9.3 Assumptions about future trends
9.3.1 Climate-change scenarios
In this section, the limits of the regions are those defined by Ruosteenoja et al. (2003). Very few regional to sub-regional climate change scenarios using regional climate models or empirical downscaling have been constructed in Africa mainly due to restricted computational facilities and lack of human resources (Hudson and Jones, 2002; Swart et al., 2002) as well as problems of insufficient climate data (Jenkins et al., 2002). Under the medium-high emissions scenario (SRES A1B, see the Special Report on Emissions Scenarios: Nakićenović et al., 2000), used with 20 General Circulation Models (GCMs) for the period 2080-2099, annual mean surface air temperature is expected to increase between 3 and 4°C compared with the 1980-1999 period, with less warming in equatorial and coastal areas (Christensen et al., 2007). Other experiments (e.g., Ruosteenoja et al., 2003) indicate higher levels of warming with the A1FI emissions scenario and for the 2070-2099 period: up to 9°C for North Africa (Mediterranean coast) in June to August, and up to 7°C for southern Africa in September to November. Regional Climate Model (RCM) experiments generally give smaller temperature increases (Kamga et al., 2005). For southern Africa (from the equator to 45°S and from 5° to 55°E, which includes parts of the surrounding oceans), Hudson and Jones (2002), using the HadRM3H RCM with the A2 emissions scenario, found for the 2080s a 3.7°C increase in summer (December to February) mean surface air temperature and a 4°C increase in winter (June to August). As demonstrated by Bounoua et al. (2000), an increase in vegetation density, leading to a cooling of 0.8°C/yr in the tropics, including Africa, could partially compensate for greenhouse warming, but the reverse effect is simulated in the case of land cover conversion, which will probably increase in the next 50 years (DeFries et al., 2002). A stabilisation of the atmospheric CO2 concentration at 550 ppm (by 2150) or 750 ppm (by 2250) could also delay the expected greenhouse gas-induced warming by 100 and 40 years, respectively, across Africa (Arnell et al., 2002). For the same stabilisation levels in the Sahel (10°-20°N, 20°W-40°E) the expected annual mean air temperature in 2071-2100 (5°C) will be reduced, respectively, by 58% (2.1°C) and 42% (2.9°C) (Mitchell et al., 2000; Christensen et al., 2007).
Precipitation projections are generally less consistent with large inter-model ranges for seasonal mean rainfall responses. These inconsistencies are explained partly by the inability of GCMs to reproduce the mechanisms responsible for precipitation including, for example, the hydrological cycle (Lebel et al., 2000), or to account for orography (Hudson and Jones, 2002). They are also explained partly by model limitations in simulating the different teleconnections and feedback mechanisms which are responsible for rainfall variability in Africa. Other factors that complicate African climatology include dust aerosol concentrations and sea-surface temperature anomalies, which are particularly important in the Sahel region (Hulme et al., 2001; Prospero and Lamb, 2003) and southern Africa (Reason, 2002), deforestation in the equatorial region (Semazzi and Song, 2001; Bounoua et al., 2002), and soil moisture in southern Africa (New et al., 2006). These uncertainties make it difficult to provide any precise estimation of future runoff, especially in arid and semi-arid regions where slight changes in precipitation can result in dramatic changes in the runoff process (Fekete et al., 2004). Nonetheless, estimations of projected future rainfall have been undertaken.
With the SRES A1B emissions scenario and for 2080-2099, mean annual rainfall is very likely to decrease along the Mediterranean coast (by 20%), extending into the northern Sahara and along the west coast to 15°N, but is likely increase in tropical and eastern Africa (around +7%), while austral winter (June to August) rainfall will very probably decrease in much of southern Africa, especially in the extreme west (up to 40%) (Christensen et al., 2007). In southern Africa, the largest changes in rainfall occur during the austral winter, with a 30% decrease under the A2 scenario, even though there is very little rain during this season (Hudson and Jones, 2002). There are, however, differences between the equatorial regions (north of 10°S and east of 20°E), which show an increase in summer (December to February) rainfall, and those located south of 10°S, which show a decrease in rainfall associated with a decrease in the number of rain days and in the average intensity of rainfall. Recent downscaling experiments for South Africa indicate increased summer rainfall over the convective region of the central and eastern plateau and the Drakensberg Mountains (Hewitson and Crane, 2006). Using RCMs, Tadross et al. (2005b), found a decrease in early summer (October to December) rainfall and an increase in late summer (January to March) rainfall over the eastern parts of southern Africa.
For the western Sahel (10 to 18°N, 17.5°W to 20°E), there are still discrepancies between the models: some projecting a significant drying (e.g., Hulme et al., 2001; Jenkins et al., 2005) and others simulating a progressive wetting with an expansion of vegetation into the Sahara (Brovkin, 2002; Maynard et al., 2002; Claussen et al., 2003; Wang et al., 2004; Haarsma et al., 2005; Kamga et al., 2005; Hoerling et al., 2006). Land-use changes and degradation, which are not simulated by some models, could induce drier conditions (Huntingford et al., 2005; Kamga et al., 2005). The behaviour of easterly jets and squall lines is also critical for predicting the impacts of climate change on the sub-region, given the potential links between such phenomena and the development of the rainy season (Jenkins et al., 2002; Nicholson and Grist, 2003).
Finally, there is still limited information available on extreme events (Christensen et al., 2007), despite frequent reporting of such events, including their impacts (see Section 9.2.1). A recent study using four GCMs for the Sahel region (3.75 to 21.25°N, 16.88°W to 35.63°E) showed that the number of extremely dry and wet years will increase during the present century (Huntingford et al., 2005). Modelling of global drought projections for the 21st century, based on the SRES A2 emissions scenario, shows drying for northern Africa that appears consistent with the rainfall scenarios outlined above, and wetting over central Africa (Burke et al., 2006). On a global basis, droughts were also estimated to be slightly more frequent and of much longer duration by the second half of the 21st century relative to the present day. Other experiments indicate that in a warmer world, and by the end of the century (2080-2100), there could also be more frequent and intense tropical storms in the southern Indian Ocean (e.g., McDonald et al., 2005). Tropical cyclones are likely to originate over the Seychelles from October to June due to the southward displacement of the Near Equatorial Trough (Christensen et al., 2007). There could very probably be an increase of between 10 and 20% in cyclone intensity with a 2-4°C SST rise (e.g., Lal, 2001), but this observation is further complicated by the fact that SST does not account for all the changes in tropical storms (McDonald et al., 2005).