1. Introduction
The Sahel region of Africa experienced a severe reduction in wet season rainfall (July–September) from the 1970s to the 1990s, from which there has been a partial recovery. Numerous studies have found that the direct cause of the drought largely arises from changes in sea surface temperature (SST) patterns, with all ocean basins being important. The role of Atlantic SSTs was revealed first, with a tendency for dry Sahel years to be associated with anomalously cool SSTs in the northern (tropical) Atlantic and anomalously warm SSTs in the South Atlantic, particularly including the Gulf of Guinea (e.g., Lamb 1978; Vizy and Cook 2001; Lau et al. 2006). Warmer SSTs in the Indian Ocean have also been linked with drought over the Sahel (see, e.g., Palmer 1986; Hagos and Cook 2008; Lu 2009), along with a large-scale SST gradient into the western Pacific (Rowell 2001). ENSO events are influential on interannual time scales (e.g., Folland et al. 1991; Janicot et al. 1996; Rowell 2001), and more recently a contribution to drought of decadal variability in the Mediterranean has been confirmed (Rowell 2003; Jung et al. 2006). These anomalous regional SST patterns also contribute to a global “interhemispheric thermal contrast” pattern of SSTs that has been linked with Sahel drought, particularly on decadal time scales (e.g., Folland et al. 1986).
A key question, however, is whether these SST patterns are due to natural variability or anthropogenic forcing. This remains an issue of some debate. Sutton and Hodson (2005), Hoerling et al. (2006), and Ting et al. (2009) suggest that Atlantic SST changes (and hence the drought) were primarily of natural origin, whereas Held et al. (2005), Biasutti and Giannini (2006), Lau et al. (2006), and Kawase et al. (2010) suggest that observed precipitation changes were due to a combination of external anthropogenic forcing and internal (natural) variability. Held et al. (2005) analyzed initial condition ensembles of two versions of a coupled model, and concluded that the drying trend over the Sahel is partly anthropogenically forced (aerosol loadings and greenhouse gas emissions) and partly due to internal variability of the ocean–atmosphere climate system. Biasutti and Giannini (2006) use 19 models from the third Coupled Model Intercomparison Project (CMIP3) archive, and reach a similar conclusion, but also estimate that the fraction of the 1930–99 drying trend that was externally forced (primarily by anthropogenic activity) was “at least 30%.” Lau et al. (2006) analyzed the same 19 models, but focused on the period of drought from the 1970s to 1990s, and speculated that the increase in both volcanic and anthropogenic aerosols may have been responsible.
Sulfate aerosol, derived from anthropogenic activity and concentrated mainly in the Northern Hemisphere (NH), increased rapidly from 1940 to 1990 and could well have contributed significantly to the cross-equatorial SST gradient, which substantially altered low-latitude circulation and rainfall (Hulme and Kelly 1993; Williams et al. 2001; Rotstayn and Lohmann 2002; Baines and Folland 2007). However, the cross-equatorial SST changes have been attributed to natural modes of ocean variability by Sutton and Hodson (2005), Hoerling et al. (2006), and Ting et al. (2009), suggesting that anthropogenic activity alone is unlikely to be the sole cause of the Sahel drying. Another process that has been discussed as playing a role in the historical drought is a direct local radiative forcing from dust aerosol, which may have amplified the SST-induced drying in the Sahel (Yoshioka et al. 2007). However, there is currently no multimodel agreement on the sign of the dust impact. This, and other processes (such as carbonaceous aerosol and changing land use), are beyond the scope of this study.
Resolving this discussion on the possible causes of the recent SST variability (that has influenced Sahel precipitation) is important, not only for attributing the prolonged drought of recent decades, but also for improving predictions of future anthropogenic change in this region (see Cook 2008). One possible explanation for the lack of scientific consensus may be due to the limited ensemble sizes of available models to attribute the natural versus forced contributions. Another may be that many of the currently available CMIP3 models do not yet represent the aerosol indirect effects very well [which have been shown to be important in Williams et al. (2001)].
Despite the strong cooling influence of sulfate aerosols on surface air temperatures, confirmation of their influence on Sahel rainfall by changing hemispheric temperatures is difficult without a large ensemble of models. The ensemble size of the studies cited earlier limits their estimates of the relative importance of different forcings, due to problems in detecting the forced signal against the background of internal variability. The climateprediction.net experiment (Stainforth et al. 2005) uses a distributed computing platform to perform tens of thousands of climate simulations on the home computers of volunteers worldwide. Therefore, the results from the climate prediction.net project are uniquely placed to separate the signal of the sulfate aerosol and greenhouse gas forcing from the effects of natural (and model internal) variability. Furthermore, the wide variety of model climatologies that are created by the “perturbed physics” approach employed by the climateprediction.net experiment also goes some way to improving on the limited availability of models in earlier studies.
