1. Introduction
Reliable regional-scale climate change prediction for marginal regions such as the Sahel is urgently needed to support adaptation planning. The purpose of this study is to provide state-of-the-art projections of summer surface temperature, precipitation, and surface moisture changes during this century due to increased atmospheric CO2, accompanied by an evaluation of confidence in the projections. Regional climate model (RCM) simulations at 30 and 90 km are analyzed along with output from coupled atmosphere–ocean global circulation models (AOGCMs) from phase 5 of the Climate Model Intercomparison Project (CMIP5) prepared in advance of the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5). Both types of models are used to take advantage of the strengths of each. The regional model simulations capture the West African climate more realistically than the AOGCMs, provide finer resolution needed for resolving intense rainfall events and topography, and offer information on space scales that are relevant for regional impacts analysis and planning. AOGCMs, on the other hand, provide global connectivity and projections of changes in SSTs that are known to influence the Sahel climate.
Model projections are presented for both the middle (2041–60) and late century (2081–2100) under the representative concentration pathway 8.5 (RCP8.5) emissions scenario from the IPCC. Confidence in the model projections is evaluated by examining differences among various models, agreement in ensemble members run using the same model, comparison of the results for the two different time periods, and an analysis of the physical processes of change.
The background section that follows provides a review of our current understanding of how Sahel climate might change in the coming decades. A description of the regional climate model simulations and the AOGCMs used is provided in section 3, and the ability of the models to simulate the current West African monsoon system is evaluated in section 4. Predictions are presented in section 5, and confidence in those predictions is discussed in section 6. Section 7 contains a summary and the conclusions.
2. Background
While the IPCC Fourth Assessment Report (AR4) model simulations (Meehl et al. 2007) projected significantly warmer surface air temperatures throughout Africa in the future, the magnitude and spatial distribution of this expected warming and its potential effects on rainfall need to be robustly established. This is especially important over the Sahel where surface temperature gradients influence the low-level monsoon circulation and hence affect rainfall. While annual and June–August multimodel IPCC AR4 precipitation projections (Meehl et al. 2007) predict rainfall to increase over much of the Sahel east of 0° longitude by late century, this increase in many locations is only observed to occur in 75% or less of the 21 models analyzed. Cook (2008) demonstrates that this uncertainty is likely overstated, as most models simulate small changes in rainfall over the Sahel and that the projected increases are likely impacted by a few outlier models.
Since IPCC AR4 many studies have been conducted to further our understanding of how climate is likely to change over this region, many of which are highlighted in Druyan's (2011) review article. Discussed below are some of the more relevant and recent studies.
By analyzing the CMIP3 AOGCMs, Biasutti et al. (2009) better understand the intermodel differences in rainfall projections and find that they cannot solely be explained in terms of differences in the global SST patterns between models. They suggest that the Saharan thermal low trough is associated with the differences among the models. Models that simulate a weak Saharan low tend to predict drying, and vice versa. They postulate that direct radiative forcing is important for the development of the thermal low trough, but it is still unclear how regional land–atmosphere interactions feed back to the thermal low's development and further impact rainfall. Skinner et al. (2012) also confirm a strong relationship between thermal low trough and the end of century Sahelian rainfall in their AGCM study with enhanced rainfall associated with a stronger thermal low trough.
Mariotti et al. (2011) use the Regional Climate Model version 3 (RegCM3) driven by an individual AOGCM for the twenty-first century under the A1B forcing scenario. Both regional and global projections predict a gradual increase of surface temperatures and a maximum warming over Sahara during boreal summer. Both models predict drying over the second half of the century over the western Sahel, while rainfall projections differ over the eastern Sahel, with no clear long-term trend in the AOGCM and a decreasing trend similar to that in western Sahel in the regional model.
Patricola and Cook (2010, 2011) generate a late-century ensemble of projections using the National Center for Atmospheric Research (NCAR) Weather Research and Forecasting model (WRF) under the A2 emission scenario. Nine AOGCMs are used to derive the future lateral boundary conditions. Instead of directly applying the conditions from the AOGCM in a traditional downscaling approach, future anomalies are calculated from the AOGCMs and applied to present-day conditions derived from available reanalyses to reduce the transfer of AOGCM errors into the regional model. Both regional and global simulations predict a significant warming by the end of the twenty-first century. Precipitation is projected to increase during July–August over the Sahel and decrease along the Guinean Coast during June–August. The wetter late-summer conditions over the Sahel are found to be associated with enhanced moisture transport by the low-level westerly flow including the West African westerly jet.
Paeth et al. (2011) compare annual rainfall projections for the first half of the twenty-first century generated by various RCMs as part of the Ensemble-Based Predictions of Climate Changes and their Impacts (ENSEMBLES) project. Coherent spatial patterns across the models are limited as are statistically significant changes. The multimodel mean predicts general drying over most of West Africa and an increase in rainfall over southeastern Niger and southern Chad.
Analyzing 12 AOGCMs from phase 3 of CMIP (CMIP3) under the A1B scenario, Fontaine et al. (2011) and Monerie et al. (2013) find a positive (negative) rainfall trend over the eastern-central (western) parts of the Sahel. Rainfall excesses expected in central Sahel are linked to an enhancement of the northern Hadley-type cell and a monsoon flux strengthening due to increased near-surface temperature and evaporation. Rainfall deficits predicted westward are associated with a strengthening of the African easterly jet and zonal circulation anomalies between the Indian and the Atlantic Oceans, favoring subsidence and moisture transport from the region. Using a multimodel approach and the “one model one vote” concept applied on eight CMIP5 AOGCMs under the RCP4.5 emission scenario and eight CMIP3 AOGCMs under the A1B forcing scenario, Monerie et al. (2012) confirm this expected dipole anomaly pattern in CMIP5 simulations and find an enhancement of the meridional surface temperature gradient linked to a strong warming over Sahara.
Diallo et al. (2012) analyze a set of four RCMs driven by two AOGCMs under the A1B forcing scenario to assess the mid-twenty-first-century climate over West Africa. Despite all models predicting warmer conditions over northern Africa, the magnitude and spatial distribution of predicted surface temperatures appears to be model dependent. Almost all models predict a significant decrease of rainfall off the western African coast and over the western Sahel associated with stronger warming found there, while southward and eastward projections are sensitive to both the AOGCM used to drive regional simulations and the RCMs themselves.
Most if not all of the modeling studies discussed above do not account for the impact of land cover changes on the future projections. Paeth and Thamm (2007) and Paeth et al. (2009) demonstrate the potential impact of accounting for land cover changes in their regional climate modeling studies for the first half of the twenty-first century. Without accounting for land cover changes, rainfall is projected to increase slightly over the Sahel; but not so when land changes are factored in. That being said, the sensitivity of the results to land cover changes could be model dependent (Wang and Alo 2012).
The studies discussed above highlight the complexity of the challenges faced by the scientific community when trying to make future climate predictions. Sensitivities to model parameterizations and/or choices of boundary condition forcings for regional models (e.g., Seth and Rojas 2003; Moufouma-Okia and Rowell 2010; Mariotti et al. 2011) can yield diverging projections. Coupled AOGCMs have difficulties properly simulating tropical Atlantic SSTs (e.g., Breugem et al. 2006; Richter and Xie 2008), which is particularly vexing for predicting climate variability over West Africa as the monsoon system is strongly tied to Atlantic SSTs (e.g., Folland et al. 1986; Druyan 1991; Rowell et al. 1995; Vizy and Cook 2001, 2002). This study attempts to bridge this gap of uncertainty by focusing on understanding the physical processes associated with the climate response and how they are represented across different models in hopes of producing a more reliable prediction of how the Sahel climate is likely to change.
