1. Introduction
A large majority of climate models project declines in rainfall over subtropical southern Africa in the twenty-first century in response to rising global temperatures (Christensen et al. 2007; Shongwe et al. 2009; Collins et al. 2013; He and Soden 2017). Robust signals emerge in models after just 2° of global warming (James and Washington 2013), and in some models the drying is appreciable by the mid-twenty-first century. The signal-to-noise ratio of the drying is particularly high in the early summer season from October to December (OND), which contributes over 40% of the annual rainfall across large parts of subtropical southern Africa (15°–30°S, 15°–40°E) including in Botswana, Zimbabwe, and South Africa. In this season models from phase 5 of the Coupled Model Intercomparison Project (CMIP5) project an average 42 mm season−1 decrease in rainfall over subtropical southern Africa by the end of the century under high emissions forcing (RCP8.5) (Table 1). Predicated in part on the confidence implied by the intermodel consensus in OND rainfall declines, climate impact studies have outlined the potentially negative consequences of climate change in the region, including its effects on agriculture, water supply, energy production, human health, and biodiversity (e.g., Schlenker and Lobell 2010; Midgley and Thuiller 2011; Gosling and Arnell 2016; Conway et al. 2015; Serdeczny et al. 2017).
CMIP5 models used in this study, their future OND absolute and percentage rainfall change over subtropical southern Africa (15°–30°S, 15°–40°E), and their precipitation bias relative to the CMAP dataset. Models are ordered by absolute precipitation change (ΔP) under RCP8.5 (2073–99 minus 1979–2005); models in bold are those included in the low-magnitude (LM) and high-magnitude (HM) change composites. Asterisks (*) indicate the 10 models with highest present-day bias; plus signs (+) indicate models with lowest present-day bias.
However, agreement between models in the sign of a climate projection is not necessarily a strong argument for confidence in that projection (Parker 2011). This is especially true under conditions where the magnitude of projected change between models varies widely (Parker 2011) and where models contain systematic biases in their estimation of present-day climate (Knutti et al. 2010). Both of these conditions are met in the case of projections of subtropical southern African drying. First, despite strong consensus among ~95% of models that OND rainfall will decline in the future, there is an order of magnitude difference in the scale of decline, from less than 10 mm per season to nearly 100 mm (Fig. 1). Second, the large intermodel spread in future projections occurs in the presence of systematic overestimations of present-day rainfall over subtropical southern Africa (Mueller and Seneviratne 2014; Dieppois et al. 2015; Lazenby et al. 2016; Munday and Washington 2017). In ~80% of models in OND, the magnitude of their present-day bias exceeds the future change, by a median factor of 2.5 (Fig. 1). The large intermodel variation in the projections and present-day climatology of rainfall suggests that the foundation on which the many impact studies rest is not secure.
CMIP5 future rainfall change (mm season−1) over subtropical southern Africa (15°–30°S, 15°–40°E) (black bars) for OND (2073–99 minus 1979–2005) and the present-day bias (red bars) in mm season−1 relative to the CMAP dataset. There is a significant negative correlation (r = −0.48, p < 0.01) between the bias and the future change: models with a larger bias tend to simulate greater rainfall declines.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
One way to build confidence in projections is to assess the mechanism of drying among models. Over southern Africa, several studies have done this for both regional and global climate models, providing a useful starting point for assessing confidence in the projections (e.g., Shongwe et al. 2009; Engelbrecht et al. 2013; Cook and Vizy 2013; Lazenby et al. 2018; Pinto et al. 2018). In a regional model driven by the CMIP3 ensemble mean to the mid-twenty-first century, for example, Cook and Vizy (2013) suggest that the rainfall decline in early summer occurs as the south Indian Ocean convergence zone (SIOCZ) shifts northeastward in association with a strengthened thermal low and weakened subtropical high. Meanwhile, Pinto et al. (2018) show that annual mean drying trends among four CMIP5 models are associated with an increase in frequency of high pressure systems across southern Africa at the expense of continental and midlatitude lows. These studies are useful in that they relate projections of rainfall change to specific climate processes, the realism of which could be evaluated in comparison with observations (e.g., James et al. 2015) or which could be tested with simpler, idealized models (Held 2005). If the processes are judged to be plausible and are consistent across a range of climate models, this would add evidence to support the reliability of the drying trend. As a contribution to this effort, the purpose of this paper is to understand in more detail the physical processes associated with the spread in early summer drying projections among 30 of the CMIP5 models, and to consider these processes in light of present-day model biases. More specifically, we aim to do the following:
Investigate the physical processes associated with OND drying over subtropical southern Africa across CMIP5 models.
Generate hypotheses for the controls on the spread in future projections of drying across CMIP5, which could be tested in idealized simulations.
Consider whether there are links between biases in the simulation of contemporary climate over subtropical southern Africa and the magnitude of projected drying.
2. Background
In this section, we briefly describe the observed climate and circulation over southern Africa during the early summer (OND), before discussing potential mechanisms of future rainfall change.
a. The southern African climate system in the early summer
The early austral summer (OND) is the time of rainfall onset over much of southern Africa apart from the southwestern Cape region (Tadross et al. 2005; Dunning et al. 2016). Onset in northern parts of southern Africa (~10°–20°S) is concomitant with the southward displacement convection over Africa, which lags the seasonal cycle in insolation by roughly one month. During this time, the low-level circulation, shown in Fig. 2, is dominated by the presence of a shallow thermal low situated over southern Angola, northern Namibia, and eastern Zambia (Mulenga 1998; Munday and Washington 2017; Howard and Washington 2018). The Angola thermal low is likely driven by high surface temperatures and can be distinguished from the (Angola) tropical lows that form later on in the summer season by the low specific humidity (Howard and Washington 2018). The thermal low, together with the subtropical highs in the adjacent south Indian and South Atlantic Oceans, sets up a strong zonal pressure gradient that is important for easterly and westerly moisture fluxes into the subcontinent. During OND, much of the subtropical portion of southern Africa is under the influence of a mid- to upper-level high pressure system: the Botswana high. The Botswana high is strongest in OND, and is likely to suppress rainfall (Matarira 1990; Unganai and Mason 2002; Reason 2016; Driver and Reason 2017) in spite of the strong moisture convergence at low levels associated with the thermal low.
Important circulation features over southern Africa in the early austral summer (OND). Topography (from the USGS GTOPO30 global digital elevation model dataset) above 1000 m is shaded; contour interval is 100 m. Features of this figure are adapted from a schematic by Neil Hart/Creative Commons.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
A second rainfall maximum in the OND season occurs to the southeast of the core of the Botswana high and extends over parts of South Africa, Botswana, and Zimbabwe. This rainfall maximum is influenced by the enhanced specific humidity associated with high evaporation from the warm Agulhas Current and by orographic triggering via the sharp topography of the eastern southern African escarpment (Rouault et al. 2002; Blamey et al. 2017). Over large parts of this subtropical region, a significant proportion, 30%–60% of the total rainfall, is associated with tropical temperature cloud bands (TTCBs) (Harrison 1984; Hart et al. 2013), which form through the interaction of tropical disturbances with westerly waves passing to the south of the continent (Harrison 1984; Washington and Todd 1999). TTCBs are particularly important in November and December (Hart et al. 2013), and can occur with or separately from mesoscale convective complexes (MCCs), which themselves account for up to 20% of the rainfall in southeastern southern Africa (Blamey and Reason 2013).
b. Potential mechanisms of future change
A number of mechanisms have been proposed to explain rainfall change in the global tropics. At a zonal mean scale, increases in water vapor content of the atmosphere due to warming may amplify existing patterns of moisture convergence [precipitation minus evaporation (P − E)], resulting in increases in rainfall in areas of climatological deep convection (the wet-get-wetter mechanism) and decreases in precipitation in climatologically dry regions (Chou and Neelin 2004; Held and Soden 2006; Chou et al. 2009; Seager et al. 2010). Although this thermodynamic increase or decrease in rainfall is likely to be offset by a (dynamical) weakening of overturning circulations (Held and Soden 2006; Vecchi et al. 2006; Ma et al. 2012; Ma and Xie 2013; Chadwick et al. 2013), it helps to explain the observed large-scale increases in P − E (E − P) in regions of climatological ascent (descent) over recent decades (Allan et al. 2010).