This study concentrates on the impacts of sulfate aerosol contrasted with the response to greenhouse gas forcing. Two types of experiment are analyzed. In section 2, we investigate the response of a slab ocean GCM to changes in greenhouse gas and sulfur dioxide (SO2) emissions. These provide a conceptual understanding of the relative roles of both aerosols and greenhouse gases in Sahel rainfall, and are used to study climate anomalies in both the present and future climates. The results from these relatively simple experiments are then used to gain insight into the twentieth-century ensemble experiments with a fully coupled ocean–atmosphere GCM, described in section 3. Here, we make use of two sets of ensemble simulations of the twentieth century to explore relationships between changes in aerosol concentration, and impacts on Sahel rainfall using two ensembles of fully coupled ocean–atmosphere GCMs. The first ensemble represents the transient climate response to historical changes in greenhouse gases and sulfate aerosols as well as volcanic and solar forcing. The second ensemble makes use of an ensemble with identical forcings except that the sulfur emissions correspond to a period shifted 70 yr into the future. The combination of all ensembles uniquely enables us to explore the dependence of Sahel drying on sulfate aerosol concentrations (over and above model internal variability) and responses to other forcing mechanisms, such as an increase in greenhouse gases. The conclusions from this study are given in section 4.
2. Slab-ocean GCM experiments
a. Sulfur cycle background and experiments
In these experiments, the third Hadley Centre Atmospheric Model is coupled to both a slab ocean model (HadSM3) and a fully interactive sulfur cycle scheme, as documented in Jones et al. (2001). However, in this study there was no representation of the second indirect effect [unlike in Jones et al. (2001)] since HadSM3 did not have a prognostic ice–cloud scheme. Two gaseous sulfur species are emitted: SO2 and dimethyl sulfide. Sulfur dioxide is derived from anthropogenic activity as well as volcanic emissions within the model, and DMS, emitted from all ocean basins, represents the sulfur produced from the metabolic processes of phytoplankton. DMS is emitted from the surface of the model, whereas anthropogenic SO2 emissions occur from both the surface and between levels 3–5 in the model (to represent “high-level chimney stack” emissions). Volcanic SO2 is emitted into the model between the levels at 870 and 422 hPa.
To form sulfate aerosol, the sulfur gases need to be oxidized. This occurs in cloud-free regions via dry oxidation or within cloud and precipitation droplets via aqueous oxidation. These oxidation processes lead to the formation of three modes of sulfate aerosol, which are Aitken mode (median radius, r = 0.024 μm), accumulation mode (r = 0.095 μm), and dissolved sulfate. Both gaseous sulfur species, as well as the sulfate aerosol, interact with the model meteorology (such as cloud and precipitation) and are transported throughout the domain using the model’s tracer advection scheme. All sulfur species are subjected to removal via dry and wet deposition (“scavenging”), except DMS, which is treated as insoluble. The balance between emissions, conversion, and deposition leads to an atmospheric loading of sulfate known as the “sulfate burden,” which will be used in the rest of this paper [for more detailed information on the climateprediction.net sulfur cycle experiment, see Ackerley et al. (2009)].
An ensemble of 15-yr experiments was run with HadSM3 within the climateprediction.net distributed computing framework. Also, a smaller ensemble of 30-yr experiments was run using the computing resource at the Rutherford Appleton Laboratory (RAL). Table 1 contains the boundary condition forcings applied to each phase of the experiment, with the numbers in parentheses denoting the number of model runs attained. The control runs (to which the subsequent forcing experiments in Table 1 are compared) consisted of one of the following features:
a preindustrial (PI) atmosphere with CO2 concentrations set to 280 ppm and natural sulfur emissions (volcanic SO2 and DMS) only or
a present day atmosphere (PDall) with CO2 concentrations set to 346 ppm and both natural and anthropogenic sulfur emissions.
The slab-ocean experiments where PI refers to a “preindustrial” control run (CO2 concentrations of 280 ppm and natural sulfur emissions) and PD refers to a “present day” control run (CO2 concentrations of 346 ppm and sulfur emissions equivalent to those of the mid-1980s). The term FUT denotes future forcings such as doubled PD CO2 concentrations, 2050 anthropogenic SO2 emissions, and both of these forcings together.
The PDall and 
The sulfate burdens in PI, PD, and 2050 
Annual and ensemble mean sulfate burdens 
Citation: Journal of Climate 24, 19; 10.1175/JCLI-D-11-00019.1
The changes in zonal mean sulfate burden for 
The zonal, annual, and ensemble mean changes in sulfate burden 
Citation: Journal of Climate 24, 19; 10.1175/JCLI-D-11-00019.1
b. Global and hemispheric surface air temperatures
Previous work, such as that by Ting et al. (2009), Hoerling et al. (2006), and Held et al. (2005), indicated that precipitation in the Sahel may be influenced by the changes in interhemispheric temperatures and so we start by analyzing the global and hemispheric 1.5-m surface air temperature (SAT) responses to each of the boundary condition forcings (see Table 1), which are given in Table 2.
The change in global, NH, and SH surface air temperatures (ΔGMT, ΔNHT, and ΔSHT, respectively) for each of the forcing experiments given in Table 1 relative to their respective control runs (K).
