Impact of Sea Surface Temperature and Soil Moisture on Seasonal Rainfall Prediction over the Sahel

Wassila M. Thiaw NOAA/NWS/NCEP Climate Prediction Center, Camp Springs, Maryland

Search for other papers by Wassila M. Thiaw in
Current site
Google Scholar
PubMed
Close
and
Kingtse C. Mo NOAA/NWS/NCEP Climate Prediction Center, Camp Springs, Maryland

Search for other papers by Kingtse C. Mo in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The ensemble rainfall forecasts over the Sahel for July–September (JAS) from the NCEP Coupled Forecast System (CFS) were evaluated for the period 1981–2002. The comparison with the gauge-based precipitation analysis indicates that the predicted Sahel rainfall is light and exhibits little interannual variability. The rain belt is shifted about 4° southward.

One major source of rainfall errors comes from the erroneous sea surface temperature (SST) forecasts. The systematic SST error pattern has positive errors in the North Pacific and the North Atlantic and negative errors in the tropical Pacific and the southern oceans. It resembles the decadal SST mode, which has a significant influence on rainfall over the Sahel. Because the systematic SST errors were not corrected during the forecasts, persistent errors serve as an additional forcing to the atmosphere.

The second source of error is from the soil moisture feedback, which contributes to the southward shift of rainfall and dryness over West Africa. This was demonstrated by the comparison between simulations (SIMs) and the Atmospheric Model Intercomparison Project (AMIP) run. Both are forced with observed SSTs. The SIMs initialized at the end of June have realistic soil moisture and do not show the southward shift of rainfall. The AMIP, which predicts soil moisture, maintains the dryness through the summer over the Sahel. For AMIP, the decreased rainfall is contributed by the decreased evaporation (E) due to the dry soil and the shift of the large temperature gradients southward. In response, the African easterly jet (AEJ) shifts southward. Since this jet is the primary source of energy for the African waves and their associated mesoscale convective systems, these too shift southward. This negative feedback contributes to increased dryness over the Sahel.

Corresponding author address: Wassila Thiaw, NOAA/NWS/NCEP Climate Prediction Center, 5200 Auth Rd., Camp Springs, MD 20746. Email: Wassila.Thiaw@noaa.gov

Abstract

The ensemble rainfall forecasts over the Sahel for July–September (JAS) from the NCEP Coupled Forecast System (CFS) were evaluated for the period 1981–2002. The comparison with the gauge-based precipitation analysis indicates that the predicted Sahel rainfall is light and exhibits little interannual variability. The rain belt is shifted about 4° southward.

One major source of rainfall errors comes from the erroneous sea surface temperature (SST) forecasts. The systematic SST error pattern has positive errors in the North Pacific and the North Atlantic and negative errors in the tropical Pacific and the southern oceans. It resembles the decadal SST mode, which has a significant influence on rainfall over the Sahel. Because the systematic SST errors were not corrected during the forecasts, persistent errors serve as an additional forcing to the atmosphere.

The second source of error is from the soil moisture feedback, which contributes to the southward shift of rainfall and dryness over West Africa. This was demonstrated by the comparison between simulations (SIMs) and the Atmospheric Model Intercomparison Project (AMIP) run. Both are forced with observed SSTs. The SIMs initialized at the end of June have realistic soil moisture and do not show the southward shift of rainfall. The AMIP, which predicts soil moisture, maintains the dryness through the summer over the Sahel. For AMIP, the decreased rainfall is contributed by the decreased evaporation (E) due to the dry soil and the shift of the large temperature gradients southward. In response, the African easterly jet (AEJ) shifts southward. Since this jet is the primary source of energy for the African waves and their associated mesoscale convective systems, these too shift southward. This negative feedback contributes to increased dryness over the Sahel.

Corresponding author address: Wassila Thiaw, NOAA/NWS/NCEP Climate Prediction Center, 5200 Auth Rd., Camp Springs, MD 20746. Email: Wassila.Thiaw@noaa.gov

1. Introduction

The Sahel is one of the world’s most vulnerable areas to climate variability on all time scales. The causes of recurrent droughts in the Sahel have been investigated for decades. However, a consensus on the mechanisms associated with Sahel rainfall variability has not yet been achieved. The Sahel rainfall responds to both sea surface temperatures (SSTs) and land surface conditions, including soil moisture and vegetation. On the interannual time scales, El Niño–Southern Oscillation (ENSO) influences the variability of Sahel rainfall (Janicot et al. 1996; Ward 1998; Thiaw et al. 1999). On the interdecadal time scales, a decadal SST mode with positive SSTs in the North Pacific and the North Atlantic and with negative SSTs in the southern oceans is essential in predicting rainfall over the Sahel (Folland et al. 1986; Barnston et al. 1996; Chelliah and Bell 2004; Thiaw and Bell 2005; Ward 1998). Giannini et al. (2003) confirmed the role of equatorial Pacific SSTs on the interannual time scales, but suggested that warming of the oceans around Africa, in particular the warming trend in the Indian Ocean, may have favored a shift of convection from land to ocean resulting in the long-term rainfall deficit in the Sahel.

It is also well known that Sahel rainfall is dependent upon the northward displacement of the intertropical convergence zone (ITCZ), which reaches its northernmost position during August. In summer, the large-scale flow is characterized by the low-level southwesterlies and an upper-level easterly flow. There are two easterly wind maxima between 10° and 14°N. The northernmost jet is the midtropospheric wind maximum at 600–700 hPa referred to as the African easterly jet (AEJ). It appears from late spring to summer and is closely related to the Sahel rainfall. Some have suggested that the strength of the AEJ modulates rainfall in the Sahel such that a wet (dry) Sahel is associated with a weak (strong) AEJ (Newell and Kidson 1984; Thorncroft and Rowell 1998; Landsea et al. 1992). Others argued that the exact position of the AEJ may be more relevant to the Sahel rainfall (Thiaw et al. 1999). A southward-displaced AEJ is often associated with dry Sahel years.

