Spatio-Temporal Characteristics of Climate Extremes in Sub-Saharan Africa and Potential Impact of Oceanic Teleconnections

Abstract

Sub-Saharan Africa (SSA) is a region vulnerable to extreme weather events due to its low level of adaptive capacity. In recent decades, SSA has been punctuated by more intense climatic phenomena that severely affect its population. Therefore, this study evaluates the performance of the ERA5 and CHIRPS datasets, and the spatio-temporal evolution of extreme weather indices and their potential relationship/response to climate variability modes in the Pacific, Indian, and Atlantic oceans, namely, the El Niño−Southern Oscillation, Indian Ocean Dipole, and Tropical Atlantic Variability (ENSO, IOD, and TAV). The CHIRPS dataset showed strong positive correlations with CPC in spatial patterns and similarity in simulating interannual variability and in almost all seasons. Based on daily CHIRPS and CPC data, nine extreme indices were evaluated focusing on regional trends and change detection, and the maximum lag correlation method was applied to investigate fluctuations caused by climate variability modes. The results revealed a significant decrease in total precipitation (PRCPTOT) in north−central SSA, accompanied by a reduction in Consecutive Wet Days (CWDs) and maximum 5-day precipitation indices (RX5DAYS). At the same time, there was an increase in Consecutive Dry Days (CDDs) and maximum rainfall in 1 day (RX1DAY). With regard to temperatures, absolute minimums and maximums (TNn and TXn) showed a tendency to increase in the center−north and decrease in the south of the SSA, while daily maximums and minimums (TXx and TNx) showed the opposite pattern. The IOD, TAV, and ENSO modes of climate variability influence temperature and precipitation variations in the SSA, with distinct regional responses and lags between the basins.

1. Introduction

Sub-Saharan Africa (SSA) is undergoing significant development and changes in extensive cultural, ecological, and climatic characteristics [1]. According to the United Nations, the population of SSA has been growing at an approximate rate of 2.6% between the years 2010 and 2020, making Africa the second most populous continent in the world. The current population is estimated to be around 1.3 billion people, a number that could double by the end of the century (ONU, 2020). The main activity in the region varies according to the countries and their specific characteristics, but prominent economic activities include agriculture, fishing, mining, oil and gas exploration, and tourism. Similar to other continents, there are social level differences between the North and South, East and West regions of Africa, which can be influenced by various factors such as economic development, infrastructure, governance, access to basic services, and living standards.

The Northern and Southern regions of Africa generally have the highest level of income per capita compared to the Eastern and Western regions [2]. Agriculture is the most present activity in Africa, especially rain-fed agriculture, which supports one-third of the population. The growth of this activity and livestock farming have been significantly influenced by the climate and extreme weather events such as droughts, floods, and temperature changes [3,4]. The Intergovenamental Panel of Climate Changes (IPCC) identified SSA as a region very vulnerable to the impacts of climate change due to its low level of economic and technological development [4]. Among others, the increase in climate-related events have led to adverse social, economic, and environmental impacts across SSA [5]. On the other hand, the simultaneous occurrence of floods and drought being present in SSA is not rare [6].

An extreme weather event is defined as the anomalous occurrence of a meteorological and climatic variable, indicating values above (or below) a certain threshold, close to the upper (or lower) end of the range of observed values for that variable [7].

It stands out that extreme weather events have been more frequent and intense during El Niño or La Niña years, potentially leading to extreme droughts and floods in certain regions of SSA. The El Niño−Southern Oscillation (ENSO) globally is the primary contributor to interannual climate variability due to its impact on disrupting the normal large-scale Walker circulation in the Pacific [8,9]. The global-scale effect of ENSO is always reflected in local phenomena, such as extreme floods and severe drought conditions [10]. Higher sea surface temperatures in the eastern equatorial Pacific warm the atmosphere during the El Niño phase, resulting in increased convection and precipitation in East Africa. In contrast, a continental high-pressure system dominates Southern Africa, suppressing atmospheric convection and precipitation regimes [11]. Conversely, during the La Niña phase (anomalous cooling of surface waters in the eastern equatorial Pacific), the opposite occurs. These changes in precipitation regimes due to ENSO directly impact rain-fed agriculture across the regions, which serves as a means of survival and subsistence for over 60% of the local population. The ENSO alters wind patterns and atmospheric and oceanic circulations, which via atmospheric bridges can modify the Tropical Atlantic Variability (TAV) and Indian Ocean Dipole (IOD) characteristics [12].

The Tropical Atlantic Variability (TAV) refers to the climatic and oceanic fluctuations that occur in the tropical Atlantic Ocean [13]. These fluctuations can have a significant impact on global climate. During the positive phase of the TAV changes in sea surface temperature (SST) in the tropical Atlantic are characterized by warming in the North Tropical Atlantic and cooling in the South Tropical Atlantic. As a consequence of this pattern, there is a displacement of the intertropical Convergence Zone (ITCZ) to the North [13]. Ref. [14] indicates that positive rainfall anomalies over the Congo basin (at interannual and multi-annual timescales) are related to anomalous westerly mid-tropospheric zonal winds, due to an anomalous SST pattern in the tropical North Atlantic. The authors of [13] argued that global warming-induced alterations in the TAV may impact the correlation strength between the TAV and the tropical Pacific. This could lead to a scenario where warming in the Atlantic diminishes warming in the eastern tropical Pacific due to modifications in the Walker circulation. Consequently, such alterations induce changes in precipitation and temperature across Africa.

