1. Introduction
One of the most prominent signals of decadal precipitation variability of the twentieth century was observed over the Sahel region of Africa (Ward 1998). This region, which extends from the Senegal coast to the Ethiopian highlands and roughly between 10° and 20°N latitudes, suffered widespread droughts through the 1970s and 1980s with major economic and social consequences to the population of the region. In the 1990s there was a partial recovery, bringing annual rainfall rates up to the climatological mean in the central and eastern Sahel. This recovery persists through the early years of this decade as well.
Figure 1a shows the 50-yr mean for the July through September (JAS) precipitation and the difference between the 1980–89 mean and the 50-yr precipitation climatology obtained from the University of East Anglia Climatic Research Unit (CRU) TS 2.1 land precipitation dataset (Mitchell and Jones 2005). It shows widespread drought across the Sahel. A comparison with the magnitude of the climatology indicates about 40% of the climatological value. Similarly, Fig. 1b displays the difference between the 1990–99 mean precipitation and the 50-yr climatology, showing that precipitation in the 1990s over the central and eastern Sahel was similar to the climatology. The recovery is even more apparent when the mean precipitation of the 1990s and 1980s is compared. Figure 1c shows the difference between these two decades.
Because of the devastating effects on the population of the region, Sahelian drought has been the subject of great concern and extensive research for more than 30 yr. In general, studies have progressed along two competing lines (Brooks 2004). The first study emphasizes localized feedbacks between land surface degradation and atmospheric radiation. According to these studies (e.g., Charney 1975), an increase in surface albedo due to human-induced change in vegetation decreased the precipitation, which in turn led to decreased vegetation and further increased surface albedo. The second study challenges this first hypothesis on the basis that many of the modeling studies used to support it exaggerated the prescribed land use changes. They suggest the observed changes in land surface conditions were more of a consequence of the low precipitation rates than a cause of the drying (Taylor et al. 2002). On the other hand, Zeng et al. (1999) uses a coupled atmosphere–vegetation model to show that land surface feedback amplifies the precipitation response to decadal-scale SST variability.
In the mid-1980s and afterward, modeling and observational analysis studies revealed an association between observed Sahel rainfall variability and tropical Atlantic Ocean SSTs (Folland et al. 1986; Rowell et al. 1995). Giannini et al. (2003) showed that the National Aeronautics and Space Administration’s (NASA) Goddard Space Flight Center in the framework of the Seasonal to Interannual Prediction Project 1 (NSIPP1) atmospheric GCM forced by observed SSTs from 1930 to 2000 captured decadal-scale variability similar to the observed SSTs over the Sahel and identified the Indian Ocean SSTs and, to a lesser extent, tropical Atlantic Ocean SSTs as the primary forcing of decadal-scale Sahel precipitation variability. Bader and Latif (2003), using the ECHAM4 atmospheric GCM, and Lu and Delworth (2005), using the Geophysical Fluid Dynamics Laboratory Climate Model version 2.0 (GFDL CM2.0) atmospheric GCM, agreed that the primary influence is Indian Ocean warming. In contrast, using an ensemble of 18 GCMs, Hoerling et al. (2006) suggested cooling in the northern tropical Atlantic Ocean and concurrent warming over the southern tropical Atlantic are the primary causes of this decadal variability.
The purpose of this study is to evaluate the physical processes responsible for the Sahel droughts of the 1980s and the partial recovery of the 1990s. It is important to understand the physics of the relationship between regional SST anomalies (SSTAs) and Sahel rainfall to improve our ability to predict for this region. For example, if the Sahel droughts of the 1970s and 1980s are indeed caused by warming in the Indian Ocean, then why was there a partial recovery in Sahel precipitation in the 1990s and early 2000s while the Indian Ocean continues to warm? Is the large spread in predictions of future Sahel rainfall from state-of-the-art coupled GCMs (Cook and Vizy 2006) due to a spread in simulations of future SSTs or an inability of the GCMs to capture the forced response to given SST distributions? If we know the physical processes responsible for precipitation variability, then we will be better able to evaluate predictions of twenty-first-century precipitation from coupled GCMs.
2. Description and evaluation of model and experiments
In this study, we use regional climate model simulations forced by observed SSTs in the ocean basins adjacent to the African continent. SSTs derived from the 50-yr climatology and the decadal means (1980s and 1990s) over the Atlantic and Indian Oceans are applied individually and in concert. Because the model captures the precipitation differences of the two decades from the climatology, the dynamical analysis reveals the reason for the dry conditions of the 1980s and the recovery in the 1990s.
a. Model description and simulation design
The tropical regional climate model (RegCM) used in this study is an adaptation of the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (PSU–NCAR MM5, version 3; Grell et al. 1994), modified according to Vizy and Cook (2002). The model is nonhydrostatic and solves the equations governing horizontal and vertical momentum, temperature, pressure, moisture, and liquid water equations on σ surfaces. The model simulations are run over a rectangular domain enclosed by 47°S, 47°N and 49°E, 100°W (Fig. 2). The grid spacing of 135 km used in these simulations allows for the resolution of the important surface and precipitation features over a domain large enough to cover the whole continent and the tropical and subtropical regions of both tropical oceans. There are 23 vertical σ levels, and the model time step is 90 s. The top of the atmosphere is fixed at 50 hPa for this tropical application and an upper radiative boundary condition is used. The effects of snow cover are neglected.