Cook (1999) suggested that the AEJ is a response to soil moisture and temperature gradients across West Africa. It acts only to maintain and provide energy to African wave disturbances that produce rainfall. Koster and Suarez (2003) studied the impact of land surface initialization on seasonal temperature and precipitation forecasts. They found that land surface initialization has a large impact on surface temperature over West Africa but only a slight impact on precipitation. Douville et al. (2001) and Douville (2002) found that the soil–precipitation feedback is positive over the Sahel. They suggested that decreased soil moisture is associated with low precipitation rates. Thus, soil moisture contributes to the interannual variability of the Sahel rainfall.

The upper-tropospheric wind maximum known as the tropical easterly jet (TEJ) is the southernmost jet and is located between 100 and 200 hPa. This jet is global in nature and is more sensitive to global SST changes than to soil moisture (Douville 2002; Rowell et al. 1995). When the TEJ extends farther to the west, it provides high vertical wind shear across the Gulf of Guinea region and enables convection to move northward into the Sahel. An eastward retraction of the TEJ is evidence for a weakening of the subtropical ridges across both hemispheres and less convection across the Sahel. It is this mechanistic complexity of the Sahel climate that makes the Sahel a unique place to test new models and to improve upon the description of our understanding of climate variability and change.

The Environmental Modeling Center (EMC) at the National Centers for Environmental Prediction (NCEP) has recently completed hindcasts from 1981 to 2004 based on the new coupled forecast system (CFS; Saha et al. 2005, manuscript submitted to J. Climate). The forecasts for the rainy season in the Sahel during July–September (JAS) show a very dry Sahel and suppressed interannual variability. Here, the Sahel is defined as the area (12.5°–17.5°N, 17.5°W–20°E), which corresponds approximately to the area chosen by Rowell et al. (1995) and Nicholson (1980). This gives us a unique opportunity to examine the influence of boundary forcing: both SSTs and soil moisture influence rainfall over the Sahel. The purposes of this paper are 1) to document the forecast errors of seasonal rainfall over the Sahel, and 2) to diagnose the causes of model errors and to understand the underlying physics that governs rainfall variability over the Sahel. This is achieved by comparing the CFS forecasts with model seasonal simulations (SIMs) and the Atmospheric Model Intercomparison Project (AMIP) runs. Both are forced with the observed SSTs. Since SSTs were not corrected during the coupled forecasts, the comparison between the CFS forecasts, the SIMs, and AMIP indicate the impact of SSTs. The SIMs starts each June and have the most current knowledge of soil moisture and surface fluxes from the initial conditions. The AMIP run is a continuous run and soil moisture is predicted by the model; therefore, it does not have the most recent information on soil moisture. The comparison between the SIMs and the AMIP allows us to examine the impact of soil moisture on rainfall over the Sahel.

All experiments studied here are discussed in section 2. The evaluation of the CFS experiments is given in section 3. The impact of SST errors on rainfall over the Sahel is examined in section 4 and the role of the soil moisture feedback is discussed in section 5. Conclusions and discussions are given in section 6.

2. Experiments

a. The CFS forecasts

The NCEP CFS is a fully coupled ocean–atmosphere model (Saha et al. 2005, manuscript submitted to J. Climate). The atmospheric general circulation model (AGCM) is the global forecast system (GFS) with the horizontal resolution T62 (approximately 200 km) with 64 vertical levels. There are 14 levels below 850 hPa so the boundary layers are well represented. (A detailed description of the GFS model can be found online at http://www.emc.ncep.noaa.gov.) Soil moisture and soil temperature are predicted by the model after initialization from the NCEP reanalysis (R2; Kanamitsu et al. 2002). The soil-related parameters are described in Chen and Dudhia (2001a, b). The model has the climatological monthly mean vegetation fraction and vegetation types. There is no vegetation–atmosphere interaction in the model. The convection scheme is the simplified Arakawa–Schubert (SAS; Arakawa and Schubert 1974) modified by Grell (1993). The same GFS model was used to perform simulations and the AMIP runs.

The ocean model is the Geophysical Fluid Dynamics Laboratory (GFDL) version 3 Modular Ocean Model (MOM3). In the Tropics, the resolution is 1° in the zonal direction and 0.3° in the meridional direction. In the midlatitudes, the resolution is 1°. There are 40 layers in the vertical. There is no flux correction and the coupling is applied once a day for the area from 74°S to 65°N. There is no ice model associated with the GFS so the monthly mean climatological ice data are supplied.

The coupled model forecasts for JAS starting in June each year from 1981 to 2001 were taken from the CFS archive. We concentrate on the JAS season because it is the rainy season over the Sahel. There are 15 members in the ensemble with three five-member clusters centered at 1 June, 11 June, and 21 June, respectively. The ensemble mean is the equal weighted mean averaged over the 15 members. The experiments were initialized from the Global Ocean Data Assimilation System (GODAS) and the R2 (Kanamitsu et al. 2002). The R2 uses observed precipitation to adjust soil moisture so the soil conditions should be realistic.

b. Ensemble simulations (SIMs)

Ensemble simulations for JAS were performed for 12 yr from 1990 to 2001. The AGCM is the same GFS T62L64 model used to perform the CFS forecasts. There are five members in the ensemble that are initialized with different initial conditions 6 h apart during 29–30 June each year. The initial conditions were taken from the R2. The AGCM is forced with the observed SSTs from the monthly mean optimum interpolation SST analysis (Reynolds and Smith 1994). The SIMs can be viewed as the atmospheric responses to perfect SST forecasts. Since the initial conditions are close to the target season, the simulations also have the current information on soil moisture and surface fluxes.

c. AMIP

The AMIP run was initiated on 1 December 1949 and continued through 31 December 2001. The AMIP is forced with the observed SSTs. The initial conditions were taken from the NCEP–National Center for Atmospheric Research (NCAR) reanalysis (R1; Kalnay et al. 1996) and later continued as the Climate Data Assimilation System (CDAS). The AGCM is the same GFS T62L64 model used in the SIMs and the CFS forecasts.

d. Corrected CFS forecasts

To isolate the influence of the systematic SST errors, a set of ensemble forecasts was performed for 12 yr from 1990 to 2001 using the GFS T62L28 model with 28 vertical levels. There are five members in the ensemble and the initial conditions are the same as the SIMs.

