Large-Scale Stability and the Greater Horn of Africa Long and Short Rains

Kevin Schwarzwald aLamont-Doherty Earth Observatory of Columbia University, Palisades, New York
bInternational Research Institute for Climate and Society, Palisades, New York

Search for other papers by Kevin Schwarzwald in
Current site
Google Scholar
PubMed
Close
,
Richard Seager aLamont-Doherty Earth Observatory of Columbia University, Palisades, New York

Search for other papers by Richard Seager in
Current site
Google Scholar
PubMed
Close
,
Mingfang Ting aLamont-Doherty Earth Observatory of Columbia University, Palisades, New York

Search for other papers by Mingfang Ting in
Current site
Google Scholar
PubMed
Close
, and
Alessandra Giannini bInternational Research Institute for Climate and Society, Palisades, New York
cLaboratoire de Météorologie Dynamique/IPSL, Ecole Normale Supérieure, PSL Research University, Sorbonne Université, École Polytechnique, IP Paris, CNRS, Paris, France

Search for other papers by Alessandra Giannini in
Current site
Google Scholar
PubMed
Close
Open access

We are aware of a technical issue preventing figures and tables from showing in some newly published articles in the full-text HTML view.
While we are resolving the problem, please use the online PDF version of these articles to view figures and tables.

Abstract

The societies of the coastal regions of the Greater Horn of Africa (GHA) experience two distinct rainy seasons: the generally wetter “long” rains in the boreal spring and the generally drier “short” rains in the boreal fall. The GHA rainfall climatology is unique for its latitude in both its aridity and for the dynamical differences between its two rainy seasons. This study explains the drivers of the rainy seasons through the climatology of moist static stability, estimated as the difference between surface moist static energy hs and midtropospheric saturation moist static energy h*. In areas and at times when this difference, hsh*, is higher, rainfall is more frequent and more intense. However, even during the rainy seasons, hsh*<0 on average and the atmosphere remains largely stable, in line with the GHA’s aridity. The seasonal cycle of hsh*, to which the unique seasonal cycles of surface humidity, surface temperature, and midtropospheric temperature all contribute, helps explain the double-peaked nature of the regional hydroclimate. Despite tropospheric temperature being relatively uniform in the tropics, even small changes in h* can have substantial impacts on instability; for example, during the short rains, the annual minimum in GHA h* lowers the threshold for convection and allows for instability despite surface humidity anomalies being relatively weak. This hsh* framework can help identify the drivers of interannual variability in GHA mean rainfall or diagnose the origin of biases in climate model simulations of the regional climate.

© 2023 American Meteorological Society. This published article is licensed under the terms of the default AMS reuse license. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Kevin Schwarzwald, kevin.schwarzwald@columbia.edu

Abstract

The societies of the coastal regions of the Greater Horn of Africa (GHA) experience two distinct rainy seasons: the generally wetter “long” rains in the boreal spring and the generally drier “short” rains in the boreal fall. The GHA rainfall climatology is unique for its latitude in both its aridity and for the dynamical differences between its two rainy seasons. This study explains the drivers of the rainy seasons through the climatology of moist static stability, estimated as the difference between surface moist static energy hs and midtropospheric saturation moist static energy h*. In areas and at times when this difference, hsh*, is higher, rainfall is more frequent and more intense. However, even during the rainy seasons, hsh*<0 on average and the atmosphere remains largely stable, in line with the GHA’s aridity. The seasonal cycle of hsh*, to which the unique seasonal cycles of surface humidity, surface temperature, and midtropospheric temperature all contribute, helps explain the double-peaked nature of the regional hydroclimate. Despite tropospheric temperature being relatively uniform in the tropics, even small changes in h* can have substantial impacts on instability; for example, during the short rains, the annual minimum in GHA h* lowers the threshold for convection and allows for instability despite surface humidity anomalies being relatively weak. This hsh* framework can help identify the drivers of interannual variability in GHA mean rainfall or diagnose the origin of biases in climate model simulations of the regional climate.

© 2023 American Meteorological Society. This published article is licensed under the terms of the default AMS reuse license. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Kevin Schwarzwald, kevin.schwarzwald@columbia.edu

1. Introduction

The societies of the coastal regions of the Greater Horn of Africa (GHA) experience two distinct rainy seasons: the “long” rains in the boreal spring, known locally as the gu in Somali or masika in Swahili, and the “short” rains in the boreal fall, known locally as the deyr in Somali or vuli in Swahili, with the long rains having, on average, higher total seasonal rainfall. These rainy seasons interrupt the long (jilaal in Somali) and short (xagaa in Somali) dry periods, forming a climate that is unique both in its aridity and for the shape of its annual rainfall climatology.

In fact, the GHA is a dramatic outlier in mean rainfall for its latitude (Fig. 1), with mean rainfall substantially below the next-driest land area of the tropics, the Brazilian sertão, and, in certain locations, similarly dry as the equatorial Pacific cold tongue regions. Authors have connected this dryness with the prevailing low-level divergence and atmospheric descent that persist in the midtroposphere even during the rainy seasons (see, e.g., Yang et al. 2015; Nicholson 2017). This descent is modulated by the global system of equatorial Walker circulations, with ascent over the Maritime Continent associated with large-scale descent of dry, stable air over the western Indian Ocean and the Horn of Africa, and climatological westerlies at the equator off the coast (King et al. 2021; Limbu and Tan 2019; Zhao and Cook 2021). While a closed circulation cell may not reach all the way to the western Indian Ocean for the whole year (Hastenrath 2000), convection over the Maritime Continent and upper-level easterlies can still modulate the stability outside of the boreal fall when the Indian Ocean Walker cell is most coherent (Liebmann et al. 2017; Zhao and Cook 2021).

Fig. 1.
Fig. 1.

Mean tropical rainfall for the period 1982–2021. (top) Meridional average across 3°S–12.5°N for three different observational data products. (bottom) Map of GPCP rainfall. Note that CHIRPS (orange dots) and GPCC (purple dots) are land-only products; for longitudes with both ocean and land grid cells, they are not expected to perfectly match the global GPCP product. Vertical red lines identify the GHA; dotted black lines in the map show the extent of meridional averaging in the top panel. The GHA is, by a substantial amount, the driest region for its latitude, with GPCP showing rainfall at some locations to be even lower than over the equatorial Pacific cold tongue region, the other driest area in the tropics.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

Further work has examined the role of the topography in maintaining the region’s aridity, with the presence of the Ethiopian and Kenyan highlands, which reach up to 4500 and 3500 m, respectively (and with individual volcanic peaks, such as Mount Kenya/Kirinyaga, reaching above 5000 m), playing a key role. These largely prevent the advection of moist air from the Congo Basin and the Atlantic Ocean, particularly compared to the climate of the Ethiopian highlands themselves (Levin et al. 2009). Correspondingly, removing topography in modeling experiments results in a wetter Horn of Africa (Sepulchre et al. 2006), even during the seasons that are currently dry [e.g., see Boos and Kuang (2010) and Wei and Bordoni (2016) for the xagaa short dry season during the South Asian monsoon]. Furthermore, the Turkana jet, an orographically forced, low-level easterly jet passing through the Turkana Channel, a gap between the Kenyan and Ethiopian highlands over Lake Turkana, diverts part of the low-level circulation in all seasons and may be associated with decreased summer rainfall in northeastern Kenya and descent and divergence throughout the region (Indeje et al. 2001; Nicholson 2016).

Intriguingly, the aridity of eastern Africa is not an immutable fact of the current position of geologic landmasses. The Holocene alone has seen several longer-term pluvial and drought periods, generally associated with continental cooling (e.g., during the Little Ice Age) or warming, respectively (Verschuren et al. 2000; Tierney et al. 2011, 2015). In other words, the regional aridity cannot be the result of topography alone but can only be understood in concert with an analysis of the dynamical drivers of the local hydroclimate.

A second notable element of the GHA climate is the presence of two distinct rainy seasons overlapping with the equinoctial months (e.g., Trewartha 1981; Herrmann and Mohr 2011; Yang et al. 2015), though both the long and short rains generally peak after the equinox. This double-peaked climate reaches to the northern tips of the Horn of Africa at 12°N, which is far beyond the northern extent of the equatorial African region with two peaks to the west of the GHA [Figs. 2b,d; the study region with a double-peaked climatology (shaded in green), farther north than surrounding areas]. While many regions along the equator experience two seasonal peaks in rainfall due to the shift in the intertropical convergence zone (ITCZ), the GHA seasons are unrelated to this phenomenon. Though definitions of the ITCZ vary, none apply to the GHA rainy seasons; the regional climate is divergent, on average, even during the wet periods (Yang et al. 2015; Nicholson 2018). Uniquely, the GHA is the only longitude at which an ITCZ cannot be defined in any season. Furthermore, paleoclimate records on changes in the position of the ITCZ tend to not match up with shifts in rainfall proxies (Tierney et al. 2015). Instead, authors have emphasized the joint role of atmospheric and oceanic circulations in the Indian Ocean Basin to explain the seasonal march of rainfall.

Fig. 2.
Fig. 2.

Region with double-peaked rainy season in the GHA. (d) Map showing the natural log ratio of the first to the second harmonic of 1981–2021 CHIRPS rainfall. The region of study is any land pixel in the green area within the black box (representing 3°S–12.5°N and east of 32°E), where the second harmonic is larger than the first (i.e., where the rainfall has a double-peaked seasonal cycle [illustrated in (b)]). This region is largely composed of the lowlands of southeastern Ethiopia, Somalia, and eastern Kenya but excludes a thin coastal strip in southern Somalia, where the rainfall climatology consists of a dominant peak in the boreal summer. Also shown are representative rainfall climatologies for surrounding single-peaked regions: (a) Ethiopia and (c) Tanzania.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

During the jilaal long dry season, lasting generally from December to March, surface winds are northeasterly and relatively dry, passing over cooler northern Indian Ocean waters and running along isolines of constant humidity, thereby limiting onshore advection (Yang et al. 2015; Liebmann et al. 2017). These winds flow down the pressure gradient from the Arabian high, which is at its strongest at this time of year (Vizy and Cook 2020). In the upper troposphere, winds are strong and easterly, consistent with a Gill-like response to off-equatorial heating in the Southern Hemisphere (Liebmann et al. 2017).