The results in Table 2 highlight the global mean cooling effects of sulfates and the warming influence of increased CO2 (when perturbed individually). The global mean temperature (GMT) responses in the combined forcing experiments are (approximately) a linear addition of the responses to each individual forcing, which suggests there is little nonlinear interaction resulting from perturbing CO2 and sulfate concentrations simultaneously, in these experiments.
Also included in Table 2 are the hemispheric changes in SAT, which highlight the inhomogeneous response of SAT to a sulfate and/or CO2 forcing. In 
The results are very different in the FUT experiments however. The sulfate cooling in 
The results in this section clearly show that the responses to sulfate and CO2 forcing differ between the NH and the SH, which will have implications for the seasonal migration of the intertropical convergence zone (ITCZ) and may therefore influence Sahel rainfall.
c. Zonal mean surface air temperature response
To show the asymmetrical properties of the surface air temperature response to each of the sulfate and/or CO2 forcings further (introduced in Table 2), the zonal, annual ensemble mean temperature responses for Hadley Centre–Climate Research Unit historical temperature data (HadCRU; Brohan et al. 2006) and each experiment are shown in Fig. 3. Figures 3a and 3b show the changes in zonal SAT from the HadCRU dataset, for the period 1860–1980, globally and in the Atlantic basin, respectively. The HadCRU data are not distributed uniformly and so do not provide a direct comparison to the model data but both plots (Figs. 3a and 3b) clearly show a preferential warming of the SH relative to the NH (south of approximately 50°N).
The zonal mean change in SAT (K) from HadCRU between 1860 and 1980 for the average (a) along all meridians and (b) for the Atlantic basin only. The model zonal mean surface air temperature responses (K) for (c) PD relative to PI globally, (d) PD relative to PI in the Atlantic basin, (e) FUT relative to PD globally, and (f) FUT relative to PD in the Atlantic basin. The ensemble mean is taken for experiments consisting of more than one run (also see Table 1).
Citation: Journal of Climate 24, 19; 10.1175/JCLI-D-11-00019.1
Having identified the larger warming in the SH compared to the NH in HadCRU, we now consider the model SAT changes for each of the forcing experiments globally and within the Atlantic basin. For 
The zonal mean surface air temperature responses to the FUT experiments globally, and in the Atlantic basin, can be seen in Figs. 3e and 3f. The strong warming of the NH due to the sea ice–albedo feedback mechanism and the hotter continental interiors can clearly be seen in 
The present-day to future hemispheric temperature response therefore is dominated by the CO2-induced warming whereas the preindustrial to present-day response is dominated by the asymmetrical influence of anthropogenic aerosol cooling the NH relative to the SH. The surface air temperature responses observed in this section will help explain the precipitation response to each of the imposed forcings.
d. Precipitation response
Figure 4 shows the zonal annual mean precipitation responses to each of the forcing experiments given in Table 1. The effects of CO2 warming in 
The zonal, annual, and ensemble mean changes in precipitation (mm day−1) for (a) PD relative to PI and (b) FUT relative to PD (see Table 1 for details on the PI and PD control runs).
Citation: Journal of Climate 24, 19; 10.1175/JCLI-D-11-00019.1
The opposite is true in the 
The Sahel is defined here to be a rectangle between 10°–18°N and 20°W–22°E (Ruosteenoja et al. 2003) and we focus on the June–July–August (JJA) period when the precipitation is high in this region. The mean change in precipitation associated with each of the forcing experiments in Table 1, relative to their respective control runs in JJA, can be seen in Fig. 5. The hemispheric impacts of increasing the sulfate loading alone from PI in 
The JJA mean precipitation response (mm day−1) to (a) 
Citation: Journal of Climate 24, 19; 10.1175/JCLI-D-11-00019.1
When considering the “future” forcing scenarios (FUT experiments in Table 1), the precipitation responses are very different. The drying associated with the sulfate cooling in 
The results of the simple thermodynamic slab-ocean model demonstrate the differing effects on tropical and Sahel precipitation and give a conceptual background as to how the Hadley Centre’s climate model responds to each forcing. The response of these slab models can be seen as a form of idealized climate response, where both the response time scale and the influence of the full ocean-driven natural variability can be neglected. When exploring the implications of drivers of historic Sahel drying however, we need to account for these processes. In the next section we use large ensembles of fully coupled ocean–atmosphere GCMs, which enable the forced climate responses to be analyzed against the background of the ocean-driven variability and response time scale.
