Africa is the second largest continent and it is bounded on the west by the Atlantic Ocean, on the north by the Mediterranean Sea, on the east by the Red Sea and the Indian Ocean while on the south by the blending waters of the Atlantic and Indian Oceans. Africa is unique continent that, almost symmetrically, straddles the equator, and hence experiences such a varied climate with both northern and southern hemispheric climatic influences [26] [27]. Agriculture is considered the largest main economic activity in Africa and it provides employment for approximately 60 percent of the African population, and more than 50 percent of GDP in some countries [28]. In Africa, the distribution of rainfall is very uneven in both space and time and the climate of much of the continent can be classified as arid and semi-arid, receiving less than 700 mm of precipitation each year [29] [30] [31]. Figure 1 indicates a description our study domains and for

this study, we select two sub-domains mainly based on rainfall amount and distribution here after referred to as R1 with a domain of longitude 9˚W - 13˚E and latitude 7 - 20˚N (West Africa) and R2 with a domain of longitude 15 - 39˚E and latitude 6 - 18˚N (East Africa).

The study employs daily rainfall from four ensemble Coordinated Downscaling Experiment (CORDEX) Africa domain for both historical (1980 to 2005) and projected (2020 to 2099) data to analyze rainfall extremes. The four regional CORDEX models which we used for this study includes CanESM2, CSIRO-Mk3.6.0, GFDLESM2 and MIROC5 and the model runs every 50 km (0.448) resolution over the CORDEX-Africa domain and accessed from https://esgf-node.llnl.gov/search/esgf-llnl/ more information on model output can be obtained from [32] [33]. The gridded data from Climatic Research Unit (CRU TS4.01) also used to examine recent variability and trends of rainfall over Africa available from 1980 to 2017 and the dataset is clearly discussed by [34] and accessed from https://data.ceda.ac.uk//badc/cru/data/cru_ts/cru_ts_4.01/data/. The seasonal sea surface temperature indexes (SST) of ENSO3.4, North Atlantic Oscillation (NAO), Southern Oscillation Index (SOI) and Atlantic oscillation (AO) obtained from NOAA of Climate Prediction Center (CPC) and additionally the Indian Ocean Dipole (IOD) also used. The composite analysis of sea level pressure and vector winds are obtained from NOAA/NCEP reanalysis data and maproom were explored from the NOAA/OAR/ESRL PSD/, Boulder, Colorado, USA, http://www.esrl.noaa.gov/psd/. The reanalysis of wind datasets are used and collected from ERA-Interim for JJA season for the periods of 1980-2017 (http://apps.ecmwf.int/datasets/) and accessed from the European Center Medium-Range Weather Forecasting (ECMWF), with a horizontal resolution of 0.75˚ × 0.75˚ [35].

Our study uses standard descriptive statistics, mean, standardized seasonal anomaly, linear regression and correlation to explore the linkage between circulations, anomalies of SST with CDD and RX5day extremes. According to the Expert Team on Climate Change Detection and Indices (ETCCDI) and World Meteorological Organization (WMO) recommended a total of 27 core extreme with primary focuses to be derived from station daily data [19] [36]. For this study we analyzed two basic rainfall extremes were selected to investigate extreme climate conditions over Africa (Table 1). The two rainfall extremes are consecutive dry days (CDD) which is important to indicate the drought intensity and frequency. Whereas monthly maximum five day precipitation amount (RX5day) helps to provide information for the occurrence of flood phenomena. Additionally RCP 4.5 and RCP 8.5 climate scenarios used in order to detect the

NB: Full definitions are available from http://etccdi.pacificclimate.org/list_27_indices.shtml.

potential impact of the projected extremes based on the recommendation of [37] [38]. A trend analysis for rainfall extremes was detected and carried out by using the Mann Kendall (MK) test as discussed in depth by [39]. The wind and rainfall anomaly is analyzed by using statistical method of empirical orthogonal function (EOF).

Understanding the spatiotemporal variability, trend, and changes in precipitation in a changing climate are crucial mainly in the agricultural and other related socio-economic sectors. The June-August (JJA) season is the most vital rainy period for the equatorial and eastern portions of African countries. Figure 2 shows that the standardized seasonal rainfall anomaly and contour based seasonal rainfall change over each region in JJA season. The temporal rainfall trend plot depicts that significant variability during 1980-2017 and which is modulated by remote forcing (Figure 2(a) and Figure 2(b)). During the JJA, inconsistent temporal variability is portrayed, and the rainfall amount is lessening temporally and spatially over each region (Figures 2(a)-(c)). The normalized plot revealed that seasonal rainfall exhibits high spatiotemporal variability with the highest variation in east African regions (Figure 2(c)). In most recent years the spatiotemporal variability of JJA seasonal rainfall over east Africa shows that constantly decreasing trend compared to West African regions (Figures 2(a)-(c). Particularly, there was observed huge shifts in JJA season during 1998, 2010, 2011, 2013 and 2015/16 which was associated with varies teleconections including ENSO episode over East African regions. This decline in JJA seasonal rainfall over the region have also been observed in earlier studies which revealed accompaniment of severe and frequent occurrence of droughts during that period (Figure 5 and Table 2). Additionally noticeable variability of JJA seasonal rainfall observed over West African regions. This result is in agreement with [40] who observed that in some portions of African countries the seasonal rainfall performance depicts both increase and decrease but in general the rainfall trends are weak. Due to the variability of the seasonal rainfall during 1980-2017 severe to extreme drought occurred over each region and such types of rainfall extremes affect varies socio-economic activities over Africa (Table 2). Changes in extreme events of droughts and floods have already been observed due to deficit and heavy precipitation events globally including Africa [41]. As indicated by

[30] the African JJA seasonal rainfall is mainly influenced by the El Niño-Southern Oscillation (ENSO) and other large scale indexes which is described in (Table 3).