About 2 weeks before the onset of the gu season, these upper-level easterlies abruptly weaken, presaging a sudden spike in CAPE and an abrupt onset of rainfall progressing from southwest (∼10–20 March, on average) to northeast (∼10–20 April) (Liebmann et al. 2017; Dunning et al. 2016). The Mascarene high strengthens, together with a breakdown of the South Indian convergence zone, reversing the geopotential height gradient in the western Indian Ocean. Consequently, low-level winds transition from northeasterly to southeasterly. During this season, SSTs off the coast of the GHA are at their warmest, and these low-level winds travel over warm SSTs, advecting moist, less stable air onto the continent (Yang et al. 2015; Vizy and Cook 2020). Occasional westerly advection of moisture from the Congo Basin can occur during the long rains; these events tend to produce the strongest rainfall in March–May for parts of Kenya in particular (Finney et al. 2019).

The xagaa short dry season begins as the large-scale circulation patterns over the Indian Ocean develop into the South Asian monsoon. Like the onset, the demise of the long rains progresses from the southwest (∼30 April–15 May) to the northeast (∼20 May–10 June). The strong, low-level, southerly Findlater or Somali jet develops along the coast, feeding into the monsoon circulation; the low-level circulation becomes more divergent (Yang et al. 2015; Vizy and Cook 2020). This jet forces strong upwelling along the coast, decreasing SSTs to their lowest annual levels and further stabilizing the lower troposphere (Slingo et al. 2005; Murtugudde et al. 2007). As with the start, the end of rainfall occurs rather abruptly (Liebmann et al. 2017), with the timing of the end of the long rains being correlated with the South Asian summer monsoon onset measured in Kerala, with a lead of 12 days, on average (Camberlin et al. 2010). Throughout this season, high-level easterlies return, consistent now with a Gill-like response to Northern Hemisphere heating (Liebmann et al. 2017).

While most of the region is dry in the boreal summer, a thin, coastal strip between roughly 42° and 45°E, including the major cities of Kismayo and Mogadishu, experiences a single annual peak in rainfall in June (see the thin strip with a single-peaked rainfall climatology by the Kenyan and southern Somali coast in Fig. 2). Older literature classifies this peak as an extension of the long rains [e.g., see Camberlin and Planchon (1997) for its occurrence farther south in Kenya]. United Nations bulletins instead refer to them as occurring during the xagaa short dry season, while some local sources identify them as part of a separate, local rainy season called the hhagaayo (e.g., Galaal 1992).

At the end of the South Asian monsoon, the Somali jet weakens, SSTs recover, and the low-level winds begin their transition from southerly to northerly. The deyr rains occur during this transition, as near-surface winds turn onshore and pass over relatively warm SSTs, bringing moist, less stable air onto the continent (Yang et al. 2015). The onset of the short rains progresses roughly from north (∼5 September) to south (∼25 October). During this season, the Indian Ocean Walker cell is most developed (Hastenrath 2000), and atmospheric ascent over the GHA is shallower than during the long rains (Zhao and Cook 2021; Schwarzwald et al. 2023). Finally, surface winds finish their transition to northeasterly, and the short rains end, bringing back the long dry season, with the demise progressing roughly north (∼25 October) to south (∼5–20 December).

Variability in both rainy seasons is substantial on interannual (e.g., Nicholson 2017; Vellinga and Milton 2018; Vizy and Cook 2020) and interdecadal (Tierney et al. 2013, 2015; Walker et al. 2020) scales, with interannual variability being greater in the short than the long rains. Long rain droughts seem to have increased in frequency over recent decades, with dramatic consequences for regional food security (Funk et al. 2008, 2018), despite global climate models (GCMs) projecting, on average, increases in GHA rainfall, a discrepancy sometimes referred to as the East African rainfall paradox in the literature (Lyon and DeWitt 2012; Lyon 2014; Rowell et al. 2015; Lyon and Vigaud 2017; Wainwright et al. 2019; Walker et al. 2020). Recent research has focused on possible anthropogenic sources for the observed drought increase, particularly on remote forcing through changes in Pacific Ocean SSTs and subsequent modulations of the Indian Ocean branch of the Walker circulation (Funk et al. 2018), which GCMs are currently largely unable to replicate when forced with historical greenhouse gas and aerosol inputs (Heede et al. 2021; Seager et al. 2022).

Disentangling the relative pathways of these influences on the climatology and variability of GHA rainfall and their interactions with each other remains a complex problem in both observations and GCMs. GCMs, in general, have difficulties simulating key aspects of the climatology of the GHA, including simulating long rains that are too weak and short rains that are too strong (Yang et al. 2014; Dunning et al. 2017; Schwarzwald et al. 2023). Understanding the seasonal cycle of rains in the GHA is a prerequisite to understanding why our state-of-the-art models are unable to accurately simulate regional processes, and understanding GHA biases in models is a prerequisite for improving GCM performance in the future so they can provide more reliable predictions of variability and projections of forced change. In this paper, we reexamine the seasonal cycle of GHA rains adopting a moist static energy (MSE) framework.

A rising parcel becomes saturated at its lifting condensation level. Since MSE is approximately conserved under adiabatic ascent and through phase changes between liquid water and water vapor, to first order, the difference between surface moist static energy (hs) and saturation moist static energy (h*), hsh*, is, therefore, proportional to thermal buoyancy of an air parcel at the height of h* (Khairoutdinov and Randall 2006). In particular, if hsh*>0 (i.e., if the MSE of a rising and now saturated parcel of air is larger than the saturation MSE in the free troposphere), the parcel will be positively buoyant and the atmosphere will be unstable, favoring convection and rainfall (Khairoutdinov and Randall 2006; Cook and Seager 2013). In particular, the MSE framework allows isolating the contributions to instability from humidity and temperature and, therefore, diagnosing the sources of convective energy in a given region.

Previous studies have looked at MSE in the GHA primarily in monthly data. Several studies, using 700 hPa as a reference level for h*, find that most of the GHA is conditionally stable (i.e., hsh*<0), even during the rainy seasons, and that stability tends to increase farther north in the region in all seasons (Yang et al. 2015; McHugh 2006; Vashisht and Zaitchik 2022). Yang et al. (2015) found that the climatology of hsh* peaks in the March–May season and is dominated by the hs component, which, in turn, is dominated by variability in surface moisture. Studies looking at h by itself have noted a similar dominance of the moisture component in its variability of the short rains (Liu et al. 2020).

In this study, we examine the seasonal cycle and variability of the difference between hs and h*. We analyze hsh* in daily GHA data and argue that the h* component may have a greater influence on regional stability than has previously been assumed. In section 2 we introduce the observational and reanalysis data sources used; in section 3, we detail calculations of hsh*. Section 4 shows the correspondence between hsh* and rainfall both in the regional climatology and at a local level; section 5 shows how the seasonal cycle of hsh* is driven by both surface and midtropospheric dynamics. Section 6 combines these findings into a working theory of the aridity and double-peaked nature of the rainfall climatology. Section 7 investigates interannual changes in rainfall and how hsh* can be used to understand these changes. Finally, section 8 summarizes results and explores how the large-scale stability framework could be applied to studies of future climate changes and climate model evaluation.

2. Data

Throughout this study, we use daily or higher-frequency data whenever possible to accurately characterize the timing of the rainy seasons (e.g., see Camberlin and Okoola 2003). Rainy seasons are often less than 2 months long and do not neatly match up with the Gregorian calendar; monthly data, therefore, often conflate both rainy and dry periods. Furthermore, subseasonal variability, apparent even in monthly data, suggests that higher temporal resolution data are needed to fully resolve the relevant dynamics (e.g., Camberlin and Philippon 2002). See Table 1 for a summary of data products used.

Table 1.

Data products used. If the resolution is not the same in both horizontal dimensions, resolution is shown as lat × lon. Unless otherwise stated, main text figures show rainfall data from CHIRPS, SST data from OISST, and circulation, humidity, and temperature data from MERRA2.

Table 1.

a. Rainfall data

We use daily precipitation data for 1981–2021 from the Climate Hazards Group Infrared Precipitation with Stations (CHIRPS) dataset (Funk et al. 2015), which combines satellite data from the Tropical Rainfall Measuring Mission (TRMM) satellite with interpolated rain gauge products and an elevation model. Evaluation of any dataset in the Horn of Africa is complicated by lack of a dense network of reference rain gauge observations (e.g., Dinku 2018). The observational network is particularly sparse in the double-peaked region of the GHA; Dinku et al. (2018), for example, for their evaluation, did not have access to station data in eastern Kenya and access to few in the Somali region of Ethiopia. Among studied data products, however, CHIRPS tends to outperform other commonly used datasets in the GHA, such as datasets based on satellite data alone (Ayehu et al. 2018; Dinku et al. 2018). CHIRPS overestimates the frequency of rainfall in parts of Ethiopia, but rainfall in those extra events tends to be minimal (Ayehu et al. 2018).

To validate CHIRPS results, we replicate our analysis using daily data from the Global Precipitation Climatology Project (GPCP; Adler et al. 2018) and from the Integrated Multi-satellitE Retrievals for GPM (IMERG) data product (Huffman et al. 2019). The former emphasizes data homogeneity in its construction, while the latter provides very high resolution data at the expense of greater changes in data sources over its available time period (Huffman 2022). IMERG uses a rotating satellite “ensemble of opportunity”; though it includes the TRMM satellite used by CHIRPS and, therefore, is not entirely independent, it assimilates data from 10 other satellites as well. All three datasets have a similar climatology over the GHA (Fig. S1 in the online supplemental material).