The wetting of regions with high climatological precipitation by enhanced moisture advection could lead to a dynamically driven rainfall decreases at convective margins (Neelin et al. 2003; Chou and Neelin 2004). In subtropical regions, such as southern Africa, increases in low-level moist static energy may be insufficient to keep pace with the near-uniform increases in dry static energy in the upper tropical troposphere due to warming of the oceans (Neelin et al. 2003). Giannini (2010) applies this moist static energy perspective to explain rainfall projections in the Sahel. She finds that ocean warming–induced increases in upper-level moist static energy affect stability from the top down and are related to rainfall declines in some CMIP3 models. This is consistent with simulations of Sahelian rainfall variability in the twentieth century, which is affected by periodic warming of the tropics during El Niño events (Held et al. 2005; Giannini 2010).
Seth et al. (2011, 2013) extend the analysis of Giannini (2010) to explain future early summer drying trends in the ensemble mean CMIP3 and CMIP5 models across a number of regions, including over subtropical southern Africa. For southern Africa, they show that upper-tropospheric warming during the early summer presents an enhanced convective barrier, which is not overcome by increases in moisture supply after a warmer and drier dry season. The enhanced stability is associated with decreased evaporation and increased moisture divergence, which combine to exacerbate the rainfall decline (Seth et al. 2013).
In addition to spatially uniform oceanic warming, which is assumed by the mechanisms discussed above, patterns of sea surface temperature (SST) change are also likely to influence future oceanic rainfall change (Xie et al. 2010; Huang et al. 2013; Ma and Xie 2013) and, through atmospheric teleconnections, could be critical for model diversity in terrestrial rainfall change in the tropics (Biasutti and Sobel 2009; Park et al. 2015; Brown et al. 2016; Rowell and Chadwick 2018). In the Sahel, for example, Park et al. (2015) show evidence that the enhanced warming of the Northern Hemisphere extratropics relative to the tropics is linked to the spread in rainfall projections in CMIP5, while Biasutti and Sobel (2009) suggest that the delay in the seasonal cycle of SSTs contributes to early summer Sahelian rainfall declines in CMIP3 models. Moreover, for the East African “short rains” (OND), Rowell and Chadwick (2018) show that the pattern of SST change in the tropics is a major driver of uncertainty in rainfall projections in CMIP5.
The pattern of sea surface temperature change in tropical oceans is also likely to be important for subtropical southern African drying projections. Present-day variability in the pattern of SSTs, including El Niño–Southern Oscillation (ENSO; Lindesay 1988; Nicholson and Kim 1997; Richard et al. 2001; Hart et al. 2018), the subtropical Indian Ocean dipole (SIOD; Fauchereau et al. 2003; Hermes and Reason 2005; Hoell et al. 2017), and Benguela Niño events in the South Atlantic (Hirst and Hastenrath 1983; Rouault et al. 2003), is associated with rainfall variability during the southern African summer. Dry conditions in southern Africa tend to prevail during negative SIOD events when the eastern tropical Indian Ocean basin is anomalously warm, and when the southwest Indian Ocean near Madagascar is relative cool. Similarly, El Niño events, when the central Pacific is anomalously warm, tend to lead to below average rainfall over southern Africa.
By investigating the future change signal over southern Africa in the CMIP5 ensemble mean model, Lazenby et al. (2018) find a northeastward shift of the SIOCZ toward East Africa and the western Indian Ocean. In decomposing the future rainfall change into dynamic and thermodynamic components following Chadwick et al. (2013), they demonstrate that the precipitation decline in OND is a consequence of both dynamic change in the location of convection and thermodynamic changes induced by decreasing relative humidity. While declines in ensemble mean rainfall are driven equally by both thermodynamic and dynamic terms, intermodel uncertainty in the rainfall decline appears to be driven mostly by the dynamical term. The authors go on to argue that uncertainty in the future evolution of dynamically driven precipitation changes may be related intermodel differences in the pattern of SST changes in the Indian Ocean, particularly related to the uncertainty in relatively enhanced warming of the eastern tropical Indian Ocean relatively low warming of the southwest subtropical Indian Ocean—similar to the SIOD anomalies.
In what follows we diagnose changes to the subtropical southern African climate system associated with early summer drying (section 4) and trace how they might arise through differences in the extent and pattern of warming of tropical oceans among models (section 5). In section 6 we comment on these results in light of present-day biases before concluding in section 7.
3. Data
We analyze future change projections in 30 models forming part of the CMIP5 dataset (Table 1; Taylor et al. 2012). Future change for each model is calculated by subtracting their present-day climatology (1979–2005) in the historical experiment (ensemble r1i1p1) from their future climatology (2073–99). The future change experiments in CMIP5 use the representative concentration pathways (RCPs) developed by van Vuuren et al. (2011). To maximize the signal-to-noise ratio for examining mechanisms of change, we use the most extreme of these pathways (RCP8.5) in which the global net radiative forcing by 2100 is 8.5 W m−2. The RCP8.5 experiment is based on the previous Coupled Model Intercomparison Project (CMIP3) SRES A2 scenario, which assumes a “business as usual” approach to climate change mitigation. This is the emissions pathway which we are currently following. All model data are remapped to a common 2° × 2° grid.
We use composites of six models that project high-magnitude (HM) drying and six models that project low-magnitude (LM) drying (OND). The models used for composites are marked in Table 1, along with their future OND rainfall change and present-day rainfall bias. The HM and LM composites and our analysis are based on the absolute rainfall change, although percentage changes (shown in Table 1) are also considered. We note that models with high (low) magnitude absolute future rainfall also tend to simulate high (low) magnitude percentage change, with a correlation of r = 0.91 (p < 0.001) between the two.
In section 6 we discuss potential links between model projections and their present-day biases in precipitation, and circulation. Our comparison dataset for precipitation is the Climate Prediction Center Merged Analysis of Precipitation (CMAP; Xie and Arkin 1997), a satellite/rain gauge product. The rainfall climatology of the CMAP dataset is in good agreement with other satellite- and rain gauge–based datasets over southern Africa (Novella and Thiaw 2013) and Africa as a whole (Maidment et al. 2014). In an intercomparison of eight different rainfall datasets over Africa, for example, CMAP agrees to within 0.2 mm day−1 with seven of those datasets for mean rainfall both in September–November (SON) and annually (Maidment et al. 2014), although there are relatively few rain gauges with which to confront the satellite datasets in tropical regions of Africa (Washington et al. 2006).
Estimates for the present-day circulation are based on the MERRA-2 reanalysis product (Gelaro et al. 2017), which has been shown to perform well over Africa (Reichle et al. 2017; Hua et al. 2016). The circulation in MERRA-2 over southern Africa is qualitatively similar to both ERA-Interim and NCEP2 reanalysis (not shown), although none of the reanalyses are well constrained by in situ data over a majority of southern Africa. CMAP rainfall and the MERRA-2 dataset are remapped to a 2° × 2° grid for comparison with CMIP5 models. The present-day climatology for CMAP is calculated over 1979–2005, and over 1980–2005 in MERRA-2.
4. Local diagnosis of southern African rainfall projections
In this section we examine changes to the vertical structure of the atmosphere, moisture circulation, and water balance over subtropical southern Africa.
a. Changes to the vertical atmospheric structure
If the remote effect of SST warming is important for determining the magnitude of rainfall decline we would expect to see larger increases in upper tropospheric MSE over subtropical subtropical southern Africa in the HM compared to LM models. Figure 3 shows that this is the case. In a longitude–height cross section averaged over 15°–30°S HM models simulate ~25%–30% larger increases in MSE between 300 and 200 hPa compared to LM models.
