1. Introduction
The Madden–Julian oscillation (MJO) is the largest mode of subseasonal climate variability in the tropics (Madden and Julian 1994). It is an eastward propagating mode centered around the equator with a characteristic time scale of 30–60 days. The MJO comprises a large-scale enhanced convective system flanked on its sides by suppressed convection that develops in the Indian Ocean, propagates eastward to the western Pacific Ocean, loses convective coupling while propagating to the eastern Pacific, and then re-emerges over the Atlantic and equatorial Africa (Madden and Julian 1971, 1994; Zhang 2005). It causes large-scale variations in zonal winds, pressure, and velocity potential, thus dynamically linking zones of enhanced and suppressed convection. MJO activity is typically diagnosed in terms of the strength and location of the convective core center of action. It is strongest in boreal winter, but it affects precipitation in regions across the globe, year-around. Although the convective core is generally most pronounced in the Indian Ocean and western Pacific Ocean, the MJO has significant precipitation impacts in other parts of the tropics, including East Africa (Madden and Julian 1994; Zhang 2013; Barlow 2012; Zaitchik 2017).
Climate drivers of precipitation variability in East Africa are of considerable interest, as the region is home to approximately 300 million people, many of whom depend on a rain-fed agricultural economy. It is a meteorologically complex region, owing in part due to its diverse topography (Spinage 2012), and is subject to significant rainfall variability from subseasonal to interannual time scales that can disrupt natural ecosystems, hydrology, and agriculture (Verdin et al. 2005; Nicholson 2016; Zaitchik 2017). This study focuses on equatorial East Africa—the portion of East Africa that lies within roughly 10° of the equator (comprising the countries Kenya, Uganda, Tanzania, Rwanda, and Burundi, as well as parts of Ethiopia and Somalia) and generally has a bimodal rainfall distribution (Nicholson 2017; Cook et al. 2020). The two rainfall seasons are the long rains during the boreal spring [March–May (MAM)] and the short rains during the boreal fall [October–December (OND)].
Previous studies have detected an MJO influence on subseasonal variability in this region. Mutai and Ward (2000) first identified eastward-propagating convective perturbations affecting equatorial East African rainfall. Omeny et al. (2006) found a significant relationship between the MJO and western Kenyan rainfall, but not with eastern Kenyan rainfall. They showed that conditions over Lake Victoria are strongly associated with the MJO, and their results suggested the possibility of skillful prediction of rainfall at intraseasonal time scales. In an analysis of annual outgoing longwave radiation (OLR), Sandjon et al. (2012) found three leading empirical orthogonal function (EOF) modes for northern Congo, Tanzania, and southern Ethiopia, whose peak spectral power indicated an MJO effect. Pohl and Camberlin (2006a) discovered a robust link between rainfall in the region and the core region of convection in the MJO, especially when the core is in the Indian Ocean, and postulated mechanisms through which the MJO drives intraseasonal variability in both long and short rains. Berhane and Zaitchik (2014) further analyzed the MJO effects for each season on a monthly basis. Vellinga and Milton (2018) showed that the MJO along with other large-scale drivers like the sea surface temperature of western Indian Ocean and quasi-biennial oscillation can impact seasonal rainfall of the long rains. Finney et al. (2020) showed that days with westerly moisture flow enhance MAM rainfall over northern Tanzania and southeastern Kenya and are more likely during MJO phases in the Indian Ocean as well as when tropical cyclones are present.
Other studies were conducted from the perspective of predictability. Hogan et al. (2015) noted that observed patterns of MJO influence matched previous studies and were generally well simulated by the operational global Met Office Unified Model (MetUM), often with a lead time of up to 5 days. Kimani et al. (2020) aimed to enhance seasonal prediction of rainfall based on amplitude of MJO phases and sea surface temperature (SST) response. MacLeod et al. (2021) evaluated model representation of MJO teleconnections and demonstrated skill of subseasonal forecasts with a lead time of 2 weeks and beyond, using reforecasts produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) and the U.K. Met Office.
MJO influence can be understood as modulation of the strength, occurrence, and spatial patterns of regional rainfall, and this modulation can vary with background climate state (Zhang 2013). El Niño–Southern Oscillation (ENSO), an interannual stationary mode in the tropical Pacific Ocean, can significantly alter the mean state of the atmosphere and ocean, affecting seasonal background flow and MJO teleconnection patterns (Roundy et al. 2010; Moon et al. 2011; Shimizu and Ambrizzi 2016; Zaitchik 2017; Arcodia et al. 2020). Inconsistent intraseasonal rainfall outcomes in East African rainfall have led to a recognition that the ENSO state interacts with MJO influence on interannual time scales (Pohl and Camberlin 2006b; Zaitchik 2017), and it can be argued that accounting for these states may afford better subseasonal-to-seasonal rainfall predictions. MJO phases can also modulate and mediate ENSO impact on rainfall at different time scales. For instance, certain phases can counteract the impacts of ENSO (Arcodia et al. 2020), and the frequency as well as amplitude of intraseasonal wet and dry periods can possibly aggregate to cause influences on interannual and long-term climate variabilities (Vigaud et al. 2017; Zaitchik 2017). Therefore, investigating the combined impacts of ENSO and the MJO is crucial to comprehending rainfall variability, which drives water management decisions.
Ideas related to this concept have been investigated in previous papers, but the findings are somewhat contradictory (Nicholson 2017) and overall fragmentary due to disparate times, locations, and specific research objectives considered. Sandjon et al. (2012) observed that over the course of an entire year, MJO activity over central-east Africa is strong during La Niña and ENSO-neutral years and weak during El Niño years. Kijazi and Reason (2005) found similar results for coastal Tanzania; de Andrade et al. (2021) found that the forecast quality of the African rainfall in the ECMWF model is linked to the strength of teleconnections of climate drivers including ENSO and the MJO. A more detailed study for East Africa by Pohl and Camberlin (2006b) found a weak association between ENSO and MJO activity. They did not find strong MJO activity for La Niña years but observed that generally MJO following the winter El Niño events was weak. The years with weaker MJO were associated with reduced rainfall and vice versa. They suggested that ENSO can, at least partially, modulate MJO rainfall association, but their study was limited to the long rains. In our study, we focus on the short rains during the boreal fall season of October–December (OND), which have a higher interannual variation associated with ENSO than long rains (Nicholson 1996; Indeje et al. 2000; Kijazi and Reason 2005) and are more spatially coherent (Moron et al. 2007; Hastenrath et al. 2011; Hirons and Turner 2018).