To place GHA rainfall in the context of rainfall across the tropics, in Fig. 1 we additionally show monthly rainfall data from the Global Precipitation Climatology Centre (GPCC; Becker et al. 2013) and the GPCP (Adler et al. 2018), spanning 1982–2021 to accommodate GPCC’s later start. Apart from Fig. 1, all other main text figures with rainfall show CHIRPS precipitation; validation results are shown in supplemental materials.

b. Atmospheric data

We use 1981–2021 daily reanalysis data from three reanalysis data products: the National Aeronautics and Space Administration (NASA) Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA-2; Bosilovich et al. 2015), fifth major global reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ERA5; Hersbach et al. 2020), and the Japanese 55-year Reanalysis (JRA-55) of the Japan Meteorological Agency (JMA) (Kobayashi et al. 2015). From all reanalysis products, daily temperature; humidity; surface radiation fluxes; top-of-atmosphere, shortwave downwelling, and shortwave fluxes; low cloud cover; and zonal winds are used. From MERRA-2, we additionally use 3-hourly temperature and humidity.

The 650-hPa pressure level is used for values in the free troposphere; this tends to be the pressure level with maximum hsh* during the long rains and close to the maximum during the short rains (Fig. S2). Results are qualitatively similar across pressure levels between 600 and 750 hPa. The lowest pressure level above each grid cell’s topography is used for near-surface values.

In addition to the paucity of surface observations mentioned above, evaluation of atmospheric variables in reanalysis products is further hindered by the lack of in situ observations at higher pressure levels. Furthermore, though a reanalysis’ precipitation can be compared to observational products such as CHIRPS as a test of a product’s performance, in general, rainfall fields in reanalyses are forecast, not assimilated, and thus are subject to additional biases compared to variables better constrained by assimilated observations. For example, ERA5’s precipitation fields are known to be substantially more biased than that of CHIRPS over the GHA (e.g., Steinkopf and Engelbrecht 2022) and are particularly wet throughout the year compared not only to CHIRPS but also the other reanalyses studied (Fig. S3). Of the three products, JRA-55 has the closest correspondence between its forecasted precipitation climatology and that of CHIRPS; however, few evaluations of the product have been published in the region.

Given the limited ability to evaluate reanalysis products in the region, the choice of any particular one remains largely arbitrary. Therefore, we repeat all analyses over the GHA using all three reanalysis products listed. In general, all produce qualitatively similar results for the analysis in this paper. Unless otherwise noted, main text figures show data from MERRA-2, with differences between reanalysis products highlighted where relevant and shown in Figs. S1 and S3.

c. SST data

We use 1981–2021 daily SST data for the Indian Ocean from the daily Optimum Interpolation Sea Surface Temperature (OISST), version 2.1, record (Huang et al. 2021). OISST is a 0.25° gridded product blending in situ ship and buoy measurements with satellite-derived estimates from the Advanced Very High Resolution Radiometer (AVHRR); though it has been found to be biased slightly low compared to in situ measurements, this discrepancy is small [e.g., ∼0.08°C vs Argo floats in Huang et al. (2021)].

d. Region of interest

Following common practice in studies of the GHA, we define an area of interest based on the extent of the region with two distinct peaks in rainfall each year. For much of this study, we consider every location in daily CHIRPS rainfall data on land between 32°E and the Indian Ocean and between 3°S and 12.5°N where the second harmonic is of greater magnitude than the first, that is, for which the 1 yr−1 frequency has less power than the 2 yr−1 frequency. This corresponds to the area shown in green in the box in Fig. 2d, and is referred to as the “double-peaked region” throughout.

e. Time frame

All analysis is conducted from 1981 to 2021 to use the longest time frame that spans all observational products analyzed, with one exception: Fig. 1 begins in 1982 to accommodate the shorter time frame of the GPCC dataset. This time span notably excludes part of the recent “triple dip” La Niña; resultant limitations are noted where relevant.

3. Methods

a. Calculating MSE

In this study, we primarily investigate the relationship between rainfall and large-scale atmospheric stability, defined as the difference between surface MSE hs and midtropospheric h*:
large-scalestabilityhsh*,wherehs=cpTs+gzs+Lυqsandh*=cpT+gz+Lυq*
for temperature T (in K), vertical height over mean sea level z, and specific humidity q, with the subscript s denoting surface values. The variable cp is the specific heat capacity of air at constant pressure 1004.6 J (kg K)−1, Lυ is the latent heat of vaporization of water 2.257 × 106 J kg−1, and g is the gravitational constant 9.807 m s−2. We calculate the saturation specific humidity q* using the approximation by Murray (1967); at a given height, q* and, subsequently, h*, is a function of T alone.

Values of h* are calculated at 650 hPa; this is comfortably above the planetary boundary layer in most locations in the region (barring isolated exceptions such as Mount Kenya) and tends to be the layer in which hsh* is at a midtropospheric maximum during both rainy seasons (Fig. S2); results are robust to choosing similar pressure levels between 600 and 750 hPa.

Since variability in z tends to be minimal at a given pressure level, we generally ignore the influence of geopotential height anomalies when discussing hsh* anomalies.

b. Defining seasonal extent

We additionally examine seasonally specific statistics of rainfall and MSE. Since the long and short rains rarely line up well with calendar months, commonly used 3-monthly averages (such as March–May for the long rains) aggregate across both wet and dry periods, which are governed by different dynamics (Riddle and Cook 2008; Schwarzwald et al. 2023).

To identify seasonal onsets and demises, we use the method of Dunning et al. (2016) based on inflection points in cumulative precipitation rate. This method was specifically designed for regions with a double-peaked rainfall climatology. For each CHIRPS grid cell in the double-peaked region, the onset and demise of the rainy seasons in each individual year from 1981 to 2021 are calculated. In sections 4 and 5, we compare across all gridcell days that are experiencing a given rainy season (long or short rains). In section 5, we additionally show tropical maps of seasonal composites of the horizontal circulation u and SSTs during the rainy and dry seasons; these use the average onset and demise across the whole double-peaked region in each year.

We compare seasonal averages for the long and short rains, applying the Dunning et al. (2016) method to March–May (MAM) and October–December (OND) averages, respectively, where relevant. Note that in section 7 and in figures analyzing interannual rainfall changes, we only use MAM and OND averages to characterize interannual differences, highlighting changes in overall seasonal strength instead of seasonal intensity.

c. Regional averages and regridding

Whenever possible, all datasets are analyzed at their native resolution. All maps show data on their original grid.

Regional averages are calculated across the CHIRPS double-peaked region, with averages weighted by the area of each grid cell. For CHIRPS rainfall data, all grid cells within this double-peaked region are used. When calculating regional averages for reanalysis products, which do not share a grid with CHIRPS, we regrid the CHIRPS ratio of the second harmonic to the first harmonic of the rainfall seasonal cycle (see section 3d) to the reanalysis grid. All grid cells for which this ratio is >1 after regridding are defined as part of the double-peaked region and are used in regional averages. In Fig. 3, reanalysis hsh* is regridded to the CHIRPS grid.

Fig. 3.
Fig. 3.

Statistics of CHIRPS rainfall in the double-peaked region, broken up by the hsh* value at a given location on a given day (a given “grid cell day”) and by season [(left) long rains; (center) short rains]. (a),(b) Histograms of hsh* (red line) and the fraction of rainfall by gridcell day hsh* (green bars). The top right of each panel shows the area-weighted fraction of unstable gridcell days in a given season. (c) The difference between the long and the short rains. (d),(e) The distribution of rainfall by gridcell day hsh*. Boxplots for a given hsh* bin are shown in dark green; black lines show the median gridcell day rainfall; dark green shading indicates the 0.25–0.75 interquartile range (IQR), and green whiskers extend out to 1.5 × IQR. Further quantiles are shown in light green shading (0.05–0.95), dashes (0.99), and dots (0.999). The maximum rainfall in a bin is written out. (f),(g) The fraction of gridcell days with P > 0 by gridcell day hsh*. For (d)–(g), note that 99.7% of gridcell days have hsh* between −25 and 5 kJ kg−1; values for bins beyond that range are calculated using very small sample sizes. See Fig. S5 for results using 3-monthly averages instead of local seasonal definitions and Fig. S6 for results using maximum daily hsh* (from 3-hourly data) instead of daily mean.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

4. Moist static instability is generally a condition for rainfall

a. Rainfall where and when hsh* is high

During both the long and the short rains, rain tends to fall where and when hsh* is higher, and particularly so when hsh* >0. The top row of Fig. 3 shows histograms of hsh* (red lines in Figs. 3a,b) across all days in each rainy season and across all grid cells within the double-peaked area of the GHA. Histograms of rainfall by hsh* value (i.e., the fraction of rain that falls on days and in grid cells with hsh* values in a certain histogram bin) are shown in green. Most rain tends to fall roughly when hsh*>5kJkg1 in both seasons. Figure S6 instead compares maximum daily hsh*, calculated from 3-hourly reanalysis data, and total daily rainfall; with this higher-frequency data, almost all rain falls on days when hsh* is confirmed to be positive for at least one 3-h period. In other words, rainfall on ostensibly stable days is likely due to averaging over subdaily instability and short-lived storms. Results are qualitatively similar when using MAM and OND averages, though an even smaller proportion of each period is unstable (Fig. S5), as would be expected when including dry periods outside of the immediate rainy seasons.

It tends to rain more (Figs. 3d,e) and more frequently (Figs. 3f,g) when hsh* is higher. Though it can rain on days and in locations with strongly negative daily mean hsh*, rainfall on those days tends to be minimal and rare. Note that few gridcell days have hsh*>5kJkg1.

The long and short rains in Fig. 3 are calculated at the gridcell level, i.e., all data points in the histograms are between each grid cell’s local onsets and demises in each year. Even within these seasons, however, the median gridcell day is stable, on average. Only 42% of gridcell days experiencing the long rains and 33% of gridcell days experiencing the short rains have average daily hsh*>0, consistent with the findings of Yang et al. (2015), who reported that the atmosphere over the GHA was, on average, stable even during the months associated with the rainy seasons. Note, however, that this does not preclude short periods of instability; the equivalent values for gridcell days when any 3-hourly period is unstable are 78% and 73% for the long and short rains, respectively (Fig. S6). The 3-hourly satellite rainfall data confirm the ephemerality of rain in the region; outside of areas directly on the coast, rainfall at a given location is largely confined to the afternoon and evening (Camberlin et al. 2018).