Longitude–height (in hPa) cross section (averaged between 15° and 30°S) of future change in moist static energy (kJ kg−1) in (left) high-magnitude and (right) low-magnitude drying sets of models.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
The increased upper-level MSE in the HM models is accompanied by large MSE anomalies (10 kJ kg−1) near the surface centered on 28°E. The increase in surface MSE is mainly associated with an increase in the dry static energy term (not shown), and lies to the east of the location of the climatological Angola heat low. This suggests a future deepening and broadening of the thermal low, in accordance with results from previous studies (Cook and Vizy 2013; Lazenby et al. 2018). The strengthened heat low sets up a steeper low-level MSE gradient between the subtropical southern African landmass and the adjacent Indian Ocean. The pattern of change is qualitatively similar in the LM models, although less intense.
To highlight the effect of the altered MSE profile on vertical motion, we examine the cross sections of vertical velocity (the pressure tendency) averaged over the same latitudes (Fig. 4). Consistent with the large increase in upper-level MSE, there is anomalous upper-level subsidence, which reaches a maximum between 400 and 300 hPa in the HM set of models. Over the continent, the anomalous subsidence overlies an increase in shallow convection near the surface. The preferential increase in shallow convection over deep convection supports the view that the continental thermal low is deepening in response to warming. The main differences between the LM and HM models follow this structure, with HM models simulating 70% larger increases in upper-level subsidence than LM models, and greater low-level uplift associated with a relatively strengthened thermal low.
Longitude–height (in hPa) cross section (averaged between 15° and 30°S) of future change in vertical velocity (102 Pa s−1) (shading) in (a) high-magnitude and (b) low-magnitude drying models, and (c) their difference. The contemporary climatology (1979–2005) in HM and LM models is shown by gray contours.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
Comparison between the present-day climatology of vertical velocity and future change raises two further points worthy of note (Fig. 4). First, in both sets of models there is an eastward shift in the location of the low-level uplift from ~18° to ~20°E, indicating an eastward shift or broadening of the thermal low (Fig. 4). Second, the contemporary climatology of vertical uplift in the HM models features strong upward motion as high as 300 hPa. Strong present-day uplift in LM models, by contrast, is mainly confined to below 600 hPa, and is consistent with the presence of the shallow (Angola) thermal low in early summer (section 2a; Adebiyi and Zuidema 2016; Munday and Washington 2017; Howard and Washington 2018). That the larger future response in vertical velocity occurs in the HM models is the first indication that the present-day bias might influence the magnitude of future drying. We consider this in more detail in section 6.
The increase in strength of the thermal low, together with the enhanced upper-level subsidence, is likely to enhance the strength of the Botswana midlevel high. Strengthening of the Botswana high in some years is associated with drier than average conditions over subtropical southern Africa (Reason 2016; Driver and Reason 2017), consistent with the future projections. Future investigation that considers in more detail changes to the strength or structure of the high, while beyond the scope of this paper, could be a useful and complementary approach to understanding the drying signal.
b. Changes in moisture circulation
The differences between HM and LM models in Figs. 3 and 4 might suggest that the enhanced stability is being set from the top down, rather than from changes in the low-level circulation. This view of precipitation declines is consistent with the mechanism set out by Giannini (2010) and applied to southern Africa by Seth et al. (2013), but could be complicated by changes in the low-level moisture circulation between models. For example, HM models may simulate reduced moisture supply into subtropical southern Africa in the future compared to LM models, thus influencing atmospheric stability from the surface up.
In both sets of models, future anomalies in 850-hPa moisture flux and convergence (Fig. 5, middle row) are superimposed on their existing pattern of convergence (Fig. 5, top row). Across both model sets, moisture convergence increases (2–5 × 108 kg kg−1 s−1) in the central subcontinent while moisture divergence increases (1–4 × 108 kg kg−1 s−1) across the eastern coast in Tanzania, Mozambique, and eastern Zimbabwe. The intensification of the existing pattern of convergence/divergence is associated with greater moisture flux from the Indian Ocean and the equatorial Congo basin region. This is likely associated with the strengthened Angola thermal low, which is evident in both model sets (Figs. 3 and 4).
Low-level (850 hPa) moisture divergence (shading; 108 kg kg−1 s−1) and moisture flux (vectors; barb in lower-right corner of plots shows scale). (top) Present-day climatology (1979–2005) in (a) high-magnitude drying models and (b) low-magnitude drying models. (middle) Future change for (c) high-magnitude drying models and (d) low-magnitude drying models. (e) The difference in future change between high- and low-magnitude drying models.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
If intermodel variation in projections of moisture circulation influences the intermodel spread in precipitation change, we would expect to see less moisture convergence in the continental interior in the HM models compared to the LM models given the increased drying evident in the HM models. This is not evident in Fig. 5. Moisture convergence over the continental interior increases more (by 2–4 × 108 kg kg−1 s−1) in the HM compared to LM models (Fig. 5, bottom row). The larger increases in moisture convergence are consistent with the greater intensification of the Angola heat low in HM compared to LM models. This emphasizes the primacy of a top-down mechanism of stabilization in controlling differences between model projections, as the additional moisture supply in HM compared to LM models is not sufficient to mitigate the rainfall decline. We do note, however, that larger increases in low-level moisture divergence in HM compared to LM models are likely to play an important role in the rainfall declines in eastern parts of southern Africa and especially in Mozambique and eastern Zimbabwe.
c. The relationship among projections of rainfall, surface temperature, and evaporation
Figures 3–5 present evidence suggesting that the intermodel spread in projections of early summer drying is controlled primarily by the degree of increase in upper-level subsidence. Since increased stability is associated with clearer skies and enhanced net surface solar radiation (Giannini 2010), one way of diagnosing this relationship at a regional scale is to examine the relationship between surface temperature change and rainfall change over subtropical southern Africa among CMIP5 models.
After normalizing OND temperature change by each model’s tropical mean change to account for different climate sensitivities, Fig. 6a shows how projections of surface temperature over subtropical southern Africa are closely associated with projections of rainfall in CMIP5 models. Models projecting high-magnitude drying also project larger surface temperature amplification over subtropical southern Africa (r = −0.60, p < 0.001). This could indicate that the uncertainty in the precipitation change is driving uncertainty in the local temperature change through cloud feedbacks, although causation cannot be demonstrated in this context. A similar intermodel association between rainfall and temperature change has been noted for future changes to the Australian monsoon in CMIP5 models (Brown et al. 2016).
For OND, the intermodel relationship between future rainfall change over subtropical southern Africa (15°–30°S, 15°–40°E) and (a) future surface temperature change (K) over subtropical southern Africa relative to the tropical mean temperature change (20°N–20°S, 180°W–180°E) and (b) future latent heat flux (LHF) change (W m−2) over subtropical southern Africa. The blue (red) dots represent models that are part of the LM (HM) composites.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
Subtropical southern Africa is a semiarid region and evapotranspiration (latent heat flux) tends to be precipitation limited (Mueller and Seneviratne 2014). This implies that as precipitation declines so should evaporation, notwithstanding the increases in moisture supply (Fig. 5). We show this relationship in Fig. 6b: there is a strong positive correlation between decreasing precipitation and decreasing evapotranspiration among models (r = 0.85, p < 0.001). In turn the reduced evapotranspiration increases the ratio of sensible to latent heat fluxes (the Bowen ratio; not shown), which could further contribute to the amplification of temperature change over subtropical southern Africa. The relationships among future projections of precipitation, temperature, and evaporation imply that uncertainty in projections of temperature and rainfall should be considered together.
5. Connections with tropics-wide projections
The enhanced atmospheric stability over subtropical southern Africa could be connected to changes in tropics-wide circulation and warming through the mechanisms outlined in the background section. Here, we investigate how the intermodel diversity in rainfall decline may be related to changes in tropical rainfall and patterns of sea surface temperature change.
a. Spatial changes in southern African and tropical rainfall
HM models project future reductions in rainfall across much of subtropical southern Africa, with the largest decreases, of up to 170 mm season−1, occurring along an axis orientated from northwest to southeast, similar to the climatological location of the SIOCZ (Fig. 7). In LM models, by contrast, there is little change in rainfall across much of subtropical southern Africa, with only small decreases in rainfall over Angola and Zambia (<40 mm season−1). As we noted in the introduction, the difference between these projections would require substantially different approaches to climate change adaptation. Interestingly, models that project high-magnitude drying over southern Africa also do so over South America and parts of northwest Australia. This could indicate the presence of a common mechanism responsible for drying across Southern Hemisphere land regions during the early summer.
