Sub-Seasonal Rainfall Variability and Atmospheric Dynamics During East African Long-Rain

Abstract

East Africa’s March–April–May (MAM) rainfall exhibits pronounced variability that strongly influences agriculture, water security, and livelihoods. This study analyzes consecutive wet day (CWD) events using CHIRPS precipitation, GridSat infrared cold-cloud brightness temperature, and ERA5 reanalysis for 1982–2023 to examine rainfall variability and its relationship with atmospheric circulation and convection. CWDs are classified into short (3–5 days), medium (6–10 days), and long (>10 days) events. Results reveal three regional activity centers: the Eastern Congo Basin, Lake Victoria, and Southwest Ethiopia. The Congo Basin emerges as the most convectively active region, sustaining frequent events across all categories and supporting long-duration rainfall through persistent moisture flow and mesoscale convection. On average, CWDs contribute 43% of total MAM rainfall across East Africa, ranging from negligible amounts in arid areas to over 90% in equatorial regions. Short-duration events dominate the seasonal total, while long-duration events, though spatially restricted, contribute up to 52% locally. Composite convection analysis shows a transition from widespread moderate activity during short events to highly localized, intense convection in long events, particularly over the equatorial Congo and Lake Victoria regions. These findings highlight the critical contribution of organized synoptic-scale systems to East Africa’s hydrological cycle, which will have implications for improving sub-seasonal rainfall forecasts.

1. Introduction

Most of the rainfall in North Africa, north of about 10° N, occurs in boreal summer, while most of the rainfall over the region south of 10° S is received during the boreal winter. Equatorial Africa within ±10° N/S receives rainfall multiple times during a year, mainly in March–May (MAM), locally referred to as long-rain, and in September–November (SON). The MAM season is critical for the region’s agriculture, water resources, and livelihoods. Yet, its variability remains challenging to predict due to the complex interaction between local and large-scale atmospheric and oceanic mechanisms, e.g., [1,2]. East African seasonal rainfall varies across multiple temporal and spatial scales (from diurnal to interannual timescales and across the region). This spatio-temporal rainfall variability is closely tied with the north–south seasonal excursion of the deep convective zone, near and remote mechanisms such as sea surface temperature anomalies and associated moisture transport, e.g., [3,4].

We focus on rainfall variability that determines this variability over East Africa, a region located east of 20° E and within ±10° N/S (see Figure 1). As noted above, the main rainfall season is boreal spring (March–May). A secondary peak of rainfall occurs in October and November, e.g., [1]. The southern part (southern Kenya and Tanzania) receives most of its annual rainfall during the boreal winter (December–February [1]).

East Africa is dominated by a complex topography ranging from above 5000 m above sea level at Mt. Kilimanjaro in Kenya/Tanzania to about 100 m below sea level at the Danakil Depression in Northeast Ethiopia (Figure 1). The region’s terrain includes the Great East African Rift Valley, stretching from Tanzania through Kenya, central Ethiopia, and meeting the Red Sea in the vicinity of Yemen. We must also note the low-lying area (below 500 m.a.s.l. (m above sea level)) between northwest Kenya and south–southwest Ethiopia, widely known as the Turkana channel or the River Omo catchment. Large water bodies such as Lake Victoria (enclosed between 1000 and 1500 m.a.s.l.) also influence rainfall patterns. Local topographic effects significantly influence precipitation and its predictability. Ogwang et al. [5], using a regional climate model, investigated the influence of topography on East African climate during the October–December season. Their results demonstrate that topographic features substantially modulate regional precipitation patterns, with the model showing significant sensitivity to terrain representation. The authors found that removing topography led to considerable changes in rainfall distribution, highlighting how complex terrain interacts with large-scale atmospheric circulation to produce the observed spatial variability in East African precipitation. These topography–circulation interactions contribute to the challenges in accurately predicting regional rainfall patterns.

The economies of East African countries depend largely on rain-fed agriculture and small-scale irrigation systems, which are sensitive to extreme fluctuations in weather and climate. The region is particularly prone to frequent droughts and heavy floods (e.g., [6]). While excessive heavy rains have been observed recently (e.g., heavy short-term rains and floods in Kenya in 2018; [7]), extended droughts are also frequent in the region [8]. For instance, Kenya experienced exceptional rainfall during the 2018 long-rain season (March–May), with many areas recording more than twice their normal precipitation and some stations reaching close to or up to three times the long-term average, leading to widespread flooding and making it one of the wettest long-rain seasons on record [7]. Similarly, the northwestern part of East Africa has experienced anomalous rainfall surpluses during some years, such as the 2019–2020 short rains season that brought above-normal precipitation to parts of the region, e.g., [1,9]. At the same time, prolonged drought conditions, for example, from 2020 to 2023, have persisted in the southern and eastern regions, including southern Ethiopia, Somalia, and northern and eastern Kenya, e.g., [10].