3. Model description and simulation design
The RCM simulations are designed to provide regional-scale information about the African climate and how it is projected to change. To achieve this objective a one-way nesting methodology is applied using WRF (Skamarock et al. 2005). A 90-km horizontal resolution outer domain, shown in Fig. 1, is chosen such that it encompasses all of Africa and most of the adjacent Atlantic and Indian Oceans. The value of 90 km is selected for use because previous studies (Patricola and Cook 2010, 2011; Cook and Vizy 2012; Vizy and Cook 2012) demonstrate that the model can realistically simulate the evolution of the climate at this resolution. Furthermore, it is close to the resolution of most of the available CMIP5 AOGCMs to foster a more direct comparison. A large domain is selected so the lateral boundaries are placed far away from Africa to minimize the effects of their constraints in the analysis region and on the development of the subtropical anticyclones over the adjacent oceans.
The 90-km model domain and mean July–September SSTAs (K) for the (a) MID21 (2041–60) and (b) LATE21 (2081–2100) simulations. Boxes denote position of the nested 30-km domain.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
To better resolve regional-scale processes a 30-km horizontal resolution domain centered over Africa is nested inside this 90-km domain. The placement of this domain is denoted by the bolded box shown in Fig. 1. Lateral boundary conditions for this domain are derived from the 90-km domain, but the 30-km resolution solution is not fed back to the coarser domain. This is done so the influence of spatial resolution on the projections can be directly evaluated.
Physical parameterizations are the same for both domains, and include the Yonsei University planetary boundary layer (Hong et al. 2006), Monin–Obukhov surface layer, new Kain–Fritsch cumulus convection (Kain 2004), Purdue–Lin microphysics (Chen and Sun 2002), Rapid Radiative Transfer Model (RRTM) longwave radiation (Mlawer et al. 1997), Dudhia shortwave radiation (Dudhia 1989), and the unified Noah land surface model (Chen and Dudhia 2001). These parameterizations are chosen based on studies that demonstrate they can produce a realistic representation of the African climate (Cook and Vizy 2012; Vizy and Cook 2012). The model time step is 3 min for the 90-km domain, and 1 min for the 30-km domain. Model output is saved every 3 h for analysis.
Three integrations are run and analyzed. The first simulation is a control (CTL) and represents conditions for 1989–2008. Initial, lateral, SSTs, and surface boundary conditions are derived from the 6-hourly National Centers for Environmental Prediction (NCEP)/Department of Energy Global Reanalysis 2 (NCEP-2; Kanamitsu et al. 2002) with spatial resolution of 2.5° for upper-air data and approximately 1.875° for surface fields. The model is initialized on 0000 UTC 15 March 1988 using NCEP-2 and is run through 0000 UTC 1 January 2009 with lateral boundary conditions on the 90-km domain being updated every 6 h from NCEP-2. The first 292 days are discarded as model spinup allowing the land surface conditions to adjust while the remaining 20 years are analyzed for both domains. The atmospheric CO2 concentration is held fixed at 367 ppmv, which is the 1989–2008 20-yr average observed at the Mauna Loa Observatory. Both domains have 32 vertical levels and the top of the atmosphere set at 20 hPa.
The second integration (MID21) represents mid-twenty-first-century (2041–60) conditions under the IPCC AR5 RCP8.5 scenario. This is a business-as-usual scenario with the largest greenhouse gas emission increases out of the four IPCC AR5 scenarios. The atmospheric CO2 concentration is increased to 546 ppmv, the 2041–60 average for the RCP8.5 scenario. Effects of other aerosols and greenhouse gases besides water vapor are not included. Boundary conditions for the MID21 simulation are derived from CMIP5 AOGCM simulations and applied as anomalies to the reanalysis boundary conditions used to force the CTL. This differs from a traditional downscaling approach where the AOGCM is used to directly force the RCM. It is done to improve the quality of the simulation by reducing the amount of AOGCM error introduced into the RCM framework that could degrade the RCM simulation (e.g., Seth and Rojas 2003; Mariotti et al. 2011). Instead of contributing a direct downscaling of the CMIP5 AOGCM results, these RCM simulations are providing a prediction that is more independent of the AOGCMs and, therefore, more valuable in assessing confidence through multimodel comparisons. It has been demonstrated (e.g., Patricola and Cook 2010, 2011, 2013a,b; Cook and Vizy 2012; Vizy and Cook 2012) as an effective way to evaluate future climate change. These future anomalies are calculated as differences between monthly-mean, RCP8.5-forced simulations averaged over 2041–60 and the CMIP5 AOGCM historical simulations averaged over 1986–2005, the last 20 years available for the historical experiment. The monthly-mean anomalies are assumed to represent the middle of the month value, and then are linearly interpolated to generate anomalies every 6 h over the annual cycle. These 6-hourly anomalies for the annual cycle are then interpolated to the NCEP-2 grid and added to the 6-hourly boundary conditions of the CTL generated from the NCEP-2.
Output from five CMIP5 AOGCMs are used to generate the future anomalies by formulating a multimodel ensemble mean difference. Multimodel mean anomalies are used to reduce the dependence of the projections on individual AOGCMs. The models used include the NCAR Community Climate System Model, version 4 (CCSM4); the Centre National de Recherches Météorologiques Coupled Global Climate Model, version 5 (CNRM-CM5); the Geophysical Fluid Dynamics Laboratory Climate Model version 3 (GFDL CM3); the Model for Interdisciplinary Research on Climate version 5 (MIROC5); and the Meteorological Research Institute Coupled GCM, version 3 (MRI-CGCM3). These models were selected for use based upon their availability of output at the time of running the RCM simulations. Table 1 lists the AOGCMs and their spatial resolutions.
List of models, their spatial resolutions, and mid- and late-century 925-hPa height scaling utilized.
The MID21 simulation is initialized on 0000 UTC 15 March 2040 and is run through 0000 UTC 1 January 2061. Again both domains have 32 vertical levels and the top of the atmosphere set at 20 hPa. MID21 soil moisture initialization values are the same as those used in the CTL initialization and the model is spun up for 292 days before analyzing the results to allow the land surface conditions time to adjust to the new climate state. The remaining 20 years of 3-hourly output are analyzed for both the 90-km and 30-km domains.
Figure 1a shows the MID21 July–September mean SST anomalies (SSTAs). Over most of the tropical Atlantic and Indian Ocean temperatures increase by 1–1.5 K. Warming is weaker in the North Atlantic north of 40°N in the deep water formation region.
The final integration represents conditions at the end of the twenty-first century (2081–2100) and is referred to as LATE21. The atmospheric CO2 concentration is increased to the 2081–2100 average value of 850 ppmv from the RCP8.5 scenario. Figure 1b shows the LATE21 annual mean SSTAs. Tropical ocean temperatures warm by 2.5–3.5 K by the late twenty-first century.
Initial and lateral boundary conditions are derived in the same manner as they were for MID21, but from using the RCP8.5 CMIP5 AOGCM anomalies averaged over 2081–2100. LATE21 is initialized on 0000 UTC 15 March 2080 and was intended to run through 0000 UTC 1 January 2101 using the same number of vertical levels (32) and top-of-the-atmosphere setting (20 hPa) as the CTL and MID21. However, the 30-km simulation began to develop model instabilities and crashed shortly after completing 2089. At this point the 30-km domain was deactivated, yielding only 9 years of integration to analyze at this resolution. A similar problem occurred for the 90-km domain after year 2092.
It turns out that model instabilities are related to the selection of the top of the atmosphere in the model. By the end of the century, convection in some instances becomes so strong that the equilibrium level for an unstably rising parcel in the upper troposphere is higher the top of the atmosphere prescribed in the RCM. Thus, the top of the atmosphere needs to be raised in the model to prevent the generation of model instabilities associated with the cumulus convective parameterization and the premature crashing of the simulation.