3. Transient GCM experiments
The results of the slab model experiments have provided a useful conceptual platform to help us understand the effects of sulfate and CO2 forcing on precipitation in the Sahel. However, the slab model uses a rather simple representation of the ocean and gives no information on how the time-evolving climate drivers impact the transient climate response. In this section we make use of a fully coupled ocean—atmosphere GCM to explore some of these transient effects. Utilizing the very large third climate configuration of the Met Office Unified Model (HadCM3L) ensemble generated within climateprediction.net (see Frame et al. 2009) enables us to look for a signal of the sulfate aerosol and CO2 forcings on Sahel precipitation against the background of large model variability (which refers to imposed natural forcing, model internal variability, and differences in model parameter settings).
a. Method
Several thousand simulations of the HadCM3L coupled ocean–atmosphere GCM were completed by volunteers on home computers (Frame et al. 2009). Each of the distributed models had a different combination of model parameters (Stainforth et al. 2005), forcing fields [well-mixed greenhouse gases, anthropogenic sulfate, and solar and volcanic aerosol; see Frame et al. (2009), but all with zero DMS emissions], and initial conditions. There were two ensemble experiments run:
standard transient runs [STR, 1920–2000 greenhouse gas concentrations, sulfur emissions, and natural forcings from Nakićenović et al. (2000)], where the simulations have SO2 emissions specified according to observations and a total of 1566 simulations are included, which have been selected such that they include the same ranges of model parameter values and forcing fields (solar or volcanic) as in the AATR ensemble (described below), and
altered aerosol transient runs [AATR, 1920–2000 greenhouse concentrations and natural forcings with IPCC SRES A1B scenario 1990–2070 sulfur emissions, from Nakićenović et al. (2000)], where the simulations have SO2 emissions appropriate for 70 yr ahead of the actual model time (see Fig. 6a) and extend on into the SRES A1B scenario for the future; these runs contain 519 models, which simulated 1920–2000 greenhouse gas and natural forcings with future SO2 emissions.
Fig. 6. (a) The total anthropogenic SO2 emissions for the STR (red) and AATR (black) ensembles. (b) The global mean radiative forcing relative to 1920–40, diagnosed for the standard model configuration for STR (red) and AATR (black). Solely to illustrate the relationships between the annual and decadal radiative forcing estimates, values are also shown (blue) for an alternative model (HadCM3) relative to 1860–80. (c) The implied hemispheric difference in radiative forcing (NH − SH) for STR (red) and AATR (black) for each decade.
Citation: Journal of Climate 24, 19; 10.1175/JCLI-D-11-00019.1
In this section, the focus will be on July–September (JAS) precipitation, as this accounts for 70% of the annual total in the Sahel and has changed significantly in the period of interest (Lau et al. 2006). For comparisons with observations, we use the CRU’s 0.5° × 0.5° precipitation dataset (New et al. 2000). We discard results prior to 1940 as there is some evidence that the models were still spinning up. Due to the distributed computing nature of this experiment, we were able to include data from runs that did not have a full dataset uploaded from the volunteer. To account for any missing model data, a “climatology” was calculated for each run, as the parameter combinations led to some simulations being systematically wetter (or drier) than others. From this a precipitation anomaly can then be calculated relative to the climatology of each ensemble member, which can then be combined with available model data to produce an ensemble of anomalies.
From the slab experiment we would expect models with high initial sulfate loadings to show a smaller-magnitude Sahel precipitation response than in those with low initial sulfate loadings.
b. Results
Figure 6a shows the total sulfur emissions in the AATR (black lines) and STR (red lines) runs. The STR ensemble starts with much lower sulfur emissions than in AATR (approximately a factor of 5) and emissions rapidly increased, between 1940 to 1980, by a factor of 5. With AATR, sulfur emissions are offset 70 yr into the future so that sulfur emissions for the same period in AATR correspond to 2010–50, while all other climate forcings are identical to STR. For comparison with the slab experiments, note that for the 1940–80 period, the AATR ensemble uses SO2 emissions, which lie closer to the 
Using diagnosed, modeled ΔTOA and ΔT, and a λ value representative of the standard model configuration (λ = 1.1 W m−2 K−1), the temporal estimate of the radiative forcing of the standard transient simulation can be found. Forcing estimates calculated in this way can be seen as indicative of forcing across the ensemble; though some variation in magnitude could be expected [intermodel forcing (Q) differences are expected to be smaller than differences in the magnitude of climate responses (λ) to these forcings]. This approach has become widely used since it was first demonstrated by Forster and Taylor (2006). The radiative forcing implied for both experiments, estimated using this method, is shown Fig. 6b. The estimates are decadal only (a number of the required diagnostics are only available for decadal means). To illustrate how decadal radiative forcing relates to the annual mean forcing [using Eq. (2)], we have also shown decadal and annual values for a higher-resolution (HadCM3) configuration of this model (as used in Collins et al. 2006) forced by the same emissions as STR, plus the inclusion of background DMS. This shows for example how the radiative forcing associated with individual large explosive eruptions (such as the Agung eruption in 1963) can have significant impacts on the decadal estimates (as can also be seen in STR and AATR).
The impacts of the different aerosol loadings on the radiative forcings in the STR and AATR ensembles are clearly evident (Fig. 6). The STR ensemble has a more negative global forcing relative to AATR, which is consistent with the inferred response from the more rapid increase in sulfur emissions from relatively clean background levels. This estimate corroborates the slab experiment findings presented in section 2.
To do this, we need to make a few assumptions. We assume that there is no net flux of energy between hemispheres and that the λ global parameter is representative for the two hemispheres. We do not expect the first of these assumptions to hold in practice as energy is likely to flow from the warmer hemisphere, and hence the values derived are likely to underestimate the true hemispheric forcing somewhat. Nevertheless, the contrasting hemispheric forcing estimates derived using this method are likely to be indicative of the sign of the hemispheric difference (even if they may underestimate the magnitude).