While past studies suggest that the long-rain season is highly variable and influenced by large-scale SST anomalies over the Indian and Pacific oceans, including atmospheric disturbances, our understanding of the interaction between local physical factors and remote mechanisms and circulation patterns that govern variability over East Africa remains limited. Understanding the variability at a range of scales and its association with mechanisms is critical for shorter and longer-term scale forecasts. Unlike other tropical regions, East Africa’s forecast skill at medium (5–10 days) and longer (>10 days) scales is low. This study attempts to fill this gap by investigating sub-seasonal processes, specifically consecutive wet days, and how the interaction between local topographic forcing and regional and global mechanisms affects sub-seasonal variability during the MAM season.

2. Materials and Methods

2.1. Data Source

This study combines high-resolution rainfall and convection data sets (CHIRPS, GRIDSAT) with reanalysis products to investigate the MAM variability.

(a)

The Climate Hazards Group Infrared Precipitation with Station data (CHIRPS) version 2 provides precipitation estimates combining thermal infrared and microwave satellite observations with station data. The dataset is available for 50° S to 50° N globally at a horizontal resolution of 0.05° × 0.05° and at daily intervals, covering the period 1981–present [11]. CHIRPS data has been used for various purposes in Africa, for drought and hydrological monitoring, and for early warning on food security by the FEWS NET (Famine Early Warning Systems Network), among others. Given our research objectives of examining sub-seasonal and interannual rainfall variability and consecutive wet days, this study utilizes CHIRPS.

(b)

The Gridded Satellite (GRIDSAT-B1) dataset provides infrared brightness temperature (Tb) measurements created by merging observations from multiple geostationary satellites, including GOES (Geostationary Operational Environmental Satellite), Meteosat, GMS (Geostationary Meteorological Satellite), and MTSAT (Multi-functional Transport Satellite) series. The dataset was constructed based on the International Satellite Cloud Climatology Project (ISCCP) version B1 data, which archives observations from international geostationary satellites at approximately 10 km resolution at 3-h intervals [12]. The dataset has a spatial resolution of approximately 0.07° and is available at a 3-h time interval, covering the period 1980–present [12].

(c)

The ECMWF ERA5 reanalysis dataset provides comprehensive atmospheric, land surface, and ocean data from 1979 to the present (e.g., [13]). The data have a spatial resolution of 0.25° × 0.25° and a temporal resolution of 1 h. Dynamic and thermodynamics variables are obtained from ERA5. It is to be noted that ERA5 has some limitations in estimating wind over the mountain chains and associated valleys [13]. However, it provides reasonable estimates on the long-term spatio-temporal scale (more details in Section 3).

2.2. Methodology

Consecutive wet days are identified using CHIRPS (1982–2023) daily rainfall data. A wet day is defined as a day with rainfall ≥ 1 mm, following standard meteorological practice [14]. Consecutive wet days are then determined by counting successive wet days without interruption. Based on the duration, consecutive wet day events are classified into three categories:

• Short duration: 3–5 consecutive wet days;
• Medium duration: 6–10 consecutive wet days;
• Long duration: >10 consecutive wet days.

Clearly, these categories are in synoptic and longer time scales and do not include diurnal cycle, although it is known that most of the rainfall variance in the tropical regions is explained by the diurnal variability [1,8]. As indicated in Section 1, the purpose of this classification is to highlight the role of extended rainfall events on seasonal rainfall variability.

For each grid point in the study region, the frequency of each category during the MAM season is calculated using the CHIRPS dataset, which provides consistent long-term precipitation records from 1982 to 2023 for climatological analysis of consecutive wet day patterns.

To quantify the rainfall contribution from each category, we calculate the percentage of total seasonal rainfall that occurs during consecutive wet-day events of each duration. This approach allows us to identify the contribution of the frequency of organized precipitation systems and their actual impact on seasonal rainfall.

The rainfall from consecutive-day categories is also matched with associated deep convection using GridSat infrared brightness temperature (Tb) data as a proxy for deep convection. Then, deep convection is identified using a threshold of Tb ≤ 235 K, which corresponds to cloud tops reaching the upper troposphere at approximately 200–300 hPa pressure levels (e.g., [15,16]). This threshold has been used in previous studies to identify deep convective systems. For example, Machado et al. [16] defined convective clusters as areas covered by adjacent cloud cells with Tb lower than the threshold. Similar approaches include Tissier and Legras [17] using 230 K threshold for deep convective clouds (see also [18]).

3. Climatological Background

To provide context to our results, we present a series of mean rainfall, sea level pressure, relative humidity, and associated wind flow at various pressure levels for the period 1981–2023 in Figure 2, Figure 3 and Figure 4.