Seasonal differences exist, however. In line with the long rains producing, on average, more rain, the distribution of hsh* is shifted toward higher values during the long rains (Fig. 3c). Furthermore, a greater percentage of rain tends to fall on days with particularly high hsh* during the long rains than the short rains.

b. The climatologies of hsh* and rainfall

Just as rainfall is higher on days when and locations where hsh*>0, the seasonal cycles of the two variables are closely linked. The climatology of the fraction of the GHA that experiences hsh*>0 on a given day matches very closely with the rainfall climatology, both in its mean state and its interannual variability (Fig. 4). During the long-rain peak, roughly 50% of the double-peaked region is unstable in the daily mean at any one time; during the short rains, which have less total seasonal precipitation, on average, only about 30% of the region is unstable at a given time. The only notable discrepancy between the two climatologies is during the xagaa short dry season, when almost no grid cells have hsh*>0, on average, but occasional, limited rainfall is present. Within the double-peaked region, this rainfall is largely concentrated around the coast of southern Somalia and Kenya, including the hhagaayo (cf. Galaal 1992) region containing the cities of Kismayo and Mogadishu, where the annual rainfall peak is in June. Both climatologies are robust to the observational (for rainfall) or reanalysis (fraction unstable) data product used (Figs. S1 and S3).

Fig. 4.
Fig. 4.

The seasonal cycle of the percentage of grid cells in the double-peaked area where hsh*>0 in MERRA-2 (black line) and CHIRPS P (blue line). Shading shows the 0.25–0.75 interquartile range across years. All lines are smoothed with a Gaussian running average with a window width of 40 days. ERA5 and JRA-55 show similar instability climatologies (Fig. S4).

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

Figure 5a shows the 1981–2021 average climatology of rainfall and hsh*. Both variables peak during the long and short rains. On average, hsh*<0 during all seasons; in line with the arid climate of the GHA hsh* only occasionally becomes positive during the rainy seasons; the interquartile range across years of hsh* includes zero in both rainy seasons but by more for the long than the short rains (Fig. 5a). Even during the long rain peak, hsh*>0 on only about 40% of grid cell days, on average (Fig. 4). A yearly minimum in hsh* occurs in the boreal winter, the driest season in the GHA, while a relative minimum occurs during the relatively dry boreal summer. The hsh* values during the boreal summer are still, on average, substantially below the levels of average daily hsh* for which substantial rainfall occurs during the rainy seasons (Fig. 3). This relative maximum in stability is enough to ensure that the boreal summer remains largely dry.

Fig. 5.
Fig. 5.

The seasonal cycle of P, h*, hs, and its components. (a) Seasonal cycle of mean double-peaked region CHIRPS precipitation (blue line) and MERRA-2 hsh* (black line). (b) Seasonal cycle of mean double-peaked region anomalies vs the annual mean of MERRA-2 hsh* (black line), h* (green), cpTs (red), and Lυqs (blue). In both panels, light brown shading shows the climatological extent of the long and short rains, calculated using the average onset and demise across all double-peaked region grid cells, and light shading in the color of each line shows the 0.1–0.9 interquartile range (IQR) of the regional average, calculated across years for each day of year. All lines are smoothed with a Gaussian running average with a window width of 40 days. ERA5 and JRA-55 show similar climatologies of hsh* and its components (Fig. S4).

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

Despite the nonlinearity of the relationship between hsh* and precipitation P, with threshold behavior at hsh*=0, a strong relationship between hsh* and average area fraction unstable can even be seen in seasonal averages (Fig. 6).

Fig. 6.
Fig. 6.

Double-peaked region average rainfall vs (a),(b) double-peaked region average hsh* and (c),(d) GHA average area fraction unstable (i.e., hsh*>0) for (a),(c) March–May averages corresponding to the long rains and (b),(d) October–December averages corresponding to the short rains. Red dots and arrow highlight the largest year-on-year increase in MAM rainfall, between 2017 and 2018, which is further analyzed in Fig. 13.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

5. The seasonal cycle of large-scale stability is composed of different seasonal cycles of its components

We have shown in the preceding section that hsh* and P are closely linked, both at the level of individual grid cells and days, and in their overall climatologies. Each component of hsh* plays a role in producing this seasonal cycle, and each component’s seasonal cycle is unique (Fig. 5b). This section individually analyzes the seasonal cycles of Ts, qs, and h* and their contributions to the hsh* seasonal cycle. We do not consider variations in geopotential height z, since, at a given pressure level, they are minimal enough to not substantially alter hsh*.

a. Surface temperature

Near-surface air temperature has a double-peaked climatology (red dot–dashed line in Fig. 5b, showing cpTs, and red lines in Fig. 7), though it is phase shifted from that of rainfall and hsh*, with two peaks roughly in line with the equinoxes right at the onset of the rainy seasons. Note that ERA5 and JRA-55 also see a small local maximum toward the end of the long rains (Fig. S4). Though incoming solar radiation is already decreasing during the long and short rains (Fig. 7b), lower temperatures are also likely facilitated through latent cooling of the surface and increased cloud cover associated with the rains (Figs. 7c,e); similar preseason peaks are often seen in monsoonal regimes (Gadgil 2003), and authors have noted geographic anticorrelations between rainfall and temperature in climatologies in the region (Ongoma and Chen 2017). Some of this cooling is counteracted by decreased sensible flux (Fig. 7f, in line with increased latent fluxes) and the advection of warm air into the region (Fig. 7d), possibly explaining the decoupling between near-surface air temperature and surface temperature in the latter half of both rainy seasons (Fig. 7a; this slight discrepancy is also not robust across reanalysis data products).

Fig. 7.
Fig. 7.

The seasonal cycle of MERRA-2 near-surface air temperature Ts (red) and (a) surface upwelling longwave radiation, (b) top-of-atmosphere (TOA) downwelling shortwave radiation, (c) surface net downwelling shortwave radiation, (d) advection of Ts into the region, (e) low cloud fraction, (f) upwelling latent heat, (g) upwelling sensible heat, and (h) OISST SSTs off the GHA coast (all in blue). The shading shows the interannual 0.25–0.75 IQR of each seasonal cycle. All atmospheric variables are averaged over the GHA double-peaked region. In (h), the solid line and shading are the average over only the closest grid cell to the GHA coast for each latitude between 3°S and 10°N; the dashed line is the average SST over a box west of 55°E and between 3°S and 10°N.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

Note that incoming solar radiation is seasonally asymmetric (Fig. 7b). The September insolation peak is weaker than the March peak, due to the eccentricity of Earth’s orbit, and the July minimum is higher than the December minimum because the study region is largely located in the Northern Hemisphere.

Despite this, the coldest season is the boreal summer during the xagaa short dry season. During this season, SSTs in the western Indian Ocean are at their annual minimum, largely due to the advection of relatively cold air over the ocean from the Southern Hemisphere winter (Ongoma and Chen 2017; Chatterjee et al. 2019) (Figs. 7d and 9e). Upwelling along the Somali coast due to the alongshore monsoon flow further cools SSTs in the northern part of the region, especially toward the start and peak of the South Asian monsoon season (Schott and McCreary 2001; Chatterjee et al. 2019) (Fig. 7h). While most of the onshore flow crosses south of the primary upwelling zone, SSTs are lower along the coast than off the coast throughout the year, especially during the peak upwelling season in the boreal summer (Fig. 7h). In addition to the cold cross-equatorial advection, low clouds driven by low SSTs block incoming solar radiation (Figs. 7c,e), further cooling and stabilizing the region. At the end of this season, temperatures rise again in anticipation of the equinox, leading into the short rains.

The second annual Ts minimum occurs during the boreal winter, corresponding to the jilaal long dry season. During this season, temperatures begin lower, after falling during the short rains; however, temperatures, on average, do not drop to the level of the boreal summer minimum, despite this being the Northern Hemisphere winter season. The annual minimum in cloud cover (Fig. 7e) increases surface insolation, especially approaching the March equinox when incoming solar radiation is at its annual maximum. Low clouds are suppressed by relatively higher SSTs, with coastal downwelling associated with the northeasterly flow partially balanced by the influx of cold, continental air (Figs. 7h and 9a) (Murtugudde et al. 2007).

Note the differences in the seasonal cycles of Ts and tropospheric temperature. The preseason peak in Ts before the long rains, therefore, sets up the instability, countering the stabilizing force of increased h*. However, during the start of the long rains, anomalies of surface and 650-hPa temperature relative to their climatological mean are nearly equal, while surface temperature decreases faster than tropospheric temperature toward the end of the long rains. Correspondingly, hsh* decreases faster than surface humidity, and rainfall peaks before surface moisture and decreases more quickly than surface moisture anomalies as well. During the short rains, surface and midtropospheric temperature move in rough opposition.

As is expected for tropical climates, interannual variability in cpTs is substantially lower than interannual variability in Lυqs, and slightly lower than the same in h* (Fig. 8).

Fig. 8.
Fig. 8.

Interannual variability of MERRA-2 seasonal mean double-peaked region anomalies vs the annual mean of the temperature (red) and humidity (blue) components of hs (gray) and anomalies of h* vs the annual mean (green). Seasonal means are calculated using local onsets and demises at every location in the double-peaked area. The filled area represents the interquartile range (IQR); the whiskers extend out to the farthest data point within 1.5 × IQR beyond the 0.25 and 0.75 quantiles, with further outliers as open circles. The white line in each boxplot represents the median year.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

b. Surface humidity

The component of hsh* with the strongest annual cycle is surface specific humidity qs (blue dashed line in Fig. 5b showing Lυqs), which shares a double-peaked climatology with hsh* and rainfall, and which correspondingly peaks during the long and short rains. Note how the qs seasonal cycle peaks slightly after rainfall during the long rains and remains elevated over its climatological mean for some time after the end of the long rains; the stabilizing effect of decreasing Ts during this time is needed to fully explain the timing of the seasonal cycle of precipitation.

During the rainy seasons, interannual variability in qs, as measured by the 0.25–0.75 interquartile range across years, is highest during the short rains (Fig. 8). Across all seasons, variability of qs is highest during the boreal winter, but, on average, strongly negative values of hsh* and low moisture availability combine to make rainfall rare. Notably, all but the lowest average long rains mean qs in the sample is higher than the mean qs in all dry seasons, but mean qs in the short rains is, on occasion, lower than the highest mean qs in either dry season.

