Understanding the Evolution and Socio-Economic Impacts of the Extreme Rainfall Events in March-May 2017 to 2020 in East Africa

This study aimed at assessing the evolution, distribution and the socio-economic impacts of extreme rainfall over East Africa during the March, April and May (MAM) rainfall season focusing on assessing the trends and contribution of MAM rainfall in mean annual rainfall across the region. It employed Principal Component Analysis (PCA) methods to capture the patterns and variability of MAM rainfall. The PCA results indicated that the first Principal Component (PC) describe 17% of the total variance, while the first six PCs account only 53.5% of the total variance in MAM rainfall, underscoring the complexity of rainfall forcing factors in the region. It has been observed that MAM rainfall accounts about 30% - 60% of the mean annual rainfall in most parts of the region, signifying its importance in agriculture, water, energy and other socio-economic sectors. MAM has been characterized by increasing variability with varying trend patterns across the region. The MAM rainfall trend is not homogeneous across the region; some areas are experiencing a slight decreasing rainfall trend, 235.1 mm of rainfall in 24 hours respectively, which are the highest amounts for these respective stations, since their establishment. Record highest 24 hours rainfall amounting to 134.9 mm and 119.4 mm were also observed at Buginyanya and Kawanda meteorological stations in Uganda on 18 th March 2018 and 7 th May 2020. On 6 th May 2020, Byimana meteorological station in Rwanda, also observed 140.6 mm of rainfall in 24 hours, the highest since its establishment. These extremes have caused multiple losses of life and proper-ty, and severe damages to infrastructure. Unfortunately, the frequency and intensity of these extremes are projected to increase under a changing regional climate patterns. It is therefore important that more studies are carried out to enhance understanding about the evolution, dynamics and predictability of these extremes in East Africa region.


Introduction
Socio-economic development and livelihood activities in most of the developing countries including East African countries are strongly affected by climate variability and change [1] [2] [3] [4] [5]. In most of the East African countries, the production and productivity in agriculture and livestock sectors are largely influenced by the availability and variability of rainfall amount of a particular season. Hydro-electricity that contributes about 30% of the total electricity consumption in Tanzania, is also strongly affected by rainfall variability associated with the extremes weather and climate events. The increase in climate variability and change is strongly manifested through the increase in frequency and intensity of extreme weather and climate events. Recent reports and records from climatological data from different parts of the world are overwhelmingly indicating an increase in the frequency and intensity of heavy precipitation and other extreme events including floods and droughts [1]- [7]. These extremes are projected to increase further at projected global warming of 1.5˚C [1]. Heavy precipitation is projected to be more intense under a global warming of 2˚C as compared to a warming of 1.5˚C [1] [2] [3]. Warming levels reached 1.1˚C in 2019 [8], and with business as usual scenario, warming is projected to reach 1.5˚C between 2030 and 2052 [1] leading to intense extremes, which will have devastating, widespread and cascading implications to the livelihoods and so- This study therefore aimed at assessing the spatial and temporal distribution of MAM rainfall for the purposes of depicting its current characteristics and trends and also depicting the evolution and assessing the impacts of extreme rainfall events in MAM 2017, 2018, 2019 and 2020. The study also aimed at establishing and understanding in the context of a changing climate, the main forcing factors responsible for the evolution of the extremes in the region.

Description of the Study Area
The East Africa (EA) region comprises of six countries namely; Tanzania, Kenya, Uganda, Burundi, Rwanda and South Sudan. However, the focus of this study is on Tanzania, Kenya, Uganda and Rwanda, located within the 5˚N and 12˚S and 29˚E and 42˚E ( Figure 1). The region is characterized by diverse climate patterns due to complex topographical features. Rainfall is one of the climate parameters that is typified by stronger spatial and temporal variations, which is also amplified by significant differences in relief and vegetation cover.
The large areas within the region receive bimodal rainfall pattern with "long rains" season in MAM and the "short rains" season in October, November and

Data Sources and Type
Daily and monthly rainfall data for the period 1961-2020 from selected synoptic stations in Tanzania, Kenya, Uganda and Rwanda were used in this study ( Figure   1). The rainfall data were obtained from Tanzania Meteorological Authority  The study also made use of two gridded datasets; Global Precipitation Climatol-

Methodology
This study made use of simple statistical tools to compute and characterize the percentage contribution of MAM rainfall into mean annual rainfall, and employed the Empirical Orthogonal Function (EOF) to [21] characterize the variability of MAM and to analyze the large-scale inter-annual variability between the mean monthly rainfall over the EA and the mean Sea Surface Temperature Anomalies (SSTA) over the western Indian and Tropical Pacific Oceans during MAM rainfall season with particular interest in MAM 2017, 2018, 2019 and 2020 rainfall season. EOF analysis is frequently applied to derive patterns and indices used to identify climate modes as expressed in state variables [22]. The approach identifies patterns in space known as EOF modes in one or multiple variables from eigenvectors of the covariance matrix for the gridded data sets. Then the original centered data are projected onto the spatial patterns to obtain time series indices (i.e., the principal components, PCs). In this study, the first EOF spatial mode of the mean MAM rainfall over the EA is taken as the dominant mode and further explains areas that were associated with enhanced/suppressed mean MAM rainfall based on the 1981 to 2010 climatology. Subsequently, ascertaining the circulation anomalies responsible for the enhanced/suppressed mean MAM rainfall over the Eastern Africa based on 1981 to 2010. Composite analysis is carried out on a number of field variables and later test for their significance with the two tailed Student's t-test. The results for the composite analysis obtained hereafter are then compared with the anomalies for the years 2017, 2018, 2019 and 2020 to reveal the likely cause of the enhanced mean MAM rainfall in these years. In this case, the composite for the enhanced mean MAM rainfall is computed when the amplitude of the first principal component (PC1) is greater than or equal to +1 (i.e., years 1981, 1985, 1989 and 2006). Meanwhile, during suppressed mean MAM rainfall over the region the amplitudes are taken to be less than or equal to −1 (i.e., years 1983, 1984, 2000 and 2007). Quantifying the likely association between the enhanced/suppressed mean MAM rainfall with ENSO and IOD indices, the present study measures their association by correlating the mean MAM rainfall anomalies (i.e., PC1) with ENSO and IOD indices, and later assesses the strength of their association. In this case, the ENSO indices are computed by averaging the normalized SST over the Nino 3.4 domain while the IOD indices are expressed in terms of Dipole Mode Index (DMI), which defines the difference in SSTA between the western equatorial Indian Ocean and the south-eastern equatorial Indian Ocean [17]. The nonparametric test, Mann-Kendall trend [23] [24], was also used to detect the trend in time series for the mean MAM rainfall over the region and tested for the corresponding significance at 95% confidence interval. All anomalies were calculated with respect to 1981-2010 climatology.