The SSTs used are the forecasted SSTs from the coupled model with the systematic errors corrected. For a given year, the model climatology was computed as the mean of predicted SST from 1981 to 2002 with that year removed. The observed SST climatology for the same years was obtained. The difference between the model climatology and the observed climatology was subtracted from the predicted SST for that year to obtain the corrected SST.

In addition to the corrected SSTs, this run differs from the CFS forecasts in two ways: (i) The CFS (corrected) forecasts have only five members. They all started from the initial conditions very close to 1 July. The CFS forecasts have 15 members and many members that start around 11 or 22 June do not have the most recent information on soil moisture and surface fluxes.

3. Precipitation and the AEJ from the CFS forecasts

In this paper, the area defined as the Sahel encompasses the region (12.5°–17.5°N, 17.5°W–20°E). The verifying rainfall dataset is the monthly gridded precipitation analysis (P analysis) based on gauge observations on a 2.5° latitude–longitude grid (Chen et al. 2002).

The mean CFS precipitation (P) JAS forecasts over West Africa for the period 1981–2001 (Fig. 1b) and the period 1990–2001 (Fig. 1c) are similar. They both show a rainband extending from the African west coast to the Sudan with two centers located near Senegal and the Guinea coast. In comparison to the P analysis (Fig. 1a), the CFS shows a 4° southward shift in the rainband. There is very little rain north of 16°N. The Sahel is very dry and the mean rainfall averaged during 1990–2001 is only 1.81 mm day−1 in comparison with 2.99 mm day−1 from the P analysis. The largest rainfall variability is located over the heavy rainfall areas. The maxima of standard deviation are collocated with the rainfall maxima (Fig. 2c). Over the Sahel, except the coastal areas, the standard deviation is only about 0.3 mm day−1 for the CFS forecasts (Fig. 2d), which is lower than the P analysis (Fig. 2a). It is clear that the suppressed rainfall variability in the CFS forecasts over the Sahel influences the P forecasts on the interannual time scale. The CFS model underestimates rainfall as well as rainfall variability (Fig. 2a, plus signs). The range of year-to-year variation is only about 1.5–2 mm day−1, while the P analysis (dark line) has a range from 2 to 4 mm day−1. The anomaly correlation between the CFS and the P analysis is 0.26 and the root-mean-square error (rmse) is 1.26 mm day−1. These are not statistically significant at the 5% level.

Both SIMs (Fig. 1f) and the CFS-corrected forecasts (Fig. 1d) show two maxima clearly separated from each other. The CFS (corrected) shows wetter West Africa than the P analysis. Both SIMs and CFS (corrected) do not have the systematic SST errors and clearly show more variability than the CFS forecasts. This implies that the systematic SST errors in the CFS serve as a constant oceanic forcing to the atmosphere. Since the errors are present each year, they may be responsible for the suppressed variability of Sahel rainfall in the CFS forecasts.

The SIMs show more rainfall than the CFS north of 16°N and compares more favorably with the P analysis (Fig. 2a, dark squares). The mean P over the Sahel for the SIMs is 2.67 mm day−1, which is close to 2.99 mm day−1 from the P analysis. The anomaly correlation between the SIMs and the P analysis is 0.57 and the rmse is 0.52, which is statistically significant at the 10% level. The CFS (corrected) has the highest mean P value of 3.91 mm day−1, but the anomaly correlation is only 0.28 and the rmse is 1.06 mm day−1, which is not statistically significant.

The fact that the SIMs and the CFS (corrected), which start from the nearest initial conditions at the end of June, do not show the southward shift of the rainband as in the AMIP and the CFS, suggests that soil moisture and surface fluxes may play a role in the southward shift of the rainband. In addition, the comparison between the AMIP and the P analysis for the period 1950–2000 reveals that the AMIP does not capture the decadal trends in the Sahel rainfall (Fig. 2b).

There is a close association between P and the AEJ. As noted by Thorncroft and Hoskins (1994a, b), African waves can develop through barotropic and baroclinic instability and these disturbances influence the P distribution (Payne and McGarry 1977; Rowell and Milford 1993). The southward shift of the rainband is accompanied by the southward shift of the AEJ. The AEJ is in general represented by the zonal wind at the 600-hPa level (Cook 1999). Unfortunately, the 600-hPa winds were not archived so the climatology of zonal wind at 700 hPa was used instead (Fig. 3c). The AEJ from R1 (Figs. 3a,b) extends from Chad, in the central Sahel, westward into the Atlantic coast of Africa with two maxima (jet streaks) located along the coast and over northern Nigeria (12°–14°N, 0°–10°E), respectively. The CFS captures the location of the maximum along the coast, but the inland jet streak, which is quite weak is shifted southward to about 8°N. The AMIP depicts a single jet streak also with an inland southward tilt. This southward tilt of the AEJ in both the CFS and the AMIP is consistent with the southward shift of the rainband. The CFS (corrected) and the SIMs both reveal a single jet streak AEJ that extends farther inland with a maximum located between 10° and 14°N. In the next two sections, the influence of SST errors and soil moisture on the CFS forecasts is discussed.

4. SST errors

The systematic SST error is defined as the difference between the predicted JAS SSTs and the SSTs from the CDAS (Fig. 4), which are the same as the observed SSTs. The largest SST errors are located in the North Pacific. Positive SST differences as large as 3°C extend from the coast of Japan to the west coast of North America. Positive errors are also located in the North Atlantic and the west coast of South America. There are cold biases as large as 2°C located in the tropical Pacific. Negative SST differences are located over the southern oceans. The interhemispheric anomaly pattern with an out-of-phase relationship between the northern and the southern oceans is known to be associated with rainfall anomalies over the Sahel (Mo et al. 2001; Rowell et al. 1995; Ward 1998).