The Kain–Fritsch scheme (Kain and Fritsch 1990) is used to parameterize convection. The closure assumes that the convective available potential energy is almost entirely removed at every time step at which convection occurs. This scheme incorporates the effects of entrainment and detrainment as well as downdrafts associated with evaporation of rain on the convection and large-scale environment. It also produces realistic precipitation distributions over northern Africa in summer (Vizy and Cook 2002).
The Medium-Range Forecast (MRF) scheme (Troen and Mahrt 1986) represents boundary layer processes. The Noah land surface model (LSM; Mitchell et al. 2000) is used for the interaction of the boundary layer with the surface. It is set to calculate soil moisture and temperature at depths of 10 and 200 cm. The Noah LSM is initialized by soil moisture and temperature fields at the two levels from the NCEP–NCAR reanalysis project (NNRP; Kalnay et al. 1996).
The radiation scheme is adapted from that of the NCAR Community Climate System Model, version 2. Its shortwave scattering/absorption is parameterized by the δ-Eddington approximation (Joseph et al. 1976), and it is applied over 18 spectral intervals. The shortwave optical properties of the clouds depend on droplet size and the liquid water path. The scheme treats longwave absorption by ozone and CO2 using the broadband absorption technique of Kiehl and Briegleb (1991). Longwave broadband emissivity of clouds is a negative exponential function of the liquid water path.
A total of seven simulations are performed (see Table 1). In the control simulation (CTL), the annual SST cycle is obtained by averaging observed monthly mean SSTs from 1950 to 1999 from the NNRP (Reynolds and Smith 1995). These monthly mean SSTs are linearly interpolated into a 12-h time series as they are fed into the model. The values for the JAS average are displayed in Fig. 2. In a similar manner, initial, lateral, and land surface boundary conditions are obtained from the monthly mean NNRP climatology. The summer vegetation data are obtained from the U.S. Geological Survey (USGS), and they are shown in Fig. 2 at the resolution of the model. Each experiment starts on 15 April.
SSTs for the decadal simulations of the 1980s and 1990s are calculated in a similar manner. As detailed in Table 1, these SSTs are applied throughout the model domain and in the Atlantic and Indian Ocean basins separately. All lateral, initial, and land surface boundary conditions as well as the physical parameterizations are the same for all the experiments.
This procedure filters out interannual SST variability, and the simulation design excludes influence from the Pacific Ocean. This is appropriate because previous work (Giannini et al. 2003) suggests that influence from the Pacific Ocean is mainly interannual. The simulation design is intended to isolate the decadal signal, enabling higher-resolution modeling that captures the dynamics of the system accurately in a computationally efficient manner.
b. Evaluation
To evaluate the performance of the model, the JAS mean precipitation from CTL (Fig. 3a) is compared with precipitation fields from three observational datasets (Figs. 3b, c). Each precipitation field presented has been interpolated to a 1° × 1° grid. Figure 3b shows the JAS 1998–2004 mean precipitation from the Tropical Rainfall Measuring Mission (TRMM) 3B42, version 5. This is a blend of microwave and infrared precipitation estimates scaled to monthly rain gauge measurements. The 1979–99 JAS mean precipitation climatology from the Global Precipitation Climatology Project (GPCP) is shown in Fig. 3c. The dataset is derived from rain gauge measurements that were merged with satellite estimates of rainfall (Huffman et al. 1997). In Fig. 3d, the 1950–99 JAS precipitation climatology from the CRU TS 2.1 (Mitchell and Jones 2005) is displayed. This dataset depends only on rain gauge measurements.
The model simulates the precipitation over Africa reasonably well. The maxima over the western coast (10°N, 10°W), the Cameroon highlands (5°N, 10°E), and the Ethiopian highlands (15°N, 35°E) are all in place. The northern edge of the modeled precipitation (e.g., the 2 mm day−1 line) is slightly farther north than in the observations, all of which are south of 18°N. Therefore, the modeled precipitation over the Sahel is greater than the observed averages by roughly 2 mm day−1. This difference will be revisited when the model’s performance in reproducing the observed precipitation variability is considered in the next section.
The African precipitation from the regional model is better than those of most of the atmospheric GCMs from the Atmospheric Model Intercomparison Project (AMIP; Gadgil and Sajani 1998). Few of the AMIP GCMs reproduce the northern edge of summer precipitation near the observed latitude, and even fewer reproduce the magnitude and location of the precipitation maxima in the continent. Currently, much GCM modeling effort is focused on coupled ocean–atmosphere models. Many of these models do not capture the strong meridional precipitation gradients that characterize the transition from rain forest to desert across the Sahel (Cook and Vizy 2006).
The model also captures the precipitation maxima over India (15°N, 75°E) and the western coast of Thailand and Burma (near 15°N, 90°E). The relatively dry condition surrounding southern India and Sri Lanka is also reproduced. However, precipitation over the equatorial Indian Ocean is excessive in the model, and it extends too far west toward the eastern coast of Africa.