The purpose of the current study is to examine modulation of East African short rains associated with the combined and potentially interacting influences of the MJO, operating at subseasonal time scales, and ENSO, operating at interannual time scales. Specifically, we consider the following questions:
-
Does the ENSO state modulate MJO influence on East African rainfall during the OND rainy season?
-
Through what mechanisms does MJO influence rainfall in OND, and how are these mechanisms influenced by the ENSO state?
-
Does the MJO increase or decrease regional rainfall anomalies associated with ENSO on either daily or seasonal time scales?
This paper is organized as follows: section 2 presents data and methodology, section 3 shows results and discussion, and section 4 summarizes results and provides conclusions along with relevant future work.
2. Data and methods
For precipitation, we use the Climate Hazards Group Infrared Precipitation with Stations (CHIRPS) data at daily time intervals. CHIRPS merges microwave and infrared satellite data with in situ gauge data to produce a daily product at a resolution of 0.05° and is available from 1981 to the present (Funk et al. 2015a). This dataset has been validated in eastern Africa and has been shown to have good performance overall (Dinku et al. 2018). National Oceanic and Atmospheric Administration (NOAA) satellite-derived outgoing longwave radiation (OLR) data, available for a daily interval from 1980 onward at a resolution of 2.5° (Liebmann and Smith 1996), are used to assess large-scale precipitation variability. Atmospheric circulation is assessed using the National Centers for Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) reanalysis available from 1948 onward on a 2.5° grid [NCEP–NCAR (R1); Kalnay et al. 1996], with variables including daily winds, specific humidity, air temperature, geopotential height, and sea level pressure (SLP).
The real-time multivariate MJO (RMM) index (Wheeler and Hendon 2004), which is calculated by applying empirical orthogonal function (EOF) analysis to combined daily mean tropical (15°N–15°S average) fields of OLR, low-level (850 mb; 1 mb = 1 hPa), and high-level (200 mb) winds, is used to compute discrete daily values of MJO amplitude and phase. Annual cycle and low-frequency variability associated with ENSO is removed prior to calculating EOFs, thus making the RMM index ideal for analysis (Wheeler and Hendon 2004; Pohl and Camberlin 2011). The RMM MJO index, available from NOAA, is composed of two separate indices, RMM1 and RMM2, which are principal component time series of EOFs and are used to define MJO amplitude (square root of the sum of the squares of RMM1 and RMM2) and MJO phase. MJO phases representing its eastward propagation across global tropics are categorized into eight phases, each corresponding to the geographical position of the MJO convective core. Phase 1 denotes the period when the center of convection is over tropical Africa. In phases 2 and 3, the core is over the western and central-eastern equatorial Indian Ocean, respectively; phase 4 is west of the Maritime Continent; phase 5 is over the Maritime Continent; phase 6 is in the western Pacific Ocean; phase 7 is around the date line; and phase 8 is in the central-eastern Pacific Ocean. Active MJO periods are defined as days with MJO amplitude greater than 1 and days with MJO amplitude less than 1 are considered inactive or without MJO activity (Barnes et al. 2019).
ENSO conditions are defined using the average OND Niño-3.4 index, which is the spatially averaged SST anomaly in the east-central tropical Pacific over 5°S–5°N, 120°–170°W (Trenberth 1997). It is calculated from 1900 to the present using the HadISST product and is available as a time series index through NOAA (Rayner et al. 2003). El Niño years are chosen as the years when the Niño-3.4 index during OND has a value higher than 1, and La Niña years are chosen for values below −1. ENSO neutral years (also called NoENSO years for brevity) are chosen as the years when the absolute value of ENSO index is less than 0.3. This ensures inclusion of clearly strong events in all categories and avoids variability within the events of a given category (e.g., to avoid differences between weak and strong types of El Niños/La Niñas). The ENSO years for our study period of 1981–2018 are 1982, 1986, 1987, 1991, 1997, 2002, 2009, and 2015 for El Niño; 1983, 1984, 1988, 1998, 1999, 2007, 2010, 2011 for La Niña; and 1981, 1989, 1990, 1993, 1996, 2001, 2003, and 2013 for ENSO neutral.
Climatology, which is the average of a given variable for a given time period, is calculated on a daily basis over the years 1981–2018. Anomalies for this time period are calculated by subtracting climatology from the value of daily variables. An average daily composite is calculated by averaging a variable for all the days of interest. These composites are made for different categories (e.g., MJO phase 3), where a given variable (e.g., daily precipitation or OLR anomaly) is averaged for that category. Composites of OLR, SLP, precipitation, moisture transport, and wind anomalies are calculated for each MJO phase under different ENSO background conditions. Furthermore, total daily composites are calculated such that the average daily values (e.g., daily precipitation or OLR anomaly) under a given category (e.g., MJO phase 3) are summed up for the entire season.
Additionally, we use the CenTrends monthly precipitation dataset, available for 1900–2014, to support our longer record analyses. CenTrends offers gridded precipitation estimates for East Africa at a resolution of 0.1°. It is a “pooled” station dataset from several institutions that is extensively quality-controlled and has been shown to be more accurate than other available long-record gridded precipitation datasets (Funk et al. 2015b). We also use the MJO reconstruction dataset (Oliver and Thompson 2011) that is a reconstruction of the RMM index based on multiple linear regression of Twentieth-Century Reanalysis (20CR) (Compo et al. 2011) surface pressure time series onto the RMM index, which is then then used to hindcast the RMM index back to 1905. This dataset has been shown to be consistent with RMM data for their common period (Oliver and Thompson 2011). It is available for 1905–2015 and is used to calculate century-long correlations of MJO parameters with rainfall and ENSO indices. The East African region with boreal fall rainfall is defined as the land area between 10°S and 8°N and between 30° and 44°E, similar to previous studies (Berhane and Zaitchik 2014; Yang et al. 2015; Nicholson 2017; Zaitchik 2017).
3. Results and discussion
a. Rainfall climatology and ENSO anomalies
Figure 1 shows OND rainfall climatology and ENSO rainfall anomalies. Seasonal rainfall (Fig. 1a) generally occurs south of about 8°N, with higher values in the west, near lakes, and over the Kenyan highlands. Similar patterns are seen in daily climatology (Fig. 1b), with values ranging from 1 to 6 mm day−1. OND El Niño rainfall anomalies are positive (Fig. 1c), indicating rainfall excess associated with the positive phase of ENSO. The anomalies are high and spatially spread out throughout the region, with values reaching up to a third to half of the climatological seasonal rainfall in some regions. OND La Niña rainfall anomalies are negative (Fig. 1d), although the intensities are weaker and not as spatially spread out as El Niño anomalies. This ENSO association has been indicated in multiple studies (Nicholson and Kim 1997; Dezfuli and Nicholson 2013; Hoell et al. 2014), and an asymmetric response of La Niña including variation of degree and spatial consistency of rainfall deficit during different La Niña events has also been previously documented (Nicholson and Kim 1997; Nicholson and Selato 2000; Hoell et al. 2014; Nicholson 2015).