These errors occur every summer and persist from one month to another. Figure 5 shows the time–longitude plots of SST anomalies over the Tropics (10°S–10°N), and over the North Pacific (35°–55°N) from the CDAS and the CFS SST forecasts, respectively. All anomalies are defined as the departures from the CDAS mean from 1981–2001. The systematic errors from the CFS are not corrected. For the CDAS, positive (negative) SSTs extend from the coast of South America to the central Pacific during warm (cold) ENSO years. The predicted SSTs show cooling over the tropical Pacific through the period with weaker positive SST anomalies during warm ENSO years (Fig. 5b). In the North Pacific, positive errors persist (Fig. 5e). SST anomalies with the systematic errors corrected have very high skill. For example, the anomaly correlations for SSTs in the Niño-3.4 area and in the North Pacific are 0.86 and 0.76, respectively (not shown). However, the systematic errors are not removed during the CFS forecasts. These errors serve as erroneous forcing to the AGCM. After the error correction, the SSTs in the Tropics extend into the central Pacific for the 1997 warm events, but the SSTs for 1999–2000 are still well below average (Fig. 5c). The largest deficiency is in the North Pacific. There are very weak SST anomalies in the North Pacific. The large negative SST anomalies in 1993 and positive anomalies in 1994 are absent (Fig. 5f). Since these errors do not appear each year, they are not corrected. Next, we will show that the CFS systematic SST errors have a large projection onto the SST mode, which has a large correlation with rainfall over the Sahel. Therefore, a part of rainfall errors can be attributed to the SST errors.

Rotated EOFs (REOFs) were computed for the JAS mean observed SSTs for the study period 1950–2001. An EOF analysis was performed on SSTAs. The anomalies were not normalized, but a latitudinal, cosine-weighting factor was included in computing the covariance matrix. The first 16 EOFs entered the VARIMAX rotation to obtain REOFs. The observed and the forecasted SST anomalies were projected onto REOFs to obtain the principal components (RPCs). The leading REOFs and corresponding RPCs are given in Fig. 6. The anomaly at each grid point is defined as the departure from the observed SST mean from 1950 to 2001. Correlations between the observed RPCs and the observed JAS rainfall over tropical West Africa were calculated (Fig. 7). Correlation values above 0.27 are statistically significant at the 5% level (Fig. 7).

The first REOF represents cold ENSO with negative anomalies in the tropical Pacific and positive anomalies over the North and South Pacific. The RPCs are positive during the cold ENSO phase and negative during the warm ENSO phase. The CFS forecasts capture the ENSO signal well as indicated by RPC 1 (Fig. 6d, dark circles), and the magnitudes are only slightly higher than the RPC 1 from the observed SSTs. If the RPC-1 mean is defined as the observed RPC-1 mean from 1981 to 2001, then the correlation between the observed RPC 1 and the CFS-predicted RPC 1 for the 1981–2001 period is 0.67. The correlation increases to 0.97 after the removal of the systematic mean error from the CFS-predicted RPC 1. The correlation map indicates that ENSO has strong influence on P over the Sahel. Large positive correlations are located over both Sudan and the Sahel (Fig. 6a) consistent with findings by Rowell et al. (1995), Ward (1998), Thiaw et al. (1999), Mo et al. (2001), among others.

REOF 2 shows negative SST anomalies in the tropical Pacific with a broad structure and an off-centered minimum. This mode resembles the decadal mode of Zhang et al. (1997). RPC 2 shows an abrupt decrease (SST warming) around 1965. The CFS captures this mode well (Fig. 6b). Except the coastal area over Senegal, the influence of this mode on the Sahel rainfall is limited.

REOF 3 shows positive SST loadings extending from Japan into the North Pacific and over the North Atlantic with negative loadings in the tropical Pacific and over the southern oceans. RPC 3 shows decadal trends similar to the rainfall trends over the Sahel (Fig. 2b, dark circles). This mode resembles the same low-frequency SST mode identified by Ward (1998), Rowell et al. (1995), and the interhemispheric SST mode documented by Barnston et al. (1996) and others. Many studies (Rowell et al. 1995; Ward 1998; Mo et al. 2001) indicated that this mode has a dominant influence on rainfall over the Sahel and Sudan. This is confirmed by the correlation map (Fig. 7c). REOF 3 has a higher correlation with rainfall over West Africa than REOF 1. REOF 3 resembles the SST systematic error pattern (Fig. 4). Therefore, the CFS forecasts have systematically higher RPC 3 than the observed SSTs (Fig. 6f). If the mean is defined as the observed RPC 3 mean, then the correlation between the observed RPC 3 and the CFS-predicted RPC 3 is 0.17. After the systematic error correction, the correlation increases to 0.71.

The systematic SST error pattern resembles the decadal SST trend mode. This mode has a large influence on rainfall over the Sahel. The errors persist from one year to another and so does this SST mode. Since the systematic errors were not corrected during the CFS forecasts, this mode serves as additional persistent forcing and may account for the lack of rainfall variability over the Sahel.

5. Feedback from soil moisture

The June volumetric soil moisture at the first layer (0–10 cm) averaged from 1990 to 2001 from the R2 is about 0.2 over the Sahel (Fig. 8a). In July, the soil moisture values increase to about 0.25 (Fig. 8b). The CFS forecasts started around 11 and 21 June do not have the most recent soil moisture and surface fluxes. As discussed in the above section, the CFS is influenced by the systematic SST errors and has a tendency to forecast a drier Sahel. The negative land surface feedback may increase the dryness over the Sahel. The CFS July forecasts are very dry and the soil moisture is about 0.1 over the Sahel (Fig. 8c). This soil moisture deficit extends to the entire area of West Africa. The SIMs are initialized with R2 at the end of June and contain the most recent information on soil moisture and surface fluxes. The model still has the tendency to lose soil moisture. The July mean for SIMs (Fig. 8f) shows that the mean soil moisture for the Sahel is about 0.17. Overall, it is higher than the CFS, but still about 0.05 less than the R2. The soil moisture depicted by the R2 does not have a strong interannual variability. For example, the mean July soil moisture for SIMs is 0.17 with a range from 0.16 to 0.18, while the mean July soil moisture for the R2 is 0.268 with a range from 0.26 to 0.27. The AMIP run does not have the most current information on soil moisture because it is a continuous run. For June and July, soil moisture is about 0.1 over the Sahel close to the values from the CFS forecasts. Since both SST and soil moisture influence rainfall over the Sahel, it is difficult to separate them. Thus, we compare the SIMs with the AMIP to demonstrate that the southward shift of both the rainband and the AEJ is influenced by soil moisture through the feedback mechanism from the land–atmosphere interaction.