For OND, future tropical rainfall change in mm season−1 (shading) averaged over the set of (a) HM models and (b) LM models, (c) is the difference in future rainfall change between HM and LM models. Black contours in (a) and (b) show the present-day rainfall climatology in HM and LM models respectively; contours start at 180 mm season−1 and the contour interval is 90 mm season−1. The stippling in (c) is a metric of model agreement; stippling indicates that all models in the HM composite simulate lower- or higher-magnitude changes relative to all models in the LM composite (e.g., over subtropical southern Africa, all models in HM composite project greater drying than those in LM composite).
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
In both sets of models, the rainfall decline over subtropical southern Africa is accompanied by large increases in rainfall in other parts of the tropics, especially to the north of the equator. In the Indian Ocean the largest increases in rainfall in both sets of models (120–160 mm season−1) occur to the northwest of the climatological maximum precipitation. In the Atlantic Ocean, there is an eastward shift in location of convection toward the Gulf of Guinea (Bakassi Peninsula), while in the Pacific Ocean there are broad increases in rainfall which maximize in the northern tropics, especially in the HM models.
The difference in rainfall projections between HM and LM models shows evidence of a north–south tropical contrast (Fig. 7, bottom). HM models simulate greater increases in precipitation over large parts of the northern tropics compared to LM models while generally projecting greater declines in subtropical regions of the Southern Hemisphere. South of the equator in the Pacific Ocean, for example, HM models project future decreases in precipitation in areas of high climatological rainfall, whereas LM models project modest increases in precipitation. Consistent with Chadwick et al. (2013) and others, this indicates that future changes in tropical rainfall cannot be understood simply in terms of a wet-get-wetter response to warming. The consistency of response between models (indicated by the stipples in Fig. 7) within HM and LM groups in areas remote from southern Africa provides some indication that rainfall declines over southern Africa could be a consequence of the large-scale adjustment of the tropical atmosphere to warming.
To link together the increases in tropical rainfall and projections of rainfall declines over subtropical southern Africa, we consider changes to Hadley circulation via examination of the vertical structure of zonal mean meridional winds in HM and LM models (Fig. 8). A strengthening of Hadley circulation, indicated by an intensification and rise in the level of meridional outflow in the upper atmosphere, has previously been shown to relate to subtropical drying in the ensemble-mean CMIP5 model (Lau and Kim 2015). Consistent with this work, we find that both model sets simulate enhanced outflow from tropical convection at upper levels of the atmosphere (200–100 hPa) and, to a lesser extent, reduced outflow from the convective region at ~300 hPa. The future anomaly in upper-level outflow to the south (between 0° and –20°S) is shifted northward and higher up in the atmosphere compared to the climatological outflow (shown in the contours). This is consistent with the intensification of rainfall north of the equator seen in Fig. 7 and with general expectation for a rising tropopause under global warming (O’Gorman and Singh 2013). The difference between HM and LM models is in the strength of this pattern. In particular, the future anomaly in 200–100-hPa southward outflow is 1 m s−1 greater in the HM models, consistent with the enhanced subsidence over subtropical southern Africa (Fig. 4). These results are similar if we restrict the longitudes to an African-only domain (20°W–55°E).
Latitude–height cross sections of zonal mean (180°W–180°E) meridional winds (m s−1) for (a) HM models and (b) LM models, and (c) HM minus LM models. The shading is the future anomaly and the contours are the present climatology. Dashed lines are northerly winds and solid lines are southerly winds.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
The areas where rainfall increases most strongly in the northern tropics, especially in the Atlantic and western Pacific, are generally displaced relative to the areas of maximum climatological rainfall. This implies a role for the pattern of future SST warming in differentiating between HM and LM models, which is what we investigate next.
b. Changes in tropical SSTs and southern African rainfall
Figure 9 shows SST warming relative to the mean tropical SST warming for OND in HM and LM models. In both model sets, the northern tropics warm more than the tropical mean while the southern tropics warm less. This is consistent with observations and historical model simulations of global annual temperature change, which show a faster rate of warming in the northern compared to Southern Hemisphere (Friedman et al. 2013), perhaps due to the greater land fraction in the Northern Hemisphere (Stouffer et al. 1989) and the cooling of the Southern Ocean by vertical mixing (Manabe et al. 1991; Xie et al. 2010).
Future SST change (K) at each grid point minus the tropical mean (20°N–20°S, 180°W–180°E) SST warming averaged over the set of (a) HM models and (b) LM models, and (c) the difference in relative future SST change between HM and LM models. Stipples are applied as in Fig. 7.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
Of particular interest here is the difference between the SST warming pattern in the HM and LM models (Fig. 9, bottom). Models projecting the highest-magnitude drying over southern Africa also simulate stronger relative warming of the northern tropics compared to the tropical mean warming. This is true across almost all longitudes in the tropical band from 10°–25°N, with particularly high agreement between models in the northern tropical Pacific and Atlantic basins. These are both regions where, as we saw in section 5a, the rainfall differences between HM and LM models are particularly pronounced—and could indicate, consistent with a growing body of work (e.g., Xie et al. 2010; Huang et al. 2013), that shifts in tropical convection are linked to the pattern of SST warming. This raises the possibility that model variability in the differential warming of the northern tropics relative to the southern tropics, and the associated adjustment of tropical convection and Hadley circulation (Figs. 7 and 8), could explain the intermodel diversity in rainfall response over southern Africa.
In support of this hypothesis, we find a negative correlation between the relative warming of the northern tropics and rainfall change over subtropical southern Africa among models (r = −0.57, p < 0.001) (Fig. 10a). Models simulating greater relative warming of the northern tropics tend to project larger precipitation declines over southern Africa. This relationship is further strengthened if the FIO-ESM model (marked with an ×) is not included in the analysis (r = −0.63, p < 0.001). In this model the northern tropics warm at a much slower rate than the tropical mean (~0.4 K), which is inconsistent with present-day observational trends (Friedman et al. 2013).
For OND, the intermodel relationship between (a) warming of northern tropical SSTs (10°–25°N, 180°W–180°E) minus tropical mean warming (20°N–20°S, 180°W–180°E) in K and rainfall change (mm season−1) over subtropical southern Africa (15°–30°S, 15°–40°E) and (b) tropical mean warming (20°N–20°S, 180°W–180°E) and rainfall change over subtropical southern Africa. Blue (red) dots indicate models that are part of the LM (HM) composites. The cross marks the outlying FIO-ESM model; note that the linear regression in (a) (green line) excludes this model.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
By definition, the rate of relative SST warming of the northern tropics depends partially on the tropical mean SST warming. To ensure that the relationship in Fig. 10a is tied to the pattern of warming and not just the average change, we also examine the relationship between tropical mean SST change and rainfall change. We find no relationship between the mean oceanic warming and rainfall change. This is evidence that the pattern of SST change is particularly important for explaining the diversity in CMIP5 rainfall projections over subtropical southern Africa. It is not, however, evidence that the mean oceanic warming is unimportant for rainfall change in individual models.
Since the models projecting strong drying in subtropical southern Africa are the same set of models projecting the strongest drying over South America (Fig. 7, bottom), we also examine the relationship between tropical SST warming and rainfall change over South America (25°–10°S, 60°–40°W) in Fig. 11. We find the same result. With FIO-ESM excluded, there is a strong and significant correlation between the differential warming of the northern tropics and rainfall change over South America (r = −0.59, p < 0.001) and no relationship between mean tropical warming and rainfall change. This suggests that understanding the drivers of the differential warming of the northern tropics is important for understanding the model diversity in early summer rainfall decline across the Southern Hemisphere.