The IOD represents interannual differences between SST in the western tropical Indian Ocean and the eastern Indian Ocean, resulting in changes in wind patterns along the equatorial Indian Ocean [15]. During a positive IOD event, there is anomalous warming of surface waters in the western Indian Ocean, while anomalous cooling is dominant in the eastern Indian Ocean [16]. The IOD weakens the Walker circulation over the Indian Ocean, and in its positive phase influences precipitation in Eastern and Southern Africa, potentially leading to extreme flood events [17]. A zonal mode of IOD variability also influences precipitation variability in many parts of Africa [18].

In years when the ENSO, IOD, or TAV occur, SSA’s vulnerability is enhanced because SST anomalies modify precipitation patterns and increase temperatures in different regions of SSA. For example, Southern and Eastern Africa may experience higher temperatures and reduced precipitation during the positive phase of the ENSO. In fact, these conditions lead these regions to high risk of extreme droughts and excessive heat. This negatively impacts agricultural production, posing a threat to food security in SSA. However, most studies on the issue deal with the precipitation response on interannual and decadal times scales, and not much effort has been put into identifying the influence of these oceanic modes on extreme precipitation and temperature indices—such as, for instance, the efforts originally proposed by the Expert Team on Sector-Specific Climate Indices (ET-SCI), or currently being carried out under the oversight of the Expert Team on Climate Information for Decision-Making (https://climpact-sci.org/indices/, accessed on 15 January 2024).

To partially alleviate this gap, this study aims to assess the pattern of extreme climate indices and their potential relationship with the ENSO, IOD, and TAV across the SSA region. This paper is organized as follows: Section 2 describes the study area, data, and the methodology used to calculate extreme indices; Section 3 discusses the results, highlighting the relationship between extreme event indices and climate indices; finally, a summary and discussion of the main findings are presented in Section 4.

2. Data and Methodology

2.1. Meteorological Reanalysis Data

To obtain the spatio-temporal distribution of extreme events, four steps were taken: data collection from reanalysis, calculations of precipitation and air temperature extremes; evaluation of trends and additional statistics; estimates of lead−lag correlations based on an auto-regressive model between the modes of climate variability (ENSO, IOD, TAV) and the weather extremes.

The datasets used were based on analysis and reanalysis methods. The analysis datasets selected were CHIRPS (Climate Hazards Group Infrared Precipitations) and CPC (Climate Prediction Center) and the reanalysis dataset was ERA 5 (5th version of the European Centre for Medium-Range Weather Forecasts model product). CHIRPS is a combination of observed in situ data and satellite products, available for the period of 1981 to the present, with a resolution of 0.05° [19]. The remote sense component of the CHIRPS data compensates for the low density of weather stations. Indeed, CHIRPS compensates for some regions in Africa, such as the Democratic Republic of Congo and central and northern Mozambique, where the conflict has greatly reduced the availability of continuous data from stations.

The CPC dataset comprises moderate-resolution (0.5° latitude/longitude) spatial data of global precipitation, based on data from more than 30,000 weather stations from institutions in WMO member countries [20].This dataset, which spans from 1979 to the present day, is subjected to a rigorous quality control process, involving comparison with data records from neighboring stations, radar/satellite observations, and numerical model forecasts. The data are organized into two databases: a historical database, covering the period from 1979 to 2005, and a real-time database, which has been continuously updated since 2006. Due to the dynamic nature of the real-time base, its data can be reprocessed periodically to ensure greater homogeneity and consistency with the historical base [21].

ERA 5 is a simulated dataset that does not directly incorporate precipitation observations−i.e., there is no mixing or adjustment with the observed data. Reanalysis datasets such as ERA 5 are therefore useful as a test bed for precipitation simulations, being based on detailed specifications of the three-dimensional atmospheric circulation. The reanalysis is a fifth-generation product developed by the European Center for Medium-Range Weather Forecasts (ECMWF) [22]. ERA 5 is available on a 0.25° latitude/longitude spatial grid and features several improvements compared to its predecessor, ERA-Interim [23], namely, higher horizontal and vertical resolutions, as well as a more accurate data assimilation scheme (

Table 1. Comparison of characteristics of reanalysis data products of the CPC, CHIRPS, and ERA5.

Characteristics	CPC	CHIRPS	ERA5
Time Interval	1979−present	1981−present	1979−present
Domain	Global	Global	Global
Model Resolution	TL550 (55 km)	TL100 (5 km)	TL639 (31 km)
Spatial resolution	0. 5° × 0.5°	0.05° × 0.05°	0.25° × 0.25°
Vertical Levels		Hybrid sigma (60 levels)	Hybrid sigma (137 levels)
Assimilation Scheme	4D-Var	4D-Var	4D-Var
Method	Analysis	Analysis	Reanalysis
Reference	[19,21]	[23]	[22]

). For a more in-depth assessment of these reanalysis datasets, the ideal would be to compare them with in situ observations. However, in Africa, the spatial distribution of weather stations is quite uneven, the available records are limited, and obtaining these data from the relevant national meteorological services is often difficult or costly.

2.2. Ranking of CHIRPS and ERA5 Dataset Performance

To facilitate the incorporation of the CHIRPS and ERA5 datasets with CPC, all variables were regridded to a common horizontal spatial resolution of 0.5° by bilinear interpolation to maintain homegenity across all datasets. The performance of the ERA5 and CHIRPS datasets was determined using the efficient Kling−Gupta metric (Equation (1)), which is a combination of three components: the correlation coefficient (r; Equation (2)), the bias ratio (β; Equation (3)), and the variability ratio (γ; Equation (4)) [24].

$$KGE=1-\sqrt{{(r-1)}^2+{(\beta -1)}^2+{(\gamma -1)}^2}$$

(1)

$$r={\displaystyle \frac{{\sum }_{i=1}^N(O_i-\overline{O})(P_i-\overline{P})}{\sqrt{{\sum }_{i=1}^N{(O_i-\overline{O})}^2}\sqrt{{\sum }_{i=1}^N{(P_i-\overline{P})}^2}}}$$

(2)

$$\beta ={\displaystyle \frac{{\mu }_p}{{\mu }_o}}$$

(3)

$$\gamma ={\displaystyle \frac{C{V}_p}{C{V}_o}}$$

(4)

where P is the modeled value, O is the reference value, N is the number of observations, $\mu$ is the average value, and CV is the coefficient of variation. The KGE and its components close to one (1.0) indicate a greater capacity of the datasets evaluated.