The location of the AEJ is clearly demonstrated from the vertical profile of the zonal wind averaged from 0° to 10°E (Fig. 9). The CDAS (Fig. 9a) shows westerlies located at the boundary layer with a maximum at 925 hPa at about 10°N. Easterlies are located above 850 hPa with a maximum at 600 hPa. Easterlies above 200 hPa are associated with the TEJ. The AEJ is centered near 12°–14°N. Both the AMIP and the SIMs show a maximum at 600 hPa, but at different latitudes. The AMIP has a maximum between 8° and 10°N. The SIMs show a maximum located between 10° and 12°N, which is closer to the jet location shown in the CDAS. Note that the midlevel easterlies outside of the jet core region are much stronger in the AMIP and the SIMs than in the CDAS. Associated with these strong easterlies, the low-level westerlies are suppressed and do not extend as far north in both the AMIP and the SIMs as in the CDAS. In contrast, the TEJ is much weaker in both the AMIP and the SIMs than in the CDAS. This results in weaker vertical wind shears over the Gulf of Guinea region and less favorable conditions for convection across the Sahel in the AMIP and the SIMs.

To demonstrate the role played by soil moisture, the mean differences of surface fluxes and temperature from the ensemble mean of SIMs and the AMIP averaged from 1990 to 2001 are plotted in Fig. 10. In Fig. 10a, we plotted the difference in total soil moisture from 0 to 200 cm between the SIMs and the AMIP. It shows that the dryness in the AMIP is not confined at the top layer of soil and the SIMs exhibit more soil moisture over the Sahel and Sudan than the AMIP (Fig. 10a). The wet soil moisture implies more evaporation (E; Fig. 10b). The SIMs evaporates more over the Sahel as indicated by the large E differences near the location of the jet maximum (12°N, 5°–10°E). Evaporation contributes to P over the area north of 14°N in the Sahel, while the contribution from the moisture convergence (EP) is mostly located south of 14°N. Cook (1999) did not find direct contribution from E to P. The discrepancy may be caused by the model differences and the detailed distribution of soil moisture. The CFS has a higher resolution and an improved physics over the GFDL model used by Cook (1999).

Evaporation is balanced by sensible heat because the difference in radiation is small. Sensible heat is closely related to surface temperature. Therefore, the AMIP has higher temperatures over the Sahel than the SIMs. The differences in soil moisture create positive surface temperature gradients between the Sahel and the equatorial Africa. For the SIMs, the tightest temperature gradients are located around 13°N while the AMIP has the tight gradients located farther south. The temperature differences also extend to the lower troposphere as indicated by the vertical profile of temperature gradients averaged from 0° to 10°E (Figs. 9d–f). The largest temperature gradients have a strong influence on the jet. The large temperature gradients in the AMIP are located farther south than in the SIMs, consistent with the southward shift of the AEJ as suggested by Cook (1999). Once the jet is shifted southward, the African wave disturbances related to the jet are also likely to shift southward. This results in wetter soils, more evapotranspiration, and stronger soil moisture gradients in the Gulf of Guinea region. The AEJ, which is a geostrophic consequence of soil moisture gradients, tends to remain to the south of the Sahel. It is this negative feedback that makes the Sahel even drier.

6. Conclusions

Precipitation forecasts over the Sahel from the NCEP coupled forecast system (CFS) model were compared to the gauge rainfall analysis. The CFS ensemble forecasts for JAS from initial conditions in June show a southward shift in the West African rainband. This leaves the Sahel very dry. The southward shift of the rainband is accompanied by the southward shift of the AEJ. The CFS forecasts also do not capture the interannual variability in the Sahel rainfall quite adequately. The suppressed interannual variability in P suggests the existence of persistent erroneous forcing. The model simulations and CFS (corrected) have better representation of the position of the AEJ and the spatial distribution of rainfall across West Africa. They also show more realistic interannual rainfall variability.

Part of the P errors comes from the SST systematic errors. For the forecasts on the seasonal time scales, SSTs have a dominant influence on rainfall over the Sahel (Rowell et al. 1995). The systematic error pattern is similar to the decadal SST mode (Rowell et al. 1995; Ward 1998). It shows positive SSTs over the North Pacific and the North Atlantic and negative errors in the tropical Pacific and the southern oceans. During the CFS forecasts, the systematic errors are not corrected so they serve as additional forcing. The persistence of the errors in the SST pattern may cause the errors in rainfall magnitudes and the suppressed variability. The CFS model does not have a realistic ice model as a subcomponent. The ice information is supplied through the mean monthly climatology and the ocean coupling is limited to the south of 65°N. These model deficiencies may contribute to the warming over the North Pacific and the North Atlantic.

In addition to the SST errors, the soil moisture feedback mechanism as proposed by Cook (1999) may also contribute to the southward shift of the AEJ and rainfall. This is demonstrated by the comparison between the AMIP run and the simulations. Both experiments are forced with the observed SSTs. The main differences are in soil moisture and surface fluxes. The SIMs are initialized from the R2 in June and have realistic information on soil moisture and surface fluxes. The AMIP, which is a continuous run, does not have such information. The AMIP run shows the southward shift of the AEJ, while the SIMs provide a better representation of the jet location and rainfall spatial pattern.

As expected, the AMIP run has less soil moisture over the Sahel and less E. Evaporation contributes to P directly, but the largest influence is indirect through the temperature gradients. The radiation differences are smaller so E is balanced by sensible heat. Less E implies more sensible heat and indeed the AMIP is warmer over the Sahel than the SIMs. This implies that the largest temperature gradients over West Africa in the AMIP are located farther south than in the SIMs. The temperature gradients reach the midtroposphere. This serves as a forcing to move the AEJ southward (Cook 1999). In response, The African wave disturbances, which account for much of the rains in the Sahel, shift southward resulting in dryness over the Sahel.