As in Fig. 10, but for future rainfall change (mm season−1) in South America (10°–25°S, 60°–40°W).
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
6. Potential connections with present-day rainfall biases
Our main aim in this paper is to consider the processes associated with the diversity in precipitation projections. Prior to summarizing and discussing our results, it is useful to briefly consider these processes in the light of present-day biases, especially given the differences in the present-day structure of vertical velocity in HM and LM models (Fig. 4).
Figure 12 shows the current-day precipitation bias in HM and LM models compared to the CMAP satellite/rain gauge product. While both sets of models tend to overestimate precipitation over southern Africa, the bias is much more severe in HM models, reaching up to 300 mm season−1 in areas of Angola and eastern South Africa. In the LM models the precipitation bias is much weaker over most of southern Africa (<100 mm season−1). The difference in severity of bias over southern Africa is made clear in the composite anomaly of the rainfall bias between HM and LM models (Fig. 12c). Interestingly, HM models also tend to simulate too little rainfall over northern Madagascar. This dipole bias is also a feature of many models in the main rainy season [January–March (JFM)] and is related to poor representation of topography and moisture circulation (Munday and Washington 2018). The large future OND rainfall declines in HM models relative to LM models result in a future rainfall climatology across models that converges toward the present-day observed climatology. The average HM bias is 143 mm season−1 and reduces by 170% over nine decades to 58.5 mm season−1, whereas there is only an 8% reduction in the average LM model bias from 64.1 to 56.6 mm season−1. Only eight models simulate a future rainfall climatology that, in an absolute sense, is drier than the observed present-day climatology.
Climatological (1979–2005) precipitation bias (mm season−1) in OND relative to the CMAP dataset averaged across (a) HM models and (b) LM models, and (c) their difference. The stipples in (c) indicate where all HM models are either wetter or drier compared to LM models (e.g., the green areas with stipples show that all HM models are wetter than all LM models in the present day).
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
In section 3 (Fig. 4) we saw that 1) HM models simulate larger decreases in uplift at upper-tropospheric levels and larger increases in lower-level uplift compared to LM models, and 2) there are important differences between LM and HM in the structure of vertical velocity in the present day. In the present day, HM models simulate an ill-defined peak in upward motion at low levels and strong uplift at upper levels of the troposphere (Fig. 4). Here, we investigate this in more detail by considering the bias in the structure of vertical velocity in the two model sets compared to MERRA-2. Figure 13 shows that both sets of models share a similar bias structure when compared with MERRA-2: uplift at low levels (<600 hPa) is underestimated, while uplift at upper levels (500–200 hPa) is overestimated. In HM models these biases are more pronounced, with uplift at 300 hPa 5–7 × 102 Pa s−1 stronger than in MERRA-2 from 20° to 35°E and this bias is 3 times the magnitude of the future subsidence anomaly. The structure of future vertical velocity anomalies is similar but of opposite sign to the present-day bias in the vertical structure: uplift decreases at upper levels and increases at lower levels (cf. Figs. 4 and 13). Together with the significant precipitation biases in HM models (Fig. 12), this gives some cause to doubt the projections of highest-magnitude drying, especially since the models with the highest-magnitude absolute rainfall changes also simulate the highest-magnitude percentage rainfall changes (Table 1).
(a) Longitude–height (in hPa) cross section (averaged between 15° and 30°S) for the present-day climatological vertical velocity (102 Pa s−1) in MERRA-2. (b),(c) The bias in vertical velocity in HM and LM models relative to MERRA-2.
Citation: Journal of Climate 32, 12; 10.1175/JCLI-D-18-0463.1
Given that HM models simulate larger present-day biases in rainfall and in vertical velocity compared to LM models; it is worth investigating whether the highest-magnitude declines in rainfall over subtropical southern Africa are only found in models with a large present-day bias. If true, this is a basis for questioning the credibility of models projecting high-magnitude change. To do this we alter our sampling base and consider the difference in rainfall projections between the third of models with the highest present-day bias over southern Africa (n = 10) and the third of models (n = 10) with the lowest present-day bias relative to CMAP. While both sets include models that project low-magnitude drying, 7 of the 10 high-bias models simulate OND rainfall declines of more than ~50 mm season−1, whereas only one model in the low-bias set projects a 50 mm season−1 or greater rainfall decrease (IPSL-CM5A-MR; −59.0 mm season−1) (Table 1). Moreover, the average projected rainfall change in the high-bias set of models (−54.7 mm season−1) is approximately double that of the low-bias set of models (−27.6 mm season−1). While this is some evidence that high-magnitude absolute rainfall declines should be treated with caution, we note that the percentage rainfall changes (approximately −15%) are similar between high- and low-bias sets of models.
Summarizing these findings together, it seems that while models projecting high-magnitude drying, both in an absolute and relative sense, simulate larger present-day biases in rainfall and circulation bias compared to models projecting low-magnitude changes, the reverse is not true. Models with pronounced rainfall and circulation biases do not necessarily simulate high-magnitude percentage rainfall changes compared to models with milder biases. This implies that more work is needed to evaluate the connections between present-day biases and future rainfall change.
7. Discussion and conclusions
a. Summary and discussion
Climate models consistently project rainfall declines in the early summer season (OND) over subtropical southern Africa, but there is an order of magnitude difference in the scale of the decline between models. In this paper we investigate the physical processes underlying differences in the projections of rainfall decline among 30 CMIP5 models.
In all models there is an increase in stability over subtropical southern Africa by the end of the twenty-first century. The increase in stability is associated with enhanced upper-level subsidence, and occurs in spite of increases in low-level moisture flux supply and convergence, which are especially pronounced in models projecting the highest-magnitude drying. The future increase in stability and in moisture flux convergence are linked with amplified surface temperatures over southern Africa relative to the tropical mean, and a strengthened Angola thermal low. These finding are consistent with previous work on the early summer drying signal (Seth et al. 2013; Cook and Vizy 2013; Lazenby et al. 2018). Seth et al. (2013) found that the rainfall decline in the multimodel mean of CMIP5 models is associated with enhanced MSE at upper levels in the atmosphere, and reduced evaporation. Meanwhile, Cook and Vizy (2013) found that a strengthened and enlarged Angola heat low was associated with rainfall decline in a regional model driven by an ensemble mean of CMIP3 models into the mid-twenty-first century.
Differences in the strength of top-down stabilization account for the key differences between models simulating high- and low-magnitude drying. We link future changes in rainfall with local surface temperature change, since increased subsidence is linked to clearer skies and higher net solar radiation. Models projecting larger precipitation declines simulate greater surface temperature amplification over subtropical southern Africa (relative to the tropical mean). The greater surface temperature amplification in models projecting drying may also arise in part from projections of decreasing evaporation and enhanced water vapor supply. In high-magnitude drying models the ratio of sensible to latent heat fluxes increases strongly, as does the water vapor supply, which could contribute to the temperature enhancement through a direct radiative effect. The coupling between future projections of precipitation and temperature has also been noted for other semiarid regions, including the Sahel (Giannini 2010) and northern Australia (Brown et al. 2016), and indicates that uncertainties in the two variables should be considered together.
We find that the magnitude of the rainfall decline among models is also linked with changing patterns of tropical sea surface temperatures. There is a negative correlation between precipitation change over subtropical southern Africa and the relative warming of northern tropical SSTs compared to the tropical mean oceanic warming. Models projecting high-magnitude drying tend to project much stronger warming of the northern tropics relative to the tropical mean, whereas model projecting low-magnitude drying simulate more spatially uniform SST warming across the tropics. The role of the pattern of SST changes indicates that dynamical mechanisms of change are important for explaining the uncertainty in future projections of rainfall, a claim consistent with studies investigating rainfall change at a tropics-wide scale (e.g., Chadwick et al. 2013; Kent et al. 2015; Huang et al. 2013). Over southern Africa, our findings provide some support for Lazenby et al.’s (2018) hypothesis that uncertainty in CMIP5 rainfall projections during the early summer is linked with changes in the north–south SST gradient in the Indian Ocean; however, we do note that the strongest gradients in HM models are over the Atlantic.