1) Correlation of seasonal rainfall with sea surface temperature (SSTs)

In this section, we analyzed different large scale atmospheric indexes and their relation with CDD and RX5DAY. The variability of Sea Surface Temperature (SST) is considered to be one of the factors that control the variability of the African monsoon rainfall. The ENSO3.4 correlates negatively and NAO positively over East African JJA seasonal rainfall. Moreover ENSO3.4 and NAO correlate positively over Western Africa regions (Figure 6(a) and Figure 6(b)). Li et al. 2012 correlates the SST to the African monsoon seasonal rain and found that more rain occurred over West African region in the years with warmer SST. The seasonal rainfall variability over Africa is mostly influenced by large scale atmospheric phenomena of ENSO3.4 and NAO. The JJA seasonal and inter-annual variability of precipitation extremes over Africa results from complex interactions of forced and free atmospheric variations [30].

Previous works of [53] depict that a complex influence of SSTs is responsible for rainfall patterns and trends in the Sahel at different frequency scales, including SSTs-ENSO in the Indian Ocean [54] and the Mediterranean Sea [55]. Furthermore, land-surface-moisture feedbacks play an important role [56].

2) Association of CDD and RX5day extremes with large scale SST anomalies

Understanding the clear relation extremes with various large scale climate indexes is very important to minimize the impact of drought and flood events.

The CDD and RX5DAY have been significantly associated with several large scale climate indexes which include ENSO3.4, SOI, NAO, AO and IOD (Figure 7, Figure 8 and Table 3). As pointed out in Figure 7 the standardized anomaly of CDD and RX5DAY shows enhancement after 1995 and is highly governed and associated by large scale indexes of ENSO, SOI, NAO, AO and IOD. The ENSO3.4, AO and IOD index has positively correlated with CDD over R1 and R2 and whereas NAO correlates negatively over both study regions (Table 3). The extent to which these large scale indexes influence the seasonal cumulative precipitation over Africa has been widely studied by [30] [57] [58]. The Southern Oscillation Index (SOI) and North Atlantic Oscillation (NAO) correlate positively and negatively with RX5day over each region respectively (Table 3). Previous studies of [59] indicated that ENSO is the primary source of variability in the tropical and global precipitation record and therefore variations in ENSO affect perceptions about changes in drought [60] [61] and their possible links to climate change and extremes. Moreover, the daily and seasonal rainfall of African region is influenced by the Inter-Tropical Convergence Zone (ITCZ) and ITCZ passes the equator twice in a year; the passages of the ITCZ coincide with the two distinct rainy seasons experienced in the area [62].

3) Atmospheric circulation features associated with extremes

Figure 9 shows the first EOF modes of 150 hPa v-wind and geo-potential height anomalies over Africa in JJA season. The mid-latitude flow which indicates that a wavelike structure propagating towards African continent approximately extends from northern Atlantic and Europe which have negative signals over eastern African portions (Figure 9). The EOF wind anomalies revealed that higher variability and the variance of the first mode is 37.4% (Figure 9(a) and Figure 9(c)) and the first mode of EOF of geo-potential height indicates 45.1%

and it shows that positive anomalies over R1 than R2 (Figure 9(b) and Figure 9(d)). Previous works of [63] showed that the drought and flood episodes in the African Sahel regions has been occurred and intensified by the intensity and frequency of various atmospheric features.

Figure 10 depicts that the dominant wind climatology at 850 mb, 700 mb, 150 mb and their anomalous years of drought and flood extremes over R1 of West Africa. The low level westerlies are the main features during JJA season which transport moisture from Tropical Atlantic towards inlands of African continent and the maximum strengthen of the wind speed is ~16 m/s (Figures 10(a)-(c)). The African Easterly Jet (AEJ) at 700 hPa located between 10˚N and 15˚N which resulted from thermal difference between the Saharan and Atlantic Ocean and it is sustained by the contrasting moist convection to the south and a dry convection to the north and the strengthening of the AEJ is usually connected with a decrease precipitation for the African summer rainfall and this situation exacerbate

increasing of rainfall extremes (Figure 10(b), Figure 10(e), Figure 11(b) and Figure 11(e)). Atmospheric circulation is an important component of the African summer monsoon [64]. It drives the monsoon inland from the coast and contributes to vertical moisture transport, a crucial requirement for African summer rainfall [65]. The JJA rainfall in the Sahel region exhibits significant variability and the region is particularly vulnerable to major droughts such as occurred in the 1970s and 1980s [66].

The anomaly field of velocity potential/divergence (convergence) associated with the composite wet and dry years are analyzed during June-August seasonal rainfall over Africa (Figure 12). The shaded lines of Omega over East Africa show that negative trend at all pressure levels and while weak positive over West Africa (Figures 12(a)-(c)). The wet events are characterized by convergence

(divergence) at low level (upper level) winds respectively (Figures 12(a)-(c)). Mainly over our study domains, thus the composite velocity potential during wet years is associated with rising/upward motion over West Africa (R1) and the dry years indicates that which are opposite to the wet years it is associated by divergence (convergence) over the study region (Figures 12(a)-(c)). The composite dry years are therefore characterized with sinking/downward motion mainly over Eastern Africa (R2).