The differences in the hemispheric radiative forcing calculated in this way are illustrated in Fig. 6c. The hemispheric difference in forcing is significantly more pronounced in the STR ensemble relative to AATR. This ties in with the results of the slab model experiments (section 2) and the discussion of the sulfur emission differences shown above. The consequences of this hemispheric forcing difference are likely to lead to a weaker warming trend in the Northern Hemisphere relative to the Southern Hemisphere, which is expected to weaken the northern migration of the ITCZ, particularly in STR.
The precipitation trend in an individual model will depend on both the forced response and on the natural internal variability represented within the model. Although we may expect the difference in hemispheric warming to lead to a drying signal in the Sahel (on the basis of the slab simulations), this signal in any full GCM may be more than offset (or enhanced) by natural variability (predominately driven by ocean processes). Particularly in a region such as the Sahel, which is prone to high coupled internal variability, it is often difficult if not impossible to tease apart the relative impacts of the forced response from a natural mode of variability. The value of the ensemble presented within this section is that we can look at the precipitation response across a large number of models. Looking at the characteristics of the ensemble as a whole enables us to identify the influence of common drivers over the impact of internal variability.
Looking first at the time period during which the historical SO2 emissions have increased at the most consistent rate (1940–80; see Fig. 6a), we investigate how the range of trends across the model ensembles compares to the observed rainfall changes. The distribution of modeled precipitation changes during this period is illustrated in Fig. 7a. The impacts of modeled process uncertainty and, perhaps more importantly, the internal climate variability are evident in the widths of the two model distributions.
The relative frequencies of the mean precipitation trends in the AATR (black) and STR (red) ensembles for the period are shown for (a) 1940–80 and (b) 1950–80. The horizontal axis indicates the decadal trend in mean rainfall (mm month−1) for JAS for the region 10°–18°N, 20°W–22°E. The observational estimate for the same period and region is marked by the blue cross. (c) Also plotted is the observational (CRU) estimate (black diamonds) for the mean precipitation change over the same region (mm month−1). The 1940–80 and 1950–80 observational trends indicated by the blue crosses in (a) and (b) is overlaid in blue.
Citation: Journal of Climate 24, 19; 10.1175/JCLI-D-11-00019.1
Differences that arise from the aerosol representation within these two ensembles are evident from the difference in the mean ensemble responses. The vast majority, although not all, of the STR ensemble produced a drying signal during this period. In contrast the AATR ensemble produces both drying and moistening responses, but no net signal emerges across the ensemble. The observed trend based on the CRU data (Fig. 7c) is also plotted in this same figure. Although a small fraction of the AATR simulations reproduce the changes, it is the large number of STR simulations that lie close to the CRU trend. Indeed, the trend in the median STR ensemble response lies close to the observed response, suggesting that much of the long-term (1940–80) change in rainfall can be explained by forced anthropogenic aerosol changes rather than a mode of internal variability.
The largest period of rainfall change, however, occurred as a subset of this longer period. The 1950s were moister in the Sahel, and the drying trend to the 1980s was more dramatic (Fig. 7c). Figure 7b illustrates the rainfall trends simulated by the ensembles for the 1950–80 period. As was the case earlier, the drying trend is evident across most of the STR ensemble, again in contrast to AATR. The shorter time period for the trends (30 yr compared to 40 in Fig. 7a) results in broader model distributions due to the larger variability on this shorter time scale. Also marked in Fig. 7b are the observed rainfall changes. Unlike the longer 1940–80 time period, the CRU observations lie in the tails of both distributions, with only a small fraction of models from either ensemble reproducing the observed trend. This suggests that a relationship between aerosols and rainfall change found in the shorter 1950–80 period is not as straightforward as the longer underlying 1940–80 change. We can now say that the observed aerosol changes are likely to have contributed to the drying from the 1950s to the 1980s although there remains a substantial fraction of the observed drying that is not explained by the forced response within this modeling framework.
There are a number of possible interpretations for the observational trend lying in the tails of the histogram of modeled trends. Perhaps the simplest of which is that the 1950–80 drying trend was a rare event. In other words, a mode (or modes) of natural variability within the climate system conspired to enhance the 1950–80 drying signal. There is well-established existing literature that supports a role for natural processes to drive the Sahel variability. Knight et al. (2006) and Ting et al. (2009), for example, argue that the Sahel rainfall changes respond to the Atlantic multidecadal oscillation (AMO) through changes in Atlantic SSTs. If we assume that the model captures the correct processes, then the interpretation that a rare mode of natural variability also contributed to the drying trend is consistent with the small fraction of modeled trends agreeing with the observed drying (Fig. 7b).