(a) OND seasonal rainfall climatology (mm); (b) OND daily rainfall climatology (mm day−1), showing the same pattern as (a) but in daily units for convenience; (c) OND El Niño rainfall anomalies (mm); and (d) OND La Niña rainfall anomalies (mm).
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
b. Average daily rainfall composites by MJO phase and ENSO state
Figure 2 shows the average daily rainfall anomaly composites during different phases of active MJO in eight columns, separated into ENSO neutral (NoENSO), El Niño, and La Niña conditions. The pattern and intensity of wetness and dryness clearly vary by both MJO phase and ENSO state. For instance, the rainfall enhancement (positive anomaly) in MJO phase 3 is greater and more widespread in El Niño than it is under La Niña or ENSO neutral state, reaching as much as 5–10 mm day−1 (which is equivalent to approximately 100%–150% increase in rainfall relative to daily climatology). Some of the wettest periods over East Africa may thus occur when MJO is positioned over the Indian Ocean during an El Niño. Phase 2 is similar, but the spatial pattern is less widespread. Note that rainfall contribution during days without MJO activity is also enhanced during an El Niño, as shown in Fig. 3a, but the spatial spread and intensity is modest. Notably, MJO phase 6 and, to a lesser extent, phase 7 are associated with substantial negative precipitation anomalies under El Niño. In fact, these negative anomalies are stronger under El Niño conditions (reaching 6 mm day−1) than under La Niña or ENSO neutral, notwithstanding the fact that El Niño is generally understood to bring wet conditions to East Africa. For La Niña, phase 7 negative anomalies are weaker than seen in El Niño, but the phase is still associated with more widespread negative anomalies than are seen for days without MJO activity (Fig. 3b).
(a)–(c) Average daily rainfall anomaly composites (mm day−1), with MJO phases in columns 1–8, and ENSO states in rows. Color shading shows results significant at the 90% confidence level. Forward slash stippling signifies where El Niño and La Niña are significantly different (90% significance) from ENSO Neutral; backward slash stippling signifies where El Niño (La Niña) plots are significantly different (90% significance) from La Niña (El Niña). Cross-stippling thus corresponds to where El Niño (La Niña) plots are significantly different for the other two ENSO categories. Number of days used in each composite are listed in parentheses.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
(left) Average daily rainfall (mm day−1) and right) OLR (W m−2) anomalies for days without an active MJO for (a),(c) El Niño and (b),(d) La Niña. Negative OLR anomaly corresponds to rainfall increase, and vice versa. Color shading shows results significant at the 90% confidence level (number of days used for El Niño plots is 290 and for La Niña plots is 262).
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
Generalizing across ENSO states, enhanced rainfall is seen for MJO phases 2 and 3 when the MJO core is in the equatorial Indian Ocean (Figs. 2a2,b2,c2,a3,b3,c3), and to a lesser degree in phase 1 (Figs. 2a1,b1,c1). Rainfall suppression is linked to MJO phases near the west Pacific, as seen in Fig. 2 for phases 6 and 7 (Figs. 2a6,b6,c6,a7,b7,c7). MJO phases 1, 2, and 3 will be referred to as wet phases and phases 6 and 7 as dry phases. Phases 4 and 5 have relatively weaker, overall insignificant, and inconsistent signals for the three rows, but in general they tend to be more wet primarily for El Niño, mixed wet and dry for La Niña, and weakly dry for NoENSO. These may be called transition phases as the signal gradually changes from wet to dry, also noted by Hogan et al. (2015), who noted a transition of upper-level velocity potential beginning in phase 4. Last, phase 8 is also weak and not significant overall. We limit our focus to the region north of 10°S, as farther southward opposite ENSO signals related to subtropical southern African influence may start to appear (Kijazi and Reason 2005).
To examine broad-scale patterns of rainfall, average daily OLR anomaly composites for different MJO phases and different ENSO background conditions are shown in Figs. 4 and 5. Negative values in these figures are indicative of enhanced rainfall (deep convective clouds) and positive values rainfall deficit (less convection) relative to climatology. These plots are smoother than rainfall plots and show broad-scale patterns on both land and nearby ocean. Phases 2 and 3 (followed weakly by phase 1) most strongly show rainfall enhancement in the study region that is associated with corresponding enhancement in the nearby Indian Ocean (Fig. 4). Phases 6 and 7, in contrast, show an increase in OLR over East Africa, accompanied by a similar suppression in the nearby Indian Ocean (Fig. 5).
Average daily OLR anomaly (W m−2) for phases 1–3. Negative OLR anomaly corresponds to rainfall increase, and vice versa. Color shading shows results significant at the 90% confidence level.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
Average daily OLR anomaly (W m−2) for phases 6 and 7. Negative OLR anomaly corresponds to rainfall increase, and vice versa. Color shading shows results significant at the 90% confidence level.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
Average daily anomalous OLR values over East Africa vary between the three ENSO states for various MJO phases (Figs. 4 and 5). Corresponding OLR anomalies for El Niño without an active MJO and La Niña without an active MJO are shown in Figs. 3c and 3d, respectively. In general, the magnitude of OLR anomalies over East Africa associated with active MJO is larger under El Niño conditions than under La Niña or ENSO neutral conditions for both wet and dry conditions (Figs. 4 and 5). These anomalies are also large relative to the OLR anomalies seen for El Niño or La Niña under inactive MJO conditions (Figs. 3c,d). This is most strongly seen for wet phases 2 and 3, and to a lesser degree for dry phases 6 and 7 (Figs. 4 and 5). Wavenumber–frequency filtered composite plots (Fig. S1) also confirm that the MJO signal is the dominant component of the patterns seen in these figures, and that this influence is present across ENSO conditions (though noticeably stronger under El Niño).
The additive influences of ENSO and the MJO on OLR signals can be thought of as interference patterns. El Niño tends to bring negative OLR anomalies (enhanced rainfall) to East Africa, and this tendency is reinforced (constructive interference) by MJO wet phases (especially phase 3), which bring significantly intensified rainfall (Figs. 2 and 4). But for the dry phases, enhanced rainfall El Niño signal is offset by the significantly suppressed rainfall (Figs. 2 and 5), leading to a destructive interference pattern. The effects are reversed for La Niña but are relatively much weaker. Phases 4, 5, and 8 tend to have overall weak or insignificant signals and will not be discussed further (shown in Fig. S2 in the online supplemental material for reference).