Ward (1998) indicated that the most important contribution to rainfall variability over the Sahel is the decadal mode. The AMIP forced with the observed SSTs does not capture the decadal changes in rainfall. The model does not have interactive vegetation fraction and does not use the information of the leaf area index (LAI). The vegetation fraction is supplied to the model through the monthly mean vegetation fraction climatology. Therefore, it is not able to simulate the changes of albedo and surface fluxes due to the greenness of vegetation. Wang and Eltahir (2000) suggested that the vegetation dynamics is a significant process in simulating rainfall over the Sahel. The decadal variability, which is the essential part of the rainfall variability over the Sahel, is better produced when the interactive vegetation is added to the model (Zeng et al. 1999). Therefore, a land surface interaction model coupled with the CFS will improve precipitation forecasts over the Sahel.

REFERENCES

  • Arakawa, A., and W. H. Schubert, 1974: Interaction of a cumulus ensemble with the large scale environment, Part I. J. Atmos. Sci., 31 , 674704.

    • Search Google Scholar
    • Export Citation
  • Barnston, G. T., W. Thiao, and V. Kumar, 1996: Long-lead forecasts of seasonal precipitation in Africa using CCA. Wea. Forecasting, 11 , 506520.

    • Search Google Scholar
    • Export Citation
  • Chelliah, M., and G. D. Bell, 2004: Tropical multidecadal and interannual climate variability in the NCEP–NCAR reanalysis. J. Climate, 17 , 17771803.

    • Search Google Scholar
    • Export Citation
  • Chen, F., and J. Dudhia, 2001a: Coupling an advanced land surface hydrology model with the Penn State/NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129 , 569586.

    • Search Google Scholar
    • Export Citation
  • Chen, F., and J. Dudhia, 2001b: Coupling an advanced land surface hydrology model with the Penn State/NCAR MM5 modeling system. Part II: Preliminary model validation. Mon. Wea. Rev., 129 , 587604.

    • Search Google Scholar
    • Export Citation
  • Chen, M., P. Xie, J. Janowiack, and P. Arkin, 2002: Global land precipitation: A 50-yr monthly analysis based on gauge observations. J. Hydrometeor., 3 , 249266.

    • Search Google Scholar
    • Export Citation
  • Cook, K. H., 1999: Generation of the African easterly jet and its role in determining West African precipitation. J. Climate, 12 , 11651184.

    • Search Google Scholar
    • Export Citation
  • Douville, H., 2002: Influence of soil moisture on the Asian and African monsoons. Part II: Interannual variability. J. Climate, 15 , 701720.

    • Search Google Scholar
    • Export Citation
  • Douville, H., F. Chauvin, and H. Broqua, 2001: Influence of soil moisture on the Asian and African monsoons. Part I: Mean monsoon and daily precipitation. J. Climate, 14 , 23812403.

    • Search Google Scholar
    • Export Citation
  • Folland, C. K., T. N. Palmer, and D. E. Parker, 1986: Sahel rainfall and world-wide sea temperatures, 1901–85. Nature, 320 , 602607.

    • Search Google Scholar
    • Export Citation
  • Giannini, A., R. Saravanan, and P. Chang, 2003: Oceanic forcing of Sahel rainfall on interannual to interdecadal timescales. Science, 302 , 10271030.

    • Search Google Scholar
    • Export Citation
  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterization. Mon. Wea. Rev., 121 , 764787.

  • Janicot, S., V. Moron, and B. Fontaine, 1996: Sahel droughts and ENSO dynamics. Geophys. Res. Lett., 23 , 515518.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kanamitsu, M., W. Ebisuzaki, J. Woolen, S. K. Yang, J. J. Hnilo, M. Fiorina, and G. L. Potter, 2002: NCEP/DOE AMIP-II reanalysis (R-2). Bull. Amer. Meteor. Soc., 83 , 16311643.

    • Search Google Scholar
    • Export Citation
  • Koster, R. D., and M. J. Suarez, 2003: Impact of land surface initialization on seasonal precipitation and temperature prediction. J. Hydrometeor., 4 , 408423.

    • Search Google Scholar
    • Export Citation
  • Landsea, C. W., P. W. Mielke Jr., and K. J. Berry, 1992: Long-term variations of western Sahelian monsoon rainfall and intense U.S. landfalling hurricanes. J. Climate, 5 , 15281534.

    • Search Google Scholar
    • Export Citation
  • Mo, K. C., G. D. Bell, and W. M. Thiaw, 2001: Impact of sea surface temperature anomalies on the Atlantic tropical storm activity and West African rainfall. J. Atmos. Sci., 58 , 34773496.

    • Search Google Scholar
    • Export Citation
  • Newell, R. E., and J. W. Kidson, 1984: African mean wind changes between Sahelian wet and dry periods. J. Climatol., 4 , 2733.

  • Nicholson, S. E., 1980: The nature of rainfall fluctuations in subtropical West Africa. Mon. Wea. Rev., 108 , 473487.

  • Payne, S. W., and M. M. McGarry, 1977: The relationship of satellite inferred convective activity to easterly waves over west Africa and the adjacent ocean during phase III of GATE. Mon. Wea. Rev., 105 , 413420.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses. J. Climate, 7 , 929948.

  • Rowell, D. P., and J. R. Milford, 1993: On the generation of African squall lines. J. Climate, 6 , 11811193.

  • 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 , 669704.

    • Search Google Scholar
    • Export Citation
  • Thiaw, W. M., and G. D. Bell, 2005: Mechanisms associated with the June–September 2003 Sahel rainfall and implications for seasonal climate forecasts. CLIVAR Exchanges, Vol. 10, No. 1, 29–31.

  • Thiaw, W. M., A. B. Barnston, and V. Kumar, 1999: Predictions of African rainfall on the seasonal time scale. J. Geophys. Res., 104 , D24,. 3158931597.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., and B. J. Hoskins, 1994a: An idealized study of African easterly waves. I. A linear view. Quart. J. Roy. Meteor. Soc., 120 , 953982.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., and B. J. Hoskins, 1994b: An idealized study of African easterly waves. II. A nonlinear view. Quart. J. Roy. Meteor. Soc., 120 , 9831015.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., and D. P. Rowell, 1998: Interannual variability of African wave activity in a general circulation model. Int. J. Climatol., 18 , 13051323.

