Depuis 4 millions d'années (Ma), le climat de l'Afrique tropicale a oscillé entre périodes humides et sèches, de l'échelle de temps interannuelle à l'échelle du million d'années (Fig. 1). La revue présentée ici, qui s'appuie sur un bon nombre d'ouvrages de synthèse, se propose d'illustrer comment ces variations climatiques sont reliées à la variabilité du climat global.

During the last 4 millions of years (Ma), tropical African climate has oscillated between markedly wetter and drier conditions. These changes have occurred on all timescales (Fig. 1). In recent times, catastrophic rainfall events in equatorial East Africa, the drastic Sahel drought in the 1970–1980s, or rapid lake-level fluctuations have illustrated the large interannual-decadal rainfall variability (Fig. 1A) and its serious impact on societies. But changes of much larger magnitude and duration have occurred in the past. On the interdecadal–centennial timescale (Fig. 1B), well-dated records combining geological, documentary and archaeological data suggest strong links between floods/droughts and changes in civilisation [30] and [31]. From ~11.5–5.5 ka BP ago (ka BP means 10³ calendar years before present), East African Rift lakes extended tens or even hundreds of metres above their present-day level (Fig. 1C), while Neolithic civilisations flourished in a wet and green Sahara [15] and [16]. On timescales of 10⁴–10⁶ yr, geological and palaeobiological data document an alternation of wet and dry episodes (Fig. 1D) [29], and a step-like increase in aridity since ~3 Ma (Fig. 1E) [9], [10] and [13]. High-amplitude climate changes have induced biome shifts, species migrations, community reorganisation, and changes in water resources of considerable importance for human populations.

The abundant literature on climate changes in Africa cannot be included here. This paper refers to several synthesis works where the reader will find original references and detailed discussions. After a brief overview of modern climates, the paper aims to illustrate how Plio-Quaternary climate changes in Africa are linked to global climate variability, from long-term oscillations, to high-frequency fluctuations that took place at the timescale of human lives.

African climates depend primarily on low altitude pressure and winds over the continent. The northern and southern ends of the continent experience extra-tropical (Mediterranean) climates. They are both flanked by deserts (Sahara and Namib deserts), dominated by the subtropical anticyclones. A wide belt of tropical climates separates the two arid zones. A zone of maximum rainfall related to the Intertropical Convergence Zone (ITCZ) follows the overhead position of the sun. The seasonal ITCZ migration (Fig. 2) results in an equatorial zone of humid climates with two seasonal rainfall maximum, flanked on the north and south by broad belts of monsoonal climates with a single summer rainy season and winter drought. The West African monsoon is driven by the pressure gradient that develops during the boreal summer between the high-pressure system over the subtropical southern Atlantic Ocean and the heated land-surface of the Sahara-Sahel; it penetrates far inland over northeastern Africa. The East African monsoon system shows reversed sub-meridional flows originating from the subtropical high cells over the southern Indian Ocean and over Arabia during the northern and southern hemisphere summer, respectively.

The zonal climate pattern is altered by sea-surface conditions. For example, the cold Benguela, and the warm Algulhas currents flowing along the western and eastern coasts of southern Africa, respectively, contribute to the east–west climate asymmetry. Interannual rainfall fluctuations in many areas of the continent are statistically linked to the El Niño Southern Oscillation (ENSO), or with sea-surface temperature (SSTs) fluctuations in the Indian and Atlantic Oceans which often occur in the context of ENSO. The sign of the ENSO-related rainfall anomalies differ, however, with the regions [30].

Regional climates also depend on orography. For example, highland areas in East Africa generate marked climatic gradients over short distances. Mountains receive high rainfall, while the basins of the East African Rift lie in the rain shadow of their bordering escarpments and are semi-arid. This topography results from the geological history of the continent at timescales longer than 10⁶ yr. Tectonics and volcanism, still active in the East African Rift system, have induced significant changes in the regional water balances during the Late Neogene. The great East African lakes also play a role on regional climate through water recycling.

Well-dated terrestrial and marine African records document an overall increase in aridity since 3–2.8 Ma with step-like shifts at 2.8–2.4, 1.8–1.6, and 1.2–0.8 Ma. These shifts were synchronous with shifts in the onset and amplification of ice sheet growth and cooling at high latitudes, suggesting coupling between low- and high-latitude climates [13].