2.3. Modes of Climate Variability

It has long been recognized that fluctuations of oceanic features away from the region of interest can induce substantial modifications of local climate/weather patterns. In this sense, the ENSO, IOD, and TAV stand out as primary candidates for inducing modifications of climate extremes Table 2, as demonstrated by [12,25,26].

The Dipole Mode Index as relevant to the Indian Ocean Dipole (IOD) and Tropical Atlantic Variability (TAV) is based on SST regional differences. X The TAV and IOD are calculated based on ERA5 SST. The IOD indicates the east−west temperature gradient across the Indian Ocean [27], between the eastern (10° S–0°, 90°–110° E) and western (10° S–10° N, 50°–70° E) basins. The TAV displays differences between SST anomalies in the tropical Atlantic between 5.5° N–23.5° N, 15° W–57.5° W and Eq-20° S, and 10° E–30° W [28]. The TAV may result from the anomaly of two modes of variability in the Tropical Atlantic: the Tropical Northern Atlantic Index and the Tropical Southern Atlantic Index [29].

Table 2. Climate modes used in this current study.

Climate Mode	Index Name	Methods	Reference
El Niño− Southern Oscillation	ENSO	First EOF of SST (20° S–20° N and 160° W–80° W)	[28]
Indian Ocean Dipole	IOD	Calculated as the monthly difference between the western (10° S–10° N, 50°–70° E) and eastern Indian Ocean (10° S–0°, 90°–108° E) SST averages	[26]
Tropical Atlantic variability	TAV	Differences between SST anomalies in the tropical Atlantic between 5.5° N–23.5° N, 15° W–57.5° W and 0–20° S, and 10° E–30° W	[28]

Table 2. Climate modes used in this current study.

Climate Mode	Index Name	Methods	Reference
El Niño− Southern Oscillation	ENSO	First EOF of SST (20° S–20° N and 160° W–80° W)	[28]
Indian Ocean Dipole	IOD	Calculated as the monthly difference between the western (10° S–10° N, 50°–70° E) and eastern Indian Ocean (10° S–0°, 90°–108° E) SST averages	[26]
Tropical Atlantic variability	TAV	Differences between SST anomalies in the tropical Atlantic between 5.5° N–23.5° N, 15° W–57.5° W and 0–20° S, and 10° E–30° W	[28]

2.4. Extreme Precipitation and Air Temperature Indices

Among the 27 extreme climate indices created by the Expert Team on Climate Change Detection and Indices (ETCCIOD), 16 are related to air temperature, and 11 to precipitation. For this study, nine representative indices including for four for temperature and five for precipitation were selected (

Table 3. Extreme precipitation and temperature indices. The temperature indices presented are temperature of the coldest nights (TNn), temperature of the hottest nights (TNx), temperature of the coldest days (TXn), and temperature of the hottest days (TXx).

Characteristics	Index	Definitions	Units
Intensity	PRCPTOT	Total precipitation index−annual total precipitation in wet days	mm
Frequence	Rx1day	Annual maximum 1-day precipitation	mm/day
	Rx5day	Annual maximum consecutive 5-day precipitation	mm/days
Time	CWD	Number of consecutive wet days	days
	CDD	Number of consecutive dry days
Intensity	TXx	Maximum value of daily maximum temperature	°C
	TXn	Minimum value of daily maximum temperature
	TNx	Maximum value of daily minimum temperature
	TNn	Minimum value of daily minimum temperature

). These indices have been extensively applied to characterize climate conditions worldwide as well as on a regional basis. For instance, Ref. [30] demonstrated that across North America, consistent warming with decreased frequency and intensity of cold extremes, such as cold days, cold nights, frost days, and ice days have decreased dramatically over the last two decades. Previous analyses by [31] found a shift in daily maximum and minimum temperatures toward a warmer mean with a faster increase in warm than cold events. To analyze trends in extreme climate indices, the Mann−Kendall test combined with the Sen linear trend estimator was calculated to determine the trend, magnitude, and statistical significance at p ≤ 0.05.

2.5. Maximum Lag Correlation

To investigate the relationship between the ENSO, IOD, and TAV and extreme weather indices, the maximum lag correlation based on the autoregressive vector was calculated, as shown by [28]. This involves calculating the cross-correlation between the ENSO, IOD, and TAV and the extreme weather indices. The maximum lag correlation coefficient between two series x(n) and y(n) is defined by Equation (1), which gives the maximum coefficient between the indices and the corresponding time lag [25].

$$La{g}_{Corr}\left(\tau \right)={\displaystyle \frac{1}{n-\tau }}.\sum _{k=1}^{n-\tau }\left\{\left[{\displaystyle \frac{x\left(k\right)-\overline{x}}{{\sigma }_x}}\right].\left[{\displaystyle \frac{y(k+\tau )-\overline{y}}{{\sigma }_y}}\right]\right\}$$

(5)

where $\tau$ is the time lag, ${\mu }_x$ and ${\sigma }_x$ are the mean and standard deviation of series x, and ${\mu }_y$ and ${\sigma }_y$ are the mean and standard deviation of series y, respectively.