These results indicate that East African rainfall modulation during active MJO phases depends on ENSO states and their interaction with MJO, and that computing MJO rainfall effects without consideration of separate ENSO states may lump a range of signals into one. This has implications for predictions of intraseasonal rainfall. During El Niño, for example, MJO activity in the Indian Ocean has a strong link to daily precipitation anomalies in East Africa, such that forecasts of emerging MJO activity in this sector can be predictive of potential for flood conditions. These links are also evident in La Niña, though weaker. For ENSO neutral conditions, phases 1 and 2 appear to be more relevant than phase 3. Subseasonal forecasts, thus, should take into consideration both the MJO forecast and the background ENSO state.
c. Associated physical mechanisms
To simplify the investigation of dynamic and thermodynamic mechanisms, we use MJO phases 3 and 6 as representative of wet and dry phases, respectively. Figure 6 shows composite plots for vertically integrated moisture transport for inactive MJO, active MJO phase 3, and active MJO phase 6 under the three ENSO categories. Row 1 is for ENSO neutral, row 2 for El Niño, and row 3 for La Niña. Other phases including phases 2 and 7 that have a weaker signal are shown in the supplemental material (see Figs. S3 and S4) and discussed. The vertical integration is performed from the surface to 500 mb. Similar results are obtained for integration up to 600 mb as the winds and specific humidity patterns in the midtroposphere resemble low-level patterns but with lower magnitudes (Fig. S5). SLP and low-level wind plots are analogous to moisture transport plots, with anomalies often reinforcing climatological winds patterns of strong easterlies on the east of the region and marginal westerlies on the west of the region (Figs. S6–S8).
Average daily vertically integrated moisture transport (qu; g kg−1 m s−1) with colors showing the dominant horizontal moisture transport over East Africa. (left) Days without active MJO, (center) MJO phase 3, and (right) MJO phase 6. (a)–(c) ENSO neutral (NoENSO), (d)–(f) for El Niño, and (g)–(i) for La Niña.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
For El Niño days without MJO activity [Fig. 6d (row 2, column 1)], anomalous easterlies from the central Indian Ocean bring moist air to the study region. However, for El Niño with active MJO phase 3 (Fig. 6e), anomalous easterlies from the Indian Ocean are weaker. Instead, a strong westerly moisture transport from western Africa (the Congo basin) is evident. Phase 2 is similar to phase 3 in terms of westerly transport, but phase 1 is accompanied by easterly transport increase from the Indian Ocean (Fig. S3). Westerly moisture transport from central-western Africa is present for MJO phases 2 and 3 under all ENSO conditions but is weakest for ENSO neutral (Fig. 6b) and strongest when exiting the region to the east for La Niña (Fig. 6h). The strengthening of winds exiting the region for phase 3 La Niña means that the enhanced westerlies are not associated with positive moisture flux convergence (MFC) the way in which they are under El Niño conditions (Figs. 7e,h). In fact, as seen in Figs. 7 and 8, both MFC and thermodynamic instability during El Niño–MJO phase 3 (Figs. 7 and 8e) are more amply increased than in La Niña–MJO phase 3 (Figs. 7 and 8h), which corresponds to rainfall signal observed in Fig. 4.
Average daily vertically integrated moisture flux convergence (MFC; 10−6 g kg−1 s−1; 90% significance). (left) ENSO neutral (NoENSO), (center) El Niño, and (right) La Niña. (a)–(c) Days without MJO activity, (d)–(f) MJO phase 3, and (g)–(i) MJO phase 6.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
As in Fig. 7, but for average daily buoyancy (103 J kg−1) obtained as a result of the difference between moist static energy at 1000 mb and saturated moist static energy at 700 mb.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
Patterns are more subtle for the dry MJO phases (represented here by phase 6). There is, however, a tendency for easterly moisture transport anomalies near the western part of the study region (Fig. 6). These anomalies are strongest for El Niño and are relatively weak for La Niña and ENSO neutral conditions. The easterly anomalies are associated with moisture flux divergence for all ENSO states. This divergence in phase 6 is the strongest in magnitude under El Niño conditions but is more consistently the same sign across the study region for La Niña (Fig. 7). The peak moisture flux divergence for El Niño with MJO phase 6 occurs in the southern highlands portion of the study region, which is consistent with OLR anomalies shown in Fig. 5. This moisture flux convergence signal is primarily a result of moisture convergence (
Additionally, changes to OLR in wet and dry phases of ENSO categories can be linked to atmospheric stability. We use lower tropospheric buoyancy (Fig. 8) to assess moist instability. This is obtained by subtracting saturated moist static energy at 700 mb from moist static energy at 1000 mb (Seager et al. 2003). El Niño (La Niña) without active MJO shows greater (less) instability, as one would expect. Instability increases for MJO phase 3 for all three ENSO categories, but for El Niño the most. This could be related to advection of anomalously warmer air from the Indian Ocean, which can make the column unstable during wet MJO phase 3, especially for El Niño (Fig. 9). Berhane and Zaitchik (2014) posited that MJO convection in the Indian Ocean releases energy and warmer temperatures that advect into East Africa, even though easterlies are not as strong for an active MJO. For phase 6, stability is generally evident under all ENSO conditions (Fig. 8).
Temperature advection anomaly (contours; K s−1) superimposed by winds (vectors; m s−1) for MJO phase 3: (a) ENSO Neutral, (b) El Niño, and (c) La Niña.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
d. Analyses at seasonal time scale
Analyses to this point have focused on daily data. To study MJO–ENSO relationships to East African precipitation over longer time periods, we now turn to correlations with MJO activity on seasonal timescales. To calculate the seasonal MJO amplitude, we average the MJO amplitude on all days in a given phase, as has been done in some previous studies (Finney et al. 2020; Vellinga and Milton 2018; Kimani et al. 2020). We use the MJO index from the MJO Reconstruction dataset and monthly CenTrends rainfall converted to seasonal rainfall (OND), for 1905–2014. Over this time period (n = 110 years) there are modest but statistically significant correlations between seasonal precipitation and the seasonally averaged reconstructed MJO RMM index amplitude for phases 1, 2, and 3 (Table 1). This result is similar to that of Kimani et al. (2020), who found a significant correlation between OND rainfall in East Africa with MJO amplitude in phases 1 and 2, although phase 3 correlation was not significant in their study. Correlation with the seasonal Niño-3.4 index is found to be significant only for MJO phase 2.
MJO amplitude-by-phase correlations with seasonal East African rainfall anomaly and Niño-3.4 index (1905–2014). Correlations significant at 95% are in bold.