No continuous continental record covering the whole Plio-Pleistocene period is available yet. However, palaeovegetation, fossil mammals, and palaeolake records provide information on past African climates with high spatial resolution. East Africa is relatively well known because of its richness in fossil records, including hominids, and radiometric chronologies based on volcanic material bounding the sedimentary sequences. At the hominid-bearing site of Hadar (Middle Awash valley, southern Afar), pollen spectra and biome reconstructions indicate a cooling of ~5 °C and a 200–300 mm yr^–1 rainfall increase just before 3.3 Ma, consistent with an initial marine δ ¹⁸O shift reflecting an increase in global ice volume [5]. Pollen data from various regions of northeastern Africa show regional shifts to cooler and drier vegetation type after ca. 2.5 Myr [6]. A comprehensive examination of the fossil mammal fauna in the Omo basin documents a remarkable decrease in closed woodland and forest species and an increase in grassland species between 3.6 and 2.4 Ma, with a marked rise between 2.6 and 2.4 Ma [4] (Fig. 3A). Stable isotopic analysis of pedogenic carbonates from the Turkana and Olduvai basins indicate gradual replacement of woodland by open savannah grasslands between 3 and 1 Myr, with step-like increases in savannah vegetation near 1.8, 1.2 and 0.6 Ma (Fig. 3B) [9] and [10]. Tectonic basins have experienced prolonged intervals of continuous lacustrine sedimentation. In southern Afar, the Bodo and Hadar palaeolakes (ca. 4.0–2.95 Ma) have stored thick sedimentary series where clays, sands, diatomaceous clays and diatomites indicate alternating swamp, shallow saline to freshwater lake environments [13]. Sedimentary bedding cycles were about 20–25 kyr in duration, suggesting that environmental fluctuations are paced by the precessional orbital cycle (in [13]). In central Afar, lacustrine deposits sandwiched between volcanic flows are dated between >2.3 and 0.8 Ma (Fig. 3C) [14]. Large, deep lake environments are represented by 4- to 8-m-thick interbeds of pure, white diatomite characterised by the dominance of large freshwater planktonic species, e.g., Stephanodiscus niagarae now living in the North American great lakes. Similar facies and diatom assemblages were found in a 40-m-thick profile that accumulated in a lake at least 100-m deep in the Nakuru-Elmenteita basin, between 1 and ~0.95 Ma [3]. These typical facies and microflora, indicators of a precipitation–evaporation balance (P–E) much higher than today, disappeared forever from Africa by the Middle Pleistocene; further lacustrine phases bear diatom flora found in modern East African lakes and reflect higher evaporation rates [14].

Continuous marine sediment sequences off the western and eastern margins of subtropical Africa have provided evidence for an increase in aridity and greater climate variability in northern Africa after ~3–2.8 Ma through their pollen spectra, clay mineral assemblages, and aeolian dust concentration ([13] and references therein). For example, the study of aeolian dust export from West and East subtropical Africa – a good indicator of continental aridity and trade-wind strength – documents step-like shifts in the amplitude and periods of aeolian variability at ~2.8, 1.7, and 1.0 Ma [13]. It shows that the orbital precession cycle (23–19 ka) was the fundamental driver of African monsoon climate throughout the Late Neogene, but high-latitude glacial cycles of 41 ka after ~1.8–1.6 Ma, and of 100 ka after ~1.2–0.8 Ma were superimposed on the precession signal after 2.8 Ma (Fig. 3D–E) [13].

African climate is thus very sensitive to high-latitude glacial conditions and has been primarily paced by earth orbital variations over the past million years. This is in good agreement with general circulation model (GCM) experiments which show cooler, drier African conditions during the glacial maximum [8].

Two time intervals are selected to illustrate interactions between orbital forcing and regional ocean–atmosphere–biosphere dynamics acting on the monsoon system during glacial (the Last Glacial Maximum, LGM, ~23–19 ka BP), and interglacial (the Early–Middle Holocene, EMH, ~11.5–5 ka BP) periods.

Due to the geometry of orbital precession, changes in summer insolation are in antiphase between hemispheres (Fig. 4A) [2]. Increased summer insolation should result in reinforced monsoon circulation by enhancing the inter-hemispheric ocean–land pressure contrast that draws the monsoon winds inland [7] and [8]. Orbital forcing predicts dry climates in the northern tropics, and enhanced monsoon rainfall in the southern hemisphere during the LGM, and an opposite response during the EMH.

Palaeoclimatic data in the northern tropics roughly follows summer insolation fluctuations [15], but a rainfall deficit also occurred in the southern tropics during the LGM, indicating that the hydrological budget has not only been paced by orbital cycles in glacial times [1] and [15] (Fig. 4). For example, decreases in P–E are suggested by an increase in diatom-inferred conductivity in Lake Massoko (Fig. 4D) – a small crater lake without surface outlet – around 20 ka BP, and by the high percentages of littoral diatoms by ~23–16 ka BP in the large Lake Malawi. Aridity in tropical Africa under LGM boundary conditions is primarily attributed to lower tropical SSTs, due to increased northward oceanic heat transport out of the tropics. GCMs simulate a global hydrological cycle weaker than today and a decrease in summer precipitation in most of the tropics during the LGM [8] and [25].