East Africa south of about 5° N does not benefit from the boreal summer rains as the peak of the intertropical convective zone shifts far to the north. Therefore, in Figure 2, we present the climatology of rainfall during March–May (MAM), September–October (SON), December–February (DJF) seasons. As noted above, the MAM is the main rain season for the study area in East Africa (long-rain), while the region also benefits from SON rains (short-rains). Our focus is on MAM season, where rainfall is dominantly shown mainly within 10° of the equator. During SON, the rainfall pattern shows a broader latitudinal spread from approximately 7° N to 5° S and west of 30° E, while in DJF, most of the rainfall is observed mainly south of 10° S.

Eastern Congo, Lake Victoria region, southern and southwestern Kenya, and Southwest Ethiopia receive high rainfall exceeding 8 mm·d⁻¹ during MAM. We note a rainfall minimum over the low-lying valley across eastern Kenya and over a southeast–northwest oriented zone over Northwest Kenya through southern South Sudan (c.f. Figure 1). In the literature, this area is referred to as the Turkana Channel (e.g., [19,20]; c.f. Figure 3). Although this zone serves as the conduit for water vapor transport from the Indian Ocean during MAM (Figure 3), it is also a zone of low-level divergence and rain-shadow area (e.g., [8,20]).

Most of the rainfall in the SON season centers across the equator, a southward migration from the MAM season. While eastern and central Congo and the Lake’s region receive high rainfall amounts, most of eastern Kenya and southwest Ethiopia receive relatively lower amounts. The maximum rainfall zone shifts further southwards during the DJF season. Southeast Congo and Tanzania receive higher amounts in DJF. In summary, the study area between ±10° receives most of the rainfall during MAM, and below we superimpose the low-level wind flow to explore the moisture sources.

Figure 3 presents the MAM rainfall, mean sea-level pressure (MSLP) and 10 m wind flow. As seen in Figure 2, peak MAM rains are associated with high terrain in East Africa. The region bordering southwestern Ethiopia and northern Kenya receives very little rainfall (under 2 mm/day). A prominent ridge running from eastern Tanzania through central Kenya (east of the peak rainfall) is seen. As also shown in Figure 3, relatively weak high pressure dominated east-central Ethiopia, east of the Rift Valley. Notable features include the dominant surface wind flow into East Africa originating from the tropical southwestern Indian Ocean and relatively weak southwesterly flow from the equatorial southeastern Atlantic Ocean towards western Congo. Similar wind patterns are observed at 925 [21] and 850 hPa levels (Figure 4), although wind estimates at 925 hPa can be biased as East Africa features a complex terrain (Figure 1). However, as shown in Figure 1, most of the region lies above 1000 m, with high mountains that are localized. Central Ethiopia, north of 10° N, exhibits mountain chains, but a large part of that is outside the study area and does not benefit from MAM rains. We note that past work (e.g., [21]) used 925 hPa to study low-level wind and geopotential to study boreal spring and fall rains over East Africa and presented reliable estimates of the larger-scale circulation.

Figure 3 also shows a notable wind convergence zone along 10° N and a strong low-level southeasterly flow between northeastern Kenya and the south–southwest Ethiopian border. As indicated above, the strong flow over this area is part of the low-level Turkana Jet [19]. The general flow also suggests topographic channeling. The peak moisture and rainfall are located to the south of the wind convergence. Additionally, the figure suggests a southwest–north–east-oriented wind confluence and channeled flow that runs through Congo. The source of moisture for this region is the southern Atlantic and in situ over the vegetated area of the Congo, evaporation from the water bodies, and the Indian Ocean. Past studies showed that the moisture transport from the Indian Ocean is critical for the MAM rainfall (e.g., [4]). Vizy and Cook [21] highlighted the importance of low-level cross-equatorial flow transitions in modulating East African rainfall, but the specific mechanisms linking atmospheric circulation patterns to consecutive wet day patterns and rainfall persistence characteristics remain unclear.

The spatial distribution of relative humidity (shaded, %) and horizontal wind flow at 850 and 700 hPa (Figure 4) show moisture availability and transport pathways originating mainly from the south Indian Ocean but also from the Atlantic Ocean. The relative humidity (RH) distribution at 850 hPa (left panel) shows enhanced low-level moisture (>80%) over eastern Tanzania, the Lake Victoria region, southern Ethiopia, and eastern Congo, with wind vectors indicating strong onshore advection into the continent. Elevated moisture availability (RH > 80%) is also observed in the middle troposphere over peak MAM rainfall areas (c.f. Figure 2 and Figure 3). As expected, the 700 hPa wind flow is mainly easterly. The figures suggest active boundary layer moisture inflow, consistent with the Indian Ocean being the primary moisture source for the East African long-rain season (e.g., [22,23]). The presence of elevated moisture at both 850 hPa and 700 hPa highlights the deepening of the moist layer and a conducive environment for deep convective activity (e.g., [24,25,26]). Taken together, these results suggest that the observed large-scale flow supports deep boundary-layer moisture transport, enhancing the likelihood of organized convection and rainfall during the MAM season.