The remaining 8 years of the LATE21 90-km domain simulation are generated by running the model with the top of the atmosphere adjusted to 10 hPa. Changing this setting results in the need to increase the number of vertical levels to 33 levels. The simulation is restarted on 0000 UTC 19 November 2092 and run through 0000 UTC 1 January 2101. The first 43 days of the restarted simulation are discarded for spinup and the last 8 years are combined with the 12 years available prior to the model crashing to produce 20 years of model output for the end of the twenty-first century. A shorter spinup is appropriate because the land surface conditions (e.g., soil moisture and soil temperature) are retained from the end point of the first 12 years. Anomaly differences from the CTL for the first 12 years and the last 8 years are compared and found to be generally similar justifying the use of this approach.
4. Model evaluation
The quality of the two RCM CTL simulations and the five CMIP5 AOGCM historical simulations is assessed to evaluate how realistically the models can simulate the evolution of the boreal summer climate system over West Africa and the Sahel. A sensible representation of the important climate features is viewed as a necessary but not sufficient condition for a confident prediction.
Figures 2a and 2b show the observed average July–September Climatic Research Unit (CRU) surface air temperature (TS3.1) and European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Re-Analysis (ERA-Interim, hereafter ERA-I) 1989–2008 climatologies. Temperatures up to 310 K are observed over the northwestern Sahara and northern Sudan. Temperatures are also relatively warmer over the northern Sahel and southern Sahara across the continent. Equatorward of 12°N, temperatures decrease below 300 K. Thus a strong meridional temperature gradient sets up across the Sahel between 10° and 20°N.
Average July–September surface air temperature (K) from the (a) 0.5° resolution CRU TS 3.1 (1989–2008), (b) 1.5° resolution ERA-I (1989–2008), (c) RCM 30-km CTL, and (d) RCM 90-km CTL, and the CMIP5 AOGCM historical experiment (1986–2005) (e) CCSM4, (f) CNRM-CM5, (g) GFDL CM3, (h) MIROC5, and (i) MRI-CGCM3 simulations.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
The RCMs at 30 km (Fig. 2c) and 90 km (Fig. 2d) generally capture this pattern including the meridional temperature gradient across the Sahel, albeit farther south than in the observations. The warm maxima over the northwestern Sahara and northern Sudan are captured reasonably well both in terms of location and magnitude. Figures 2d–i show the summer surface air temperatures from five different CMIP5 AOGCMs. Each AOGCM captures the warm maxima over the Sudan. The meridional temperature gradient is also simulated, although it is displaced farther equatorward over the southern Sahel in most models.
The seasonal evolution of surface air temperatures over the Sahara is also evaluated, but is not shown for brevity. Results from this assessment indicate that the RCM at 90- and 30-km resolution is able to realistically simulate the observed temperature evolution over the Sahara. In contrast there is a wide spread in the AOGCMs with some models (CNRM-CM5 and MRI-CGCM3) systematically colder than the observations by 1–5 K from March to December, while other models (CCSM4 and GFDL CM3) are realistic during July–August but systematically cooler at other times. MIROC5 is able to simulate the seasonal evolution of Saharan temperature but is warmer by 2–3 K during July–September.
Figures 3a and 3b show the average July–September precipitation during the height of the West African monsoon season from the National Aeronautics and Space Administration (NASA) Tropical Rainfall Measuring Mission (TRMM) 3B42V6 and Global Precipitation Climatology Project (GPCP) v2.2 satellite-derived products, respectively. At this time the primary rainfall band over land is centered along 10°N with rainfall maxima over the Guinean Highlands (10°N, 13°W) and the Cameroon Highlands (7°N, 11°E).
Average July–September precipitation (mm day−1) from the (a) 0.25° resolution TRMM 3B42V6 product (1998–2010), (b) 2.5° resolution GPCP v2.2 (1989–2008), (c) RCM 30-km CTL, (d) RCM 90-km CTL, and the CMIP5 AOGCM historical experiment (1986–2005) (e) CCSM4, (f) CNRM-CM5, (g) GFDL CM3, (h) MIROC5, and (i) MRI-CGCM3 simulations. The boxes in (a) indicate the Guinean Coast (4°–7°N, 12°W–6°E) and Sahel (10°–13°N, 12°W–6°E) averaging regions used in the analysis.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
The RCM CTL runs at 30 km (Fig. 3c) and 90 km (Fig. 3d) capture this observed structure well, but with rainfall maxima larger than typically observed. Simulated rainfall rates over the Sahel north of 15°N are less than 2 mm day−1, which is lower than observed. For comparison, results from the 5 CMIP5 AOGCMs are shown in Figs. 3e–i. There is a wide spread in the ability of the AOGCMs as some models are able to simulate a generally realistic summer precipitation, structure such as the CCSM4 (Fig. 3e) and the MIROC5 (Fig. 3h), while others tend to produce too much rainfall over the Gulf of Guinea and/or fail to move the rainfall maxima northward over West Africa into the continental interior (Figs. 3f,g,i). MIROC5 appears to be systematically wetter over much of West Africa. While the RCM is also wetter than the observations, it is more spatially confined. CCSM4 does not capture the observed Cameroon Highland maximum well.
Figure 3 does not assess how well the models capture the seasonal evolution of rainfall over West Africa, including the transition of rainfall from the Guinean Coast into the continental interior in early summer and its equatorward retreat back to the Guinean Coast in early fall. To evaluate this evolution the climatological daily rainfall is area-averaged over two regions, the Guinean Coast (4°–7°N, 12°W–6°E) and the Sahel (10°–13°N, 12°W–6°E), smoothed using a 5-day running mean filter, and plotted in Fig. 4. The locations of these two averaging regions are denoted by the boxes in Fig. 3a. From these time series plots the onset of the Sahel rainy season can be simply defined as when the Sahel precipitation becomes greater than the Guinean Coast rainfall. Likewise the demise of the rainy season can be defined as when the Sahel rainfall falls below the Guinean Coast rainfall.
Average daily precipitation (mm day−1) area averaged over the Guinean Coast (gray) and the Sahel (black) regions for the (a) TRMM 3B42V6 product, (b) GPCP v2.2, (c) RCM 30-km CTL, (d) RCM 90-km CTL, and the CMIP5 AOGCM historical experiment (e) CCSM4, (f) CNRM-CM5, (g) GFDL CM3, (h) MIROC5, and (i) MRI-CGCM3 simulations.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
Figures 4a and 4b show the annual Guinean Coast and Sahel rainfall time series for the 1998–2010 TRMM 3B42V6 and 1997–2008 GPCP v2.2 products. The observations suggest an onset occurring around 5–8 July and a demise of the rainy season on 23 September.
Climatological daily rainfall time series plots for the RCM CTL at 30 and 90 km are shown in Figs. 4c and 4d, respectively. While the magnitudes of the rainfall are larger than the observations during the peak wet periods for each region, the timing of the simulated onset and demise agrees well with the observations. The higher rainfall magnitudes in the RCM may be associated with the assumptions made in the cumulus parameterization that are designed to reduce widespread light rainfall in marginally unstable environments (Kain 2004) and/or related to errors resulting from the interactions between the cumulus and planetary boundary layer schemes (e.g., Flaounas et al. 2011).