3. Results

3.1. Performance of Datasets

Based on a methodology proposed by [32], an attempt was made to evaluate the capacity of datasets to estimate rainfall in different subregions of Africa, such as West Africa (represented by the Central West Coast basin), Central Africa (Congo basin), East Africa (Central East Coast), and Southern Africa (Orange). It was possible to explain the systematic biases, discrepancies, and similarities identified between the datasets evaluated in this study, based on the inter-annual variability of the seasonality and annual cycle of rainfall in the representative river basins.

3.2. Spatial Analysis of Performance of Chirps Dataset and ERA 5 in Relation to CPC

The analysis of the monthly rainfall bias using the CHIRPS, CPC, and ERA5 datasets revealed distinct patterns in different regions of SSA (Figure 1). A significant positive bias was observed between CHIRPS and CPC in Central and West Africa, suggesting that CHIRPS overestimates rainfall in these regions, while in Southern Africa CHIRPS tends to underestimate rainfall relative to CPC, although the biases are not statistically significant in all areas (Figure 1a). When comparing CPC and ERA5, a significant positive bias (>5 mm) was observed in several regions of northern SSA, while southern Africa showed a negative bias (Figure 1b). These results indicate that the choice of precipitation dataset can significantly influence climatic and hydrological analyses in SSA.

Evaluating the performance of the CHIRPS and ERA5 datasets for estimating rainfall in SSA revealed that, in general, both products showed good fits (KGE > 0.5, Figure 1d–f). However, significant discrepancies were observed in some specific regions, such as northern East Africa, central Central Africa, northwest southern Africa and the west coast of West Africa, where KGE values were lower than 0.5, especially for the ERA5 ensemble (Figure 1e). It is well localized and on a small scale for the CHIRPS ensemble (Figure 1d). Analysis of the KGE components revealed that these low values are associated with high biases (Figure 1a,b) (modeled values lower than those observed), high variability (Figure 1c,d), and low correlation (Figure 1e,f) with the reference dataset (CPC). The low correlation indicates that the CHIRPS and ERA5 datasets show discrepant trends and extreme values in relation to the CPC dataset, as pointed out by [24].

3.3. Monthly Precipitation Climatology in Congo, West Coast, Central East Coast and Orange River Basins of Sub-Saharan Africa

The analysis of the ERA5, CHIRPS, and CPC climate data, shown in Figure 2, reveals the seasonal rainfall patterns in the Congo, Central East Coast, West Coast, and Orange river basins. The Congo (Figure 2a) basin shows a well-defined hydrological cycle, with two rainy periods separated by a dry season, characteristic of regions under the influence of the Intertropical Convergence Zone. The three climate models reproduce this pattern, but with variations in the magnitudes and exact periods of the rainfall peaks. The discrepancies between the models can be attributed to differences in spatial resolution, calibration periods, and data sources. In the Central East Coast basin (Figure 2b), the unimodal pattern, with a rainy period concentrated between November and March, is influenced by sea breezes. The models also capture this pattern, but with some discrepancies in the rainfall estimates.

The West Coast basin (Figure 2c) has a seasonal rainfall pattern characterized by two annual peaks, concentrated in the transition periods between seasons, under the influence of the Intertropical Convergence Zone (ITCZ) in the Northern Hemisphere. The Orange basin (Figure 2d), located in a subtropical climate region, displays a unimodal pattern, with a rainfall peak during the austral summer. The ERA5, CHIRPS, and CPC climate models manage to reproduce these general patterns, but with significant differences in the magnitudes and timing of the rainfall peaks. The comparative analysis between the models indicates that CHIRPS and CPC tend to present more concordant estimates, especially in the West Coast and Central East Coast basins. In the Congo basin, ERA5 is closer to CPC, while CHIRPS shows some discrepancies, as observed by [33]. These differences can be explained by factors such as the spatial resolution of the models, the methodologies used to estimate rainfall and the quality of the input data, as highlighted by [34].

Figure 3 shows a detailed analysis of the temporal variability of rainfall in the Congo basin, dividing the period into four seasons and comparing the estimates of the ERA5, CHIRPS, and CPC climate models. The graphs reveal significant inter-annual variability in rainfall, both in terms of the magnitude and timing of peaks, for all seasons. Although the three models capture the characteristic seasonal patterns of the basin, with rainfall peaks concentrated in the periods from March to May and October to December, discrepancies are observed in the maximum rainfall values and in the exact timing of these peaks. The comparison between the models indicates greater agreement between CHIRPS and CPC, especially in the rainfall peaks, suggesting that these datasets can more accurately capture the variability of rainfall in the Congo basin. However, the differences observed can be attributed to factors such as the spatial resolution of the models, the methodologies used to estimate precipitation, and the quality of the input data, as pointed out by [35].

The analysis of the temporal variability of rainfall in the Central East Coast basin, shown in Figure 3, reveals significant inter-annual variability, with the ERA5, CHIRPS, and CPC models presenting different rainfall estimates for each season. The three models agree on the basin’s seasonal patterns, with more intense rainfall peaks during the summer (DJF) and fall (MAM) periods. However, there are discrepancies in the amplitude of the inter-annual variation, i.e., the difference between the driest and wettest years. These differences can be attributed to factors such as the spatial resolution of the models, the methodologies used to estimate rainfall, and differences in the geographical coverage of the datasets. Despite these discrepancies, the CPC and ERA5 models show greater similarity in the temporal variability of rainfall over the historical series.

The analysis of the temporal variability of rainfall in the West Coast basin, shown in Figure 4, reveals significant inter-annual variability, with the ERA5, CHIRPS, and CPC models indicating the occurrence of wetter and drier years in all seasons. The three models manage to reproduce the basin’s seasonal patterns, with rainfall peaks in all seasons except summer (DJF). However, there are differences in the amplitude of inter-annual variation and the timing of rainfall peaks between seasons and between models. These differences can be attributed to factors such as the spatial resolution of the models, the methodologies used to estimate precipitation, and differences in the geographical coverage of the datasets. The influence of the Intertropical Convergence Zone (ITCZ) is evident in the basin’s seasonal patterns, especially in the occurrence of rainfall peaks during periods of greater ITCZ activity. Despite the discrepancies found, the CHIRPS and CPC models show greater similarity in the temporal variability of rainfall over the historical series.