4. Results and Discussion

In this section, we begin the rainfall analysis by identifying dry (rainfall < 1 mm/d) from wet (rainfall $\ge 1 \mathrm{m}\mathrm{m}/\mathrm{d})$ periods during the long-rain (MAM) season over East Africa, as, for example, in Panthou et al. [27], who identified wet and dry days using this threshold. The variability of extended wet (consecutive wet periods) at various timescales was studied to highlight the nature and changes over time of MAM rainfall over the 42-year period.

Previous studies have established that rainfall in Africa is predominantly produced by mesoscale convective systems (MCS), along with synoptic and larger-scale atmospheric disturbances playing a secondary role (e.g., [28,29,30]). The MCSs exhibit a characteristic feature of temporal clustering that creates extended wet and dry periods, which impact regional hydrology and agricultural systems. The persistence of these rainfall patterns reflects interactions between large-scale circulation features and local forcing mechanisms, and categorization of consecutive days is consistent with MCS clustering. Our analysis focuses on consecutive wet days (rainfall ≥ 1 mm/d) to address knowledge gaps regarding sub-seasonal processes and rainfall mechanisms.

4.1. Wet and Dry Days

The spatial distribution of average wet days during the MAM season shows pronounced variability across East Africa, with values ranging from fewer than 10 days (<10% of the time) in climatologically drier regions to over 70 days (raining about 76% of the time) in the wettest areas (Figure 5).

Consistent with Figure 2 and Figure 3, wet day frequencies are maximum over eastern Congo, Lake Victoria region and southwestern Ethiopia (Figure 5). A strong north–south and west–east wet day gradient is also observed.

A dominant feature in Figure 5 is “hotspots”, regions of peak frequencies (see also [29]). Peak wet day frequencies occur in three distinct areas that are either in the vicinity of elevated terrain (Lake region and SW Ethiopia) or enhanced moisture areas (eastern Congo; c.f. [31]. While elevated areas show higher wet day frequencies than surrounding lowlands, the East African Rift Valley creates notable discontinuities in the distribution pattern. As summarized in Section 2, the Rift Valley area over eastern Kenya and the Turkana valley is a rain–shadow area (Figure 2 and Figure 3). Further, coastal regions along the Indian Ocean exhibit moderate wet day frequencies (20–35 days), although closer to the moisture source, but may be limited by absence of lifting mechanisms. The northern regions (Somalia, eastern Ethiopia) show the lowest frequencies (5–15 days), consistent with their location in low-lying areas and areas of divergence (Figure 3).

The primary wet days maximum occurs around the eastern half of Congo, with the highest frequency of wet days exceeding > 70 days per year, representing nearly 85% of the of the MAM season, which also indicates almost daily rainfall occurrence. Frequent rain events reflect a conducive environment for daily convective activity for rainfall production. This result is consistent with past findings that suggested that the Congo region is one of the most convectively active regions in the world (e.g., [32]). It is an area of high evapotranspiration and easterly moisture inflow (Figure 3 and Figure 4) that interacts with the local environment (see Figure 1, higher elevation over the eastern border; [31]). Also, as shown in previous studies (e.g., [33,34]), the Congo Basin’s elevated wet day frequencies are associated with intense afternoon convection driven by differential heating processes [32,33].

Another local maximum with 50–60 wet days per year is seen around Lake Victoria. The wet days maximum around Lake Victoria is associated with local lake–mountain contrasts compared to the surrounding areas. We note the complex topography surrounding the Lake, with high hills flanking the eastern and western shores creating a natural basin that influences moisture transport patterns (e.g., [35]). The mountain chain to the east (elevations ~ 3000 m) and to the west of Lake Victoria (2000–2500 m; the highlands of western Uganda and Rwanda) create orographic barriers that channel and focus atmospheric moisture towards the lake basin through enhanced orographic lifting processes. Consistent with Figure 4, Anyah & Semazzi [31] suggested peak rains occur over the western (upwind) sides of rift escarpments, where peak frequencies of daily rain events occur. Reduced frequencies in Rift floor areas over central and eastern Kenya indicate consistent rain shadow effects [31].