Our results show that there is a strong coherence between models in the rainfall change over subtropical southern Africa and the rest of the tropics (Fig. 7). Models projecting high-magnitude changes over subtropical southern Africa also project consistent declines over South America and the southern tropical Pacific, while projecting consistent increases in rainfall north of the equator in the Sahel, Indian Ocean, Atlantic Ocean, and Pacific Ocean. This pattern of rainfall response to climate change could be evidence of a broadening and potential northward shift of Hadley circulation during the early austral summer. Meanwhile, for the CMIP3 ensemble, Chou et al. (2007) and Biasutti and Sobel (2009) suggest that this pattern of rainfall change is due to a delay in the global seasonal cycle of SSTs under warming. Examining the plausibility of the delay in seasonal cycle in SSTs, as well as how it relates to rainfall over specific regions, should be a focus of future work.
The projections of future rainfall decline could be affected by large present-day biases in precipitation over subtropical southern Africa (Figs. 1 and 12 and Table 1). The biases precipitation are especially pronounced in the models that project high-magnitude drying over subtropical southern Africa (and South America) relative to those projecting low-magnitude drying and is associated with a misrepresentation of the structure of uplift/subsidence over subtropical southern Africa. In HM models, upper-level uplift is overestimated in models compared to reanalysis, while the shallow convection associated with the Angola thermal low is underestimated. Importantly, this bias structure is similar but of opposite sign to the future anomalies in vertical velocity whereby upper-level uplift decreases, and low-level uplift increases. The juxtaposition of present-day biases and future anomalies gives some reason to be sceptical of high-magnitude future drying in OND over subtropical southern Africa. This call for caution is supported by observational evidence, which indicates that there is no clear early summer drying trend in the last 30 years over southern Africa (Maidment et al. 2015).
b. Conclusions
Large ensemble mean rainfall reductions in early summer are driven primarily by the response of more extreme members of the CMIP5 ensemble, while a third of the models project modest declines of less than 30 mm across the whole October–December season. This paper provides an insight into the controls on the intermodel diversity in early summer rainfall declines. We find that members of the ensemble projecting high-magnitude rainfall decreases simulate large increases in tropospheric stability associated with strong relative warming of the northern tropics compared to the tropical mean. The link between this pattern of SST warming and enhanced stability over subtropical southern Africa could be mediated by a northward shifted and broadened Hadley circulation during OND, especially in HM models (Fig. 8).
The rainfall declines across models occur in the context of systematic rainfall biases in the present-day climatology across the ensemble. We have shown here that that models that simulate extreme future drying (both in an absolute and relative sense), simulate larger present-day biases in vertical velocity and rainfall over subtropical southern Africa compared to those models that simulate low-magnitude rainfall changes. This is not decisive evidence that high-magnitude drying is implausible, but does suggest that further analysis of these extreme projections, and their relationship with present-day climate dynamics, could help to constrain the range of model projections of early summer drying. This is the subject of ongoing work.
Acknowledgments
The first author is funded through a Met Office CASE studentship (ACR00400) and through the doctoral training partnership of the U.K. Natural Environment Research Council (NERC) (NE/L002612/1). Richard Washington is supported by the NERC and Department for International Development (DfID) funded Future Climate for Africa (FCFA) UMFULA (NE/M020207/1) and IMPALA (NE/M017206) African climate programs. We source the climate model data from the Earth System Grid Federation (ESGF) (https://pcmdi.llnl.gov/). CMAP data are downloaded from NOAA/OAR/ESRL PSD, Boulder, Colorado, USA at http://www.esrl.noaa.gov/psd/ and MERRA-2 data are downloaded from http://gmao.gsfc.nasa.gov/products/. USGS GTOPO30 global digital elevation model data are sourced from http://earthexplorer.usgs.gov/. The authors thank Dr Amy Creese, Dr Neil Hart, Emma Howard, and Dr. Rachel James for useful discussions. We also thank three anonymous reviewers whose comments enriched the manuscript.
REFERENCES
Adebiyi, A. A., and P. Zuidema, 2016: The role of the southern African easterly jet in modifying the southeast Atlantic aerosol and cloud environments. Quart. J. Roy. Meteor. Soc., 142, 1574–1589, https://doi.org/10.1002/qj.2765.
Allan, R. P., B. J. Soden, V. O. John, W. Ingram, and P. Good, 2010: Current changes in tropical precipitation. Environ. Res. Lett., 5, 025205, https://doi.org/10.1088/1748-9326/5/2/025205.
Biasutti, M., and A. H. Sobel, 2009: Delayed Sahel rainfall and global seasonal cycle in a warmer climate. Geophys. Res. Lett., 36, L23707, https://doi.org/10.1029/2009GL041303.
Blamey, R. C., and C. J. C. Reason, 2013: The role of mesoscale convective complexes in southern Africa summer rainfall. J. Climate, 26, 1654–1668, https://doi.org/10.1175/JCLI-D-12-00239.1.
Blamey, R. C., C. Middleton, C. Lennard, and C. J. C. Reason, 2017: A climatology of potential severe convective environments across South Africa. Climate Dyn., 49, 2161–2178, https://doi.org/10.1007/s00382-016-3434-7.
Brown, J. R., A. F. Moise, R. Colman, and H. Zhang, 2016: Will a warmer world mean a wetter or drier Australian monsoon? J. Climate, 29, 4577–4596, https://doi.org/10.1175/JCLI-D-15-0695.1.
Chadwick, R., I. Boutle, and G. Martin, 2013: Spatial patterns of precipitation change in CMIP5: Why the rich do not get richer in the tropics. J. Climate, 26, 3803–3822, https://doi.org/10.1175/JCLI-D-12-00543.1.
Chou, C., and J. D. Neelin, 2004: Mechanisms of global warming impacts on regional tropical precipitation. J. Climate, 17, 2688–2701, https://doi.org/10.1175/1520-0442(2004)017<2688:MOGWIO>2.0.CO;2.
Chou, C., J. Y. Tu, and P. H. Tan, 2007: Asymmetry of tropical precipitation change under global warming. Geophys. Res. Lett., 34, L17708, https://doi.org/10.1029/2007GL030327.
Chou, C., J. D. Neelin, C.-A. Chen, and J.-Y. Tu, 2009: Evaluating the “rich-get-richer” mechanism in tropical precipitation change under global warming. J. Climate, 22, 1982–2005, https://doi.org/10.1175/2008JCLI2471.1.
Christensen, J. H., and Coauthors, 2007: Regional climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 847–940.
Collins, M., and Coauthors, 2013: Long-term climate change: Projections, commitments and irreversibility. Climate Change 2013: The Physical Science Basis, T. Stocker et al., Eds., Cambridge University Press, 1029–1136.
Conway, D., and Coauthors, 2015: Climate and southern Africa’s water–energy–food nexus. Nat. Climate Change, 5, 837–846, https://doi.org/10.1038/nclimate2735.
Cook, K. H., and E. K. Vizy, 2013: Projected changes in East African rainy seasons. J. Climate, 26, 5931–5948, https://doi.org/10.1175/JCLI-D-12-00455.1.
Dieppois, B., M. Rouault, and M. New, 2015: The impact of ENSO on Southern African rainfall in CMIP5 ocean atmosphere coupled climate models. Climate Dyn., 45, 2425–2442, https://doi.org/10.1007/s00382-015-2480-x.
Driver, P., and C. J. C. Reason, 2017: Variability in the Botswana High and its relationships with rainfall and temperature characteristics over southern Africa. Int. J. Climatol., 37, 570–581, https://doi.org/10.1002/joc.5022.
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, 11 405–11 424, https://doi.org/10.1002/2016jd025428.
Engelbrecht, C. J., F. A. Engelbrecht, and L. L. Dyson, 2013: High-resolution model-projected changes in mid-tropospheric closed-lows and extreme rainfall events over southern Africa. Int. J. Climatol., 33, 173–187, https://doi.org/10.1002/joc.3420.