The seasonal cycle of surface humidity is largely a function of the seasonal cycle of the surface circulation (Fig. 9). The moisture supply for the GHA is primarily the Indian Ocean; consequently, changes in the atmospheric circulation over the Indian Ocean Basin throughout the year modulate the moisture transport and moisture convergence or divergence over the GHA. During the boreal winter, strong northeasterly winds transport limited moisture over the study region (Figs. 9a,b). This air originates in dry continental Asia and only passes briefly over relatively cool SSTs before passing by the GHA, where the circulation is divergent and its onshore component is weak (Fig. 9a); (Yang et al. 2015; Vizy and Cook 2020). Similarly, during the boreal summer, the monsoon winds and the Somali jet transport moisture over the study region (Figs. 9e,f) but tend to diverge, with part of the circulation passing through the Turkana Gap into the Congo Basin (Yang et al. 2015; Nicholson 2016). Thus, in both dry seasons, the flow over the GHA is from the winter hemisphere with cooler SSTs and drier air, reducing hs and favoring stability and low precipitation.

Fig. 9.
Fig. 9.

Seasonal average MERRA-2 (left) surface circulation u and (right) surface moisture transport uq, with seasons defined using the average onset and demise dates across the double-peaked region. Shading shows hsh* anomalies vs annual mean (over land, left color bar) and OISST SSTs (over ocean, right color bar). The red contour shows the double-peaked study area in the GHA, for reference. White space over the Himalayas shows no hsh* data since the topography is higher than the reference 650-hPa pressure level used for h*.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

In contrast, during the rainy seasons in between the boreal winter and summer monsoons, moisture transport has a strong onshore component. During the long rains, this wind arrives after passing over relatively warm SSTs over the southern Indian Ocean at the end of its summer, while during the short rains, the flow is similarly onshore but from cooler SSTs following Southern Hemisphere winter (Figs. 9c,g and 7h). The seasonal cycles of flow directions and SSTs in the western Indian Ocean thus explain the seasonal cycle of surface humidity; these seasonal cycles are fundamentally caused by the swing between the northern summer and winter Asian monsoons.

The surface moisture climatology combines with the surface temperature climatology to produce seasonal maxima of hs during the long and short rains. During the long rains in particular, this average 5.3 kJ kg−1 anomaly over the annual mean is geographically coherent with positive hs anomalies over much of the tropics in the Atlantic and Indian Ocean basins (Fig. 10b). During the short rains, the pattern is more localized, while much of the study region experiences mild positive hs anomalies on the order of roughly 2 kJ kg−1 (in line with similar signals in the Sahel and southwestern Africa), the Indian Ocean, Congo Basin, and the tropical and South Atlantic all have negative hs anomalies (Fig. 10d).

Fig. 10.
Fig. 10.

Maps of surface h and u anomalies vs the annual mean, showing the 1981–2021 average over each season, with seasons defined using the average onset and demise dates across the double-peaked region. Green contour shows double-peaked region, for reference.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

c. Tropospheric temperature and h*

If a near-surface parcel’s moist static energy hs is greater than h*, which, at a given height, is a function of temperature alone, then that parcel will have positive thermal buoyancy if displaced to that height. In other words, h* sets a tropospheric limit to atmospheric instability, and though it can be affected by local convection, tends to have weak horizontal gradients across large scales in the tropics (e.g., Sobel et al. 2001).

In contrast to the surface, local thermodynamic variables analyzed above, h* has a single-peaked climatology, with a maximum during the long rains and a minimum during the short rains. The relative minimum in the seasonal cycle of h* during the boreal fall is crucial for hsh* to reliably reach positive values during the short rains.

Interannual variability in h* is strongest during the short rains and the boreal winter (Fig. 8); this is the time of year when the magnitude of ENSO events tends to be strongest. Given known teleconnections between ENSO and the tropical troposphere (e.g., Yulaeva and Wallace 1994; Chiang and Sobel 2002), this connection is unsurprising.

Though convection can increase tropospheric temperature and, therefore, h* through latent heating (e.g., Sobel et al. 2002), hs is already increasing months before the long rains begin, making it unlikely that the annual peak is solely a reaction to the rainy season. Similarly, h* decreases as the short rains begin and only increases once again as the rains weaken. Given the weak horizontal gradients of temperature in the free troposphere, the remote forcing must arise from the seasonal cycle in surface temperatures and moist convection throughout the tropics.

This connection is, however, seasonally dependent and can be geographically limited; the GHA seasonal cycle for h* can differ substantially from that in surrounding regions (Fig. 11). The GHA shares the boreal spring timing of the annual peak of h* during the long rains with tropical Africa, South America, and the Indian and Atlantic Oceans; seasonal average h* values across the tropics are highly correlated with GHA h* during this season (Fig. S7). The climatological minimum during the short rains seen in the GHA is more localized. In fact, the short rain h* anomaly is the lowest in the entire tropics between 15°S to 15°N during any season. Nearby regions to the northwest in the Sahara and to the north in the Arabian Peninsula also experience a climatological minimum of h* during the GHA short rains; of those regions, however, only northeast Africa shows a substantial daily and interannual correlation with the GHA during this season (Figs. S7 and S8). The correlation with most of the Arabian Peninsula is even weakly, but insignificantly (at the p = 0.05 level; see Fig. S7 for details), negative for seasonal means, suggesting that these regions are not directly dynamically linked in the midtroposphere during the rainy seasons. Instead, h* over the GHA during the short rains is strongly correlated with h* over the eastern Indian Ocean, a region with known teleconnections to the GHA through the Indian Ocean dipole and the Walker circulation. Understanding the local and remote drivers of h* variability is crucial to understanding the drivers of the hsh* seasonal cycle; more research will be needed to identify them.

Fig. 11.
Fig. 11.

Maps of 650-hPa h* and u anomalies vs the annual mean, showing the 1981–2021 average over each season, with seasons defined using the average onset and demise dates across the double-peaked region. Green contour shows double-peaked region, for reference. Note the smaller range of the color bar compared to Fig. 10, in line with lower spatial or temporal variability in midtropospheric temperature compared with surface values.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

6. A working theory of the seasonal cycle of precipitation in the GHA

These results allow us to present a general theory about the two biggest peculiarities of the GHA climate: its aridity and its two distinct rainy seasons.

The double-peaked region’s aridity can be explained by the fact that hsh*<0 in both the climatological average and, on average, in even the rainy seasons. During the peak of the long rains, usually the wettest part of the year, only 50% of the land area of the region is unstable on a given day in the daily average; during the long dry season, that proportion is usually less than 5%.

Even the rainy seasons are drier than would be expected for this latitude. During the long rains, convection is made more difficult by a seasonal maximum in h*, while during the short rains, hs is relatively low because of the cool SSTs over which the incoming moist circulation originates. Consequently, the two distinct rainy seasons are a product of the interplay between the seasonal cycles of surface temperature, surface humidity, and midtropospheric temperature.

We now pursue this analysis for each season separately, beginning with the boreal winter dry period. At the height of the jilaal long dry season in January, surface moisture is at its annual minimum, while temperature at the surface and at 650 hPa is at its annual average. Dry northeasterly winds, which have a continental origin, carry little moisture into the GHA; consequently, hsh* almost never reaches 0 and rainfall is at its rarest. Since this season falls during the Northern Hemisphere winter, tropical positive hs anomalies compared to local climatology are almost exclusively confined to the Southern Hemisphere. Though the region warms up as the dry season progresses, both surface and 650 hPa temperature increase in tandem, without a net impact on hsh*. For the rainy season to start, this balance must shift, with hs increasing or h* decreasing.

The long rains thus begin when the northeasterlies weaken and surface moisture sharply increases, increasing hs such that hsh* can be positive, and peak when surface moisture reaches its peak in late April and early May. During this time, weaker onshore winds originating from the Southern Hemisphere advect moist air from the western Indian Ocean, which is now at its warmest temperatures of the year at the end of austral summer. This surface state would clearly favor high precipitation. However, 650-hPa temperature and, therefore, h* are at their yearly maximum not only over the GHA but across much of the tropical Africa and the Atlantic and Indian Ocean basins; this maximum offsets, but is not enough to outweigh, the destabilizing impact of the surface moisture maximum. Not only is hs is at its yearly maximum in the GHA but also in the equatorial Indian Ocean and central Africa. By the second half of the long rains, surface temperature begins to fall while 650-hPa temperature remains high; correspondingly, hsh* and average rainfall decrease faster than does the surface specific humidity.

During the xagaa short dry period in the boreal summer, during the peak of the south Asian monsoon, surface humidity remains substantially higher than during the boreal winter. However, surface temperature is anomalously low, with southeasterly inflow from the winter hemisphere where strong winds have cooled the SST, while 650-hPa temperature remains anomalously high. Therefore, hsh* almost never reaches zero, and rainfall is rare and confined to a thin coastal strip.

Shortly before the onset of the short rains, 650-hPa h* begins decreasing to its lowest value of the year, lowering the tropospheric threshold for convection. Onshore winds are once more from the southeast and originate from SSTs that remain cool following austral winter and the strong winds associated with the peak boreal summer monsoon. Surface moisture anomalies are once again above the climatological average but remain substantially lower than during the long rains. It is only the yearly minimum in h*, the largest negative anomaly for its latitude, that allows hsh* to regularly cross zero in this season. Subsequently, rainfall is lower and more variable than during the long rains.

7. Interannual changes in hsh* explain interannual changes in regional rainfall

A theory of the seasonal cycle of precipitation in the GHA can be tested to see if it explains differences in the seasonal cycle between different periods. Figure 12 shows year-on-year changes in rainfall and both hsh* and the average area fraction of the double-peaked area that is unstable in each season. Note that unlike in previous sections, we define seasons using MAM and OND averages for the long and short rains, respectively. This is because, since hsh* is associated with rainfall, using locally defined onsets and demises would merely show hsh* contingent on rainfall and, therefore, highlight interannual differences in rainfall intensity instead of differences in overall seasonal strength. Interannual changes are closely correlated with changes in both hsh* and fraction of the area unstable (Fig. 12; results are robust to data products used; see Fig. S9 for other reanalysis and rainfall data products). Surprisingly, this relationship is strongly linear, despite the nonlinear relationship between rainfall and hsh* at the level of individual grid cells and gridcell days shown in Fig. 3. We also note the difference in slope between the short and long rains; the same decrease in stability produces a greater change in rainfall during the short than the long rains.