Southwest Ethiopia presents a third regional maximum with about 45–55 wet days per year. Camberlin [36] suggested that this region is positioned at the convergence zone of multiple moisture sources that create favorable conditions for sustained convective activity. The higher elevation of the southwest Ethiopian Highlands also promotes orographic lifting of southerly-southwesterly moisture transported from equatorial regions, while the southern rift valley system channels moisture inflow that is particularly important across southern Ethiopia (e.g., [37,38]). Additionally, the transition period dynamics between the northward and southward migration of the convective zones during MAM contribute to extended wet periods over this region.

4.2. Spatial and Temporal Variability of Consecutive Wet Days

In this section, we explore the variability of the three categories of extended wet days. Figure 6 reveals distinct regional patterns across East Africa similar to wet frequencies (c.f. Figure 5), with three main maximum centers over the duration categories.

Eastern Congo represents the most prominent feature across all duration categories. For short-duration events (3–5 days, left panel), this region exhibits the highest frequencies, exceeding 7 events per year. Significant activity in medium-duration events (6–10 days, middle panel) is also observed with 4–6 events per year and shows notable persistence in long-duration events (>10 days, right panel) with 2–3 events per year. This sustained activity across all categories reflects the region’s position among the maximum convection areas in the tropics. As also seen in Figure 4, the lower- and middle-level troposphere, on average, is moist (RH > 80%), consistent with past studies that documented that the Congo area is moisture-rich, which supports consistent deep convection. It is also to be noted that the area is a moisture source for monsoon rains in East Africa (e.g., [39]).

Lake Victoria region shows the second major center of action, with pronounced maxima in short-duration events (7+ events per year) around the western shores of the lake. The region maintains enhanced frequencies in medium-duration events (4–8 events per year) but shows reduced activity in long-duration events, primarily concentrated in small areas around the lake margins. This, in part, suggests the short-scale influence of the high terrain in the region.

Southwest Ethiopia presents the third center, showing significant activity in short-duration events (up to 5 events per year of uninterrupted 3–5 days of rainfall). Reduced activity is observed in the 6–10 days consecutive duration events (approximately 1 event per year), while the long-duration events seem to be rare.

As summarized above, the three regions are the peak rainfall activity areas, but there is also a regional difference in the duration of events that highlights the importance of proximity to abundant moisture and differences in local drivers. Southwest Ethiopian convection and rainfall benefit from seasonal moisture availability during MAM but lack the persistent moisture sources necessary for extended (>10 days) wet periods that characterize the Congo basin.

The systematic decrease in spatial coverage from short to long-duration events demonstrates the increasing selectivity of atmospheric conditions required for sustained rainfall persistence. While short-duration events (3–5 days) show widespread occurrence across East Africa, long-duration events (>10 days) are restricted primarily to the equatorial Congo Basin and limited areas around Lake Victoria, where moisture (Figure 4) and atmospheric boundary conditions can sustain extended convective organization.

Further, the persistence of consecutive wet days is fundamentally linked to the organization and lifecycle of mesoscale convective systems (MCSs). However, it is important to distinguish between MCS intensity and duration. While Taylor et al. [18] documented increased intensity of MCSs in the Congo Basin during February, intense convective systems do not necessarily translate to longer-lasting rainfall events. Extended wet periods (>10 days) more likely result from the clustering of multiple shorter-lived convective events stirred by large-scale atmospheric disturbances, such as Kelvin waves or specific phases of the MJO [40], rather than individual long-lived convective systems [41].

In summary, the eastern Congo basin is a densely forested area in the tropics and is characterized by high temperature, which drives a high rate of evapotranspiration. This induces a favorable environment for sustained convective development. Thus, two primary mechanisms may explain extended wet periods: persistent light precipitation (and evaporation that keeps the atmosphere unstable) and moisture influx from near water bodies, and the repeated regeneration of MCS (e.g., [18,42,43]). Laing and Carbone [43] suggested that the longer persistence via MCS clustering is in response to larger-scale atmospheric forcing. Therefore, it is no surprise that eastern Congo is the rainiest area with a short and long duration of consecutive wet events. The areas in central and eastern Kenya and the area just south of Ethiopia are low-rainfall area and this is related to local differences in elevation and moisture availability for convection.

Figure 7 shows the interannual variability of consecutive wet day frequencies during the three primary activity centers identified in Figure 6. The figure summarizes short-duration (3–5 days), medium-duration (6–10 days), and long-duration (>10 days) precipitation events normalized by grid point density. The figures provide additional quantitative substantiation of the spatial patterns documented in Figure 5, revealing both regional differences and their tendencies over the 42 year period.