The analysis of the temporal variability of seasonal rainfall in the Orange basin, shown in Figure 4, reveals significant inter-annual variability, with the ERA5, CHIRPS, and CPC models indicating the occurrence of wetter and drier years in all seasons. The three models manage to reproduce the basin’s seasonal patterns, with rainfall peaks concentrated in the hottest periods of the year. However, there are differences in the amplitude of the inter-annual variation and in the exact timing of these peaks. These differences can be attributed to factors such as the spatial resolution of the models, the methodologies used to estimate rainfall, and differences in the geographical coverage of the datasets. The comparison between the models indicates greater similarity between ERA5 and CPC in representing the temporal variability of precipitation over the historical series.

The comparative analysis of the ERA5, CHIRPS, and CPC models carried out for the river basins of sub-Saharan Africa revealed good overall agreement in the representation of rainfall patterns. Although CHIRPS and ERA5 underestimated rainfall in some cases, CHIRPS was closer to CPC in the West Coast and Central East Coast basins. In the Congo basin, ERA5 was closer to the CPC. These results corroborate previous studies, such as [36,37], which highlighted the good performance of CHIRPS in capturing the spatial and temporal variability of precipitation in Africa. The comparative analysis between the CPC, CHIRPS, and ERA5 datasets indicated that CHIRPS showed the highest agreement(kge) with the CPC, considered as a reference, both on a basin and continental scale. Although some discrepancies were observed, such as differences in the timing of rainfall peaks, CHIRPS proved to be more reliable for analyzing rainfall variability in sub-Saharan Africa. Given the comprehensiveness of the data, the next climate analyses will focus on the CHIRPS datasets for precipitation and the CPC for temperature.

3.4. Spatial Distribution of Extreme Weather Indices for Precipitation (Based on CHIRPS Data) and Temperature (Based on CPC Data) in Sub-Saharan Africa

Figure 5 shows the spatial distribution of extreme climate indices for rainfall and temperature calculated using the CHIRPS and CPC datasets. The equatorial regions and West Africa, characterized by the influence of the Intertropical Convergence Zone (ITCZ), show the highest values of precipitation indices, such as the total number of days with precipitation (PRCPTOT) (Figure 5a), the duration of consecutive wet days (CWD) (Figure 5b), and the maximum amount of precipitation in 5 days (RX5DAYS) (Figure 5e). The frequency of heavy rainfall events, represented by the RX1DAY and RX5DAYS (Figure 5d,e) indices, is particularly high on the west coast of Africa and in some regions of southern and central Africa, where values in excess of 150 mm and 240 mm are observed, respectively. These events are associated with long periods of consecutive wet days (CWDs) (Figure 5b), which can exceed 32 days in some regions. In contrast, southern Africa, especially the western region, has a greater number of consecutive dry days (CDDs) (Figure 5c), indicating more arid climatic conditions. The combination of these climatic factors results in great spatial variability of rainfall in sub-Saharan Africa, with well-defined wet and dry regions.

From the CPC data, we can see that the southern region of SSA, characterized by mountainous terrain, had the lowest absolute minimum and maximum temperatures (TNn and TXn), with values around 9 °C (Figure 5g,f). In contrast, the center-north recorded values of over 24 °C for these same indices. As for the daily maximum and minimum temperatures (TXx and TNx), we observed values below 40 °C in the Central and Southern Africa sub-regions (Figure 5i,h), while the north showed values above this limit. The spatial variation of these indices reflects the influence of geographical and climatic factors, such as altitude and latitude.

3.5. Trend Analysis of CHIRPS and CPC Data

The analysis of trends in extreme weather indices shown in Figure 6 reveals contrasting patterns in sub-Saharan Africa. The central region of the continent shows a declining trend in total annual rainfall (PRCPTOT) (Figure 6a), accompanied by a reduction in the intensity and frequency of heavy rainfall events (RX1DAY and RX5DAYS) (Figure 6d,e), and in the duration of consecutive wet periods (CWDs)(Figure 6b). At the same time, there has been an increase in the number of consecutive dry days (CDDs) (Figure 6c), indicating an increased risk of drought. These results corroborate the findings of [38], who also identified a decrease in PRCPTOT and CWDs associated with an increase in CDDs in Central Africa. On the other hand, the northern and southern regions of sub-Saharan Africa show an upward trend in PRCPTOT, associated with an increase in the intensity and frequency of heavy rainfall events and in the duration of consecutive wet periods. However, an increase in the number of consecutive dry days (CDDs) is also observed in these regions. These results suggest an increased risk of flooding in the northern and southern regions, corroborating the findings of [39], who identified a trend of increasing PRCPTOT associated with a decrease in CWDs in West Africa.

The analysis of the trends in the daily minimum and maximum temperature indices, shown in Figure 6, reveals complex spatial patterns in sub-Saharan Africa. The minimum maximum temperature (TXn) (Figure 6f) shows a declining trend over most of the region, with the exception of some areas in the center and south. On the other hand, the minimum temperature (TNn) (Figure 6g) shows an upward trend across sub-Saharan Africa. The highest maximum (TXx) (Figure 6h) and minimum (TNx) (Figure 6i) temperature indices show an increase in the western and southern regions, while decreasing in the central and eastern regions. These results corroborate the findings of [39], who identified increasing trends in TXx and TXn, as well as TNn and TNx in West Africa, associating these trends with an increased risk of droughts. Further, Ref. [40] also identified significant upward trends in TNn. The results of this analysis indicate an intensification of temperature extremes in sub-Saharan Africa, with significant implications for ecosystems, agriculture, and human health.