Figure 4 shows the mean summer atmospheric circulation from CTL and the NNRP at the 925- and 700-hPa pressure levels. The model realistically simulates the low-level southwesterly monsoon flow over West Africa and the flows associated with the North Atlantic and South Atlantic highs (Figs. 4a, b). Consistent with the fact that the model produces excessive precipitation over the equatorial Indian Ocean, the model’s Somali jet is weaker than that in the NNRP. Similarly, the African easterly jet is weaker than in the reanalysis, but the westerly flow across the South Atlantic and Indian Oceans is well represented (Figs. 4c, d). For an extensive discussion on the model’s performance in simulating the seasonal cycle, specifically the onset and dynamics of the monsoon, the reader is referred to Hagos and Cook (2007).
Because the model captures the important features of the circulation and precipitation fields over regions relevant to the purpose of the study, it builds confidence in the subsequent analysis.
3. Analysis
Figure 5a shows the modeled anomalous precipitation when the 1980s SSTs are prescribed over both ocean basins, that is, ATL+IND80 minus CTL. Precipitation rates over the western and central Sahel are reduced by an average of about 1 mm day−1. Comparison of Fig. 5a with the precipitation from CTL (Fig. 3a) shows that the 1980s SSTs introduce about a 50% reduction in precipitation over the Sahel, which is similar to the observed percentage reduction (Fig. 1a). The magnitude of the drying is greatest over the western Sahel where the observed and simulated climatological precipitation rates are also the largest. The drying extends across Africa to the Ethiopian highlands, with a minor discontinuity over the central Sahel.
The precipitation difference between ATL+IND90 and CTL (Fig. 5c) shows that prescribing the 1990s decadal mean SSTs introduces somewhat wetter conditions over the central Sahel and Guinean coast, although there are still localized dry regions over western Sahel and Ethiopian highlands (Fig. 5b). Overall, the 1990s anomalies are smaller than those of the 1980s as shown in the observations.
Figure 5c shows the difference between the 1990s and the 1980s SSTs in the model, which are analogous to Fig. 1c for the observations. In both the model and the observations, the 1990s SSTs are significantly wetter over much of the Sahel.
Clearly, the model captures the decadal precipitation variability over the Sahel. The physical processes for the 1980s drought and the partial recovery during the 1990s are analyzed below.
a. Mechanisms of the 1980s drought
To isolate the processes responsible for the 1980s dryness, the results from the experiments with prescribed 1980s SSTs over the individual ocean basins are analyzed. Figures 6a and 6b show SST and precipitation differences between IND80 and CTL, respectively. The widespread Indian Ocean warming of about 0.4 K imposed in IND80 (Fig. 6a) introduces drought over the western and central Sahel as well as the Ethiopian highlands, whereas the Guinean coast is mostly wetter. The magnitude of the response is comparable to that associated with the warming over of both the Atlantic and Indian Ocean basins (cf. Figs. 5a and 6b), but the former is not as widespread as the latter. Prescribing the 1980s Atlantic Ocean SSTs (Figs. 6c, d) introduces dry conditions along the western coast, but the response in the continental interior is not as pronounced.
The simulations indicate that the 1980s droughts were the result of a combined warming of both ocean basins. Even though Indian Ocean warming is primarily responsible for the drying in the continental interior, the concurrent tropical Atlantic Ocean warming significantly enhanced it, both in areal coverage and magnitude.
Figures 7a–d show the lhs and the first three terms on the rhs of Eq. (1); the last two terms on the rhs are negligible. The main source of moisture for condensation over the Sahel is horizontal moisture convergence (cf. Figs. 7a, b). The contribution of diffusion (Fig. 7c; i.e., direct evaporation from the surface) is small, and it does not reflect the structure of the condensation field. This shows that much of the moisture over the Sahel is transported from the adjacent oceans.
On the other hand, evaporation is widespread over the oceans. The near-surface vertical transport of moisture (Fig. 7d), that is, resolved vertical mixing, is downward so it reduces condensation. Therefore, oceanic evaporation is the main source of moisture in the domain, whereas condensation and vertical mixing over both the oceans and land are the sinks. Hence, to understand the influence of SSTs on Sahel precipitation, the nature of the moisture transport into the region needs to be analyzed.
Figure 8a shows the JAS mean vertically integrated χ and moisture flux vectors from CTL. The main source of moisture in the domain is the tropical southern Atlantic Ocean (centered near 10°S, 25°W), whereas the Indian Ocean and the African Sahel are sinks. The strongest mean moisture flow into the continent occurs across the Guinean coast between 10°E and 15°W. Figure 8b shows the vertical structure of the meridional moisture flux at 5°N. Figures 8a and 8b together demonstrate that the low-level southerly moisture flux across the Guinean coast is the main supply of moisture for the African Sahel, as expected. The moisture transport vectors suggest some net moisture transport from the Mediterranean Sea as well.
To understand how Indian Ocean warming influences Sahel precipitation, the anomalous moisture flux field is considered (Fig. 9a). Along the western coast of the Sahel, the anomalous moisture flux is zonal and directed away from the continent. A comparison with that of CTL (Fig. 9b) shows a moisture loss of up to 50% from the Sahel region. Figure 9b shows a latitude–height cross section of the anomalous zonal moisture flux at 15°W. The moisture flow away from the continent occurs above the boundary layer and is strongest near 700 hPa. Near the surface, there is a shallow anomalous moisture transport into the continent. This strengthening of the westerly flow onto the continent is consistent with the fact that the reduced precipitation and cloud cover warms up the land surface (not shown).