Fig. 12.
Fig. 12.

Interannual changes in double-peaked region average rainfall vs interannual changes in (a),(b) double-peaked region average hsh* or (c),(d) average area fraction unstable (i.e., hsh*<0) for (a),(c) March–May averages corresponding to the long rains and (b),(d) October–December averages corresponding to the short rains. Numbers shown in upper-left corner of each panel are Pearson’s correlation coefficients between rainfall and hsh* or fraction unstable changes shown in that panel. See Fig. S9 for correlations calculated with other rainfall and reanalysis data products. Red dots highlight the largest year-on-year increase in MAM rainfall, between 2017 and 2018, which is further analyzed in Fig. 13.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

The linearity of this relationship is not necessarily mirrored at the level of individual locations. Interannual changes in hsh* may have a different effect on rainfall depending on hsh* in a given year and at a given location. In the driest locations, particularly in the northeastern tip of the Horn of Africa, humidity may be too low to allow for substantial rainfall even if hsh*<0. Figure 13 shows an example map of rainfall and hsh* changes for the strongest year-on-year increase in the long rains in the data analyzed, from 2017 to 2018 (marked in red in Fig. 12). Though hsh* has the same sign change as the rainfall change over the double-peaked region as a whole, several regional differences are seen. Rainfall in the northeastern tip of the Horn of Africa decreased while hsh* strongly increased, and the maximum hsh* change is over the central border between Somalia and Ethiopia, as opposed to over central Kenya, like the rainfall maximum.

Fig. 13.
Fig. 13.

Illustration of interannual variations in rainfall and hsh* and its components. Shown are 2017 vs 2018 differences between (a) mean seasonal precipitation and (b)–(e) mean seasonal hsh* and its components in the long rains, corresponding to the years with the largest year-on-year increase in rainfall in the data analyzed (highlighted in red in Fig. 12). Note that changes in h* are shown with their sign reversed such that positive values represent decreased stability, in line with the other panels. Geographic anticorrelations between rainfall and temperature have previously been noted in climatologies in the region by Ongoma and Chen (2017), in line with expectations of latent cooling of the surface.

Citation: Journal of Climate 36, 20; 10.1175/JCLI-D-23-0126.1

Nevertheless, regional average changes in hsh* seem to explain well regional average changes in rainfall. Trends in hsh* can be significant even if small in absolute terms; for example, a uniform increase in hsh* of 1 kJ kg−1 in historical data (the order of magnitude of interannual changes shown in Fig. 12 and less than many of the unusually large changes shown in Fig. 13 in both surface and midtropospheric variables) would increase the proportion of unstable gridcell days from 42% (cf. Fig. 3) to 60% in the long rains. Even variations in h*, i.e., tropospheric temperature, which tend to be low in absolute magnitude but comfortably span more than 1 kJ kg−1 across years (Fig. 8), can substantially impact the local stability of the air column.

8. Discussion and conclusions

In seeking to explain the aridity and double-peaked seasonal cycle of rainfall in the GHA, we have shown close correspondences between large-scale stability and rainfall. In areas and at times when hsh* is higher, the probability of rainfall is higher, as is the magnitude of average daily rainfall. In line with the unusual aridity of the GHA, mean seasonal hsh* remains below 0 in both rainy seasons; its daily distribution reaches positive values more often during the long than the short rains and does so rarely, if at all, during the dry seasons. The seasonal cycle of hsh*, in which the unique seasonal cycles of all three components (Lυqs, cpTs, h*) play a role, corresponds closely with the seasonal cycle of rainfall and explains well the double-peaked nature of the regional hydroclimate.

However, though powerful, the hsh* framework cannot capture all of the regional dynamics at play in the GHA. Recent research has solidified the case for a significant influence of Congo Basin westerlies on the strength of the long rains; since the Rift Mountains and easterly flow through the Turkana Channel tend to block low-level westerly transport, this moisture transport is expected to occur in the midtroposphere instead (Finney et al. 2019; Walker et al. 2020). Dyer and Washington (2021) correspondingly find that differences between wet and dry long rain composites of h are stronger in the midtroposphere than at the surface and suggest a positive feedback between midtropospheric moisture and rainfall. This effect would work via the influence of midtropospheric moisture on the buoyancy of entraining convective plumes and would not be captured by focusing only on hs and h*, the latter of which only depends on midtropospheric temperature. In line with this, Schwarzwald et al. (2023) find a weaker interannual correlation between average seasonal hsh* and average seasonal rainfall during the long rains than the short rains.

Indeed, several further questions about the relationship between hsh* and the rainy seasons remain unanswered. The drivers of the different seasonal cycles of midtropospheric h* and surface temperature in the GHA also have not fully been investigated. The seasonal large-scale maximum of h* during the long rains prevents them from being stronger, while the more-localized minimum in the short rains prevents them from being even weaker. This variability in midtropospheric temperature is influenced by the entire seasonal cycle of the tropics due to the inability of the tropical troposphere to sustain large horizontal temperature gradients.

Nevertheless, the hsh* framework conveniently explains several well-known features of the hydroclimate of the study region. In line with the region’s arid climate, hsh* is negative in the climatological mean and even remains negative during much of the rainy seasons themselves. The general aridity is not just a product of the two dry seasons but also is due to the fact that the equinoctial rainy seasons, whose existence would otherwise be typical for an equatorial region, are quite dry. For the long rains, this is because h* is high during this season even as warm moist air flows in from the southern Indian Ocean at the end of its warm season. For the short rains, this is so because, even though h* is low, the winds inflowing from the southern Indian Ocean come from waters that have been cooled during the winter by strong winds associated with the Asian summer monsoon. The long rains are, on average, stronger than the short rains; correspondingly, hsh* is positive more often during the boreal spring than the boreal fall. Interannual variability in the short rains is more strongly coupled to remote teleconnections; during this time, hs and h* anomalies are balanced enough that a forcing to either can have an outsized impact on regional stability.

Many studies of the tropical climate assume a form of the weak temperature gradient (WTG), which assumes that, due to the low strength of the Coriolis force in the tropics, h* above the boundary layer is expected to be relatively uniform in space at a given pressure level (Sobel et al. 2001). Though our results are consistent with this observation (note the significantly larger range of hs anomalies in Fig. 10 than h* anomalies in Fig. 11), we also show that, in certain cases, even small variations in h* can have substantial impacts on local atmospheric stability. The relationship between hsh* and rainfall shows strong threshold effects, with rainfall becoming much more likely when hsh*>0. Thus, a complete and improved theory of the seasonal cycle of precipitation in the GHA will require a theory of the seasonal cycle of the spatially uniform component of the tropical tropospheric temperature and its small but important local variations. Furthermore, it would be of interest to investigate whether the relationship between GHA hsh* and rainfall found in this study is generalizable to other convective regimes in the tropics and elsewhere, and whether it is similar over both land and ocean, for example. Some work in this direction has already been conducted, for example, by Cook and Seager (2013) on the North American monsoon.

Further studies could also apply the hsh* framework to better understand the drivers of interannual variability in the region. For example, even though El Niño years tend to be wet in the GHA (e.g., see the discussion in Nicholson 2017), a recent study found that positive El Niño conditions can cause drying in the GHA if not accompanied by certain Indian Ocean SST patterns that often, but not always, occur during positive El Niño years (Wenhaji Ndomeni et al. 2018). This result is consistent with El Niño’s role in warming the tropical troposphere (e.g., Sobel et al. 2002), which would increase h* and the threshold for convection. Identifying physical drivers of such changes, particularly if they affect hs and h* differently or independently, may be crucial to understanding the impact of climate change on future changes in GHA rainfall as well.

Finally, hsh* can be a powerful tool to investigate GCM behavior in the GHA. Schwarzwald et al. (2023), for example, found that CMIP6 biases in the rainy seasons of the double-peaked region of the GHA could partially be explained by models advecting too much surface moisture into the region and having tropical tropospheric temperatures that are too cool during the short rains, though models systemically overestimated hsh* even when producing too little rainfall. In regions such as the GHA, where rainfall is affected by a complex confluence of local and remote forcings, analyzing model behavior through hsh* can help identify the underlying processes causing model biases.

Acknowledgments.

The authors are grateful for insightful conversations with Adam Sobel, David Rowell, and Sarah Smith and for constructive comments from three anonymous reviewers. The authors thank the International Research Institute for Climate and Society (IRI) for institutional and computational support. This work is partially undertaken as part of the Columbia World Project, ACToday, Columbia University, New York, NY. K.S. was also supported by the Graduate School of Arts and Sciences and the Department of Earth and Environmental Sciences at Columbia University. R.S. was supported by NSF Award OCE-22-19829.

Data availability statement.

Data needed to replicate main text figures and analysis are available at https://doi.org/10.5281/zenodo.8092600. All code needed to replicate this study is available at https://doi.org/10.5281/zenodo.8092685; the latest version is stored at https://github.com/ks905383/gha_stability. All other data are available by request.

REFERENCES

  • Adler, R. F., and Coauthors, 2018: The Global Precipitation Climatology Project (GPCP) monthly analysis (new version 2.3) and a review of 2017 global precipitation. Atmosphere, 9, 138, https://doi.org/10.3390/atmos9040138.

    • Search Google Scholar
    • Export Citation
  • Ayehu, G. T., T. Tadesse, B. Gessesse, and T. Dinku, 2018: Validation of new satellite rainfall products over the upper Blue Nile basin, Ethiopia. Atmos. Meas. Tech., 11, 19211936, https://doi.org/10.5194/amt-11-1921-2018.

    • Search Google Scholar
    • Export Citation
  • Becker, A., P. Finger, A. Meyer-Christoffer, B. Rudolf, K. Schamm, U. Schneider, and M. Ziese, 2013: A description of the global land-surface precipitation data products of the Global Precipitation Climatology Centre with sample applications including centennial (trend) analysis from 1901–present. Earth Syst. Sci. Data, 5, 7199, https://doi.org/10.5194/essd-5-71-2013.