Figures 4e–i show the simulated results from the CMIP5 AOGCMs using the last 20 years of the historical experiment for each model. Overall, the majority of AOGCMs cannot properly simulate the onset and demise dates. CCSM4 (Fig. 4e) simulates an onset on 12 August, which is over a month later than the observed onset, and a demise 30 September, which is about 1 week later. There is only a short period of time at the beginning of September when Sahel rainfall is larger than the Guinean Coast rainfall in CNRM-CM5 (Fig. 4f). Likewise there is only a 4-day period in the beginning of June when the Sahel rainfall is greater than the Guinean Coast precipitation in GFDL CM3 (Fig. 4g). MIROC5 (Fig. 4h) simulates a realistic onset date of 8 July, but the demise is 11 days later than observations. Additionally, rainfall rates are systematically too wet during the summer in MIROC5. The Sahel rainfall never is greater than the Guinean Coast rainfall in MRI-CGCM3 (Fig. 4i).
Figure 5 evaluates the model's ability to simulate the July–September surface evaporation. Figures 5a and 5b show the surface evaporation from the ERA-I and NCEP-2. Evaporation rates are low over the Sahara, generally less than 1 mm day−1. Over the Sahel there is a strong meridional gradient in evaporation rates associated with the availability of moisture. The RCM (Figs. 5c,d) does capture this meridional gradient, but it is farther south than the observed position. Evaporation rates south of the Sahel east of 10°E are also slightly larger than in the reanalyses. The AOGCMs (Figs. 5d–i) also simulate this gradient, but its position in individual models varies depending on the model's ability to simulate the summer migration of rainfall northward into the Sahel (Fig. 4). It should be kept in mind that evaporation values from the reanalyses are highly model dependent and may not capture actual spatial details due to the coarseness of the resolution.
Average July–September evaporation rate (mm day−1) from the (a) 1.875° resolution NCEP-2 (1989–2008), (b) 1.5° resolution ERA-I (1989–2008), (c) RCM 30-km CTL, (d) RCM 90-km CTL, and the CMIP5 AOGCM historical experiment (1986–2005) (e) CCSM4, (f) CNRM-CM5, (g) GFDL CM3, (h) MIROC5, and (i) MRI-CGCM3 simulations.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
The ability of the models to simulate the summer low-level circulation is assessed by analyzing Fig. 6, which shows the July–September 925-hPa geopotential heights and winds for the ERA-I (Fig. 6a) and NCEP-2 (Fig. 6b), the RCM CTL at 90- (Fig. 6c) and 30-km (Fig. 6d) resolution, and the five CMIP5 AOGCMs (Figs. 6e–i).
July–September 925-hPa geopotential heights (m) and winds (m s−1) from the (a) 2.5° resolution NCEP-2 (1989–2008), (b) 1.5° resolution ERA-I (1989–2008), (c) RCM 30-km CTL, (d) RCM 90-km CTL, and the CMIP5 AOGCM historical experiment (1986–2005) (e) CCSM4, (f) CNRM-CM5, (g) GFDL CM3, (h) MIROC5, and (i) MRI-CGCM3 simulations.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
The dominant summertime height feature over northern Africa is a thermal low trough centered along approximately 20°N. The reanalyses (Figs. 6a,b) indicate the minimum 925-hPa heights range around 760–780 m. Equatorward of this trough is a meridional height gradient over the Sahel and Guinean Coast. Low-level flow is southerly over the Gulf of Guinea becoming southwesterly–westerly around 10°N on the southern flank of the thermal low. This includes westerly flow at approximately 10°N from the Atlantic into West Africa associated with the West African westerly jet (WAWJ; Grodsky et al. 2003; Pu and Cook 2010, 2012).
The RCM CTL 30-km (Fig. 6c) and 90-km (Fig. 6d) simulations also simulate 925-hPa minimum heights of the thermal low trough around 760–770 m, but the position of the trough is further south than in the reanalyses. The meridional height gradient is captured over the Sahel and Guinean Coast, although it is slightly stronger associated with the equatorward shift in position of the thermal low trough. The low-level flow agrees with the reanalyses, but the westerly–southwesterly flow near 10°N is stronger associated with the enhanced meridional height gradient relative to the reanalyses.
Not surprising, the CMIP5 AOGCMs that best simulate the summer rainfall distribution also simulate the low-level circulation and Saharan surface air temperature evolution most realistically. CCSM4 (Fig. 6e) and to some extent MIROC5 (Fig. 6h) capture all of the low-level circulation features fairly well, including the thermal low trough, the meridional height gradient over the Sahel and Guinean Coast, the WAWJ, and the southwesterly–westerly flow over the Sahel. Interestingly, the thermal low in MIROC5 is approximately 10–20 m deeper than observed and the meridional height gradient and southwesterly flow is stronger, consistent with increased convection simulated by this model relative to the observations. The thermal low trough is too strong in CNRM-CM5 (Fig. 6f) and too weak and located farther south in MRI-CGCM3 (Fig. 6i). The GFDL CM3 (Fig. 6g) simulates the appropriate depth for the thermal low trough, but the height gradient is more zonal and less meridional over the Sahel and Guinean Coast compared to the observations and is associated with reduced southwesterly flow.
Overall the RCM is able to realistically simulate the seasonal evolution of the precipitation over West Africa and the associated low-level circulation patterns, thus warranting using this model to predict how the Sahel climate will change in the future under increased greenhouse gas forcing.
5. Predictions
a. Surface air temperature
Figure 7 shows the projected midcentury July–September surface temperature anomalies from the RCM at 30 km (Fig. 7a) and 90 km (Fig. 7b) and the five CMIP5 AOGCMs (Figs. 7c–g). At midcentury, northern Sahel–southern Saharan surface temperatures are projected to warm by 2.5–3.5 K with the strongest warming positioned over northern Chad and Sudan in the RCM. This warming diminishes equatorward, with temperatures increasing only 1–2 K south of 15°N, resulting in an intensification of the meridional temperature gradient over the Sahel (see Fig. 2). The AOGCMs also predict warming over the Sahel/Sahara ranging from around 2 K in CNRM-CM5 (Fig. 7d) and MRI-CGCM3 (Fig. 7e) to over 4 K in the GFDL CM3 model (Fig. 7e). Only the CNRM-CM5 (Fig. 7d) and MRI-CGCM3 (Fig. 7g) simulations do not increase the meridional temperature gradient. The response is consistent for the entire July–September period for all models, with little month-to-month variations in the anomalies.
Midcentury (2041–60) July–September surface air temperature (K) anomalies for the RCM (a) 30- and (b) 90-km simulations, and the CMIP5 RCP8.5 (c) CCSM4, (d) CNRM-CM5, (e) GFDL CM3, (f) MIROC5, and (g) MRI-CGCM3 AOGCM simulations.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
In the late-century simulations, the projected summertime temperature increases are enhanced. Figures 8a and 8b show the predicted temperature anomalies from the RCM at 30- and 90-km resolution. Temperatures across the northern Sahel–southern Sahara warm by over 6 K in some regions of northern Sudan and Niger, and by 3–4 K between the equator and 10°N. With only 3–4 K warming along the Guinean coast, the meridional temperature gradient across the Sahel is quite large in these simulations.
As in Fig. 7, but for late century (2081–2100).
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
Figures 8c–g show the late-century temperature changes predicted by the CMIP5 AOGCMs. Each of the AOGCMs predicts continued warming over all of northern Africa with the greatest warming over the Sahara. All of the models also predict an intensification of the Sahel meridional temperature gradient, although the change is relatively small in CNRM-CM5 (Fig. 8d) and MRI-CGCM3 (Fig. 8g) and is considerably stronger in the GFDL CM3 model (Fig. 8e) compared to the RCM and the other AOGCMs.
b. Hydrology
Figure 9 shows predicted midcentury July–September precipitation anomalies (color shading); anomalies that are not statistically significant at the 95% confidence interval are not shown (see section 6 for details on the calculation of statistical significance). Also shown are 925-hPa wind vectors and scaled geopotential height anomalies for the RCM 30- (Fig. 9a) and 90-km (Fig. 9b) simulations and the AOGCMs (Figs. 9c–g). The geopotential height anomalies are scaled to emphasize the changes in the geopotential height field structure associated with circulation differences in the warmer climate state. For each model the scaling consists of calculating the domain-averaged difference in geopotential height (using the 90-km RCM domain shown in Fig. 1) and then subtracting that value from the full difference field. The purpose is to remove the overall change in geopotential heights in the warmer climate to focus on changes in gradients. Table 1 shows the scaling values for each model. The midcentury scaling value generally falls between 5 and 7 m, with the GFDL CM3 and MRI-CGCM3 appearing to be outliers.