Figure 9c shows the anomalous circulation at 700 hPa over northwestern Africa. This circulation is a Rossby response to the wind divergence to its immediate east (Fig. 9d), which is a strengthening of the Saharan high. The associated anticyclonic circulation enhances the African easterly jet and the anomalous easterly moisture flux. This divergence, in turn, along with the pair of cyclonic circulations over northeastern and southeastern Africa, constitutes the Rossby response to the large-scale convergence forcing over the equatorial Indian Ocean.
In their regional modeling study, Paeth and Hense (2006) evaluate the validity of a linear analysis of the tropical African climate response to a prescribed forcing. They find that, to a large extent, the atmospheric circulation associated with SST anomalies can be represented by the first 10 Matsuno–Gill modes. However, they also find that the processes that give rise to the atmospheric heating field are nonlinear. Therefore, the primary nonlinear processes that determine the atmospheric response to SSTs and the role of the background climatology deserve some attention.
The response of the modeled Sahel precipitation to the 1980s Atlantic Ocean warming (Figs. 6c, d) is also analyzed. Atlantic Ocean warming also introduces strong moisture convergence off the western coast of the continent, (Fig. 10a) and it poses as a direct competition for the available moisture. The anomalous anticyclonic circulation, which was shown to drive moisture away from the continent in the case of Indian Ocean warming, also exists in this case. However, here it extends to the surface (Fig. 10b). Once again, the strong forcing over the northern equatorial Atlantic, despite the fact that much of the SST warming is south of the equator, can be explained by the fact that the former is the region where the background mixing ratio [and the first term on the rhs of Eq. (3)] is maximum (Fig. 10d).
b. The 1990s recovery
To understand the modeled and observed partial recovery of Sahel precipitation in the 1990s (Figs. 1c and 5c), the results from the experiments in which the 1990s SSTs are individually imposed over the two oceans are considered (see Table 1). Figure 11 shows the prescribed summer mean SST differences between IND90 and CTL as well as between ATL90 and CTL, and the respective precipitation differences. In the 1990s, the northern tropical Atlantic Ocean was warmer than during the 1980s (cf. Figs. 11c and 6c), but the rest of the tropical Atlantic Ocean was cooler. The Atlantic in both decades was generally warmer than the climatology. Similarly, a comparison of Figs. 11a and 6a demonstrates that the tropical Indian Ocean was anomalously warm during both the 1980s and 1990s, but the warming was stronger and zonally broader during the 1990s than during the 1980s.
The SST anomalies over both ocean basins individually introduce widespread precipitation recovery (Figs. 11b, d), but the positive precipitation anomalies associated with ATL90 SSTs are more widespread. Therefore, the recovery of the 1990s is primarily due to Atlantic Ocean SST patterns of that decade.
Figure 12a shows differences in the moisture flux potential [Eq. (2)] and the moisture flux vectors for ATL90 minus CTL, and Fig. 12b displays the irrotational wind and the mixing ratio differences. The warm SST anomalies off the western coast of Africa (Fig. 11c) are seen to be coincident with moist conditions, and these anomalies introduce localized moisture convergence along the coast. The associated cyclonic wind pattern favors moisture transport into the continent that is greatest at 700 hPa (Figs. 12c, d).
Whereas the circulation patterns associated with 1990s and 1980s Atlantic Ocean warming are similar in that they both have a pair of cyclonic circulations, which are typical responses to local convergence, they differ in their strength and areal coverage (Figs. 10c and 12c). The westerly part of the cyclonic circulation pattern of the 1990s favors moisture flow into the continent, while 1980s is fully over the ocean. This suggests that the scale of the forcing is an important factor in determining the direction of moisture transport between the ocean and continent, whereas this response is amplified by the circulations associated with the condensation (and convergence) anomalies over the continent.
Figures 13a and 13b show the circulation in response to the 1990s warming in the Indian Ocean. Subsidence is induced over the equatorial Atlantic, with anomalous irrotational westerly flow onto the continent. The nondivergent flow onto the continent is also westerly, which is associated with the expected pair of cyclones that straddle the equator (Fig. 13b).
The reason for the strikingly different response in the 1990s compared with the 1980s, even though the forcing over the Indian Ocean is quite similar, becomes apparent when the horizontal scales of the forcing above the boundary layer, that is, the moisture convergence, are compared. Although many of the circulation features of the 1980s and 1990s are similar, the 1990s convergence over the Indian Ocean covers a wider area and the subsequent divergence occurs over the equatorial Atlantic Ocean. In contrast, the moisture divergence associated with Indian Ocean warming in the 1980s occurs over western and central Africa (cf. Figs. 13a and 9a). Consequently, although the anomalous moisture flow across the western coast of Africa (15°W) is directed away from the continent in the 1980s, it is directed onto the continent in the 1990s.
The difference in the scales of the forcing is also apparent in the nondivergent wind fields at 700 hPa (and Figs. 13b and 9c). Although the anomalous cyclonic circulations over northeastern and southeastern Africa are similar in both simulations, the 1990s circulations extend far into western Africa and favor westerly moisture flow from the eastern equatorial Atlantic.
c. Comparison with observations
To further validate the conclusions drawn from the model simulations, decadal variability in the NNRP wind and geopotential height fields is considered. We examine the full wind fields and the geopotential heights from the reanalysis because these are assimilated variables and, therefore, close to observations. The relatively low resolution of the reanalysis (2.5° × 2.5°) and the fact that moisture fields are not assimilated makes it inadvisable to decompose the moisture convergence field as in the analysis of the model output. However, we can confidently examine the full fields for evidence of the processes identified in the model analysis.