3.6. Role of ENSO, IOD, and TAV on Precipitation Extremes

The distribution of maximum correlation between the ENSO, IOD, TAV, CDDs, and CWDs based on cross-correlation calculations are shown in Figure 7. This analysis is justified by the fact that modes of climate variability impact weather on timescales that vary from sub-seasonal to days, as demonstrated by [41]. Previous results demonstrated that ENSO-related soil moisture affects African agricultural production in distinct geographical location and crop types [11]. In addition, SST features related to IOD characteristics can modify the photosynthesis rate in response to anomalies in rainfall and vegetation greenness [42].It is not speculative to argue that all the above changes may be linked to changes in oceanic conditions across the Atlantic, Indian, and Pacific oceans.

As shown by [43], African rainfall variability also responds to tropical southeast Atlantic SSTs. In this sense, Lutz et al. (2015) demonstrated that lower SST leads to a seasonal decrease in rainfall along the west African coast.

Analysis of Figure 7a reveals a strong correlation between the ENSO Index and the number of consecutive dry days (CDDa) in West Africa and the equatorial belt of sub-Saharan Africa. El Niño events are associated with a substantial increase in CDDs in these regions, indicating a greater risk of prolonged droughts. However, this relationship is less pronounced in the eastern and southern regions of the continent. The pattern is anti-correlated with annual rainfall, as demonstrated by [44].

Analysis of Figure 7b reveals that the Indian Ocean Index (IOD) exerts a significant influence on the variability of the number of consecutive dry days (CDDs) in sub-Saharan Africa. There is a positive correlation between the warm phase of the IOD and CDDs in eastern and equatorial Africa, similar to the pattern observed during El Niño events. This relationship indicates that the warm phase of the IOD contributes to an increased risk of drought in these regions. However, in the central and southern regions, the relationship is negative, with the warm phase of the IOD associated with an increase in precipitation after a delay of approximately 6 months (Appendix A, Figure A1b). The co-occurrence of El Niño and IOD events, which show a positive correlation, can intensify rainfall extremes in sub-Saharan Africa, with significant consequences for agriculture and the population’s food security [28].

The analysis in Figure 7c shows that the Tropical Atlantic Area Activity Index (TAV) has a notable influence on the number of consecutive dry days (CDDs) in sub-Saharan Africa. A negative relationship is observed between TAV and CDDs in the central region of the continent, indicating that positive phases of TAV are associated with an increase in rainfall. However, in the southern region, the relationship is positive, suggesting that positive TAV phases contribute to an increase in the number of dry days. In the equatorial belt and the east coast of Africa, the TAV shows a positive correlation with the CDDs, with a lag of approximately 2 to 4 months (Appendix A, Figure A1c). On the other hand, on the west coast and inland central and southern regions, the relationship is negative, indicating that positive phases of the TAV are associated with rainy events. The TAV, which represents sea surface temperature anomalies in the tropical Atlantic, affects the CDD in South Africa with a delay of 4 to 6 months (Appendix A, Figure A1c). This temporal and spatial variability in the relationship between the TAV and the CDD highlights the complexity of climate teleconnections in sub-Saharan Africa.

Figure 7d shows that the El Niño−Southern Oscillation (ENSO) phenomenon has a significant influence on the number of consecutive wet days (CWDs) in sub-Saharan Africa. There is a negative correlation between ENSO and CWDs in West Africa, East Africa, and Southeast Southern Africa, indicating that La Niña events are associated with an increase in the number of wet days. However, in some regions of Central and Southern Africa, the relationship is positive. The Indian Ocean Index (IOD) shows a similar pattern to ENSO, with a negative correlation between the IOD and CWDs over much of the continent. Analysis of the correlation between the Tropical Atlantic Zone Activity Index (TAV) and CWDs (Figure 7f) reveals a positive relationship in southern Africa, with the exception of the east coast. However, the intensity of this relationship is lower compared to the correlations between ENSO and the IOD and CWDs. Analysis of the cross-correlation between ENSO and CWDs indicates a delay of 2 to 4 months in the response of CWDs to ENSO (Appendix A, Figure A1d,e). This faster response of the CWDs compared to the CDDs suggests that the increase in the frequency of days with precipitation in response to ENSO occurs faster than the establishment of dry spells in most of sub-Saharan Africa (Appendix A, Figure A1a–d).

In the same way, Figure 7 shows the spatial distribution of the maximum correlations and their respective lags between the ENSO, IOD, and TAV indices and the extreme rainfall indicators RX1DAY (maximum rainfall in one day) and PRCPTOT (total annual rainfall in wet days). There is a strong negative correlation between ENSO and IOD (Figure 7g,h) and RX1DAY in West, Southern, and East Africa, indicating that the cold phases of these indices are associated with a significant decrease in maximum rainfall in one day, suggesting extreme drought events. Figure 7h shows a positive relationship between the TAV index and RX1DAY in southern and eastern Africa, indicating a remote impact of sea surface temperature variability in the Tropical Atlantic on extreme precipitation events in this region. The RX5DAYS index (maximum rainfall in five days), not shown in the figure, shows a similar pattern to RX1DAY. The correlation values are higher for total annual precipitation (PRCPTOT), which is to be expected considering that oceanic climate variability modes act on longer time scales. Thus, the results indicate that both extreme daily precipitation events and accumulated seasonal and sub-seasonal precipitation in sub-Saharan Africa are influenced by the ENSO, IOD, and TAV modes of climate variability, with different spatial and temporal patterns (Figure 7j–l).