Contribution of MAM Rainfall to the Mean Annual Rainfall
Characterization and quantification of the percentage contribution of the MAM rainfall in the mean annual rainfall is very important for climate monitoring and research related to climate variability and change. It provides a good benchmark for detecting shift in seasonal rainfall patterns. The percentage contribution of mean MAM rainfall in mean annual rainfall for the East African region is provided in Figure 2. For most parts of the region, MAM rainfall contributes between 30% and 50% of the mean annual rainfall underscoring its significant role in rain-fed agricultural activities and in water sectors. The distribution of MAM  rainfall in East Africa exhibit a well defined bipolar orientation with areas over the eastern side of the region and along the coast areas featuring between 40% and 50% of the annual rainfall, while areas over western side of the region featuring between 30% and 40% slightly lower contribution as compared to the eastern side of the region. The higher percentage contribution over the eastern side of the region could be amplified by moisture influx from of Indian Ocean and the dynamics of the Indian Ocean Dipole (IOD). In few areas in Tanzania, including areas around Dar es Salaam, Kilimanjaro, Moshi and Arusha, MAM contribution in the mean annual rainfall is more pronounced; it ranges between 50% to 60%. The remaining seasons combined, including the October-December (OND), January-February (JF) and June, July, August and September (JJAS) add up to the remaining 40% to 50%.         It can be noted that, the sequential MK trend test applied in this study graphically illustrates the forward, u (F), and backward, u (B), trends of heavy rainfall over few selected countries over the Eastern Africa (Tanzania, Uganda and Kenya) as indicated in Figure 7. When a set of two series, a forward and backward one, cross each other and diverge beyond the specific threshold value (i.e., α = 0.05 significant level for this study), the point is said to be a significant change point. [25] assumed that the point is the approximate year when the trend begins, while [26] referred to it as the period at which the critical point of change occurs. [27] noted that, extreme weather events (i.e., heavy rainfall events) reveals an alarming increasing in trend over most area in Tanzania.

Recent Trend of March, April and May Rainfall Season
However, the typical sequential MK trend test over Tanzania in Figure 7(a) shows that, even though extreme heavy rainfall events are in increasing trend [27] but generally the mean MAM rainfall indicates the decrease in trend at 95% confidence interval. It reveals period of abrupt decline in trend from 1995 to 2019 which became significant at 0.05 confidence levels in 2020. Subsequently, the mean MAM rainfall over Kenya (Figure 7(c)) indicates a decrease in trend though insignificant at 95% confidence interval. Generally, the large climatological window (1981 to 2012), revealed the decrease in trend of the mean MAM rainfall over Uganda (Figure 7(b)) and thereafter leaned towards increasing trend (i.e., from 2012) though insignificant at 95% confidence interval.

Evolution and Distribution of Rainfall in MAM 2017, 2018, 2019, and 2020 Rainfall Season in East Africa
In recent years, the distribution of MAM rainfall in East Africa has been charac-

Principal Component Analysis of MAM Rainfall in East Africa
Using CHIRPS dataset, the Principal Component Analysis (PCA) of the MAM rainfall in the region was performed to capture the patterns and variability of the MAM rainfall for the period 1981-2010. The PCA analysis indicates that the first Principal Components accounts only about 17% of the total MAM rainfall variance ( Figure 11), while the first six Principal Components only accounts for 53.5% of the total MAM rainfall variance underscoring the complexity associated with rainfall patterns in the region. The second and third Principal Components are presented in Figure 12, and they account for 12.4% and 6.9% of the total variance respectively. It is evident that, unanimous positive loading anomalies (Figure 11(a)) are concentrated over the bimodal areas of Tanzania (northern sector), many areas of Rwanda and Burundi, the central to southern Kenya and the southern Uganda (along the Lake Victoria basin). Meanwhile, EOF2 and EOF3 ( Figure 12) still reveal the same scenario with positive coefficients over the areas recaptured by EOF1. Notably, the results in Figure 11 and Figure      Meanwhile, the in-phase correlation analysis between the mean MAM rainfall (PC1 in Figure 11(b)) over the EA with ENSO and IOD indices (figures are not included) revealed the weak positive correlation.

Socio-Economic Impacts of Extreme Rainfall in 2017, 2018, 2019 and 2020 MAM Rainfall Season in East Africa
Over the last ten years, an upsurge in the frequency and intensity of extreme events has been widespread and consequential (Tables 1-4 (Tables 1-4).
According to World Bank study [28], it was estimated that the household losses More details about impacts of floods in Dar es Salaam are described in [4].

Conclusion and Recommendation
On average MAM rainfall accounts about 30% to 60% of the mean annual rainfall in most parts of East Africa, underscoring the significant importance at-