    • Search Google Scholar
    • Export Citation
  • Wang, G., and E. A. B. Eltahir, 2000: Role of vegetation dynamics in enhancing the low frequency variability of the Sahel rainfall. Water Resour. Res., 36 , 10131021.

    • Search Google Scholar
    • Export Citation
  • 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 , 31673190.

    • Search Google Scholar
    • Export Citation
  • Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability 1900–93. J. Climate, 10 , 10041020.

  • 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 , 15371540.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Mean precipitation for JAS from (a) the P analysis from 1990 to 2001, (b) ensemble mean CFS forecasts from 1981 to 2001, (c) ensemble mean CFS forecasts from 1990 to 2001, (d) ensemble mean CFS (corrected) forecasts from 1990 to 2001, (e) AMIP mean from 1990 to 2001, and (f) ensemble mean SIMs from 1990 to 2001. Contour intervals are 1, 2, 4, 6, 8, 10, 12, 15, and 20 mm day−1. Values greater than 8 (12) mm day−1 are shaded light (dark).

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 2.
Fig. 2.

(a) Seasonal JAS mean precipitation over the Sahel (12.5°–17.5°N, 17.5°W–20°E) from the gauge analysis (solid line), CFS forecasts (plus signs), CFS (corrected; open circles), and SIMs (dark squares). (b) Same as in (a) but for the AMIP (open circles) and the gauge analysis (dark circles) from 1950 to 2001. Std dev for JAS for the period 1990–2001 from (c) the gauge analysis, (d) CFS forecasts, and (e) CFS (corrected). Contour interval is 0.3 mm day−1. Values greater than 1.2 (1.5) mm day−1 are shaded light (dark).

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 3.
Fig. 3.

(a) Mean zonal wind at 600 hPa from the CDAS, averaged from 1990 to 2001. Contour interval 2 m s−1. Values less than −8 m s−1 are shaded. (b) Same as in (a), but for the mean zonal wind at 700 hPa from the CDAS. (c) Same as in (b), but from the ensemble CFS forecasts from 1990–2001. (d) Same as in (a), but ensemble mean from the CFS (corrected) forecasts. (e) Same as in (a) but for SIMs. (f) Same as in (a) but for the AMIP.

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 4.
Fig. 4.

SST difference between the ensemble SST forecasts for JAS with the initial conditions in Jun averaged from 1981 to 2001 and the corresponding SSTs from the CDAS. Contour interval is 1°C. Positive values are shaded.

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 5.
Fig. 5.

(a) Time–longitude plot of the CDAS SST JAS mean difference averaged from 10°S to 10°N and the CDAS mean for the period 1981–2001. Contour interval is 0.5°C. Positive values greater than 1°C are shaded. (b) Same as in (a), but for the difference between the CFS ensemble forecasts and the CDAS mean. (c) Same as in (a), but for the CFS (corrected) ensemble forecasts. (d) Same as in (a), but for the SST difference averaged from 35° to 55°N. (e) Same as in (d), but for the difference between the CFS forecasts and the CDAS mean. (f) Same as in (d), but for the CFS (corrected) ensemble forecasts.

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 6.
Fig. 6.

(a) REOF 1, (b) REOF 2, and (c) REOF 3. Contour interval is 0.5 nondimensional units. Negative values are shaded. (d) RPC 1, (e) RPC 2, and (f) RPC 3 for the observed SSTs (open circles) and for the SST anomalies from the CFS forecasts after removing the observed SST mean (dark circles).

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 7.
Fig. 7.

Correlation between the observed rainfall and RPC 1 from observed SSTs for (a) RPC 1, (b) RPC 2, and (c) RPC 3 for the period 1950–2001. Contour interval is 0.1. Values statistically significant at the 5% level are shaded.

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 8.
Fig. 8.

(a) Mean volumetric soil moisture from 0 to 10 cm for Jun averaged from 1990 to 2001 from the R2. Contour interval is 0.05. Values greater than 0.3 are shaded. (b) Same as in (a), but for Jul. (c) Same as in (a), but for the CFS ensemble forecasts in Jul. (d) Same as in (a), but for AMIP. (e) Same as in (b), but for the AMIP. (f) Same as in (c), but for SIMs in Jul.

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 9.
Fig. 9.

Vertical profile of the zonal wind averaged over 0°–10°E from the (a) CDAS, (b) AMIP, (c) SIMs. Contour interval is 2 m s−1. Values less than 6 m s−1 are shaded. (d), (e), (f) Same as in (a)–(c), but for dT day−1 averaged over 5°W–15°E. Contour interval is 2 × 10−6 K m−1. Values greater than 8 × 10−6 K m−1 are shaded.

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Fig. 10.
Fig. 10.

(a) Difference in soil moisture from 0 to 200 cm between the SIMs and AMIP for JAS averaged from 1990 to 2001. Contour interval is 0.02. Values greater than 0.06 are shaded. (b) Same as in (a), but for evaporation difference. Contour interval is 0.5 mm day−1. Values greater than 0.5 mm day−1 are shaded. (c) Same as in (b), but for P. (d) Same as in (b), but for EP. (e) Ensemble JAS mean temperature at 2 m from the SIMs for 1990–2001. Contour interval is 2°C. Values greater than 32°C are shaded. (f) Same as in (e), but for AMIP.

Citation: Journal of Climate 18, 24; 10.1175/JCLI3552.1

Save
  • Arakawa, A., and W. H. Schubert, 1974: Interaction of a cumulus ensemble with the large scale environment, Part I. J. Atmos. Sci., 31 , 674704.

    • Search Google Scholar
    • Export Citation
  • Barnston, G. T., W. Thiao, and V. Kumar, 1996: Long-lead forecasts of seasonal precipitation in Africa using CCA. Wea. Forecasting, 11 , 506520.

    • Search Google Scholar
    • Export Citation
  • Chelliah, M., and G. D. Bell, 2004: Tropical multidecadal and interannual climate variability in the NCEP–NCAR reanalysis. J. Climate, 17 , 17771803.

    • Search Google Scholar
    • Export Citation
  • Chen, F., and J. Dudhia, 2001a: Coupling an advanced land surface hydrology model with the Penn State/NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129 , 569586.