Analyses between ENSO and PRCPTOT show that they are not strongly correlated in West, Central, and Southern Africa, indicating that during the cold phase, there is a reduction in PRCPTOT (Figure 7j). The opposite is observed in East Africa, where the oceanic mode induces an increase in annual precipitation. Figure 7j shows a positive (negative) correlation between the IOD mode and PRCPTOT in almost all of SSA (with the exception of some portions such as central and southern Mozambique, the west coast of Namibia and Angola, and northern Congo), indicating that El Niño is accompanied by extreme precipitation/flood events (drought). The TAV index does not oppose the patterns presented by the IOD. With regard to the analysis of the temporal response, it can be seen that the PRCPTOT changes in phase with the ENSO, but is led for 5–6 months on the east coast, south and in the center of SSA by the IOD and TAV (Appendix A, Figure A1j–l). The above results may shed light on the influence of these oceanic modes on flooding due to the increase in the amount of rainfall over a short period–between 1 and 5 consecutive days. In line with what was shown by [45], it is stated that the IOD and Atlantic modes play a similar role to ENSO in driving changes in flood frequency in sub-Saharan Africa.

3.7. Role of ENSO, IOD and TAV on Temperature Extremes (TNn, TNx, TXx, TXn)

In association with precipitation, temperature plays a fundamental role in the livelihoods of people, due to its influence on water availability. In the last decades, furthermore, heatwaves have been longer, hotter, and more frequent, which lead to human mortality, heat stress, and reduction in crop production [46]. Thus, it is crucial to investigate the effect of these oceanic modes on temperature indices.

Figure 8 shows the spatial distribution of the correlation coefficients between the oceanic modes ENSO, IOD, and TAV and the indices of lowest daily minimum temperature (TNn) and highest daily maximum temperature (TXx) in sub-Saharan Africa. The analysis reveals a predominantly negative correlation between ENSO and TNn, indicating a reduction in minimum temperature during the cold phases of ENSO (La Niña) over a large part of the region (Figure 8a). However, a region located on the west coast of southern Africa, covering Angola and Namibia, shows a significant positive correlation, suggesting an increase in TNn during the warm phases of ENSO. The IOD and TAV indices, in turn, show a positive correlation with TNn in the southern, central and western regions of SSA, indicating an increase in minimum temperature during the warm phases of these modes of climate variability (Figure 8b,c). Analysis of the lag time reveals a delay of approximately 6 months between the IOD and TNn variability over almost the whole of SSA (Appendix A, Figure A2b), while the TAV has a more immediate impact, especially in southern and western Africa (Appendix A, Figure A2c). The ENSO, on the other hand, has an intermediate lag time, varying between 3 and 6 months (Appendix A, Figure A2a). The highest maximum temperature (TXx) shows a similar pattern to TNn in relation to the ENSO (Figure 8d), with a significant negative correlation in southern and northern Africa. However, the relationship between TXx and the IOD and TAV indices is more complex. While IOD and TAV show a positive correlation with TXx in the equatorial belt, the relationship becomes negative for TAV in the central and southern regions of SSA (Figure 8e,f). Analysis of the lag time reveals that the IOD and TAV influence TXx with a delay of approximately 5–6 months in the south and central regions (Appendix A, Figure A2e,f), while ENSO induces changes in TXx in a shorter timeframe, ranging from 2 to 4 months, especially in the south and north of SSA (Appendix A, Figure A2d).

The relationship between ENSO and TNx is positive in Central Africa, indicating an increase in minimum temperature during El Niño events (Figure 8g). The IOD shows a similar pattern to ENSO, with an increase in TNx during warm phases (Figure 8h). However, in the northern and southern parts of SSA, both ENSO and IOD induce a reduction in TNx. TAV, on the other hand, has a positive relationship with TNx throughout the SSA (Figure 8i). With regard to TXn, ENSO and IOD show a negative correlation in the equatorial band (Figure 8j,k), while TAV has a negative correlation throughout the SSA (Figure 8l). Analysis of the lag time reveals that ENSO and IOD dominate the changes in TNx in central and northwestern SSA, with a lag time of 4–6 months (Appendix A, Figure A2g,h). TAV, on the other hand, dominates the changes in TXn in the south and center of SSA, with a lag time of 2–3 months (Appendix A, Figure A2l).

Figure 9 shows additional details represented by river basins in sub-Saharan Africa and their respective inter-annual variability in extreme climate indices, namely CDDs, CWDs, TXx, and TNn. The choice of basins follows the study by [47]. A general picture reveals that CWDs is decreasing in most basins, which is highlighted in the Congo basin (Figure 9a). On the other hand, these basins are reacting to a substantial upward trend in CDDs, which is also evident in the other basins (Central-East and Orange). These changes in CDDs and CWDs largely reproduce the results based on global climate models [47] adjusted for biases.

As for temperature extremes, a positive trend can be seen in these basins (Figure 9e–h) for TNn and TXx, indicating that minimum and maximum temperatures are increasing. It is interesting to note that in the Congo and Orange basins, there are upward trends in TNn and more pronounced upward trends in TXx in the Orange basin. In all the basins, the inter-annual changes are smoother and are not statistically significant at 95%. These results indicate that reduced water availability related to reduced precipitation and rising temperatures affects water resources for hydroelectricity, fisheries, and water supply for domestic, industrial, and agricultural operations.

4. Discussion

Analysis of the temporal and spatial classification of precipitation performance revealed that the CHIRPS dataset showed better agreement with ERA5 data in the SSA catchments. These results are in line with the findings of [36], which demonstrated that CHIRPS outperforms ARC2 in terms of accuracy and lower bias.