Eastern Congo is the dominant center of consecutive rainfall activity across all temporal scales, with mean frequencies of 5.2 ± 0.8 events per grid point per year for short-duration events, 2.6 ± 0.3 for medium-duration events, and 1.0 ± 0.3 for long-duration events. These normalized frequencies directly correspond to the highest activity zones (7+ events per year) observed in Figure 5, consistent with high moisture flow into East Africa (e.g., [44]). The 3–5 days events over eastern Congo appear to be mostly above average, while the 6–10 and >10 events are within the average. The short-term events show an increasing tendency, although not very significant, and this may be similar to Taylor et al.’s [29] findings of MCS increasing and intensifying.

Lake Victoria demonstrates intermediate activity levels with mean frequencies of 4.7 ± 0.9, 2.0 ± 0.4, and 0.6 ± 0.3, respectively, for 3–5, 6–10, and >10 days periods. This is consistent with the result of moderate-to-high activity zones (4–7 events annually; Figure 5) concentrated around the lake margins, while the temporal variability (coefficient of variation ~20%) reflects the region’s sensitivity to interannual climate fluctuations driven by complex lake–atmosphere interactions and surrounding topographic influences [45]. While the time tendencies for short and medium terms remain within the average limits, the >10 days events show some increase, albeit small, after 2005.

Southwest (SW) Ethiopia exhibits the lowest mean frequencies (3.8 ± 0.7, 1.2 ± 0.3, and 0.2 ± 0.1, respectively, for 3–5, 6–10, and >10 days categories). The short-term events per area are nearly unchanged over time in SW Ethiopia. However, longer time events are relatively rare as shown above (e.g., [46]). The above analysis leads us to examine the interannual variability of rainfall at the three regions in more detail.

4.3. Temporal Variability of Consecutive Wet Days

The spatial analysis presented in Figure 5, Figure 6 and Figure 7 reveals three key findings about East African MAM rainfall patterns: (1) wet day frequencies show pronounced spatial heterogeneity, with peak values (>70 wet days per year) concentrated in the Eastern Congo Basin, moderate-to-high frequencies (50–60 days per year) around Lake Victoria, and another maxima (45–55 days per year) over Southwest Ethiopia (Figure 5); (2) consecutive wet day patterns demonstrate distinct regional characteristics across duration categories, with the Eastern Congo Basin showing the highest frequencies across all durations (3–5, 6–10, and >10 days), Lake Victoria maintaining strong activity in short and medium durations, and Southwest Ethiopia showing activity primarily in shorter duration events (Figure 6); and (3) long-duration consecutive wet day events (>10 days) are spatially restricted to equatorial regions, particularly the Congo Basin and limited areas around Lake Victoria, indicating that sustained rainfall persistence requires specific conditions most consistently present near the equator.

Furthermore, past work based on observations and future projections of rainfall for East Africa noted a contrasting result between observational rainfall trends and projections, leading past work to characterize this discrepancy as an “East African rainfall paradox” (e.g., [2,47]). It is to be noted, however, that past work mostly contained analyses based on an extended region consisting of several countries with different climate zones. As shown in Figure 5 and Figure 6, the rainfall climate over East Africa is heterogeneous. This leads us to consider rainfall analysis based on similar characteristic patterns such as “hotspots” or homogenous zones (e.g., [27,30,48]). Therefore, building on these spatial patterns, interannual variability analysis examines rainfall variability for the three core regions or hotspots during MAM (Figure 8).

Figure 8 shows strong interannual fluctuation. In general, the number of below and above average rainfall years were nearly equal and within one standard deviation ($\sigma$). However, extreme wet and dry years were also observed. For example, 2008 was a very deficient year over eastern Congo, and 1987 was the wettest (above 2$\sigma$). Over the Lake Victoria region, 1984 was the driest (around −2$\sigma$), while 2018 was the wettest (near 2σ above average). The driest year over SW Ethiopia was 2003, while the wettest year was 2014 (>2$\sigma$). Extreme variability in rainfall has been reported in the recent literature. For example, Kilavi et al. [7] reported a record rainfall during March–May 2018 over Central Kenya. Overall, we note a common pattern over the East African regions in that rainfall shows an increasing trend since 2010, and some years were very wet.

The analysis reveals remarkable regional contrasts during notable wet and dry periods. The period 2019–2022 stands out as particularly anomalous across all three regions, with SW Ethiopia experiencing the most severe reductions (rainfall anomalies below −1.0$\sigma$), while Eastern Congo showed more moderate but persistent negative anomalies. Conversely, 2018 shows coherent positive anomalies across all three regions, with Lake Victoria region recording exceptional values approaching more than 2$\sigma$. Similarly, 2020 demonstrates strong positive anomalies in Lake Victoria and SW Ethiopia, while eastern Congo experienced negative anomalies. In the literature, e.g., [7], those extreme wet periods were linked to the impacts of the intraseasonal oscillation, particularly, the Madden–Julian Oscillation, and to active tropical cyclone activity in the Southwest Indian Ocean. Persistent dry conditions, particularly the 1980s dry periods, were linked to anomalous eastern Pacific SSTs and Southern Oscillation events that correspond to severe food shortages reminiscent of the devastating famine of that decade (e.g., [49,50,51]).