Midcentury (2041–60) July–September precipitation (colors, mm day−1), 925-hPa scaled geopotential height (m), and 925-hPa wind (m s−1) anomaly projections for the RCM (a) 30- and (b) 90-km simulations, and the CMIP5 RCP8.5 for (c) CCSM4, (d) CNRM-CM5, (e) GFDL CM3, (f) MIROC5, and (g) MRI-CGCM3 AOGCM simulations. Only precipitation anomalies statistically significant at the 95% level of significance according to a Student's t test are shaded.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
At midcentury, rainfall is projected to significantly increase across the Sahel east of 10°W by the RCM. The magnitude of the change differs depending on the model resolution, ranging from 1 to 6 mm day−1 for the 30-km domain and from 1 to 3 mm day−1 for the 90-km domain. The RCM does not project much change in the length of the Sahel rainy season (not shown), indicating that a change in intensity of the wet season rains is responsible for the precipitation increase. The AOGCMs also indicate an increase in summer rainfall in the Sahel, generally with smaller magnitudes and less spatially contiguous. By midcentury, CCSM4 (Fig. 9c) and GFDL CM3 (Fig. 9e) project a widespread enhancement of rainfall in the Sahel. MIROC5 (Fig. 9f) and MRI-CGCM3 (Fig. 9g) also hint at a region of increased rainfall in the Sahel, but the signal is not spatially coherent. No significant rainfall changes are projected by CNRM-CM5 (Fig. 9d).
Low-level geopotential heights generally decrease over the northern Sahel/southern Sahara in association with a deepening of the thermal low trough that is related to the enhanced surface warming shown in Fig. 8. The magnitude of the difference ranges from 1 to 4 m in CNRM-CM3 (Fig. 9d), MIROC5 (Fig. 9f), and MRI-CGCM3 (Fig. 9g), from 3 to 5 m in the RCM simulations (Figs. 9a,b), from 4 to 8 m in CCSM4 (Fig. 9c), and over 10 m in GFDL CM3 (Fig. 9e). South of 10°N there is little change in the scaled anomalous height field, mimicking the meridional structure of the surface warming (Fig. 8). As was the case for surface temperature, the meridional height gradient across Sahel within 10°–20°N increases as does the anomalous westerly low-level flow. Such an increase in the low-level westerly flow over the Sahel and in the WAWJ on the west coast is associated with increases of Sahel rainfall on interannual and, especially, decadal times scales in observations (Pu and Cook 2010, 2012) and paleoclimate simulations of the African Humid Period (Patricola and Cook 2007). There is little change in the southerly monsoon flow onto the Guinean Coast.
Projections from the GFDL CM3 discussed above are interesting as this model projects significant rainfall increases across the Sahel comparable to the RCM despite not being able to capture the present-day transition of rainfall into the Sahel (Fig. 4g). This occurs because the projected deepening of the thermal low in this model is about twice as strong as in the RCM and most of the other AOGCMs. Thus, while not being able to simulate the present-day climate accurately, the GFDL CM3 projects a future increase in Sahel rainfall for the same physical reasons as the RCM and other AOGCMs, and the height anomalies are likely larger than the other projections because of the deficiencies in the present-day climate state.
Models that simulate an increase in the meridional surface temperature gradient in the Sahel (Fig. 8) also simulate an increase in the meridional geopotential height gradient, the westerly flow (and westerly moisture advection), and precipitation in the Sahel, while models that miss the meridional gradient intensification do not produce significant precipitation enhancements. This indicates that the basic mechanisms of Sahelian precipitation enhancement in the models are the same (i.e., an intensification of the WAWJ and its enhanced penetration into the continental interior). There is one exception, and that is the MIROC5 simulation. This model projects increases in the meridional surface temperature gradient (Fig. 8f) but not the meridional geopotential height gradient (Fig. 9f). This suggests that there may be a difference in how this model ventilates heat and, perhaps, moisture through the atmospheric boundary layer.
Simulated precipitation differences indicate increased moisture in the Sahel for midcentury, but also increased surface temperatures which will increase evaporation rates. Implications for moisture availability at the surface are evaluated by examining the simulated differences in precipitation and evaporation (P − E). Neglecting differences in runoff and land use, we refer to these as differences in soil moisture.
Contours in Fig. 10 show midcentury differences in P − E for the RCM simulations (Figs. 10a,b) and the CMIP5 AOGCMs (Figs. 10c–g). Statistically significant differences are shaded. The precipitation increases simulated in the RCM (Figs. 9a,b) lead to statistically significant soil moisture increases in the western and central Sahel, to about 20°E, and the far eastern Sahel. Enhanced surface warming increases evaporation in the northern Sahel, minimizing P − E, implying decreases in soil moisture over this region. Thus, soil moisture increases are confined to the central and southern Sahel south of about 15°N.
Midcentury (2041–60) July–September P − E (mm day−1), anomalies for the RCM (a) 30- and (b) 90-km simulations, and the CMIP5 RCP8.5 (c) CCSM4, (d) CNRM-CM5, (e) GFDL CM3, (f) MIROC5, and (g) MRI-CGCM3 AOGCM simulations. Red contours denote negative anomalies while blue contours denote positive anomalies. The P − E anomalies statistically significant at the 95% level of significance according to a Student's t test are shaded.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
Despite the often significant changes in precipitation shown in Fig. 9, the CMIP5 AOGCMs do not produce significant soil moisture differences at the 95% confidence level. GFDL CM3 (Fig. 10e) and MIROC5 (Fig. 10f) hint at significant increases in the central Sahel, but the responses are not spatially coherent.
Figure 11 shows the predicted late-century July–September precipitation, 925-hPa winds, and scaled geopotential height anomalies for each model. The RCM at both 30- (Fig. 11a) and 90-km (Fig. 11b) resolution predicts a further intensification of rainfall compared with the midcentury simulations (Figs. 9a,b), and significant increases in precipitation extend farther north into the Sahara, although the increase in rainfall over the Sahara may be overstated as the RCM tends to undersimulate summer rainfall over this region in the CTL. Summer rainfall rates increase by 2–10 mm day−1 in the 30-km domain and 1–6 mm day−1 in the 90-km domain. As in the midcentury projections, the onset of the Sahel wet season does not change much by late century. The wet season demise is projected to be 6 days later.
As in Fig. 9, but for late century (2081–2100).
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
The CMIP5 AOGCMs evaluated here (Figs. 11c–g) all predict significant rainfall increases in the Sahel by late century. MIROC5 (Fig. 11f) produces the largest magnitude increases, but this model is also systematically wet (Figs. 3h and 4h). The CCSM4 (Fig. 11c), MRI-CGCM3 (Fig. 11g), and to a lesser extent MIROC5 simulations produce wet conditions preferentially in the continental interior with a potential for reduced rainfall over the western Sahel west of 10°W.
The 925-hPa geopotential height scaled value further increases by late century (Table 1). The values from the RCM, CNRM-CM5, and MIROC5 range between 13 and 15 m with the CCSM4 a little lower at 10.46 m. The GFDL CM3 and MRI-CGCM3 again are outliers.