    • Search Google Scholar
    • Export Citation
  • Boos, W. R., and Z. Kuang, 2010: Dominant control of the South Asian monsoon by orographic insulation versus plateau heating. Nature, 463, 218222, https://doi.org/10.1038/nature08707.

    • Search Google Scholar
    • Export Citation
  • Bosilovich, M. G., and Coauthors, 2015: MERRA-2: Initial evaluation of the climate. Technical Report Series on Global Modeling and Data Assimilation, Vol. 43, NASA Tech. Rep. NASA/TM-2015-104606, 145 pp., https://gmao.gsfc.nasa.gov/pubs/docs/Bosilovich803.pdf.

  • Camberlin, P., and O. Planchon, 1997: Coastal precipitation regimes in Kenya. Geogr. Ann., 79A, 109119, https://doi.org/10.1111/j.0435-3676.1997.00010.x.

    • Search Google Scholar
    • Export Citation
  • Camberlin, P., and N. Philippon, 2002: The East African March–May rainy season: Associated atmospheric dynamics and predictability over the 1968–97 period. J. Climate, 15, 10021019, https://doi.org/10.1175/1520-0442(2002)015<1002:TEAMMR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Camberlin, P., and R. E. Okoola, 2003: The onset and cessation of the “long rains” in eastern Africa and their interannual variability. Theor. Appl. Climatol., 75, 4354, https://doi.org/10.1007/s00704-002-0721-5.

    • Search Google Scholar
    • Export Citation
  • Camberlin, P., B. Fontaine, S. Louvet, P. Oettli, and P. Valimba, 2010: Climate adjustments over Africa accompanying the Indian monsoon onset. J. Climate, 23, 20472064, https://doi.org/10.1175/2009JCLI3302.1.

    • Search Google Scholar
    • Export Citation
  • Camberlin, P., W. Gitau, O. Planchon, V. Dubreuil, B. M. Funatsu, and N. Philippon, 2018: Major role of water bodies on diurnal precipitation regimes in eastern Africa. Int. J. Climatol., 38, 613629, https://doi.org/10.1002/joc.5197.

    • Search Google Scholar
    • Export Citation
  • Chatterjee, A., B. P. Kumar, S. Prakash, and P. Singh, 2019: Annihilation of the Somali upwelling system during summer monsoon. Sci. Rep., 9, 7598, https://doi.org/10.1038/s41598-019-44099-1.

    • Search Google Scholar
    • Export Citation
  • Chiang, J. C. H., and A. H. Sobel, 2002: Tropical tropospheric temperature variations caused by ENSO and their influence on the remote tropical climate. J. Climate, 15, 26162631, https://doi.org/10.1175/1520-0442(2002)015<2616:TTTVCB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Cook, B. I., and R. Seager, 2013: The response of the North American monsoon to increased greenhouse gas forcing. J. Geophys. Res. Atmos., 118, 16901699, https://doi.org/10.1002/jgrd.50111.

    • Search Google Scholar
    • Export Citation
  • Dinku, T., 2018: Overcoming challenges in the availability and use of climate data in Africa. ICT Update, Vol. 87, 14, https://cgspace.cgiar.org/handle/10568/96855.

  • 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, 292312, https://doi.org/10.1002/qj.3244.

    • Search Google Scholar
    • Export Citation
  • Dunning, C. M., E. C. L. Black, and R. P. Allan, 2016: The onset and cessation of seasonal rainfall over Africa. J. Geophys. Res. Atmos., 121, 11 40511 424, https://doi.org/10.1002/2016JD025428.

    • Search Google Scholar
    • Export Citation
  • Dunning, C. M., R. P. Allan, and E. Black, 2017: Identification of deficiencies in seasonal rainfall simulated by CMIP5 climate models. Environ. Res. Lett., 12, 114001, https://doi.org/10.1088/1748-9326/aa869e.

    • Search Google Scholar
    • Export Citation
  • Dyer, E., and R. Washington, 2021: Kenyan long rains: A subseasonal approach to process-based diagnostics. J. Climate, 34, 33113326, https://doi.org/10.1175/JCLI-D-19-0914.1.

    • Search Google Scholar
    • Export Citation
  • Finney, D. L., J. H. Marsham, D. P. Walker, C. E. Birch, B. J. Woodhams, L. S. Jackson, and S. Hardy, 2019: 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, 647664, https://doi.org/10.1002/qj.3698.

    • Search Google Scholar
    • Export Citation
  • Funk, C., M. D. Dettinger, J. C. Michaelsen, J. P. Verdin, M. E. Brown, M. Barlow, and A. Hoell, 2008: Warming of the Indian Ocean threatens eastern and southern African food security but could be mitigated by agricultural development. Proc. Natl. Acad. Sci. USA, 105, 11 08111 086, https://doi.org/10.1073/pnas.0708196105.

    • Search Google Scholar
    • Export Citation
  • Funk, C., and Coauthors, 2015: 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.

    • Search Google Scholar
    • Export Citation
  • Funk, C., and Coauthors, 2018: Examining the role of unusually warm Indo-Pacific sea-surface temperatures in recent African droughts. Quart. J. Roy. Meteor. Soc., 144, 360383, https://doi.org/10.1002/qj.3266.

    • Search Google Scholar
    • Export Citation
  • Gadgil, S., 2003: The Indian monsoon and its variability. Annu. Rev. Earth Planet. Sci., 31, 429467, https://doi.org/10.1146/annurev.earth.31.100901.141251.

    • Search Google Scholar
    • Export Citation
  • Galaal, M. H. I., 1992: Stars, Seasons Weather in Somali Pastoral Tradition. Centre for Linguistic and Historical Studies, 90 pp.

  • Hastenrath, S., 2000: Zonal circulations over the equatorial Indian Ocean. J. Climate, 13, 27462756, https://doi.org/10.1175/1520-0442(2000)013<2746:ZCOTEI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Heede, U. K., A. V. Fedorov, and N. J. Burls, 2021: A stronger versus weaker Walker: Understanding model differences in fast and slow tropical Pacific responses to global warming. Climate Dyn., 57, 25052522, https://doi.org/10.1007/s00382-021-05818-5.

    • Search Google Scholar
    • Export Citation
  • Herrmann, S. M., and K. I. Mohr, 2011: A continental-scale classification of rainfall seasonality regimes in Africa based on gridded precipitation and land surface temperature products. J. Appl. Meteor. Climatol., 50, 25042513, https://doi.org/10.1175/JAMC-D-11-024.1.

    • Search Google Scholar
    • Export Citation
  • Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 19992049, https://doi.org/10.1002/qj.3803.

    • Search Google Scholar
    • Export Citation
  • Huang, B., C. Liu, V. Banzon, E. Freeman, G. Graham, B. Hankins, T. Smith, and H.-M. Zhang, 2021: Improvements of the Daily Optimum Interpolation Sea Surface Temperature (DOISST) version 2.1. J. Climate, 34, 29232939, https://doi.org/10.1175/JCLI-D-20-0166.1.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., 2022: IMERG precipitation algorithm and the Global Precipitation Measurement (GPM) mission: Expert developer guidance. NCAR, https://climatedataguide.ucar.edu/climate-data/gpm-global-precipitation-measurement-mission.

  • Huffman, G. J., E. F. Stocker, D. T. Bolvin, E. J. Nelkin, and J. Tan, 2019: GPM IMERG final precipitation L3 1 day 0.1 degree × 0.1 degree, version 6 (GPM_3IMERGDF). Goddard Earth Sciences Data and Information Services Center, accessed 30 May 2023, https://doi.org/10.5067/GPM/IMERGDF/DAY/06.

  • Indeje, M., F. H. M. Semazzi, L. Xie, and L. J. Ogallo, 2001: Mechanistic model simulations of the East African climate using NCAR regional climate model: Influence of large-scale orography on the Turkana low-level jet. J. Climate, 14, 27102724, https://doi.org/10.1175/1520-0442(2001)014<2710:MMSOTE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Khairoutdinov, M., and D. Randall, 2006: High-resolution simulation of shallow-to-deep convection transition over land. J. Atmos. Sci., 63, 34213436, https://doi.org/10.1175/JAS3810.1.

    • Search Google Scholar
    • Export Citation
  • King, J. A., R. Washington, and S. Engelstaedter, 2021: Representation of the Indian Ocean Walker circulation in climate models and links to Kenyan rainfall. Int. J. Climatol., 41, E616E643, https://doi.org/10.1002/joc.6714.

    • Search Google Scholar
    • Export Citation
  • Kobayashi, S., and Coauthors, 2015: The JRA-55 Reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 548, https://doi.org/10.2151/jmsj.2015-001.

    • Search Google Scholar
    • Export Citation
  • Levin, N. E., E. J. Zipser, and T. E. Cerling, 2009: Isotopic composition of waters from Ethiopia and Kenya: Insights into moisture sources for eastern Africa. J. Geophys. Res., 114, D23306, https://doi.org/10.1029/2009JD012166.

    • Search Google Scholar
    • Export Citation
  • Liebmann, B., and Coauthors, 2017: Climatology and interannual variability of boreal spring wet season precipitation in the eastern Horn of Africa and implications for its recent decline. J. Climate, 30, 38673886, https://doi.org/10.1175/JCLI-D-16-0452.1.

    • Search Google Scholar
    • Export Citation
  • Limbu, P. T. S., and G. Tan, 2019: Relationship between the October–December rainfall in Tanzania and the Walker circulation cell over the Indian Ocean. Meteor. Z., 28, 453469, https://doi.org/10.1127/metz/2019/0939.

    • Search Google Scholar
    • Export Citation
  • Liu, W., K. H. Cook, and E. K. Vizy, 2020: Influence of Indian Ocean SST regionality on the East African short rains. Climate Dyn., 54, 49915011, https://doi.org/10.1007/s00382-020-05265-8.

    • Search Google Scholar
    • Export Citation
  • Lyon, B., 2014: Seasonal drought in the Greater Horn of Africa and its recent increase during the March–May long rains. J. Climate, 27, 79537975, https://doi.org/10.1175/JCLI-D-13-00459.1.