Figure 14a shows the climatological mean 700-hPa wind and geopotential heights from the reanalysis, and Fig. 14b shows the difference between the 1980s mean and the climatology. The high geopotential heights and the associated anticyclonic circulation anomaly over the central Sahel (Fig. 14b) are the main features of the decade, representing a southward extension of the Saharan high and the African easterly jet. This agrees with the regional model simulations of the 1980s, which resulted from both Indian and Atlantic Oceans’ SSTAs (Figs. 9c and 10c). In agreement with our modeling results, the NNRP data also suggest the droughts of 1980s are related to the anticyclonic circulation and strengthened easterly flow. Compared with the reanalysis, the model shifts the anticyclonic circulation anomaly to the west, and the precipitation anomaly is also farther west than in the observations (cf. Figs. 1a and 5a).
A similar geopotential high and anticyclonic circulation anomaly is also observed in the 1990s in the reanalysis (Fig. 14c). The associated easterly flow over western Africa, however, is weaker, implying that the moisture loss from the continent is also weaker. The differences between the 1990s and 1980s in the reanalysis highlight the distinction (Fig. 14d). Thus, consistent with our modeling results, the reanalysis suggests that the precipitation recovery of the 1990s over the Sahel is related to the anomalous cyclonic circulation (relative to the 1980s) associated with an anomalous low off the northwestern coast of Africa.
Finally, the modeled relationship between Indian Ocean SSTs and Sahel precipitation is indirectly verified by comparing the 1990s minus 1980s precipitation from the model with those from Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP; Xie and Arkin 1996) and GPCP datasets, which contain precipitation data over the ocean for those decades; Fig. 15 shows the differences. Although the differences between the model and the observation, and between the observations themselves, are significant, they agree that the precipitation of the 1990s is higher over equatorial the Indian Ocean, as are the SSTs. The relatively cool SSTs in the southern Indian Ocean (cf. Figs. 6a and 11a) are accompanied by relatively dry conditions both in the model and CMAP and, to a lesser extent, in the GPCP datasets. This suggests that the precipitation distribution (as well as the associated convergence) is a consequence of the SST warming, not vice versa.
4. Discussion and conclusions
This study shows that combined Indian and Atlantic Ocean SST anomalies are responsible for the 1980s droughts over the Sahel. Warming over the Indian Ocean during the 1980s introduced local convergence that forced subsidence over central Africa and an anomalous anticyclonic circulation over the western coast as part of the Rossby wave response to that convergence. This anticyclonic circulation enhances the westward transport of moisture from the Sahel to the tropical eastern Atlantic, desiccating the Sahel (Figs. 9 and 14). The Atlantic Ocean warm anomaly of the 1980s introduced anomalous moisture convergence over the northern tropical Atlantic, further depleting the moisture supply into the continental interior (Fig. 10).
Important nonlinear processes are identified. The location and extent of the forcing (i.e., anomalous moisture convergence in the vicinity of the SST anomalies) not only depend on the SST anomalies but also on the distribution of background atmospheric moisture. For example, whereas the strongest SST anomalies in the 1980s Atlantic Ocean are south of the equator (Fig. 6c), the strongest anomalous moisture convergence over the ocean is near 10°N, close to the latitude of maximum background mixing ratio (Figs. 10a, d). This is also true for the remote response. The Sahel’s response to the forcing over both ocean basins is strongest because the climatological moisture content there is at its maximum during northern summer. In other words, the larger the background moisture content, the larger the changes will be in moisture convergence for the same anomalous winds convergence [Eq. (3)].
The modeled 1990s recovery is related to changes in the scale and distribution of the forcing. The warm ocean surface off the northwestern coast of Africa during the 1990s is mainly responsible for the recovery of the model precipitation (Figs. 11c, d). The relatively strong cyclonic circulation over the northern tropical Atlantic favors anomalous moisture transport into the continent. Comparison of the moisture convergence responses to the 1980s and 1990s Indian Ocean SST anomalies shows that the response to the latter has a larger scale, so the associated low-level divergence is located over the tropical Atlantic (cf. Figs. 9 and 13). Therefore, instead of an anomalous anticyclonic circulation driving away moisture from the continent, as was the case for 1980s SST forcing, the flow is cyclonic and favors moisture transport into the continent for 1990s SST forcing.
It is important to note that, whereas the primary forcing is external to the Sahel region, the local responses are amplified by moist processes. For example, the divergence over the Sahel in the 1980s is enhanced by the anomalous moisture transport that desiccates the region just as much as the latter is a (Rossby wave) “response” to the former. The same applies to the convergence and the associated anomalous cyclonic circulations, which supply the moisture during the 1990s recovery.
Some of these results agree with previous studies by various authors. Giannini et al. (2003), Bader and Latif (2003), and Lu and Delworth (2005) also show in their numerical studies that the subsidence and drought over the Sahel region are Rossby wave responses to Indian Ocean warming. In addition to that, however, our model results show that the associated anticyclonic circulation to the immediate west of the subsidence plays an important role in driving moisture away from the Sahel region.