By late century the thermal low over the northern Sahel and southern Sahara deepens by 8–12 m in the RCM (Figs. 11a,b), further intensifying the low-level meridional height gradient, the WAWJ, and the westerly flow across the Sahel. All of the AOGCMs except CNRM-CM5 also predict an intensification of the thermal low by late century accompanied by an intensification of the meridional gradient, the WAWJ, and westerly flow across the Sahel. The reduced rainfall over the western Sahel west of 10°W predicted by the CCSM4, MIROC5, and MRI-CGCM3 models are associated with a stronger intensification of the thermal low over the western Sahara of northern Mauritania and Mali. In contrast the strongest intensification of the thermal low trough in the RCM occurs farther east over the central and eastern Sahara. This westward positioning of the anomaly center is associated with anomalous low-level northwesterly flow over the far western Sahel that is relatively dry. SST forcing differences between the individual AOGCMs may also be important, as this region is known to be sensitive to regional Atlantic and Indian Ocean SST forcing (Folland et al. 1986; Ward 1998; Moron et al. 2008; Hagos and Cook 2008). For example, present-day variability studies have associated drier Sahel conditions with warm tropical Atlantic SST forcing (e.g., Druyan 1991; Rowell et al. 1995; Vizy and Cook 2002; Hastenrath and Polzin 2011), but the role of Atlantic SST forcing can be modified by Indian Ocean SSTAs depending on the scale and magnitude of the warming (Hagos and Cook 2008). How the different ocean basins warm in the future relative to each other and also between the tropics and high latitudes will likely influence Sahel rainfall variability.
Figure 12 shows differences in the late-century soil moisture, calculated as P – E, from the RCM at 30-km (Fig. 12a) and 90-km (Fig. 12b) resolution, and the CMIP5 AOGCMs (Figs. 12c–g). By the end of the century the positive significant soil moisture anomalies seen at midcentury (Fig. 10) over the Sahel become stronger in the RCM, GFDL CM3, and MIROC5 predictions. A significant response over southern Niger also begins to emerge in the CCSM4 and CNRM-CM5 projections. This response is more spatially coherent at the 90% significance level (not shown). No significant soil moisture response is projected by the MRI-CGCM3 (Fig. 12g). Over southern Niger, northern Nigeria, and western Chad the RCM predicts soil moisture increases of 2–3 mm day−1 which is approximately 4 times the simulated present-day value. Only 17% of the available rainfall remains on the surface after evaporation in the present-day simulation in this region, but 35% remains in the late-century simulation, an 18% increase in water availability. The MIROC5 simulation projects the water availability to increase by 20%–40%, GFDL CM3 by 5%–10%, and CCSM4 by 5%–20%.
As in Fig. 10, but for late century (2081–2100).
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
6. Evaluating confidence in the projections
Climate predictions are more useful when they are accompanied by an evaluation of confidence. In the previous section, several techniques were used to estimate the accuracy of the projections. Each of those techniques is discussed briefly below, along with some specific examples how they were applied here and what the results of the application were.
The first step in evaluating the accuracy of the future projections from a given model is to examine the quality of the simulation of the present-day climate. This examination should extend deeper than just comparing a few fields with observations and reanalysis, and should also include a consideration of the regional climate dynamics. Here we present a succinct examination of the present-day climate in the RCM and AOGCM simulations. Additional evaluation of the RCM at 90-km resolution is presented in Cook and Vizy (2012), who show that the simulated present-day climate is sufficiently accurate to produce a realistic distribution of growing season days across Africa. Also, Vizy and Cook (2012) examine the simulation of extreme events in the RCM at 90-km resolution and show that temperature ranges, the distributions of heat wave and dry days, and rainfall intensities are consistent with reanalyses and observations.
Evaluation of the individual AOGCMs beyond the few figures presented here is conducted by the modeling groups (e.g., Delworth et al. 2006; Cook et al. 2012).
While a high value should be placed on the accuracy of a model's representation of the present-day climate, including its variability, this does not guarantee that the model's projections will be accurate. Different forcing factors that are not tested, or not dominant, in the current climate emerge as more important in the future. For this reason, a reasonable simulation of the present-day climate can be seen as a necessary, but not sufficient, condition for an accurate simulation of future climate.
An ensemble approach is also used here to aid in the evaluation of confidence. For both the RCM and AOGCM simulations, 20 years of model integration is treated as a 20-member ensemble. That is, we consider the simulation of 2041–60 as a 20-member ensemble representation of the mid-twenty-first century. For each model, the 20-member ensemble is used to calculate statistical significance, and only greenhouse gas–forced responses that emerge from the climate noise at the 95th percent confidence level are presented in the previous section.
We also employ a multimodel ensemble approach, using the five AOGCMs and two simulations with a regional model at two different resolutions as a seven-member ensemble. Agreement among different models supports confidence because it suggests that the results are not highly dependent on the details of parameterizations that differ from one model to the next.
The analysis of physical processes in the response to greenhouse gas forcing is also used as a tool for evaluating confidence. As a start, one can simply ask if the response is reasonable given our current understanding of the region's climate dynamics. The climate response can be compared with known modes of variability on decadal, interannual, and even intraseasonal time scales, especially in simulations of the near future. In the previous section, enhanced rainfall in the Sahel in the twenty-first century is associated with an intensification of the westerly flow onto and across the continent, similar to a mode of variability that is prominent in present-day climate variability in the region. In addition, a similar mechanism occurs in each of the seven simulations in both the midcentury and late-century time periods.
It is most useful for planning purposes to have projections for the coming few decades, as opposed to the end of the century, but this is difficult to accomplish with confidence because the climate system is so noisy. To improve our confidence in midcentury projections, we compare them with late-century projections. In the absence of nonlinearities, the late-twentieth-century simulations provide information about climate change signals that may be just emerging and of borderline statistical significance in the mid-twenty-first-century simulations. As an example, Fig. 13 shows midcentury precipitation differences with 925-hPa scaled geopotential height and wind differences as in Fig. 9, but with the criterion of statistical significance at 80% instead of 95%. Figure 14 is a similar plot but for soil moisture. The differences between Figs. 9 and 13, and between Figs. 10 and 14, are not large, but on the space scales of different countries and other planning units there are some potentially important differences. For example, moisture increases over the northern Sahel and Sahara are more definite at midcentury using the 80% level of significance. Reference to the projections at the end of the century (Figs. 11 and 12) support the idea that precipitation increases are coming to these regions and support the evaluation of the midcentury projections at the 80% confidence level.
As in Fig. 9, but only precipitation anomalies statistically significant at the 80% level of significance according to a Student's t test are shaded.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
As in Fig. 10, but for P – E anomalies statistically significant at the 80% level of significance according to a Student's t test are shaded.
Citation: Journal of Climate 26, 13; 10.1175/JCLI-D-12-00533.1
7. Summary and conclusions
State-of-the-art regional climate model simulations at 90- and 30-km resolutions are run and analyzed along with output from CMIP5 AOGCMs to predict how the Sahel summer precipitation, surface temperature, and surface moisture are likely to change at the middle and late century due to increased atmospheric CO2 concentrations. Both regional models and AOGCMs are used to take advantage of strengths of each, as the regional model provides finer resolution needed for resolving intense rainfall events and orography and offers information on spatial scales needed for regional impacts analysis, while the AOGCMs provide a global connectivity.
Three RCM simulations are generated and analyzed. The first simulation is a control, representative of 1989–2008, and is used to assess the ability of the regional model to simulate the current evolution of the summer West African monsoon. The other two runs are representative of mid- (2041–60) and late-twenty-first-century (2081–2100) climate under the CMIP5 RCP8.5 CO2 emissions scenario forcing. These simulations are designed to produce high-resolution projections relevant for mid- and long-range impact analysis.