The correlations shown in Table 1 capture only one component of seasonal-scale ENSO–MJO–rainfall relationships. In addition to the average amplitude of each MJO phase, total precipitation in a season might be influenced by the number of days that MJO is active in each phase. Figure 10 shows the average total precipitation anomaly for days in each MJO phase over the OND season (i.e., rainfall on each day for phase X is summed over OND and averaged across years for each ENSO condition). This plot uses the daily precipitation data that are available only for the more recent period (1981 onward). For El Niño (row 2), anomalies are highest for MJO wet phases, while the anomalies in dry phase are much lower. For La Niña (row 3), dry phase anomalies in phase 6 seem to be somewhat higher than wet phases, although this is not as convincing as El Niño. It should also be noted that the total anomalies in phase 3 (and to a lesser extent phase 2) for El Niño are larger than the total anomalies for inactive MJO days during El Niño (Fig. 11a) and to some degree, phase 6 anomalies for La Niña are slightly larger than total anomalies for inactive MJO days during La Niña (Fig. 11b).
Total daily precipitation anomalies (mm; 90% confidence level) for each phase and for the different ENSO categories [(top) NoENSO, (middle) El Niño, (bottom) La Niña]. Numbers in parentheses show number of days in each subplot.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
(a) Total daily rainfall (mm) for days without MJO activity during an average El Niño year; (b) as in (a), but for La Niña; and (c) duration (in number of days) of active (>1 amplitude) MJO in an average El Niño (blue), La Niña (orange), and NoENSO (gray) year.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
Given that certain wet phases stand out in El Niño and dry phases in La Niña, along with the fact that El Niño is associated with rainfall excess and La Niña with rainfall deficit in the region, we hypothesize that seasonal impacts of ENSO extremes may, at least partly, arise from MJO wet and dry phases. The number of days in each MJO phase for all three ENSO categories are stated in parentheses in Fig. 10. Plotting these durations (averaged to get values for an average year under each ENSO category) on a line plot in Fig. 11c shows that for wet phases 1, 2, and 3, there are a higher (lower) number of active MJO days during El Niño (La Niña); for dry phases 6 and 7, there are a lower (higher) number of active MJO days during El Niño (La Niña). Phase 5 has a higher number of days for La Niña, and phase 8 has higher number of days for El Niño, but these do not have consistently significant signals in daily composites, so will not be discussed further. Similar results are obtained using active MJO days from the MJO reconstruction dataset (Fig. S11).
Correlations of regional rainfall and Niño-3.4 with MJO duration in each phase reveal moderately high significant correlations for phases 1, 2, 3, and 6 (Table 2). Surprisingly, phase 8 has a slightly lower but still significant correlation, but we do not attempt to interpret this result because of the mixed signal seen in OLR plots (Fig. 5). It is also worth noting that the slope of linear relation between MJO duration and rainfall (as well as Niño-3.4) is higher for wet phases than dry phase 6 (not shown; average slope for wet phases is 4.2 and for dry phases is −3.0 when phase duration is used as a predictor). Nonetheless, we join these phases’ durations to quantify an MJO duration metric called Pmjo, which is the sum of the wet phases’ duration minus the dry phases’ duration (Pmjo = P1 + P2 + P3 − P6). Linear correlation of this combined phase metric slightly increases over individual phases (Table 2) and is 0.45 with rainfall and 0.56 with Niño-3.4. Using this metric, we perform multiple linear regression with East African rainfall as a predictand. Pmjo alone explains about 20% of the rainfall variability and the coefficient of Pmjo is significant (p value = 2 × 10−7), Niño-3.4 alone explains 23% of the variability and coefficient of Niño-3.4 is significant (p value = 2 × 10−6). When they are both added as predictors, the percentage of variability explained does not go up significantly (it goes up by 4% from 23% due to Niño-3.4 alone). It is worth noting that the coefficient of Niño-3.4 becomes nonsignificant (p value = 0.30), while the interaction term between Niño-3.4 and Pmjo becomes significant (p value = 0.07) at 90% (Fig. 12). This could be interpreted as statistical evidence that the observed impact of ENSO on East African OND rainfall depends in part on its interaction with the MJO.
Results of linear regression of ENSO and Pmjo indices with East African precipitation anomaly as a predictand. Arrows point to coefficients and p values of individual and combined predictors [an asterisk (*) shows 90% significance and a double asterisk (**) shows 95% significance using a two-sided t test].
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
MJO duration-by-phase correlations with seasonal East African rainfall anomaly and the Niño-3.4 index (1905–2014). Correlations significant at 95% are in bold.
Figure 13a shows the scatterplot between Niño-3.4 and the Pmjo metric showing a correlation of 0.56. This plot is divided into four quadrants. The y axis of this plot is partitioned vertically based on a mean of 14 from the histogram of Pmjo (Fig. 13b). A high (low) Pmjo value means a high (low) number of active MJO days in wet phases relative to dry phases. Along the x axis of the scatterplot, El Niños (La Niñas) are considered above (below) a 0.5 (−0.5) threshold (the 0.5 threshold was chosen in order to include a higher number of El Niños and La Niñas than one would obtain from a stricter threshold). Quadrants 1 and 2 are for El Niño, and quadrants 3 and 4 are for La Niña. Quadrant 1 is El Niño years with lower-than-mean Pmjo; quadrant 2 is El Niño years with higher-than-mean Pmjo; quadrant 3 is La Niña years with higher-than-mean Pmjo; quadrant 4 is La Niña years with lower-than-mean Pmjo. A box plot of East African seasonal precipitation anomaly (Fig. 13c) shows that quadrant 1 is significantly different from quadrant 2, but quadrants 3 and 4 are not significantly different from each other (t test at 95% significance). While the El Niño quadrants are different statistically, we note that they have unequal sample sizes (e.g., quadrant 1 is limited to 5 years and quadrant 2 has 23 years). A majority of El Niño years in quadrant 2 are found to be in the upper tercile of seasonal rainfall, whereas none of the El Niño years in quadrant 1 are in the upper tercile of seasonal rainfall. The higher end of quadrant 2 has greater positive seasonal rainfall anomalies than that of quadrant 1. This means that high rainfall excesses under El Niño are more likely to be observed in the presence of a high number of active MJO days over Africa (phase 1) or the Indian Ocean (phases 2 and 3). Similar analysis for the more recent MJO RMM data shows increased correlation of 0.65 between Niño-3.4 and Pmjo (Fig. S12), but due to the smaller sample size, the two quadrants of El Niño were not determined to be statistically significantly different from each other.