Although our results also agree with those of Hoerling et al. (2006) in their conclusion that Atlantic Ocean SSTs play an important role in Sahel decadal variability, the physical mechanisms inferred from the model results differ somewhat. In particular, Hoerling et al. (2006) suggest that a modulation of the latitudinal location of the ITCZ by meridional SST gradients in the Atlantic is the main cause of Sahel decadal precipitation variability. Our results, however, suggest that the influence of SST anomalies is most prominent over regions of maximum mixing ratio. Therefore, the anomalous convergence over the Atlantic Ocean is at nearly the same latitude as the Sahel region, rendering the competition for moisture between the Sahel and the Atlantic a predominantly zonal process (Fig. 10a). The role of northern Atlantic SST warming in the partial recovery of precipitation is, however, apparent in our model as well. In this regional modeling study, only the influence of the adjacent oceans is considered; a global modeling work by Bader and Latif (2003) suggests a significant role of the Pacific Ocean over eastern Sahel precipitation.
The importance of background moisture distributions and the scale of the forcing, as demonstrated by the vastly different response of Sahel precipitation to 1980s and 1990s SSTs, have some implications for analysis and projections of twenty-first-century Sahel precipitation. This study suggests that the Sahel precipitation response to SST variations in this century may be very sensitive to changes in the distribution of both atmospheric moisture and SST.
Acknowledgments
The authors thank Dr. E. K. Vizy, who provided many of the analysis tools used in this study, and the anonymous reviewers whose comments greatly improved the paper. This work was supported by NSF Award ATM-0415481.
REFERENCES
Bader, J., and M. Latif, 2003: The impact of decadal-scale Indian Ocean sea surface temperature anomalies on Sahelian rainfall and the North Atlantic Oscillation. Geophys. Res. Lett., 30 .2169, doi:10.1029/2003GL018426.
Brooks, N., 2004: Drought in the African Sahel: Long term perspectives and future prospects. Tyndall Centre for Climate Change Research Working Paper 61, 37 pp.
Charney, J. G., 1975: Dynamics of deserts and drought in the Sahel. Quart. J. Roy. Meteor. Soc., 101 , 193–202.
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.
Folland, C. K., T. N. Palmer, and D. E. Parker, 1986: Sahel rainfall and worldwide sea temperatures. Nature, 320 , 602–607.
Gadgil, S., and S. Sajani, 1998: Monsoon precipitation in the AMIP runs. Climate Dyn., 14 , 659–689.
Giannini, A., R. Saravanan, and P. Chang, 2003: Oceanic forcing of Sahel rainfall on interannual to interdecadal time scales. Science, 302 , 1027–1030.
Grell, G., J. Dudhia, and D. Stauffer, 1994: A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+STR, 117 pp.
Hagos, S. M., and K. H. Cook, 2007: Dynamics of the West African monsoon jump. J. Climate, 20 , 5264–5284.
Hoerling, M., 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.
Huffman, G. J., and Coauthors, 1997: The Global Precipitation Climatology Project (GPCP) combined precipitation dataset. Bull. Amer. Meteor. Soc., 78 , 5–20.
Jaluria, Y., and K. Torrance, 2003: Computational Heat Transfer. 2nd ed. Taylor and Francis, 544 pp.
Joseph, J. H., W. J. Wiscombe, and J. A. Weinman, 1976: The delta-Eddington approximation for radiative flux transfer. J. Atmos. Sci., 33 , 2452–2459.
Kain, J. S., and J. M. Fritsch, 1990: A one-dimensional entraining/detraining plume model and its application in convective parameterization. J. Atmos. Sci., 47 , 2784–2802.
Kalnay, E. M., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437–471.
Kiehl, J. T., and B. P. Briegleb, 1991: A new parameterization of the absorptance due to the 15-μm band system of carbon dioxide. J. Geophys. Res., 96 , 9013–9019.
Lu, J., and T. L. Delworth, 2005: Oceanic forcing of the late 20th century Sahel drought. Geophys. Res. Lett., 32 .L22706, doi:10.1029/2005GL023316.
Mitchell, K. E., and Coauthors, 2000: Recent GCIP-sponsored advancements in coupled land-surface modeling and data assimilation in the NCEP Eta mesoscale model. Preprints, 15th Conf. on Hydrology, Long Beach, CA, Amer. Meteor. Soc., P1.22. [Available online at http://ams.confex.com/ams/annual2000/techprogram/paper_63.htm.].
Mitchell, T., and P. Jones, 2005: An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol., 25 , 693–712.
Paeth, H., and A. Hense, 2006: On the linear response of tropical African climate to SST changes deduced from regional climate model simulations. Theor. Appl. Climatol., 83 , 1–19.
Reynolds, R. W., and T. M. Smith, 1995: A high-resolution global sea surface temperature climatology. J. Climate, 8 , 1571–1583.
Rowell, D. P., C. K. Folland, K. Maskell, and M. Ward, 1995: Variability of summer rainfall over tropical North Africa (1906–92): Observations and modelling. Quart. J. Roy. Meteor. Soc., 121 , 669–704.