4.4. Rainfall Contribution from Consecutive Events

The previous sections examined the frequency of consecutive wet day events and their impact on seasonal rainfall. It is also important to understand precipitation systems that are organized in short and extended periods and the contribution they make to the seasonal rainfall. Rainfall analysis based on this approach will also give insight as to whether a recent observed rainfall increase (more above average rains since 2010 across the regions) can be explained by is a contribution from an extended rainfall period as opposed to diurnal or very short-term (<3 days) rainfall periods. As described above, an extended period of rainfall is a result of organized, longer-lived active convection. As suggested in Mekonnen and Rossow [52,53] and Rossow et al. [54], most of tropical rainfall is derived from well-organized, larger area covered deep convective systems (mesoscale convective clusters), while scattered and isolated convective systems contribute much less rainfall. Consecutive rains of more than 3 days are produced by a well-organized and large area deep convective systems [52,53,54]. Figure 9 presents the contribution of consecutive events to the MAM rainfall.

As may be expected, short-duration events contribute most significantly to total seasonal rainfall, with contributions in the range 25–50% particularly in highland and coastal regions, while longer-duration events show more limited contributions as they are much less frequent (see Figure 5, Figure 6 and Figure 7). The highest contributions (40–50%) occur over the Eastern Congo Basin, extending eastward through the Lake Victoria region and western highlands of Kenya and Uganda. This pattern reflects the dominance of short-term convective events, which typically maintain organized convection for 3–6 days before dissipating (e.g., [55,56]).

Medium-duration events (6–10 days) contribute 20–25% over the moist areas of eastern Congo and high terrain around Lake Victoria, indicating preferential areas of orographically induced convection. Long-duration consecutive events (>10 days) show highly localized contributions, primarily concentrated in the equatorial Congo Basin, where they account for 10–20% of seasonal rainfall. These patterns are indicators of the influence of longer time-scale atmospheric disturbances such as intraseasonal oscillations, which modulate synoptic and longer convective time scales (e.g., [1]). We note that outside the Congo Basin, long-duration events contribute less than 5% of seasonal rainfall, indicating their relative rarity in most of East Africa.

4.5. Composite Analysis of Convective Activity

To further examine the consecutive rain events, we compute composites of deep convection to better understand convective organization across East Africa. Figure 10 presents the deep convective activity during consecutive events in short, medium, and longer time-scale categories.

The left panel shows that short-duration consecutive events are associated with widespread but relatively moderate deep convective activity (Tb ≤ 235 K occurring 6–18% of the time during these events). The spatial pattern exhibits broad coverage across the equatorial belt, with enhanced activity over the Eastern Congo Basin (12–18%) and moderate activity extending through the Lake Victoria region and into the Ethiopian Highlands (9–15%).

The pattern becomes more spatially spread during medium-duration events (middle panel), with enhanced deep convective activity primarily concentrated in the Eastern Congo Basin (18–24% frequency). The Lake Victoria region maintains moderate activity, but the spatial extent of significant convection is notably reduced compared to shorter events.

For long-duration (right panel), we observe mostly spatial concentration and intensity, with convective activity reaching maximum frequencies (36–48%) in highly localized areas within the Eastern Congo Basin and extending into southwestern Ethiopia. Additional patches of intense activity appear in the northern Ethiopian Highlands and scattered locations around the equatorial belt. We note that the percentage contribution from frequent cold-cloud top temperatures in long-duration events (>10 days of Tb ≤ 235 K) is greater than medium- or short-duration frequencies, while the rainfall contribution from shorter and medium range is much more than long-duration events (see, e.g., Figure 9). We remind that cold-cloud top temperatures ≤ 235 K include thick anvil clouds that may not rain out as suggested in Mekonnen and Rossow [51,52].

Additionally, we investigated the interannual variability of frequencies of deep convection over the three key regions (Figure 11). As shown in Figure 11, eastern Congo exhibits frequencies of convective variability ranging from −3$\sigma$ in 1984 to +2$\sigma$ in 1987 and 1991, consistent with, respectively, very dry and very wet rainfall years found earlier (c.f. Figure 8). Positive convective anomalies during 1987–1988, 1997–1998, and 2015–2018 in eastern Congo correspond with the largest positive rainfall anomalies in Figure 6. These periods coincide with documented El Niño events and enhanced Atlantic–Indian Ocean moisture transport interactions (e.g., [1,57]).