(a) Scatterplot between Niño-3.4 and the Pmjo metric showing a correlation of 0.56, divided into four quadrants. (b) The y axis of this plot is sectioned vertically based on a mean of 14 from the histogram of Pmjo. A high (low) Pmjo value means relatively high (low) number of days in wet phases relative to the dry phase. Along the x axis of the scatterplot, El Niños (La Niñas) are considered above (below) a 0.5 (−0.5) threshold. (c) Box plots show rainfall distributions for different quadrants with n = 4 (quadrant 1), 23 (quadrant 2), 9 (quadrant 3), and 22 (quadrant 4). Quadrant 1 is significantly different from quadrant 2, but quadrants 3 and 4 are not significantly different from each other (t test at 95% significance).
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
Figure 14 shows total daily OLR anomaly for all phases during an average El Niño year. During MJO phases 2 and 3, a significant negative OLR anomaly (which generally corresponds to a positive precipitation anomaly) occurs throughout the region. These phases contribute approximately 25% and 30% respectively to the net seasonal OLR value (Table 3). A total of about 62% contribution comes from the sum of phases 1, 2, and 3. Negative OLR anomalies for all days without MJO activity contribute about 32% to the net seasonal OLR value (Table 3). The negative contribution from wet phase 6 is only about 5%. Thus, increased MJO activity in the Indian Ocean during an El Niño can mediate the El Niño impact on East African rainfall by increasing the net seasonal amount and spatial influence (i.e., extending the spatial influence farther inland).
Total daily OLR anomalies (10−1 MJ m−2) for an average El Niño year (obtained by averaging total OLR anomalies over all El Niño years to show an average El Niño year value), showing non-MJO and active MJO phases. Negative OLR anomaly corresponds to rainfall increase, and vice versa.
Citation: Journal of Climate 35, 6; 10.1175/JCLI-D-21-0583.1
Percent contribution of each phase to total seasonal OLR anomaly; phases contributing the most are in bold.
e. Nonstationarity
The use of an extended reconstruction of 110 years offers a number of advantages when studying ENSO–MJO interactions, but the behavior of these climate modes has not been stationary for the entire period of analysis. Major shifts in magnitude and variability of short rains and associated circulation parameters occurred around 1920, 1961, 1982, and 1997 (Nicholson 2015). To quantify nonstationarity in our relationships of interest, we calculate correlations between Pmjo and seasonal East African rainfall and Niño-3.4 for sequential 20-yr time periods (Table 4; first and last time periods have a few missing years). The correlations are notably stronger after the 1960s, a pattern that roughly corresponds to the increase in 20-yr sliding correlations that Nicholson (2015) found between ENSO and a number of circulation features relevant to East African rainfall (her Fig. 7). It is possible that MJO plays a role in this regime shift, but we defer that question to a future study.
Table showing correlations for different time periods of roughly 20 years corresponding to a common time period denominator of regime shifts documented in previous literature. Correlations with Pmjo significant at 90% for each period are in bold. Correlations showing significant decadal variability (90% bootstrap confidence level) are italicized.
4. Conclusions
We have presented the combined impacts of the MJO and ENSO on East African rainfall during boreal fall. We show that rainfall response depends on ENSO states and their interaction with MJO phases. Daily rainfall intensification during MJO wet phases is much stronger during an average El Niño year than an average La Niña or ENSO neutral year. Stronger rainfall response for these phases during El Niño is linked to an increased westerly moisture flux convergence accompanied by high moist instability over East Africa. This is related to enhanced rainfall in the region, particularly over the highlands, including increased potential for daily rainfall excesses. Intraseasonally, intensities of wet anomalies during El Niño and dry anomalies during La Niña depend upon the MJO phase. Understanding rainfall modulation for a given active MJO phase during different ENSO states may improve subseasonal rainfall predictions. Where subseasonal forecasts based on MJO state alone inadvertently lump together a range of signals from different ENSO–MJO interactions, a forecast that incorporates knowledge of ENSO, MJO, and their interaction may yield more meaningful predictions.
Over an entire season, we observe more (fewer) days of the MJO in phases that favor precipitation in East Africa for El Niño (La Niña), and vice versa for the phases that suppress precipitation. During El Niño (La Niña) seasons, MJO activity in wet-favoring phases occur an average of 52% (25%) of active MJO days, and an average of 33% (15%) of all days in the season. Multiple linear regression with Pmjo and the Niño-3.4 index as predictors shows that there may be an interaction between the MJO and ENSO in influencing short rains. Total daily rainfall for wet MJO phases show strongly enhanced positive signals during El Niño. In fact, a seasonal El Niño composite (Fig. 1) appears to be more similar to total daily MJO wet phases 2 and 3 composites (Figs. 10b2,b3) than the composite of El Niño without MJO activity (Fig. 11a). These wet phases contribute up to about 60% to the seasonal OLR anomalies during an El Niño. The signals are weak for La Niña, and MJO phase durations do not seem to have a significant impact on seasonal rainfall. All of this points to a meaningful role of the MJO in the communication of ENSO SST anomalies to the East African short rains.
We recognize limitations of this study, including that the number of years used in precipitation composite plots is limited because a reliable and consistent record of gridded daily precipitation estimates is only available from 1981. Composite analysis and multiple linear regression may be limited due to the fact that MJO and ENSO interactions could be nonlinear, and it is not known if linear or nonlinear dynamics dominate teleconnections (Moon et al. 2011; Arcodia et al. 2020). While we focused on ENSO in the tropical Pacific Ocean, variability in the Indian Ocean and to some degree the Atlantic Ocean also have been documented to impact the boreal fall rainfall (Black et al. 2003; Onyutha and Willems 2015). The Indian Ocean dipole (IOD) is an important interannual variability that was not a focus of this study. The IOD years often occur concurrently with ENSO, and because of the limited observational daily precipitation dataset (1981–2018) it was not possible to separate out pure IOD or ENSO years for a separate analysis. We did test statistical relationships with the IOD and found that seasonal rainfall variability explained by a linear regression on the IOD index was 33% (the ENSO index was 23%; this is for the century-scale analysis), but correlations of MJO amplitude and durations with IOD index were weaker than those with the ENSO index. The regression interaction term between the MJO, IOD, and ENSO was also insignificant. Future work, including numerical modeling experiments, needs to be done to elucidate three-way interactions between the influences that ENSO, the IOD, and the MJO have on each other and on the East African short rains.
Acknowledgments.
We would like to acknowledge funding from NSF Award BCS-1639214.
Data availability statement.
Datasets used are publicly available through Climate Hazards Group (https://www.chc.ucsb.edu/data) and NOAA (https://psl.noaa.gov/data/index.html).