An alternative interpretation, however, is that the model does not capture the full strength of the precipitation response to the underlying climate drivers. It has previously been suggested by Lau et al. (2006) and Biasutti and Giannini (2006) that HadCM3 (the higher ocean resolution version of this model) reproduces only a weak Sahel response to SST anomalies in the Indian and Atlantic Oceans. Indeed, a weak SST response appears to be a general limitation of current climate models driven by observed SST changes (Scaife et al. 2009). The implication of this would be that the model may underestimate the magnitude of the rainfall changes as a response to either natural variability in SSTs or from SST changes resulting from the forced response. If the model is underrepresenting the rainfall impacts from natural variability, then this would suggest that the width of the distributions of trends (in Figs. 7a and 7b) is too small and hence natural variability is more likely to explain the Sahel 1950–80 drying. If the model is underrepresenting the rainfall impacts from anthropogenically forced SSTs, then this might explain why the ensemble underestimates the 1950–80 drying signal. If we accept this interpretation, a similar experiment with a model, which was able to capture stronger coupling, may conclude that the 1950–80 drying trend is much more consistent with the forced response. We are not able to resolve this issue here, and this remains a question for the future literature.
A further possibility is that the model may fail to fully capture the strength of some important local feedbacks. These could operate on any time scale from days to years, and might include, for example, dust aerosol emissions (Yoshioka et al. (2007)), soil moisture feedbacks, or vegetation feedbacks (Los et al. (2006)). If this is the case, then improved modeling of these processes would also have the effects of broadening the distributions of trends (due to enhanced natural variability) and enhancing the response to anthropogenically forced SST changes.
4. Conclusions
Simulations with a slab-ocean model have shown that NH tropical precipitation is sensitive to both greenhouse gas and sulfate aerosol forcing. Increases in greenhouse gases cause an increase in precipitation and increases in aerosol loading cause a reduction. We suggest that a hemispherical asymmetry in temperature change leads to a shift in the seasonal migration of the ITCZ.
When the sulfate and greenhouse gas forcings are combined, tropical precipitation depends both on the relative magnitude of the imposed forcings and on the location of the ITCZ. For the sorts of combinations of aerosol and greenhouse gas forcings characteristic of the second half of the twentieth century (such as the 1980s), a reduction in precipitation is seen. The model suggests that, historically, aerosol changes dominate. In contrast, when simulating the change between 1980 and 2050, the future Sahel rainfall increases because the future change is influenced more by greenhouse gas concentrations.
The slab model simulations presented here enable us to identify the model response to changes in atmospheric composition within a framework that minimizes atmospheric variability by explicitly neglecting major modes of ocean internal variability (and using multiyear averages). This enables us to identify hemispheric warming differences as a driver of ITCZ position and seasonal rainfall as an important physical mechanism (within this model structure). The slab experiment presented here provides a conceptual framework against which we can understand the modeled Sahel precipitation changes in ensembles of fully coupled ocean–atmosphere GCMs.
In light of the slab model results, it is not surprising that, in a large fully coupled ocean–atmosphere GCM ensemble simulation (STR) of past climate, we see a fall in Sahel precipitation. This reduction may be attributed to hemispheric forcing differences associated with large increases in historic NH, anthropogenic SO2 emissions over the period, ameliorated by the effect of a rise in greenhouse gas emissions. This drying signal is not evident in a second large ensemble (AATR) with identical climate forcings apart from the sulfur emissions.
The differences between historical rainfall trends in the two ensembles illustrate the important role that aerosol changes are likely to have played in the Sahel drought.
Previous studies have also suggested that aerosol changes may have played an important role in governing Sahel precipitation, but the small number of models available in these studies and the large variability inherent in this region has made it difficult to link the observed changes either to variability or the forced response. In this paper we have shown that representing the historical aerosol changes within the models drives a consistent drying trend across most of the ensemble. The absence of this signal in the ensemble with the altered sulfur emissions emphasizes the importance of the historical aerosol changes as part of our understanding of the observed drying trend.
The unique perturbed physics, boundary forcing, and initial condition ensemble approach used here allow us to confirm that historical SO2 emissions are likely to explain most of the 1940–80 rainfall changes and a significant proportion of the more pronounced 1950–80 drying. The use of a perturbed-physics ensemble means that our conclusions are not reliant on any single version of the Hadley Centre climate model, but they do, of course, rely on the underlying model structure. If this structure correctly represents the relevant processes in the real world, it is very unlikely that the observed drought was entirely due to natural climate variability.
Acknowledgments
This work was funded by the Natural Environmental Research Council Grant NER/T/S/2001/00871 and E-Science Grant NER/S/G/2003/1193, and enabled by the climateprediction.net team, particularly Tolu Aina, Carl Christensen, and Nick Faull. The work undertaken by Ben Booth and David Rowell was funded by the Joint DECC, Defra, and MoD Integrated Climate Program (GA01101 CBC/2B/0417_Annex C5), and (for David Rowell) the DFID Climate Science Research Programme. We are also extremely grateful for collaboration with the BBC, to all the volunteers who donated computer time to this project, and the Rutherford Appleton Laboratory for use of their supercomputing capabilities. Christopher Knight helped with statistical analyses and Mark New supplied the CRU data.