Lake Victoria presents convective variability in the range −2$\sigma$ to +1.5$\sigma$. The recent period (2015–2020) shows correspondence between sustained positive convective anomalies and rainfall increases (+1.5$\sigma$ to +1.8$\sigma$), representing some of the largest positive precipitation anomalies in the 42-year record. This recent increase in rainfall and deep convection over eastern Congo and Lake’s region indicates the dominant role of large-scale mechanisms such as El Niño events.

SW Ethiopia demonstrates the strongest coupling between convective activity and rainfall variability (c.f., Figure 6). The convective anomalies during 1988, 1996–1997, and 2018–2020 correspond exactly with the largest positive rainfall anomalies (+1$\sigma$ to +2$\sigma$).

Figure 6 and Figure 11 show near or above average rainfall and deep convective events in the 2015–2020 period over the three regions. Note that the region does not show a particular trend, in contrast to past work that shows increased incidences of mesoscale systems.

Further, sea surface temperature (SST) anomalies in the Indian Ocean, particularly the Indian Ocean Dipole (IOD), are known to significantly modulate MAM rainfall [58,59]. Past work (e.g., [58,59]) documented the IOD as an important driver of the East African rainfall variability. Liebmann et al. [3] suggested substantial spatial heterogeneity in IOD influence across East African highlands that show stronger sensitivity to IOD variability, with positive IOD phases typically enhancing orographic precipitation through increased moisture flux, while negative phases lead to significant rainfall reductions. Coastal areas demonstrate more complex responses, as they are influenced by closer moisture sources from local sea-breeze circulations and are less dependent on the large-scale Indian Ocean moisture transport that dominates highland precipitation patterns (e.g., [31]).

5. Summary and Final Remarks

This study provides insight into the sub-seasonal to interannual rainfall variability and mechanisms that influence the variability over East Africa during March–April–May, the long rainy season, through systematic analysis of consecutive wet day events, convective activity patterns, and their relationship with atmospheric factors. Using CHIRPS precipitation data, GRIDSAT brightness temperature measurements, and ERA5 reanalysis products, we employed spatial pattern analysis, event-based classification, and composite techniques to characterize rainfall persistence and its atmospheric drivers.

Based on natural breaks in the duration and distinct hydrological/agricultural impacts, we established three event categories, short-duration (3–5 days) events that provide soil moisture recharge and are linked to wave disturbances and mesoscale convective systems (MCSs); medium-duration (6–10 days) events critical for crop establishment and associated with propagating features like Kelvin waves; and long-duration (>10 days) events that lead to both beneficial deep soil saturation and potentially harmful flooding conditions, linked to intraseasonal modes of variability. This classification framework enables a mechanistic understanding of how different atmospheric processes contribute to East Africa’s seasonal water supply. Key findings include

• Substantial contribution to seasonal totals: Consecutive wet day events (≥3 days) contribute 43% of total MAM rainfall across East Africa on average, ranging from near 0% in arid regions to 93% in areas of maximum organization. Short-duration events provide the largest contribution (29% domain average, affecting 94% of the region), medium-duration events contribute 12% on average (up to 42% locally), and long-duration events, though spatially restricted, contribute up to 52% in optimal locations.
• Spatial heterogeneity in rainfall organization: Consecutive wet days exhibit distinct regional patterns, with enhanced frequencies over complex topography (Eastern Congo, Ethiopian Highlands) and around Lake Victoria. Long-duration events (>10 days) concentrate in equatorial regions with significant elevation gradients, while short-duration events are more uniformly distributed across the domain.
• Relationship between convective organization and rainfall variability: Analysis of brightness temperature (Tb ≤ 235 K) reveals that Eastern Congo exhibits the highest frequency convective variability (standardized anomalies from −3.0 to +2.5 K), yet shows the lowest relative rainfall variability among three core regions, indicating a buffered regime with high precipitation efficiency. SW Ethiopia demonstrates the strongest coupling between convective activity and rainfall variability, confirming a threshold-dependent precipitation regime.
• Increasing interannual variability: Recent decades (particularly 2015–2020) show elevated year-to-year variability in both event frequencies and convective organization across all regions, with coordinated positive anomalies suggesting fundamental shifts in regional atmospheric dynamics.

These findings advance our understanding of East African climate dynamics and provide a foundation for improving sub-seasonal prediction capabilities. The demonstrated relationships between large-scale patterns, local topographic effects, and rainfall persistence characteristics offer practical applications for agricultural planning, water resource management, and disaster risk reduction.

Future research should focus on incorporating these process-based insights into dynamical forecasting systems and exploring the potential for sub-seasonal to seasonal prediction. Additionally, investigation of soil moisture feedback and their role in sustaining consecutive wet days represents an essential avenue for advancing process understanding and hydrologic and agricultural prediction capabilities.