REFERENCES
Arcodia, M. C., B. P. Kirtman, and L. S. Siqueira, 2020: How MJO teleconnections and ENSO interference impact U.S. precipitation. J. Climate, 33, 4621–4640, https://doi.org/10.1175/JCLI-D-19-0448.1.
Barlow, M., 2012: Africa and West Asia. Intraseasonal Variability in the Atmosphere–Ocean Climate System, 2nd ed., W.-K. Lau and D. E. Waliser, Eds., Springer, 477–495.
Barnes, E. A., S. M. Samarasinghe, I. Ebert‐Uphoff, and J. C. Furtado, 2019: Tropospheric and stratospheric causal pathways between the MJO and NAO. J. Geophys. Res. Atmos., 124, 9356–9371, https://doi.org/10.1029/2019JD031024.
Berhane, F., and B. Zaitchik, 2014: Modulation of daily precipitation over East Africa by the Madden–Julian oscillation. J. Climate, 27, 6016–6034, https://doi.org/10.1175/JCLI-D-13-00693.1.
Black, E., J. Slingo, and K. R. Sperber, 2003: An observational study of the relationship between excessively strong short rains in coastal East Africa and Indian Ocean SST. Mon. Wea. Rev., 131, 74–94, https://doi.org/10.1175/1520-0493(2003)131<0074:AOSOTR>2.0.CO;2.
Compo, G. P., and Coauthors, 2011: The Twentieth Century Reanalysis Project. Quart. J. Roy. Meteor. Soc., 137 (654), 1–28, https://doi.org/10.1002/qj.776.
Cook, K. H., R. G. Fitzpatrick, W. Liu, and E. K. Vizy, 2020: Seasonal asymmetry of equatorial East African rainfall projections: Understanding differences between the response of the long rains and the short rains to increased greenhouse gases. Climate Dyn., 55, 1759–1777, https://doi.org/10.1007/s00382-020-05350-y.
de Andrade, F. M., M. P. Young, D. MacLeod, L. C. Hirons, S. J. Woolnough, and E. Black, 2021: Subseasonal precipitation prediction for Africa: Forecast evaluation and sources of predictability. Wea. Forecasting, 36, 265–284, https://doi.org/10.1175/WAF-D-20-0054.1.
Dezfuli, A. K., and S. E. Nicholson, 2013: The relationship of rainfall variability in western equatorial Africa to the tropical oceans and atmospheric circulation. Part II: The boreal autumn. J. Climate, 26, 66–84, https://doi.org/10.1175/JCLI-D-11-00686.1.
Dinku, T., C. Funk, P. Peterson, R. Maidment, T. Tadesse, H. Gadain, and P. Ceccato, 2018: Validation of the CHIRPS satellite rainfall estimates over eastern Africa. Quart. J. Roy. Meteor. Soc., 144, 292–312, https://doi.org/10.1002/qj.3244.
Finney, D. L., J. H. Marsham, D. P. Walker, C. E. Birch, B. J. Woodhams, L. S. Jackson, and S. Hardy, 2020: The effect of westerlies on East African rainfall and the associated role of tropical cyclones and the Madden–Julian oscillation. Quart. J. Roy. Meteor. Soc., 146, 647–664, https://doi.org/10.1002/qj.3698.
Funk, C., and Coauthors, 2015a: The Climate Hazards Infrared Precipitation with Stations—A new environmental record for monitoring extremes. Sci. Data, 2, 150066, https://doi.org/10.1038/sdata.2015.66.
Funk, C., S. E. Nicholson, M. Landsfeld, D. Klotter, P. Peterson, and L. Harrison, 2015b: The centennial trends Greater Horn of Africa precipitation dataset. Sci. Data, 2, 150050, https://doi.org/10.1038/sdata.2015.50.
Hastenrath, S., D. Polzin, and C. Mutai, 2011: Circulation mechanisms of Kenya rainfall anomalies. J. Climate, 24, 404–412, https://doi.org/10.1175/2010JCLI3599.1.
Hirons, L., and A. Turner, 2018: The impact of Indian Ocean mean-state biases in climate models on the representation of the East African short rains. J. Climate, 31, 6611–6631, https://doi.org/10.1175/JCLI-D-17-0804.1.
Hoell, A., C. Funk, and M. Barlow, 2014: La Niña diversity and northwest Indian Ocean rim teleconnections. Climate Dyn., 43, 2707–2724, https://doi.org/10.1007/s00382-014-2083-y.
Hogan, E., A. Shelly, and P. Xavier, 2015: The observed and modelled influence of the Madden–Julian oscillation on East African rainfall. Meteor. Appl., 22, 459–469, https://doi.org/10.1002/met.1475.
Indeje, M., F. H. Semazzi, and L. J. Ogallo, 2000: ENSO signals in East African rainfall seasons. Int. J. Climatol., 20, 19–46, https://doi.org/10.1002/(SICI)1097-0088(200001)20:1<19::AID-JOC449>3.0.CO;2-0.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77: 437–472, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Kijazi, A. L., and C. J. C. Reason, 2005: Relationships between intraseasonal rainfall variability of coastal Tanzania and ENSO. Theor. Appl. Climatol., 82, 153–176, https://doi.org/10.1007/s00704-005-0129-0.
Kimani, M., J. C. B. Hoedjes, and Z. Su, 2020: An assessment of MJO circulation influence on air–sea interactions for improved seasonal rainfall predictions over East Africa. J. Climate, 33, 8367–8379, https://doi.org/10.1175/JCLI-D-19-0296.1.
Liebmann, B., and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77, 1275–1277, https://doi.org/10.1175/1520-0477-77.6.1274.
MacLeod, D. A., and Coauthors, 2021: Drivers and subseasonal predictability of heavy rainfall in equatorial East Africa and relationship with flood risk. J. Hydrometeor., 22, 887–903, https://doi.org/10.1175/JHM-D-20-0211.1.
Madden, R. A., and P. R. Julian, 1971: Detection of a 40–50 day oscillation in the zonal wind in the tropical Pacific. J. Atmos. Sci., 28, 702–708, https://doi.org/10.1175/1520-0469(1971)028<0702:DOADOI>2.0.CO;2.
Madden, R. A., and P. R. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122, 814–837, https://doi.org/10.1175/1520-0493(1994)122<0814:OOTDTO>2.0.CO;2.
Moon, J. Y., B. Wang, and K. J. Ha, 2011: ENSO regulation of MJO teleconnection. Climate Dyn., 37, 1133–1149, https://doi.org/10.1007/s00382-010-0902-3.
Moron, V., A. W. Robertson, M. N. Ward, and P. Camberlin, 2007: Spatial coherence of tropical rainfall at the regional scale. J. Climate, 20, 5244–5263, https://doi.org/10.1175/2007JCLI1623.1.