REFERENCES
Ackerley, D., E. J. Highwood, and D. J. Frame, 2009: Quantifying the effects of perturbing the physics of an interactive sulfur scheme using an ensemble of GCMs on the climateprediction.net platform. J. Geophys. Res., 114, D01203, doi:10.1029/2008JD010532.
Ackerman, A. S., O. B. Toon, and P. V. Hobbs, 1995: Numerical modeling of ship tracks produced by injections of cloud condensation nuclei into marine stratiform clouds. J. Geophys. Res., 100 (D24), 7121–7133.
Albrecht, B., 1989: Aerosols, cloud microphysics and fractional cloudiness. Science, 245, 1227–1230.
Baines, P. G., and C. K. Folland, 2007: Evidence for a rapid global climate shift across the late 1960s. J. Climate, 20, 2721–2744.
Biasutti, M., and A. Giannini, 2006: Robust Sahel drying in response to late 20th century forcings. Geophys. Res. Lett., 33, L11706, doi:10.1029/2006GL026067.
Brohan, P., J. J. Kennedy, I. Harris, S. F. B. Tett, and P. D. Jones, 2006: Uncertainty estimates in regional and global observed temperature changes: A new dataset from 1850. J. Geophys. Res., 111, D12106, doi:10.1029/2005JD006548.
Collins, M., B. B. B. Booth, G. R. Harris, J. M. Murphy, D. M. H. Sexton, and M. J. Webb, 2006: Towards quantifying uncertainty in transient climate change. Climate Dyn., 27, 127–147, doi:10.1007/s00382-006-0121-0.
Cook, K. H., 2008: The mysteries of Sahel droughts. Nat. Geosci., 1, 647–648.
Cook, K. H., and E. K. Vizy, 2006: Coupled model simulations of the West African monsoon system: Twentieth- and twenty-first-century simulations. J. Climate, 19, 3681–3703.
Curry, J. A., J. L. Schramm, and E. E. Ebert, 1995: Sea ice–albedo climate feedback mechanism. J. Climate, 8, 240–247.
Feichter, J., E. Roeckner, U. Lohmann, and B. Liepert, 2004: Nonlinear aspects of the climate response to greenhouse gas and aerosol forcing. J. Climate, 17, 2384–2398.
Folland, C. K., T. N. Palmer, and D. E. Parker, 1986: Sahel rainfall and worldwide sea temperatures, 1901–85. Nature, 320, 602–607.
Folland, C. K., J. Owen, M. N. Ward, and A. Colman, 1991: Prediction of seasonal rainfall in the Sahel region using empirical and dynamical methods. J. Forecast., 10, 21–56, doi:10.1002/for.3980100104.
Forster, P. M. F., and K. E. Taylor, 2006: Climate forcings and climate sensitivities diagnosed from coupled climate model integrations. J. Climate, 19, 6181–6194.
Frame, D. J., and Coauthors, 2009: The climateprediction.net BBC climate change experiment: Design of the coupled model ensemble. Philos. Trans. Roy. Soc., 367A, 855–870, doi:10.1098/rsta.2008.0240.
Hagos, S. M., and K. H. Cook, 2008: Ocean warming and late-twentieth-century Sahel drought and recovery. J. Climate, 21, 3797–3814.
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, doi:10.1073/pnas.0509057102.
Hoerling, M. J., J. Hurrell, J. Eischeid, and A. Phillips, 2006: Detection and attribution of twentieth-century northern and southern African rainfall change. J. Climate, 19, 3989–4008.
Houghton, J. T., Y. Ding, D. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson, Eds., 2001: Climate Change 2001: The Scientific Basis. Cambridge University Press, 892 pp.
Hulme, M., and M. Kelly, 1993: Exploring the links between desertification and climate change. Environment, 5, 5–11 , 39–45.
Janicot, S., V. Moron, and B. Fontaine, 1996: Sahel droughts and ENSO dynamics. Geophys. Res. Lett., 23, 515–518.
Jones, A., D. Roberts, M. Woodage, and C. Johnson, 2001: Indirect sulphate aerosol forcing in a climate model with an interactive sulphur cycle. J. Geophys. Res., 106 (D17), 20 293–20 310.
Jung, T., L. Ferranti, and A. M. Tompkins, 2006: Response to the summer of 2003 Mediterranean SST anomalies over Europe and Africa. J. Climate, 19, 5439–5454.
Kawase, H., M. Abe, Y. Yamada, T. Takemura, T. Yokohata, and T. Nozawa, 2010: Physical mechanism of long-term drying trend over tropical North Africa. Geophys. Res. Lett., 37, L09706, doi:10.1029/2010GL043038.
Knight, J. R., C. K. Folland, and A. A. Scaife, 2006: Climate impacts of the Atlantic Mulitdecadal Oscillation. Geophys. Res. Lett., 33, L17706, doi:10.1029/2006GL026242.
Kristjánsson, J. E., T. Iversen, A. Kirkevåg, and Ø. Seland, 2005: Response of the climate system to aerosol direct and indirect forcing: Role of cloud feedbacks. J. Geophys. Res., 110, D24206, doi:10.1029/2005JD006299.