Confidence in the model projections is evaluated based on the model's ability to simulate the evolution of the current West African monsoon climate, a necessary but not sufficient condition, and by examining differences among various models, agreement in ensemble members run using the same model, comparison of the results for the two different time periods, and an analysis of the physical processes of change.
It is shown here that the RCM can simulate the boreal summer evolution of the West Africa monsoon (see section 4). This includes realistically simulating the timing of the seasonal shift of convection from the Guinean Coast northward into the Sahel at the beginning of summer and the southward retreat in late September, as well as the development of the thermal low trough over northern Africa. In contrast most AOGCMs have difficulties simulating the seasonal cycle of the rainfall (Fig. 4) and the development of the thermal low trough and its associated meridional height gradient and circulation over northern and sub-Saharan Africa (Fig. 6).
The future climate projections for the evaluated fields are summarized in the following subsections.
a. Surface temperature
The model projections analyzed here indicate a strong likelihood for increased summertime surface air temperatures over Africa over the century. All models are in agreement that temperatures will increase by midcentury (2041–60), and all but one model indicate that the warming will be somewhere between 2 and 3.5 K over the Sahara and Sahel with the strongest warming over the Sahara and reduced warming closer to the equator. The outlier model (Fig. 7e) projects the warming to be 4–5 K depending on the location. By late century (2081–2100), surface air temperatures will warm even more, generally by 3–6 K depending on the location and the model. Again the warming predicted by the outlier model is more severe, ranging from 6 to 10 K over the Sahara (Fig. 8e).
b. Precipitation
The RCM model predictions analyzed indicate the strong likelihood that rainfall rates over the Sahel will increase. The analysis indicates that the strongest intermodel agreement at the 95% confidence level occurs in the late century over the central Sahel between 0° and 20°E as all seven model simulations analyzed project a significant increase (Fig. 11). This precipitation increase at the 95% confidence level is also predicted at midcentury in the RCM simulations and in most of the AOGCMs in some form (Fig. 9). However, the signal is more discernible at midcentury in the AOGCMs if an 80% confidence level is applied instead (Fig. 13). The predicted changes in rainfall are not found to be associated with a significant change in the length of the Sahel wet season. Instead it is associated with an increase in intensity of the rainfall during the wet season.
The projected rainfall response over the Sahel west of 0° is less certain for both the middle and late century. The RCM simulations and one AOGCM predict wetter conditions at midcentury, while the other AOGCMs project either drier or no statistically significant changes. By late century there is still disagreement among the models, with the RCM and two of the AOGCMs predicting significantly wetter conditions and the other three AOGCMs indicating the potential for significantly drier conditions. The intensification of the thermal low trough over the western Sahara is found to be stronger in the models that predict drying over the western Sahel. Associated with this intensification of the thermal low over the western Sahara is anomalous northwesterly low-level flow, which is relatively drier compared to westerly and/or southwesterly flow over the North Atlantic due west of the Sahel. Based upon the analysis presented, the wetter conditions predicted by the RCM are judged to be a more likely outcome for this region by the end of the century, but more work is still needed to increase confidence in this prediction.
c. P – E (surface soil moisture availability)
Moisture availability at the surface is evaluated by examining the simulated differences in P − E since increased surface temperatures in the future will be associated with increased evaporation rates. The RCM model projections analyzed here indicate the likelihood for increased summertime moisture availability at midcentury over the western Sahel, to about 20°E, and the far eastern Sahel (Fig. 10). Some AOGCMs hint at significant increases over the Sahel, but the response is not spatially coherent. By late century (Fig. 12) positive significant soil moisture increases over the Sahel become more spatially robust in all but one of the models. A relatively large increase in moisture availability is predicted over southern Niger, northern Nigeria, and western Chad ranging between 10% and 20% in the RCM, GFDL CM3, and CCSM4 models, and between 20% and 40% in MIROC5. Overall, there is good agreement among the models that by late century there will be an increase in moisture availability over this region, but the magnitude of the increase appears to be model dependent.
These climate projections are found to be robust across different models. To first order, RCM studies (e.g., Patricola and Cook 2010, 2011) and an analysis of the CMIP5 AOGCMs generally predict similar rainfall changes by late century (Monerie et al. 2012). Furthermore, unlike for CMIP3 where AOGCM Sahelian precipitation predictions were indicating minimal changes by the end of the century except for two outlier models (e.g., Cook and Vizy 2006; Cook 2008; Biasutti et al. 2009), the new RCM and CMIP5 projections indicate an improved agreement amongst the different models regarding how rainfall will change, and suggests greater reliability in these predictions. What is less coherent from the different model projections is exactly when the increases will first commence and their relative magnitude. The RCM, CCSM4, and GFDL CM3 indicate that significant increases at the 95% confidence level will be prevalent by midcentury, while the CNRM-CM5, MIROC5, and MRI-CGCM3 are less confident at this time. This disagreement among the different model simulations likely highlights the sensitivity of the model projections to the physical parameterizations and model resolutions used. The three CMIP5 AOGCMs that do not simulate the rainfall increase at midcentury also do not simulate strong warming over the Sahara for both middle and late century relative to the other models (Figs. 7 and 8), which may indicate that the differences are related to the simulation of cloud cover, SSTs, or potentially land surface–atmosphere interactions associated with the land surface model parameterizations. A better understanding of why some of the models are slower to simulate a future rainfall increase is needed first, before simulations can better pinpoint the exact timing of when the change will first become significant. Additionally this may also aid in reconciling why earlier modeling studies (e.g., Held et al. 2005; Cook and Vizy 2006; Mariotti et al. 2011) predict a drier Sahel by the late twenty-first century.
A final caveat of this study is that the RCM and CMIP5 AOGCM simulations analyzed here do not consider land use or vegetation changes. Other studies (e.g., Douville et al. 2000; Paeth and Thamm 2007; Paeth et al. 2009, 2011; Wang and Alo 2012) suggest that such changes could have a considerable impact over sub-Saharan Africa at regional scales. Future projections from models that factor such changes are needed to evaluate the impact that vegetation and land use changes may have and improve predictions.
Acknowledgments
Support from the National Science Foundation (Award ATM-1036604) and the U.S. Army Research Laboratory Minerva Project (Contract W911NF-09-1-0077) is gratefully acknowledged. The Texas Advanced Computing Center (TACC) at the University of Texas at Austin provided the high performance computing and database resources. We also gratefully acknowledge the GCM modeling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI), and the World Climate Research Program's Working Group on Coupled Modeling (WGCM) for their roles in making available the WCRP CMIP5 multimodel dataset. Support of this dataset is provided by the Office of Science, U.S. Department of Energy.
REFERENCES
Biasutti, M., and A. H. Sobel, 2009: Delayed Sahel rainfall and global seasonal cycle in a warmer climate. Geophys. Res. Lett., 36, L23707, doi:10.1029/2009GL041303.
Biasutti, M., A. H. Sobel, and S. J. Camargo, 2009: The role of the Saharan low in summertime Sahel rainfall variability and change in the CMIP3 models. J. Climate, 22, 5755–5771.
Breugem, W.-P., W. Hazeleger, and R. J. Haarsma, 2006: Multimodel study of tropical Atlantic variability and change. Geophys. Res. Lett., 33, L23706, doi:10.1029/2006GL027831.
Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model description and implementation. Mon. Wea. Rev., 129, 569–585.
Chen, S.-H., and W.-Y. Sun, 2002: A one-dimensional time dependent cloud model. J. Meteor. Soc. Japan, 80, 99–118.
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.
Cook, K. H., and E. K. Vizy, 2012: Impact of climate change on mid-twenty-first century growing seasons in Africa. Climate Dyn., 38, 2937–2955, doi:10.1007/s00382-012-1324-1.