Fauchereau, N., S. Trzaska, Y. Richard, P. Roucou, and P. Camberlin, 2003: Sea-surface temperature co-variability in the southern Atlantic and Indian Oceans and its connections with the atmospheric circulation in the Southern Hemisphere. Int. J. Climatol., 23, 663–677, https://doi.org/10.1002/joc.905.
Friedman, A. R., Y. T. Hwang, J. C. H. Chiang, and D. M. W. Frierson, 2013: Interhemispheric temperature asymmetry over the twentieth century and in future projections. J. Climate, 26, 5419–5433, https://doi.org/10.1175/JCLI-D-12-00525.1.
Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 5419–5454, https://doi.org/10.1175/JCLI-D-16-0758.1.
Giannini, A., 2010: Mechanisms of climate change in the semiarid African Sahel: The local view. J. Climate, 23, 743–756, https://doi.org/10.1175/2009JCLI3123.1.
Gosling, S. N., and N. W. Arnell, 2016: A global assessment of the impact of climate change on water scarcity. Climatic Change, 134, 371–385, https://doi.org/10.1007/s10584-013-0853-x.
Harrison, M. S. J., 1984: A generalized classification of South African summer rain-bearing synoptic systems. J. Climatol., 4, 547–560, https://doi.org/10.1002/joc.3370040510.
Hart, N. C. G., C. J. C. Reason, and N. Fauchereau, 2013: Cloud bands over southern Africa: Seasonality, contribution to rainfall variability and modulation by the MJO. Climate Dyn., 41, 1199–1212, https://doi.org/10.1007/s00382-012-1589-4.
Hart, N. C. G., R. Washington, and C. J. C. Reason, 2018: On the likelihood of tropical–extratropical cloud bands in the South Indian convergence zone during ENSO events. J. Climate, 31, 2797–2817, https://doi.org/10.1175/JCLI-D-17-0221.1.
He, J., and B. J. Soden, 2017: A re-examination of the projected subtropical precipitation decline. Nat. Climate Change, 7, 53–57, https://doi.org/10.1038/nclimate3157.
Held, I. M., 2005: The gap between simulation and understanding in climate modeling. Bull. Amer. Meteor. Soc., 86, 1609–1614, https://doi.org/10.1175/BAMS-86-11-1609.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrologic cycle to global warming. J. Climate, 19, 5686–5699, https://doi.org/10.1175/JCLI3990.1.
Held, I. M., T. L. Delworth, J. Lu, K. L. Findell, and T. R. Knutson, 2005: Simulation of Sahel drought in the 20th and 21st centuries. Proc. Natl. Acad. Sci. USA, 102, 17 891–17 896, https://doi.org/10.1073/pnas.0509057102.
Hermes, J. C., and C. J. C. Reason, 2005: Ocean model diagnosis of interannual coevolving SST variability in the South Indian and South Atlantic Oceans. J. Climate, 18, 2864–2882, https://doi.org/10.1175/JCLI3422.1.
Hirst, A. C., and S. Hastenrath, 1983: Atmosphere–ocean mechanisms of climate anomalies in the Angola–tropical Atlantic sector. J. Phys. Oceanogr., 13, 1146–1157, https://doi.org/10.1175/1520-0485(1983)013<1146:AOMOCA>2.0.CO;2.
Hoell, A., C. Funk, J. Zinke, and L. Harrison, 2017: Modulation of the Southern Africa precipitation response to the El Niño Southern Oscillation by the subtropical Indian Ocean dipole. Climate Dyn., 48, 2529–2540, https://doi.org/10.1007/s00382-016-3220-6.
Howard, E., and R. Washington, 2018: Characterizing the synoptic expression of the Angola low. J. Climate, 31, 7147–7165, https://doi.org/10.1175/JCLI-D-18-0017.1.
Hua, W., L. Zhou, H. Chen, S. Nicholson, A. Raghavendra, and Y. Jiang, 2016: Possible causes of the Central Equatorial African long-term drought. Environ. Res. Lett., 11, 124002, https://doi.org/10.1088/1748-9326/11/12/124002.
Huang, P., S.-P. Xie, K. Hu, G. Huang, and R. Huang, 2013: Patterns of the seasonal response of tropical rainfall to global warming. Nat. Geosci., 6, 357–361, https://doi.org/10.1038/ngeo1792.
James, R., and R. Washington, 2013: Changes in African temperature and precipitation associated with degrees of global warming. Climatic Change, 117, 859–872, https://doi.org/10.1007/s10584-012-0581-7.
James, R., R. Washington, and R. Jones, 2015: Process-based assessment of an ensemble of climate projections for West Africa. J. Geophys. Res. Atmos., 120, 1221–1238, https://doi.org/10.1002/2014JD022513.
Kent, C., R. Chadwick, and D. P. Rowell, 2015: Understanding uncertainties in future projections of seasonal tropical precipitation. J. Climate, 28, 4390–4413, https://doi.org/10.1175/JCLI-D-14-00613.1.
Knutti, R., R. Furrer, C. Tebaldi, J. Cermak, and G. Meehl, 2010: Challenges in combining projections from multiple climate models. J. Climate, 23, 2739–2758, https://doi.org/10.1175/2009JCLI3361.1.
Lau, W. K. M., and K. Kim, 2015: Robust Hadley circulation changes and increasing global dryness due to CO2 warming from CMIP5 model projections. Proc. Natl. Acad. Sci. USA, 112, 3630–3635, https://doi.org/10.1073/pnas.1418682112.
Lazenby, M. J., M. C. Todd, and Y. Wang, 2016: Climate model simulation of the South Indian Ocean Convergence Zone: Mean state and variability. Climate Res., 68, 59–71, https://doi.org/10.3354/cr01382.
Lazenby, M. J., M. C. Todd, R. Chadwick, and Y. Wang, 2018: Future precipitation projections over central and southern Africa and the adjacent Indian Ocean: What causes the changes and the uncertainty? J. Climate, 31, 4807–4826, https://doi.org/10.1175/JCLI-D-17-0311.1.
Lindesay, J. A., 1988: South African rainfall, the Southern Oscillation and a Southern Hemisphere semi-annual cycle. J. Climatol., 8, 17–30, https://doi.org/10.1002/joc.3370080103.
Ma, J., and S. P. Xie, 2013: Regional patterns of sea surface temperature change: A source of uncertainty in future projections of precipitation and atmospheric circulation. J. Climate, 26, 2482–2501, https://doi.org/10.1175/JCLI-D-12-00283.1.
Ma, J., S. P. Xie, and Y. Kosaka, 2012: Mechanisms for tropical tropospheric circulation change in response to global warming. J. Climate, 25, 2979–2994, https://doi.org/10.1175/JCLI-D-11-00048.1.
Maidment, R., D. Grimes, R. P. Allan, E. Tarnavsky, M. Stringer, T. Hewison, R. Roebling, and E. Black, 2014: The 30 year TAMSAT African Rainfall Climatology and Time series (TARCAT) data set. J. Geophys. Res. Atmos., 116, 10 619–10 644, https://doi.org/10.1002/2014JD021927.
Maidment, R., R. P. Allan, and E. Black, 2015: Recent observed and simulated changes in precipitation over Africa. Geophys. Res. Lett., 42, 8155–8164, https://doi.org/10.1002/2015GL065765.
Manabe, S., R. J. Stouffer, M. J. Spelman, and K. Bryan, 1991: Transient responses of a coupled ocean–atmosphere model to gradual changes of atmospheric CO2. Part II: Seasonal response. J. Climate, 4, 785–817, https://doi.org/10.1175/1520-0442(1991)004<0785:TROACO>2.0.CO;2.
Matarira, C. H., 1990: Drought over Zimbabwe in a regional and global context. Int. J. Climatol., 10, 609–625, https://doi.org/10.1002/joc.3370100605.
Midgley, G. F., and W. Thuiller, 2011: Potential responses of terrestrial biodiversity in Southern Africa to anthropogenic climate change. Reg. Environ. Change, 11, 127–135, https://doi.org/10.1007/s10113-010-0191-8.