Taylor, C. M., E. F. Lambin, N. Stephenne, R. J. Harding, and R. L. H. Essery, 2002: The influence of land use change on climate in the Sahel. J. Climate, 15 , 3615–3629.
Troen, I., and L. Mahrt, 1986: A simple model of the atmospheric boundary layer: Sensitivity to surface evaporation. Bound.-Layer Meteor., 37 , 129–148.
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.
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.
Xie, P., and P. A. Arkin, 1996: Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9 , 840–858.
Zeng, N., J. D. Neelin, K-M. Lau, and C. J. Tucker, 1999: Enhancement of interdecadal climate variability in the Sahel by vegetation interaction. Science, 286 , 1537–1540.
JAS precipitation (mm day−1) from CRU TS 2.1: (a) 1980s minus climatological, (b) 1990s minus climatological, and (c) 1990s minus 1980s.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS precipitation (mm day−1) from CRU TS 2.1: (a) 1980s minus climatological, (b) 1990s minus climatological, and (c) 1990s minus 1980s.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS precipitation (mm day−1) from CRU TS 2.1: (a) 1980s minus climatological, (b) 1990s minus climatological, and (c) 1990s minus 1980s.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Simulation domain with JAS mean of the prescribed SSTs (K) and the vegetation distribution. The numbers indicate USGS vegetation categories as follows: 1 = shrubland, 2 = grassland, 3 = savanna, 4 = desert, and 5 = evergreen (rain) forest.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Simulation domain with JAS mean of the prescribed SSTs (K) and the vegetation distribution. The numbers indicate USGS vegetation categories as follows: 1 = shrubland, 2 = grassland, 3 = savanna, 4 = desert, and 5 = evergreen (rain) forest.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Simulation domain with JAS mean of the prescribed SSTs (K) and the vegetation distribution. The numbers indicate USGS vegetation categories as follows: 1 = shrubland, 2 = grassland, 3 = savanna, 4 = desert, and 5 = evergreen (rain) forest.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS mean precipitation (mm day−1) from the (a) RegCM simulation and observations, (b) TRMM (3B42V5; 7-yr mean), (c) GPCP (21-yr mean), and (d) CRU TS 2.1 (50-yr mean).
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS mean precipitation (mm day−1) from the (a) RegCM simulation and observations, (b) TRMM (3B42V5; 7-yr mean), (c) GPCP (21-yr mean), and (d) CRU TS 2.1 (50-yr mean).
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS mean precipitation (mm day−1) from the (a) RegCM simulation and observations, (b) TRMM (3B42V5; 7-yr mean), (c) GPCP (21-yr mean), and (d) CRU TS 2.1 (50-yr mean).
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Winds (m s−1) and geopotential heights (gpm) from the (a) NNRP at 925 hPa, (b) RegCM at 925 hPa, (c) NNRP at 700 hPa, and (d) RegCM at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Winds (m s−1) and geopotential heights (gpm) from the (a) NNRP at 925 hPa, (b) RegCM at 925 hPa, (c) NNRP at 700 hPa, and (d) RegCM at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Winds (m s−1) and geopotential heights (gpm) from the (a) NNRP at 925 hPa, (b) RegCM at 925 hPa, (c) NNRP at 700 hPa, and (d) RegCM at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS precipitation (mm day−1) from the RegCM simulations: (a) ATL+IND80 minus CTL, (b) ATL+IND90 minus CTL, and (c) ATL+IND90 minus ATL+IND80.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS precipitation (mm day−1) from the RegCM simulations: (a) ATL+IND80 minus CTL, (b) ATL+IND90 minus CTL, and (c) ATL+IND90 minus ATL+IND80.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS precipitation (mm day−1) from the RegCM simulations: (a) ATL+IND80 minus CTL, (b) ATL+IND90 minus CTL, and (c) ATL+IND90 minus ATL+IND80.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
RegCM precipitation response to 1980s Indian and Atlantic Ocean SST variations: (a) SST (K) IND80 minus CTL, (b) precipitation (mm day−1) IND80 minus CTL, (c) SST ATL80 minus CTL, and (d) precipitation ATL80 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
RegCM precipitation response to 1980s Indian and Atlantic Ocean SST variations: (a) SST (K) IND80 minus CTL, (b) precipitation (mm day−1) IND80 minus CTL, (c) SST ATL80 minus CTL, and (d) precipitation ATL80 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
RegCM precipitation response to 1980s Indian and Atlantic Ocean SST variations: (a) SST (K) IND80 minus CTL, (b) precipitation (mm day−1) IND80 minus CTL, (c) SST ATL80 minus CTL, and (d) precipitation ATL80 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
The terms in the vertically integrated moisture budget (mm day−1) [Eq. (1)]: (a) condensation, (b) horizontal convergence, (c) diffusion (surface evaporation), and (d) vertical advection at the surface.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
The terms in the vertically integrated moisture budget (mm day−1) [Eq. (1)]: (a) condensation, (b) horizontal convergence, (c) diffusion (surface evaporation), and (d) vertical advection at the surface.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
The terms in the vertically integrated moisture budget (mm day−1) [Eq. (1)]: (a) condensation, (b) horizontal convergence, (c) diffusion (surface evaporation), and (d) vertical advection at the surface.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
(a) Moisture flux potential and moisture flux [g m (kg s)−1] from CTL and (b) cross section of meridional moisture flux at 5°N [g m (kg s)−1].
