trends included in the analysis have their probability exceeding the 95% significance level.

The Climatology over West Africa
Most recent research on the intraseasonal variability of the West African monsoon's (time scale of 10 to 90 days) has centered on the Madden-Julian Oscillation (MJO). Generally, its origin is over the Pacific warm pool, from which an eastward-propagating Kelvin wave and a westward-propagating Rossby wave emanate (Zebaze et al., 2015). A similar phenomenon can also occur over the Indian Ocean providing a link between the Indian and African monsoons. Occurring in the boreal summer, these waves meet over West Africa, where they spawn convection and modulate easterly wave activity, the wind regime, and  (Nicholson, 2013). Figure 2 shows that the maximum rainfall effectively occurs in mid-August over West Africa. The spatial evolution during the summer is presented in Figure 3. The maximum rainfall can be found in the West Africa coastal zone and particularly around the Fouta-Jalon Mountain. climatological monthly means of precipitation (1979-2014) over West Africa from GPCP reanalysis can be found in Ndao et al. (2019).
The trend of the precipitation presents a maximum at the center of Senegal in August, which corresponds to the presence of ICTZ in this region as presented in Figure 4. Depending on the local rainfall peak, Thorncroft et al. (2003) defined four phases of the West African monsoon: oceanic, coastal, transitional, and Sahelian. During the oceanic phase, between November and mid-April, a broad rainbelt lies just north of the equator. During the subsequent coastal phase, which generally prevails to mid-June, peak rainfall lies over the ocean but in the near-coastal region around 4˚ to 5˚ North. The transition phase, when a decrease in rainfall is observed, occurs in early July. Le Barbé et al. (2002) refer to these first three phases collectively as the oceanic regime. The Sahelian phase lasts from mid-July to September. Throughout this last phase, the rainfall peak is  with African wave activity is consistent. Recent studies suggest that convective wave initiation is favoured by the ITCZ when it is closer to the equation during these months (Nguyen & Duvel, 2008;Zebaze et al. 2017   GPCP). 95% confidence limits and Red noise were determined using the theoretical Markov spectrum methods. Spectral peaks that are statistically significant at the 95% confidence level will stand above the upper limit of these curves. Over the West Africa, four main spectral peaks for precipitation exist at the synoptic time scales (Figure 6). One spectral peak is centred near a period of 3 days, and another peak is centred around a period of 5 days. The 3-day peak is presumably in association with African easterly waves (Reed et al., 1977;Lenouo et al., 2005) and with convective activity over land. The 5-day peak corresponds to 5 -6 days synoptic variability found by Nguyen and Duvel (2008). Figure 7 shows the mean wavelet spectrum averaged for the period 1980-2009 of unfiltered OLR anomalies over the Northern Senegal (14˚N -16.5˚N and 12˚W -16.5˚W). The largest oscillations of convection are observed in June -August with synoptic variance between 2 and 5 days. In northern winter, depending on the regions, high variances are also present. In summer, the ITCZ migrates northward and the synoptic AEW becomes the main feature of rainfall in this region. The relationship between convection and AEWs is complex. It was originally thought that they principally organize convection. Recently, research has shown feedback relationships in which convection can modulate waves, which in turn modulate convection. Nicholson (2013) suggested that the mesoscale convection systems (MCSs) that bear most of West Africa's precipitation are often imbibed in the northerly flow ahead of the wave trough. More recent work has also shown the relationship between AEWs and convection to be much more complex (Lenouo et al., 2016).    1979-1981, 1984-1985, 1987-1990 and 1997-1998. We can note that the negative phase 1982-1983 correspond to the dry years over this region. These components account for about 25.3% of the total variance. The PC2 of GPCP precipitation was in its negative phase between 1979-1997; while the PC3 of GPCP precipitation was in its positive phase in June to August and in its negative phase respectively between years 1980-1999 and 2002-2010. The PC2 and PC3 account respectively for around 20.5% and more than 14% of the total variance. The first corresponding three EOFs are presented in Figure 9. The EOF1 presents the permanent monsoon zone around the West Africa coast where the EOF2 shows this penetration in the continent and the Mountains areas convection.

Conclusion
Using upper air soundings, satellites, meteorological radar and instrumented aircrafts, convective activities over West African Sahel associated with the Afri- The synoptic times-scales precipitation in West Africa during the summer 1980-2009 is dominated by variability on both 3 -5 and 5 -6 day times-scales.
The 3-day peak is presumably in association with African easterly waves and with convective activity over land whereas the 5-day peak presumably corresponds to 5 -6 days synoptic variability. Over the Northern Senegal (14˚N -16.5˚N and 12˚W -16.5˚W), the largest oscillations of convection are observed during June-August with synoptic variance between 2 and 5 days. It also shows the first three PC and EOF of GPCP precipitations over West Africa during the summer (JAS) from 1979 to 2010. The PC1 accounts for 25.3% of the total variance. The PC2 of GPCP precipitation was generally in its negative phase between A. Dia-Diop et al. 1979-1997 while the PC3 of GPCP precipitation was in its positive phase in June and August and in its negative phase in July and September respectively between years 1980-1999 and 2002-2010. The PC2 and PC3 account respectively for around 20.5% and more than 14% of the total variance. Future research over West Africa Sahel should focus on examining additional precipitation events in different developing and/or non-developing AEWs and