Lamb, P., 1978: Large-scale tropical Atlantic surface circulation patterns associated with sub-Saharan weather anomalies. Tellus, 30, 240–251.
Lau, K. M., S. S. P. Shen, K.-M. Kim, and H. Wang, 2006: A multimodel study of the twentieth-century simulations of Sahel drought from the 1970s to 1990s. J. Geophys. Res., 111, D07111, doi:10.1029/2005JD006281.
Los, S. O., G. P. Weedon, P. R. J. North, J. D. Kaduk, C. M. Taylor, and P. M. Cox, 2006: An observation-based estimate of the strength of rainfall–vegetation interactions in the Sahel. Geophys. Res. Lett., 33, L16402, doi:10.1029/2006GL027065.
Lu, J., 2009: The dynamics of the Indian Ocean sea surface temperature forcing of Sahel drought. Climate Dyn., 33, 445–460, doi:10.1007/s00382-009-0596-6.
Nakićenović, N., and Coauthors, 2000: Special Report on Emissions Scenarios: A Special Report of Working Group III of the Intergovernmental Panel on Climate Change. Cambridge University Press, 599 pp.
New, M., M. Hulme, and P. Jones, 2000: Representing twentieth-century space–time climate variability. Part II: Development of 1901–96 monthly grids of terrestrial surface climate. J. Climate, 13, 2217–2238.
Palmer, T., 1986: Influence of the Atlantic, Pacific and Indian Oceans on Sahel rainfall. Nature, 322, 251–253.
Romanou, A., B. Liepert, G. A. Schmidt, W. B. Rossow, R. A. Ruedy, and Y. Zhang, 2007: Twentieth-century changes in surface solar irradiance in simulations and observations. Geophys. Res. Lett., 34, L05713, doi:10.1029/2006GL028356.
Rotstayn, L. D., and U. Lohmann, 2002: Tropical rainfall trends and the indirect aerosol effect. J. Climate, 15, 2103–2116.
Rotstayn, L. D., B. F. Ryan, and J. E. Penner, 2000: Precipitation changes in a GCM resulting from the indirect effects of anthropogenic aerosols. Geophys. Res. Lett., 27, 3045–3048.
Rowell, D., 2001: Teleconnections between the tropical Pacific and the Sahel. Quart. J. Roy. Meteor. Soc., 127A, 1683–1706.
Rowell, D., 2003: The impact of Mediterranean SSTs on the Sahelian rainfall season. J. Climate, 16, 849–862.
Ruosteenoja, K., T. R. Carter, K. Jylha, and H. Tuomenvirta, 2003: Future climate in world regions: An intercomparison of model-based projections for the new IPCC emissions scenarios. The Finnish Environment 644, Finnish Environment Institute, 83 pp.
Scaife, A. A., and Coauthors, 2009: The CLIVAR C20C project: Selected twentieth century climate events. Climate Dyn., 33, 603–614, doi:10.1007/s00382-008-0451-1.
Stainforth, D. A., and Coauthors, 2005: Uncertainty in predictions of the climate response to rising levels of greenhouse gases. Nature, 433, 403–406.
Stanhill, G., and S. Cohen, 2001: Global dimming: A review of the evidence for a widespread and significant reduction in global radiation with a discussion of its probable causes and possible agricultural consequences. Agric. For. Meteor., 107, 255–278.
Sutton, R. T., and D. Hodson, 2005: Atlantic Ocean forcing of North American and European summer climate. Science, 309, 115–118, doi:10.1126/science.1109496.
Sutton, R. T., B. Dong, and J. M. Gregory, 2007: Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparisons with observations. Geophys. Res. Lett., 34, L02701, doi:10.1029/2006GL028164.
Ting, M., Y. Kushnir, R. Seager, and C. Li, 2009: Forced and internal twentieth-century SST trends in the North Atlantic. J. Climate, 22, 1469–1481.
Twomey, S. A., 1977: The influence of pollution on the shortwave albedo of clouds. J. Atmos. Sci., 34, 1149–1152.
Vizy, E., and K. Cook, 2001: Mechanisms by which Gulf of Guinea and eastern North Atlantic sea surface temperature anomalies can influence African rainfall. J. Climate, 14, 795–821.
Williams, K. D., A. Jones, D. L. Roberts, C. A. Senior, and M. J. Woodage, 2001: The response of the climate system to the indirect effects of anthropogenic sulfate aerosol. Climate Dyn., 17, 845–856.
Yoshioka, M., N. M. Mahowald, A. J. Conley, W. D. Collins, D. W. Fillmore, C. S. Zender, and D. B. Coleman, 2007: Impact of desert dust radiative forcing on Sahel precipitation: Relative importance of dust compared to sea surface temperature variations, vegetation changes, and greenhouse gas warming. J. Climate, 20, 1445–1467.
Zhang, Z. B., F. W. Zwiers, G. C. Hegerl, F. H. Lambert, N. P. Gillet, S. Solomon, P. A. Stott, and T. Nozawa, 2007: Detection of human influence on twentieth-century precipitation trends. Nature, 448, 461–465.