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
(a) Moisture flux potential and moisture flux [g m (kg s)−1] from CTL and (b) cross section of meridional moisture flux at 5°N [g m (kg s)−1].
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
(a) Moisture flux potential and moisture flux [g m (kg s)−1] from CTL and (b) cross section of meridional moisture flux at 5°N [g m (kg s)−1].
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
IND80 minus CTL, JAS mean (a) vertically integrated irrotational moisture flux [g m (kg s)−1], (b) zonal moisture flux at 15°W [g m (kg s)−1], (c) nondivergent wind at 700 hPa (m s−1), and (d) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
IND80 minus CTL, JAS mean (a) vertically integrated irrotational moisture flux [g m (kg s)−1], (b) zonal moisture flux at 15°W [g m (kg s)−1], (c) nondivergent wind at 700 hPa (m s−1), and (d) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
IND80 minus CTL, JAS mean (a) vertically integrated irrotational moisture flux [g m (kg s)−1], (b) zonal moisture flux at 15°W [g m (kg s)−1], (c) nondivergent wind at 700 hPa (m s−1), and (d) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
ATL80 minus CTL, JAS mean (a) vertically integrated irrotational moisture flux [g m (kg s)−1], (b) zonal moisture flux at 15°E [g m (kg s)−1], (c) nondivergent wind at 700 hPa (m s−1), and (d) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
ATL80 minus CTL, JAS mean (a) vertically integrated irrotational moisture flux [g m (kg s)−1], (b) zonal moisture flux at 15°E [g m (kg s)−1], (c) nondivergent wind at 700 hPa (m s−1), and (d) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
ATL80 minus CTL, JAS mean (a) vertically integrated irrotational moisture flux [g m (kg s)−1], (b) zonal moisture flux at 15°E [g m (kg s)−1], (c) nondivergent wind at 700 hPa (m s−1), and (d) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Model precipitation response to 1990s Indian and Atlantic Ocean SST variations: (a) SST (K) IND90 minus CTL, (b) precipitation (mm day−1) IND90 minus CTL, (c) SST ATL90 minus CTL, and (d) precipitation ATL90 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Model precipitation response to 1990s Indian and Atlantic Ocean SST variations: (a) SST (K) IND90 minus CTL, (b) precipitation (mm day−1) IND90 minus CTL, (c) SST ATL90 minus CTL, and (d) precipitation ATL90 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Model precipitation response to 1990s Indian and Atlantic Ocean SST variations: (a) SST (K) IND90 minus CTL, (b) precipitation (mm day−1) IND90 minus CTL, (c) SST ATL90 minus CTL, and (d) precipitation ATL90 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
ATL90 minus CTL, JAS mean (a) vertically integrated moisture flux [g m (kg s)−1], (b) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa, (c) nondivergent wind at 700 hPa, and (d) zonal moisture flux at 15°E [g m (kg s)−1].
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
ATL90 minus CTL, JAS mean (a) vertically integrated moisture flux [g m (kg s)−1], (b) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa, (c) nondivergent wind at 700 hPa, and (d) zonal moisture flux at 15°E [g m (kg s)−1].
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
ATL90 minus CTL, JAS mean (a) vertically integrated moisture flux [g m (kg s)−1], (b) irrotational wind (m s−1) and mixing ratio (g kg−1) at 700 hPa, (c) nondivergent wind at 700 hPa, and (d) zonal moisture flux at 15°E [g m (kg s)−1].
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
IND90 minus CTL, JAS mean (a) vertically integrated moisture flux [g m (kg s)−1] and (b) nondivergent wind (m s−1) at 700 hPa IND90 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
IND90 minus CTL, JAS mean (a) vertically integrated moisture flux [g m (kg s)−1] and (b) nondivergent wind (m s−1) at 700 hPa IND90 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
IND90 minus CTL, JAS mean (a) vertically integrated moisture flux [g m (kg s)−1] and (b) nondivergent wind (m s−1) at 700 hPa IND90 minus CTL.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS mean wind (m s−1) and geopotential height (gpm) at 700 hPa from the NNRP: (a) climatological (1950–2002) average, (b) 1980s minus climatological, (c) 1990s minus climatological, and (d) 1990s minus 1980s.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS mean wind (m s−1) and geopotential height (gpm) at 700 hPa from the NNRP: (a) climatological (1950–2002) average, (b) 1980s minus climatological, (c) 1990s minus climatological, and (d) 1990s minus 1980s.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
JAS mean wind (m s−1) and geopotential height (gpm) at 700 hPa from the NNRP: (a) climatological (1950–2002) average, (b) 1980s minus climatological, (c) 1990s minus climatological, and (d) 1990s minus 1980s.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Mean difference between 1990s and 1980s JAS precipitation (mm day−1) from (a) the RegCM, (b) CMAP, and (c) GPCP.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Mean difference between 1990s and 1980s JAS precipitation (mm day−1) from (a) the RegCM, (b) CMAP, and (c) GPCP.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Mean difference between 1990s and 1980s JAS precipitation (mm day−1) from (a) the RegCM, (b) CMAP, and (c) GPCP.
Citation: Journal of Climate 21, 15; 10.1175/2008JCLI2055.1
Description of the regional model experiments.