Cook, K. H., G. A. Meehl, and J. M. Arblaster, 2012: Monsoon regimes and processes in CCSM4. Part II: African and American monsoon systems. J. Climate, 25, 2609–2621.
Delworth, T. L., and Coauthors, 2006: GFDL's CM2 global coupled climate models. Part I: Formulation and simulation characteristics. J. Climate, 19, 643–674.
Diallo, I., M. B. Sylla, F. Giorgi, A. T. Gaye, and M. Camara, 2012: Multimodel GCM-RCM ensemble-based projections of temperature and precipitation over West Africa for the early 21st century. Int. J. Geophys., 2012, 972896, doi:10.1155/2012/972896.
Douville, H., S. Planton, J.-F. Royer, D. B. Stephenson, S. Tyteca, L. Kergoat, S. Lafont, and R. A. Betts, 2000: Importance of vegetation feedbacks in doubled-CO2 climate experiments. J. Geophys. Res., 105, 14 841–14 861.
Druyan, L. M., 1991: The sensitivity of sub-Saharan precipitation to Atlantic SST. Climatic Change, 18, 17–36.
Druyan, L. M., 2011: Studies of 21st century precipitation trends over West Africa. Int. J. Climatol., 31, 1415–1424.
Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3107.
Flaounas, E., S. Bastin, and S. Janicot, 2011: Regional climate modeling of the 2006 West African monsoon: Sensitivity to convection and planetary boundary layer parameterization using WRF. Climate Dyn., 36, 1083–1105, doi:10.1007/s00382-010-0785-3.
Folland, C. K., T. N. Palmer, and D. E. Parker, 1986: Sahel rainfall and worldwide sea temperatures. Nature, 320, 602–607.
Fontaine, B., P. Roucou, and P. A. Monerie, 2011: Changes in the African monsoon region at medium-term time horizon using 12 AR4 coupled models under the A1B emissions scenario. Atmos. Sci. Lett., 12, 83–88.
Grodsky, S. A., J. A. Carton, and S. Nigam, 2003: Near surface westerly wind jet in the Atlantic ITCZ. Geophys. Res. Lett., 30, 2009, doi:10.1029/2003GL017867.
Hagos, S. M., and K. H. Cook, 2008: Ocean warming and late-twentieth-century Sahel drought and recovery. J. Climate, 21, 3797–3814.
Hastenrath, S., and D. Polzin, 2011: Long-term variations of circulation in the tropical Atlantic sector and Sahel rainfall. Int. J. Climatol., 31, 649–655.
Held, I. M., T. L. Delworth, J. Lu, K. 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.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341.
Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170–181.
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S. K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP–DOE AMIP-II Reanalysis (R2). Bull. Amer. Meteor. Soc., 83, 1631–1643.
Mariotti, L., E. Coppola, M. B. Sylla, F. Giorgi, and C. Piani, 2011: Regional climate model simulation of projected 21st century climate change over an all-Africa domain: Comparison analysis of nested and driving model results. J. Geophys. Res., 116, D15111, doi:10.1029/2010JD015068.
Meehl, G. A., and Coauthors, 2007: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–845.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmosphere: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682.
Monerie, P. A., B. Fontaine, and P. Roucou, 2012: Expected future changes in the African monsoon between 2030 and 2070 using some CMIP3 and CMIP5 models under a medium-low RCP scenario. J. Geophys. Res., 117, D16111, doi:10.1029/2012JD017510.
Monerie, P. A., P. Roucou, and B. Fontaine, 2013: Mid-century effects of climate change on African monsoon dynamics using the A1B emission scenario. Int. J. Climatol., 33, 881–896. doi:10.1002/joc.
Moron, V., A. W. Robertson, M. N. Ward, and O. Ndiaye, 2008: Weather types and rainfall over Senegal. Part I: Observational analysis. J. Climate, 21, 266–287.
Moufouma-Okia, W., and D. P. Rowell, 2010: Impact of soil moisture initialization and lateral boundary conditions on regional climate model simulations of the West African monsoon. Climate Dyn., 35, 213–229.
Paeth, H., and H.-P. Thamm, 2007: Regional modeling of future African climate north of 15°S including greenhouse warming and land degradation. Climatic Change, 83, 401–427.
Paeth, H., K. Born, R. Girmes, R. Podzun, and D. Jacob, 2009: Regional climate change in tropical and northern Africa due to greenhouse forcing and land use changes. J. Climate, 22, 114–132.
Paeth, H., and Coauthors, 2011: Progress in regional downscaling of West African precipitation. Atmos. Sci. Lett., 12, 75–82.
Patricola, C. M., and K. H. Cook, 2007: Dynamics of the West African monsoon under mid-Holocene precessional forcing: Regional climate model simulations. J. Climate, 20, 694–716.
Patricola, C. M., and K. H. Cook, 2010: Northern African climate at the end of the 21st century: Integrated application of regional and global climate models. Climate Dyn., 35, 193–212.
Patricola, C. M., and K. H. Cook, 2011: Sub-Saharan northern African climate at the end of the twenty-first century: Forcing factors and climate change processes. Climate Dyn., 37, 1165–1188.
Patricola, C. M., and K. H. Cook, 2013a: Mid-twenty-first century warm season climate change in the central United States. Part I: Regional and global model predictions. Climate Dyn., 40, 551–568.
Patricola, C. M., and K. H. Cook, 2013b: Mid-twenty-first century warm season climate change in the central United States. Part II: Climate change processes. Climate Dyn., 40, 569–583.
Pu, B., and K. H. Cook, 2010: Dynamics of the West African westerly jet. J. Climate, 23, 6263–6276.
Pu, B., and K. H. Cook, 2012: Role of the West African westerly jet in Sahel rainfall variations. J. Climate, 25, 2880–2896.
Richter, I., and S.-P. Xie, 2008: On the origin of equatorial Atlantic biases in coupled general circulation models. Climate Dyn., 31, 587–598.
Rowell, D. P., C. K. Folland, K. Maskell, and M. N. Ward, 1995: Variability of summer rainfall over tropical North Africa (1906–92): Observations and modeling. Quart. J. Roy. Meteor. Soc., 121, 669–704.
Seth, A., and M. Rojas, 2003: Simulation and sensitivity in a nested modeling system for tropical South America. Part I: Reanalyses boundary forcing. J. Climate, 16, 2437–2453.
Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, and J. G. Powers, 2005: A description of the Advanced Research WRF version 2. NCAR/TN-408+STR, 88 pp. [Available from NCAR Information Services, P.O. Box 3000, Boulder, CO 80307.]
Skinner, C. B., M. Ashfaq, and N. S. Diffenbaugh, 2012: Influence of twenty-first-century atmospheric and sea surface temperature forcing on West African climate. J. Climate, 25, 527–542.
Vizy, E. K., and K. H. 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.
Vizy, E. K., and K. H. Cook, 2002: Development and application of a mesoscale climate model for the tropics: Influence of sea surface temperature anomalies on the West African monsoon. J. Geophys. Res., 107, 4023, doi:10.1029/2001JD000686.
Vizy, E. K., and K. H. Cook, 2012: Mid-twenty-first-century changes in extreme events over northern and tropical Africa. J. Climate, 25, 5748–5767.
Wang, G., and C. A. Alo, 2012: Changes in precipitation seasonality in West Africa predicted by RegCM3 and the impact of dynamic vegetation feedback. Int. J. Geophys., 2012, 597205, doi:10.1155/2012/597205.
Ward, M. N., 1998: Diagnosis and short-lead time prediction of summer rainfall in tropical North Africa at interannual and multidecadal timescales. J. Climate, 11, 3167–3191.