The spatial distribution of trends in extreme weather precipitation indices, shown in Figure 6 reveals a general downward trend in total annual precipitation (PRCPTOT) (Figure 6a) and heavy precipitation events, represented by the RX1DAY (Figure 6d) and RX5DAYS (Figure 6e) indices. This trend is most pronounced in the central−northern region of sub-Saharan Africa. On the other hand, there is an increase in the number of consecutive dry days (CDDs). These results are in line with previous studies, such as [48,49], which identified a significant decrease in PRCPTOT and heavy rainfall events in East Africa. In addition, Ref. [50] observed a spatial distribution of PRCPTOT in West Africa, with a decrease in total annual rainfall as it moved from south to north, with the exception of some mountainous regions.

Evaluation of the trends in extreme temperature indices indicates that the daily minimum (TNn) and maximum (TXn) lower air temperatures show a gradually increasing pattern towards the north and center of SSA, while in the south, a decreasing trend is observed. In contrast, the highest daily maximum (TXx) and minimum (Tnx) air temperatures show an increase in the southern regions of the SSA and a decrease in the central regions. These results corroborate previous studies by [39,40,49], which also identified increasing trends in maximum and minimum temperatures in different regions of Africa.

Using the maximum-lag correlation method, it was possible to identify the relationships between extreme weather indices and the ENSO, IOD, and TAV modes of climate variability. The results indicate that the ENSO and IOD indices exert a similar influence on extreme precipitation indices, with a reduction in total annual precipitation (PRCPTOT) and heavy precipitation events (RX1DAY and RX5DAYS) in the equatorial belt. However, these same indices are associated with an increase in the number of consecutive wet days (CWDs) and a reduction in the number of consecutive dry days (CDDs) in the central and southern regions of SSA, suggesting an increase in the frequency of heavy rainfall events and flooding in these areas, corroborating the findings of [27]. With regard to temperature, the ENSO is associated with a general reduction in temperatures in SSA, while IOD induces an increase in temperatures in most of the region, with the exception of East Africa. The TAV index, on the other hand, shows a more complex relationship with temperatures, with increases in the lowest daily minimum temperature (TNn) and the highest daily minimum temperature (TNx) and reductions in the lowest daily maximum temperature (TXn) and the highest daily maximum temperature (TXx) in Central and Southern Africa.

With all of this, it is expected that in sub-Saharan Africa, especially in the agricultural sector, which is particularly sensitive to the mentioned changes, a negative impact on crop productivity may be expected. This will occur due to thermal stress and water shortage, which lead to conditions more conducive to the emergence of abiotic stress and the proliferation of pests and diseases. Additionally, changes in precipitation patterns can result in water scarcity for irrigation, impairing plant growth and leading to crop failure. Some areas currently suitable for the cultivation of certain crops may become unsuitable. This will require the implementation of new varieties of crops more resistant to heat and drought. In summary, the aforementioned changes in temperature and precipitation indices pose significant challenges to agriculture in sub-Saharan Africa, demanding adaptations and mitigation strategies to ensure food security, and the livelihoods of people in the region, which primarily rely on agriculture.

5. Conclusions

Analysis of the data reveals a downward trend in total annual rainfall on wet days (PRCPTOT) in central sub-Saharan Africa (SSA), due to a reduction in both the length of consecutive rainy periods (CWDs) and the maximum intensity of rainfall events, both over one day (RX1DAY) and five days (RX5DAYS). At the same time, there was an increase in daily minimum temperatures, both the lowest (TNn) and the highest (TXn), in the same region. This combination of decreasing rainfall and increasing temperature, together with a reduction in the number of consecutive dry days (CDDs), suggests an increase in the frequency and intensity of extreme drought events in the center of SSA.

Analysis of the correlation between extreme weather indices and the ENSO, IOD, and TAV modes of climate variability reveals that these oceanic modes exert significant control over the spatial and temporal distribution of extreme events in sub-Saharan Africa. The results indicate that the ENSO and IOD indices have a similar influence on the temperature (TNn, TNx, TXx, and TXn) and precipitation (PRCPTOT, CWD, RX1DAY, RX5DAYS, and CDD) indices, especially in the central and southern regions of the continent. On the other hand, the TAV index has the opposite influence on these indices, suggesting different teleconnection mechanisms. This heterogeneity in the response of extreme climate indices to the different modes of climate variability is consistent with the findings of [28], which demonstrate the low correlation between TAV and IOD/ENSO.

Specifically, the western Sahel’s main rainy season (July–September) is shown to be affected by the growth phase of El Niño through (I) a lack of neighboring North Atlantic sea surface warming; (II) the absence of an atmospheric column water vapor anomaly over the North Atlantic and western Sahel; and (III) higher atmospheric vertical stability over the western Sahel, resulting in the suppression of mean seasonal rainfall as well as a number of wet days. In contrast, the short rainy season (October–December) of tropical eastern Africa is impacted by the mature phase of El Niño through (I) neighboring Indian Ocean sea surface warming; (II) positive column water vapor anomalies over the Indian Ocean and tropical eastern Africa; and (III) higher atmospheric vertical instability over tropical eastern Africa, leading to an increase in the mean seasonal rainfall as well as in the number of wet days. While the modulation of the frequency of wet days and seasonal mean accumulation is statistically significant, daily rainfall intensity (for days with rainfall > 1 mm.day⁻¹), whether mean, median, or extreme, does not show a significant response in either region. Hence, the variability in seasonal mean rainfall that can be attributed to the El Niño–Southern Oscillation phenomenon in both regions is likely due to changes in the frequency of rainfall.

These conclusions highlight the importance of understanding and monitoring changes in extreme weather patterns as well as verifying continuously in advance the influence of oceanic climate modes in sub-Saharan Africa to develop adaptation and mitigation strategies, to deal with challenges arising from current and future climate change in the region.