One of the most striking features of the African climate history has been the wetting of the Sahara during the EMH [15] and [16]. Data indicate a northward migration of tropical rainfall belts as far as 20–24°N (Fig. 5). This suggests a northward migration of the summer ITCZ and of the upper level jets which are crucial in the development of rain-bearing disturbance and of summer rainfall [16] and [17]. All GCMs simulations show a strengthening and a northward penetration of the African monsoon front over northern Africa at 6 ka BP, in response to increase summer insolation over the northern hemisphere and an enhanced north–south temperature gradient over the Sahara that reinforced the ocean-land pressure gradient [7] and [8] (Fig. 5). However, atmospheric GCMs simulations underestimated the observed northward shift; vegetation feedbacks amplify and modify the monsoon system's response to orbital forcing; in coupled ocean–atmosphere models, ocean feedbacks enhance the African monsoon, lengthen the rainfall season, and shift the belt of maximum precipitation further north ([7] and references therein). The best consistency between observations and simulations is found with coupled ocean–atmosphere–vegetation models [7] and [17] (Fig. 5).

On timescales shorter than 1000 years, severe climatic disruptions are identified in many records [15], [16], [17], [18], [30] and [31], and correspond to widespread climate events that have different regional expressions.

For example, dry spells interrupt the generally wet EMH period at ~9–8, 7–6.6, and ~6–5 ka BP in the northern tropics (Fig. 1 and Fig. 4B–C). The ~9–8-ka BP event was rapidly followed by a return to ‘initial’ conditions. In the African and Indian monsoon domains, it is a short, but pronounced dry interval identified in several lake, pollen and speleothem records [15], [18] and [23]. Summer monsoons over the Tibet Plateau, the Arabian Sea and tropical Africa weaken dramatically. In the North Atlantic, this event is a significant short-lived cooling. This interval has been generally cool over much of the northern hemisphere [23]. Several hypotheses have been proposed to explain this worldwide rapid climate change: (i) the outbreak of the glacial Lake Agassiz in the Laurentide at 8.2 ka BP may have abruptly slowed down the oceanic thermohaline circulation, (ii) increased volcanic aerosols in the northern hemisphere may have caused a cooling and a weakening of the Afro-Asian monsoon circulation; (iii) small fluctuations in solar output have been invoked as a contributing mechanism to Holocene monsoon variability ([23] and references therein). Whatever its causes, the ~9–8-ka-BP event reflects linkages between low- and high-latitudes climates, and associates cool poles with dry tropics, as for the glacial–interglacial variations.

Conversely, the rapid offset of the EMH wet period at ~5.5 ka BP can be regarded as a step-over shift to a new millennial-duration background climate, from relatively stable wet to dry climate at the Middle–Late Holocene boundary. Late Holocene aridity is consistent with a decrease in summer insolation, but the abruptness of the climate response to gradual insolation forcing requires strongly non-linear feedback processes. GCMs studies have invoked vegetation and ocean temperature feedbacks to explain this abrupt switch [12].

Determining the exact nature and duration of short-lived events in tropical Africa remains, however, difficult because high-resolution records are scarce, and because of the regional responses to large-scale climate events. For example, a high-resolution δ ¹⁸O speleothem record from the Makapansgat Valley in South Africa [30] and a similar resolution lake-level record for Lake Naivasha in Kenya [31] (Fig. 1B) are in antiphase over much of the last thousand years [30]. Presently, ENSO-induced climate variability is inversely correlated between equatorial East Africa and subtropical southern Africa [30]. The anti-correlation between the two records suggests that climate in these two regions has been paced by multidecadal–centennial shifts in the ENSO mode. The resulting latitudinal gradient of change may have been a significant factor in promoting southward migration of people from equatorial Africa when environments in the north were deteriorating and those in the south were ameliorating [30].

The past million years represent a period of profound global climate shifts well expressed in the African climate. Long-term fluctuations have been primarily paced by orbital variations of the Earth and were closely linked to the onset and amplification of the high-latitude glacial cycles. Within a glacial or interglacial period, orbital forcing alone cannot account for the magnitude of observed climate changes. The response to orbital forcing is modulated by coupled vegetation–albedo and surface ocean temperature–moisture transport feedback. Higher frequency variations can be induced by solar output variability, or by internal variations in the climate system caused by major volcanic eruptions or changes in major atmospheric circulation modes. At all timescales, climate changes in the African tropics are related to widespread, if not global, disruptions in the Earth's climate system.