Mutai, C. C., and M. N. Ward, 2000: East African rainfall and the tropical circulation/convection on intraseasonal to interannual timescales. J. Climate, 13, 3915–3939, https://doi.org/10.1175/1520-0442(2000)013<3915:EARATT>2.0.CO;2.
Nicholson, S. E., 1996: A review of climate dynamics and climate variability in Eastern Africa. The Limnology, Climatology and Paleoclimatology of the East African Lakes, E. O. Odada and D. O. Olago, Eds., Springer, 25–56.
Nicholson, S. E., 2015: Long‐term variability of the East African ‘short rains’ and its links to large‐scale factors. Int. J. Climatol., 35, 3979–3990, https://doi.org/10.1002/joc.4259.
Nicholson, S. E., 2016: An analysis of recent rainfall conditions in eastern Africa. Int. J. Climatol., 36, 526–532, https://doi.org/10.1002/joc.4358.
Nicholson, S. E., 2017: Climate and climatic variability of rainfall over eastern Africa. Rev. Geophys., 55, 590–635, https://doi.org/10.1002/2016RG000544.
Nicholson, S. E., and J. Kim, 1997: The relationship of the El Niño–Southern Oscillation to African rainfall. Int. J. Climatol., 17, 117–135, https://doi.org/10.1002/(SICI)1097-0088(199702)17:2<117::AID-JOC84>3.0.CO;2-O.
Nicholson, S. E., and J. C. Selato, 2000: The influence of La Niña on African rainfall. Int. J. Climatol., 20, 1761–1776, https://doi.org/10.1002/1097-0088(20001130)20:143.0.CO;2-W.
Oliver, E. C., and K. R. Thompson, 2011: A reconstruction of Madden–Julian oscillation variability from 1905 to 2008. J. Climate, 25, 1996–2019, https://doi.org/10.1175/JCLI-D-11-00154.1.
Omeny, P. A., L. Ogallo, R. Okoola, H. Hendon, and M. Wheeler, 2006: East African rainfall variability associated with the Madden–Julian oscillation. J. Kenya Meteor. Soc., 2, 105–114.
Onyutha, C., and P. Willems, 2015: Spatial and temporal variability of rainfall in the Nile Basin. Hydrol. Earth Syst. Sci., 19, 2227–2246, https://doi.org/10.5194/hess-19-2227-2015.
Pohl, B., and P. Camberlin, 2006a: Influence of the Madden–Julian Oscillation on East African rainfall. I: Intraseasonal variability and regional dependency. Quart. J. Roy. Meteor. Soc., 132, 2521–2539, https://doi.org/10.1256/qj.05.104.
Pohl, B., and P. Camberlin, 2006b: Influence of the Madden–Julian Oscillation on East African rainfall: II. March–May season extremes and interannual variability. Quart. J. Roy. Meteor. Soc., 132, 2541–2558, https://doi.org/10.1256/qj.05.223.
Pohl, B., and P. Camberlin, 2011: Intraseasonal and interannual zonal circulations over the equatorial Indian Ocean. Theor. Appl. Climatol., 104, 175–191, https://doi.org/10.1007/s00704-010-0336-1.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, https://doi.org/10.1029/2002JD002670.
Roundy, P. E., K. MacRitchie, J. Asuma, and T. Melino, 2010: Modulation of the global atmospheric circulation by combined activity in the Madden–Julian oscillation and the El Niño–Southern Oscillation during boreal winter. J. Climate, 23, 4045–4059, https://doi.org/10.1175/2010JCLI3446.1.
Sandjon, A. T., A. Nzeukou, and C. Tchawoua, 2012: Intraseasonal atmospheric variability and its interannual modulation in central Africa. Meteor. Atmos. Phys., 117, 167–179, https://doi.org/10.1007/s00703-012-0196-6.
Seager, R., R. Murtugudde, N. Naik, A. Clement, N. Gordon, and J. Miller, 2003: Air–sea interaction and the seasonal cycle of the subtropical anticyclones. J. Climate, 16, 1948–1966, https://doi.org/10.1175/1520-0442(2003)016<1948:AIATSC>2.0.CO;2.
Shimizu, M. H., and T. Ambrizzi, 2016: MJO influence on ENSO effects in precipitation and temperature over South America. Theor. Appl. Climatol., 124, 291–301, https://doi.org/10.1007/s00704-015-1421-2.
Spinage, C. A., 2012: The changing climate of Africa. Part I: Introduction and eastern Africa. African Ecology, Springer, 57–141.
Trenberth, K. E., 1997: The definition of El Niño. Bull. Amer. Meteor. Soc., 78, 2771–2777, https://doi.org/10.1175/1520-0477(1997)078<2771:TDOENO>2.0.CO;2.
Vellinga, M., and S. F. Milton, 2018: Drivers of interannual variability of the East African “long rains”. Quart. J. Roy. Meteor. Soc., 144, 861–876, https://doi.org/10.1002/qj.3263.
Verdin, J., C. Funk, G. Senay, and R. Choularton, 2005: Climate science and famine early warning. Philos. Trans. Roy. Soc. London, B360, 2155–2168, https://doi.org/10.1098/rstb.2005.1754.
Vigaud, N., B. Lyon, and A. Giannini, 2017: Sub‐seasonal teleconnections between convection over the Indian Ocean, the East African long rains and tropical Pacific surface temperatures. Int. J. Climatol., 37, 1167–1180, https://doi.org/10.1002/joc.4765.
Wheeler, M. C., and H. H. Hendon, 2004: An all-season real-time multivariate MJO index: Development of an index for monitoring and prediction. Mon. Wea. Rev., 132, 1917–1932, https://doi.org/10.1175/1520-0493(2004)132<1917:AARMMI>2.0.CO;2.
Yang, W., R. Seager, M. A. Cane, and B. Lyon, 2015: The annual cycle of East African precipitation. J. Climate, 28, 2385–2404, https://doi.org/10.1175/JCLI-D-14-00484.1.
Zaitchik, B. F., 2017: Madden–Julian oscillation impacts on tropical African precipitation. Atmos. Res., 184, 88–102, https://doi.org/10.1016/j.atmosres.2016.10.002.
Zhang, C., 2005: Madden–Julian oscillation. Rev. Geophys., 43, RG2003, https://doi.org/10.1029/2004RG000158.
Zhang, C., 2013: Madden–Julian oscillation: Bridging weather and climate. Bull. Amer. Meteor. Soc., 94, 1849–1870, https://doi.org/10.1175/BAMS-D-12-00026.1.