Mueller, B., and S. I. Seneviratne, 2014: Systematic land climate and evapotranspiration biases in CMIP5 simulations. Geophys. Res. Lett., 41, 128–134, https://doi.org/10.1002/2013GL058055.
Mulenga, H., 1998: Southern African climatic anomalies, summer rainfall and the Angola low. Ph.D. thesis, University of Cape Town, 261 pp., http://hdl.handle.net/11427/19649.
Munday, C., and R. Washington, 2017: Circulation controls on southern African precipitation in coupled models: The role of the Angola Low. J. Geophys. Res. Atmos., 122, 861–877, https://doi.org/10.1002/2016JD025736.
Munday, C., and R. Washington, 2018: Systematic climate model rainfall biases over southern Africa: Links to moisture circulation and topography. J. Climate, 31, 7533–7548, https://doi.org/10.1175/JCLI-D-18-0008.1.
Neelin, J. D., C. Chou, and H. Su, 2003: Tropical drought regions in global warming and El Niño teleconnections. Geophys. Res. Lett., 30, 2275, https://doi.org/10.1029/2003GL018625.
Nicholson, S. E., and E. Kim, 1997: The relationship of the El Nino–Southern oscillation to African rainfall. Int. J. Climatol., 17, 117–135, https://doi.org/10.1002/(SICI)1097-0088(199702)17:2<117::AID-JOC84>3.0.CO;2-O.
Novella, N. S., and W. M. Thiaw, 2013: African rainfall climatology version 2 for famine early warning systems. J. Appl. Meteor. Climatol., 52, 588–606, https://doi.org/10.1175/JAMC-D-11-0238.1.
O’Gorman, P. A., and M. S. Singh, 2013: Vertical structure of warming consistent with an upward shift in the middle and upper troposphere. Geophys. Res. Lett., 40, 1838–1842, https://doi.org/10.1002/grl.50328.
Park, J. Y., J. Bader, and D. Matei, 2015: Northern-hemispheric differential warming is the key to understanding the discrepancies in the projected Sahel rainfall. Nat. Commun., 6, 5985, https://doi.org/10.1038/ncomms6985.
Parker, W. S., 2011: When climate models agree: The significance of robust model predictions. Philos. Sci., 78, 579–600, https://doi.org/10.1086/661566.
Pinto, I., C. Jack, and B. Hewitson, 2018: Process-based model evaluation and projections over southern Africa from Coordinated Regional Climate Downscaling Experiment and Coupled Model Intercomparison Project Phase 5 models. Int. J. Climatol., 38, 4251–4261, https://doi.org/10.1002/joc.5666.
Reason, C. J. C., 2016: The Bolivian, Botswana, and Bilybara highs and Southern Hemisphere drought/floods. Geophys. Res. Lett., 43, 1280–1286, https://doi.org/10.1002/2015GL067228.
Reichle, R., Q. Liu, R. Koster, C. Drapar, S. Mahanama, and G. Partyka, 2017: Land surface precipitation in MERRA-2. J. Climate, 30, 1643–1664, https://doi.org/10.1175/JCLI-D-16-0570.1.
Richard, Y., N. Fauchereau, I. Poccard, M. Rouault, and S. Trzaska, 2001: 20th century droughts in southern Africa: Spatial and temporal variability, teleconnections with oceanic and atmospheric conditions. Int. J. Climatol., 21, 873–885, https://doi.org/10.1002/joc.656.
Rouault, M., S. A. White, C. J. C. Reason, J. R. E. Lutjeharms, and I. Jobard, 2002: Ocean–atmosphere interaction in the Agulhas Current region and a South African extreme weather event. Wea Forecasting, 17, 655–669, https://doi.org/10.1175/1520-0434(2002)017<0655:OAIITA>2.0.CO;2.
Rouault, M., P. Florenchie, N. Fauchereau, and C. J. C. Reason, 2003: South East tropical Atlantic warm events and southern African rainfall. Geophys. Res. Lett., 30, 8009, https://doi.org/10.1029/2002GL014840.
Rowell, D. P., and R. Chadwick, 2018: Causes of the uncertainty in projections of tropical terrestrial rainfall change: East Africa. J. Climate, 31, 5977–5995, https://doi.org/10.1175/JCLI-D-17-0830.1.
Schlenker, W., and D. B. Lobell, 2010: Robust negative impacts of climate change on African agriculture. Environ. Res. Lett., 5, 014010, https://doi.org/10.1088/1748-9326/5/1/014010.
Seager, R., N. Naik, and G. A. Vecchi, 2010: Thermodynamic and dynamic mechanisms for large-scale changes in the hydrological cycle in response to global warming. J. Climate, 23, 4651–4668, https://doi.org/10.1175/2010JCLI3655.1.
Serdeczny, O., and Coauthors, 2017: Climate change impacts in sub-Saharan Africa: From physical changes to their social repercussions. Reg. Environ. Change, 17, 1585–1600, https://doi.org/10.1007/s10113-015-0910-2.
Seth, A., S. A. Rauscher, M. Rojas, A. Giannini, and S. J. Camargo, 2011: Enhanced spring convective barrier for monsoons in a warmer world? Climatic Change, 104, 403–414, https://doi.org/10.1007/s10584-010-9973-8.
Seth, A., S. A. Rauscher, M. Biasutti, A. Giannini, S. J. Camargo, and M. Rojas, 2013: CMIP5 projected changes in the annual cycle of precipitation in monsoon regions. J. Climate, 26, 7328–7351, https://doi.org/10.1175/JCLI-D-12-00726.1.
Shongwe, M. E., G. J. van Oldenborgh, B. J. J. M. van den Hurk, B. de Boer, C. A. S. Coelho, and M. K. van Aalst, 2009: Projected changes in mean and extreme precipitation in Africa under global warming. Part I: Southern Africa. J. Climate, 22, 3819–3837, https://doi.org/10.1175/2009JCLI2317.1.
Stouffer, R. J., S. Manabe, and K. Bryan, 1989: Climatic response to a gradual increase of atmospheric carbon dioxide. Nature, 342, 660–662, https://doi.org/10.1038/342660a0.
Tadross, M. A., B. C. Hewitson, and M. T. Usman, 2005: The interannual variability of the onset of the maize growing season over South Africa and Zimbabwe. J. Climate, 18, 3356–3372, https://doi.org/10.1175/JCLI3423.1.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498, https://doi.org/10.1175/BAMS-D-11-00094.1.
Unganai, L. S., and S. J. Mason, 2002: Long-range predictability of Zimbabwe summer rainfall. Int. J. Climatol., 22, 1091–1103, https://doi.org/10.1002/joc.786.
van Vuuren, D. P., and Coauthors, 2011: The representative concentration pathways: An overview. Climatic Change, 109, 5–31, https://doi.org/10.1007/s10584-011-0148-z.
Vecchi, G. A., B. J. Soden, A. T. Wittenberg, I. M. Held, A. Leetmaa, and M. J. Harrison, 2006: Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing. Nature, 441, 73–76, https://doi.org/10.1038/nature04744.
Washington, R., and M. Todd, 1999: Tropical–temperate links in southern African and southwest Indian Ocean satellite-derived daily rainfall. Int. J. Climatol., 19, 1601–1616, https://doi.org/10.1002/(SICI)1097-0088(19991130)19:14<1601::AID-JOC407>3.0.CO;2-0.
Washington, R., and Coauthors, 2006: African climate change: Taking the shorter route. Bull. Amer. Meteor. Soc., 87, 1355–1366, https://doi.org/10.1175/BAMS-87-10-1355.
Xie, P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78, 2539–2558, https://doi.org/10.1175/1520-0477(1997)078<2539:GPAYMA>2.0.CO;2.
Xie, S. P., C. Deser, G. A. Vecchi, J. Ma, H. Teng, and A. T. Wittenberg, 2010: Global warming pattern formation: Sea surface temperature and rainfall. J. Climate, 23, 966–986, https://doi.org/10.1175/2009JCLI3329.1.