    • Search Google Scholar
    • Export Citation
  • Lyon, B., and D. G. DeWitt, 2012: A recent and abrupt decline in the East African long rains. Geophys. Res. Lett., 39, L02702, https://doi.org/10.1029/2011GL050337.

    • Search Google Scholar
    • Export Citation
  • Lyon, B., and N. Vigaud, 2017: Unraveling East Africa’s climate paradox. Climate Extremes: Patterns and Mechanisms, Geophys. Monogr., Vol. 226, Amer. Geophys. Union, 265–281, https://doi.org/10.1002/9781119068020.ch16.

  • McHugh, M. J., 2006: Impact of South Pacific circulation variability on East African rainfall. Int. J. Climatol., 26, 505521, https://doi.org/10.1002/joc.1257.

    • Search Google Scholar
    • Export Citation
  • Murray, F. W., 1967: On the computation of saturation vapor pressure. J. Appl. Meteor., 6, 203204, https://doi.org/10.1175/1520-0450(1967)006<0203:OTCOSV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Murtugudde, R., R. Seager, and P. Thoppil, 2007: Arabian Sea response to monsoon variations. Paleoceanography, 22, PA4217, https://doi.org/10.1029/2007PA001467.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S., 2016: The Turkana low-level jet: Mean climatology and association with regional aridity. Int. J. Climatol., 36, 25982614, https://doi.org/10.1002/joc.4515.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S., 2017: Climate and climatic variability of rainfall over eastern Africa. Rev. Geophys., 55, 590635, https://doi.org/10.1002/2016RG000544.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S., 2018: The ITCZ and the seasonal cycle over equatorial Africa. Bull. Amer. Meteor. Soc., 99, 337348, https://doi.org/10.1175/BAMS-D-16-0287.1.

    • Search Google Scholar
    • Export Citation
  • Ongoma, V., and H. Chen, 2017: Temporal and spatial variability of temperature and precipitation over East Africa from 1951 to 2010. Meteor. Atmos. Phys., 129, 131144, https://doi.org/10.1007/s00703-016-0462-0.

    • Search Google Scholar
    • Export Citation
  • Riddle, E. E., and K. H. Cook, 2008: Abrupt rainfall transitions over the Greater Horn of Africa: Observations and regional model simulations. J. Geophys. Res., 113, D15109, https://doi.org/10.1029/2007JD009202.

    • Search Google Scholar
    • Export Citation
  • Rowell, D. P., B. B. B. Booth, S. E. Nicholson, and P. Good, 2015: Reconciling past and future rainfall trends over East Africa. J. Climate, 28, 97689788, https://doi.org/10.1175/JCLI-D-15-0140.1.

    • Search Google Scholar
    • Export Citation
  • Schott, F. A., and J. P. McCreary Jr., 2001: The monsoon circulation of the Indian Ocean. Prog. Oceanogr., 51 (1), 1123, https://doi.org/10.1016/S0079-6611(01)00083-0.

    • Search Google Scholar
    • Export Citation
  • Schwarzwald, K., L. Goddard, R. Seager, M. Ting, and K. Marvel, 2023: Understanding CMIP6 biases in the representation of the Greater Horn of Africa long and short rains. Climate Dyn., 61, 12291255, https://doi.org/10.1007/s00382-022-06622-5.

    • Search Google Scholar
    • Export Citation
  • Seager, R., N. Henderson, and M. Cane, 2022: Persistent discrepancies between observed and modeled trends in the tropical Pacific Ocean. J. Climate, 35, 45714584, https://doi.org/10.1175/JCLI-D-21-0648.1.

    • Search Google Scholar
    • Export Citation
  • Sepulchre, P., G. Ramstein, F. Fluteau, M. Schuster, J.-J. Tiercelin, and M. Brunet, 2006: Tectonic uplift and eastern Africa aridification. Science, 313, 14191423, https://doi.org/10.1126/science.1129158.

    • Search Google Scholar
    • Export Citation
  • Slingo, J., H. Spencer, B. Hoskins, P. Berrisford, and E. Black, 2005: The meteorology of the western Indian Ocean, and the influence of the East African Highlands. Philos. Trans. Roy. Soc., 363A, 2542, https://doi.org/10.1098/rsta.2004.1473.

    • Search Google Scholar
    • Export Citation
  • Sobel, A. H., J. Nilsson, and L. M. Polvani, 2001: The weak temperature gradient approximation and balanced tropical moisture waves. J. Atmos. Sci., 58, 36503665, https://doi.org/10.1175/1520-0469(2001)058<3650:TWTGAA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sobel, A. H., I. M. Held, and C. S. Bretherton, 2002: The ENSO signal in tropical tropospheric temperature. J. Climate, 15, 27022706, https://doi.org/10.1175/1520-0442(2002)015<2702:TESITT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Steinkopf, J., and F. Engelbrecht, 2022: Verification of ERA5 and ERA-interim precipitation over Africa at intra-annual and interannual timescales. Atmos. Res., 280, 106427, https://doi.org/10.1016/j.atmosres.2022.106427.

    • Search Google Scholar
    • Export Citation
  • Tierney, J. E., S. C. Lewis, B. I. Cook, A. N. LeGrande, and G. A. Schmidt, 2011: Model, proxy and isotopic perspectives on the East African humid period. Earth Planet. Sci. Lett., 307, 103112, https://doi.org/10.1016/j.epsl.2011.04.038.

    • Search Google Scholar
    • Export Citation
  • Tierney, J. E., J. E. Smerdon, K. J. Anchukaitis, and R. Seager, 2013: Multidecadal variability in East African hydroclimate controlled by the Indian Ocean. Nature, 493, 389392, https://doi.org/10.1038/nature11785.

    • Search Google Scholar
    • Export Citation
  • Tierney, J. E., C. C. Ummenhofer, and P. B. deMenocal, 2015: Past and future rainfall in the Horn of Africa. Sci. Adv., 1, e1500682, https://doi.org/10.1126/sciadv.1500682.

    • Search Google Scholar
    • Export Citation
  • Trewartha, G. T., 1981: The Earth’s Problem Climates. University of Wisconsin Press, 371 pp.

  • Vashisht, A., and B. Zaitchik, 2022: Modulation of East African boreal fall rainfall: Combined effects of the Madden–Julian Oscillation (MJO) and El Niño–Southern Oscillation (ENSO). J. Climate, 35, 20192034, https://doi.org/10.1175/JCLI-D-21-0583.1.

    • Search Google Scholar
    • Export Citation
  • Vellinga, M., and S. F. Milton, 2018: Drivers of interannual variability of the East African “long rains.” Quart. J. Roy. Meteor. Soc., 144, 861876, https://doi.org/10.1002/qj.3263.

    • Search Google Scholar
    • Export Citation
  • Verschuren, D., K. R. Laird, and B. F. Cumming, 2000: Rainfall and drought in equatorial east Africa during the past 1,100 years. Nature, 403, 410414, https://doi.org/10.1038/35000179.

    • Search Google Scholar
    • Export Citation
  • Vizy, E. K., and K. H. Cook, 2020: Interannual variability of East African rainfall: Role of seasonal transitions of the low-level cross-equatorial flow. Climate Dyn., 54, 45634587, https://doi.org/10.1007/s00382-020-05244-z.

    • Search Google Scholar
    • Export Citation
  • Wainwright, C. M., J. H. Marsham, R. J. Keane, D. P. Rowell, D. L. Finney, E. Black, and R. P. Allan, 2019: ‘Eastern African Paradox’ rainfall decline due to shorter not less intense long rains. npj Climate Atmos. Sci., 2, 34, https://doi.org/10.1038/s41612-019-0091-7.

    • Search Google Scholar
    • Export Citation
  • Walker, D. P., J. H. Marsham, C. E. Birch, A. A. Scaife, and D. L. Finney, 2020: Common mechanism for interannual and decadal variability in the East African long rains. Geophys. Res. Lett., 47, e2020GL089182, https://doi.org/10.1029/2020GL089182.

    • Search Google Scholar
    • Export Citation
  • Wei, H.-H., and S. Bordoni, 2016: On the role of the African topography in the South Asian monsoon. J. Atmos. Sci., 73, 31973212, https://doi.org/10.1175/JAS-D-15-0182.1.

    • Search Google Scholar
    • Export Citation
  • Wenhaji Ndomeni, C., E. Cattani, A. Merino, and V. Levizzani, 2018: An observational study of the variability of East African rainfall with respect to sea surface temperature and soil moisture. Quart. J. Roy. Meteor. Soc., 144, 384404, https://doi.org/10.1002/qj.3255.

    • Search Google Scholar
    • Export Citation
  • Yang, W., R. Seager, M. A. Cane, and B. Lyon, 2014: The East African long rains in observations and models. J. Climate, 27, 71857202, https://doi.org/10.1175/JCLI-D-13-00447.1.

    • Search Google Scholar
    • Export Citation
  • Yang, W., R. Seager, M. A. Cane, and B. Lyon, 2015: The annual cycle of East African precipitation. J. Climate, 28, 23852404, https://doi.org/10.1175/JCLI-D-14-00484.1.

    • Search Google Scholar
    • Export Citation
  • Yulaeva, E., and J. M. Wallace, 1994: The signature of ENSO in global temperature and precipitation fields derived from the microwave sounding unit. J. Climate, 7, 17191736, https://doi.org/10.1175/1520-0442(1994)007<1719:TSOEIG>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zhao, S., and K. H. Cook, 2021: Influence of Walker circulations on East African rainfall. Climate Dyn., 56, 21272147, https://doi.org/10.1007/s00382-020-05579-7.

    • Search Google Scholar
    • Export Citation

Supplementary Materials

Save
  • Adler, R. F., and Coauthors, 2018: The Global Precipitation Climatology Project (GPCP) monthly analysis (new version 2.3) and a review of 2017 global precipitation. Atmosphere, 9, 138, https://doi.org/10.3390/atmos9040138.

    • Search Google Scholar
    • Export Citation
  • Ayehu, G. T., T. Tadesse, B. Gessesse, and T. Dinku, 2018: Validation of new satellite rainfall products over the upper Blue Nile basin, Ethiopia. Atmos. Meas. Tech., 11, 19211936, https://doi.org/10.5194/amt-11-1921-2018.

    • Search Google Scholar
    • Export Citation