    • Search Google Scholar
    • Export Citation
  • Chen, F., and J. Dudhia, 2001b: Coupling an advanced land surface hydrology model with the Penn State/NCAR MM5 modeling system. Part II: Preliminary model validation. Mon. Wea. Rev., 129 , 587604.

    • Search Google Scholar
    • Export Citation
  • Chen, M., P. Xie, J. Janowiack, and P. Arkin, 2002: Global land precipitation: A 50-yr monthly analysis based on gauge observations. J. Hydrometeor., 3 , 249266.

    • Search Google Scholar
    • Export Citation
  • Cook, K. H., 1999: Generation of the African easterly jet and its role in determining West African precipitation. J. Climate, 12 , 11651184.

    • Search Google Scholar
    • Export Citation
  • Douville, H., 2002: Influence of soil moisture on the Asian and African monsoons. Part II: Interannual variability. J. Climate, 15 , 701720.

    • Search Google Scholar
    • Export Citation
  • Douville, H., F. Chauvin, and H. Broqua, 2001: Influence of soil moisture on the Asian and African monsoons. Part I: Mean monsoon and daily precipitation. J. Climate, 14 , 23812403.

    • Search Google Scholar
    • Export Citation
  • Folland, C. K., T. N. Palmer, and D. E. Parker, 1986: Sahel rainfall and world-wide sea temperatures, 1901–85. Nature, 320 , 602607.

    • Search Google Scholar
    • Export Citation
  • Giannini, A., R. Saravanan, and P. Chang, 2003: Oceanic forcing of Sahel rainfall on interannual to interdecadal timescales. Science, 302 , 10271030.

    • Search Google Scholar
    • Export Citation
  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterization. Mon. Wea. Rev., 121 , 764787.

  • Janicot, S., V. Moron, and B. Fontaine, 1996: Sahel droughts and ENSO dynamics. Geophys. Res. Lett., 23 , 515518.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kanamitsu, M., W. Ebisuzaki, J. Woolen, S. K. Yang, J. J. Hnilo, M. Fiorina, and G. L. Potter, 2002: NCEP/DOE AMIP-II reanalysis (R-2). Bull. Amer. Meteor. Soc., 83 , 16311643.

    • Search Google Scholar
    • Export Citation
  • Koster, R. D., and M. J. Suarez, 2003: Impact of land surface initialization on seasonal precipitation and temperature prediction. J. Hydrometeor., 4 , 408423.

    • Search Google Scholar
    • Export Citation
  • Landsea, C. W., P. W. Mielke Jr., and K. J. Berry, 1992: Long-term variations of western Sahelian monsoon rainfall and intense U.S. landfalling hurricanes. J. Climate, 5 , 15281534.

    • Search Google Scholar
    • Export Citation
  • Mo, K. C., G. D. Bell, and W. M. Thiaw, 2001: Impact of sea surface temperature anomalies on the Atlantic tropical storm activity and West African rainfall. J. Atmos. Sci., 58 , 34773496.

    • Search Google Scholar
    • Export Citation
  • Newell, R. E., and J. W. Kidson, 1984: African mean wind changes between Sahelian wet and dry periods. J. Climatol., 4 , 2733.

  • Nicholson, S. E., 1980: The nature of rainfall fluctuations in subtropical West Africa. Mon. Wea. Rev., 108 , 473487.

  • Payne, S. W., and M. M. McGarry, 1977: The relationship of satellite inferred convective activity to easterly waves over west Africa and the adjacent ocean during phase III of GATE. Mon. Wea. Rev., 105 , 413420.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses. J. Climate, 7 , 929948.

  • Rowell, D. P., and J. R. Milford, 1993: On the generation of African squall lines. J. Climate, 6 , 11811193.

  • 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 , 669704.

    • Search Google Scholar
    • Export Citation
  • Thiaw, W. M., and G. D. Bell, 2005: Mechanisms associated with the June–September 2003 Sahel rainfall and implications for seasonal climate forecasts. CLIVAR Exchanges, Vol. 10, No. 1, 29–31.

  • Thiaw, W. M., A. B. Barnston, and V. Kumar, 1999: Predictions of African rainfall on the seasonal time scale. J. Geophys. Res., 104 , D24,. 3158931597.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., and B. J. Hoskins, 1994a: An idealized study of African easterly waves. I. A linear view. Quart. J. Roy. Meteor. Soc., 120 , 953982.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., and B. J. Hoskins, 1994b: An idealized study of African easterly waves. II. A nonlinear view. Quart. J. Roy. Meteor. Soc., 120 , 9831015.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., and D. P. Rowell, 1998: Interannual variability of African wave activity in a general circulation model. Int. J. Climatol., 18 , 13051323.

    • Search Google Scholar
    • Export Citation
  • Wang, G., and E. A. B. Eltahir, 2000: Role of vegetation dynamics in enhancing the low frequency variability of the Sahel rainfall. Water Resour. Res., 36 , 10131021.

    • Search Google Scholar
    • Export Citation
  • 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 , 31673190.

    • Search Google Scholar
    • Export Citation
  • Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability 1900–93. J. Climate, 10 , 10041020.

  • 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 , 15371540.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Mean precipitation for JAS from (a) the P analysis from 1990 to 2001, (b) ensemble mean CFS forecasts from 1981 to 2001, (c) ensemble mean CFS forecasts from 1990 to 2001, (d) ensemble mean CFS (corrected) forecasts from 1990 to 2001, (e) AMIP mean from 1990 to 2001, and (f) ensemble mean SIMs from 1990 to 2001. Contour intervals are 1, 2, 4, 6, 8, 10, 12, 15, and 20 mm day−1. Values greater than 8 (12) mm day−1 are shaded light (dark).

  • Fig. 2.

    (a) Seasonal JAS mean precipitation over the Sahel (12.5°–17.5°N, 17.5°W–20°E) from the gauge analysis (solid line), CFS forecasts (plus signs), CFS (corrected; open circles), and SIMs (dark squares). (b) Same as in (a) but for the AMIP (open circles) and the gauge analysis (dark circles) from 1950 to 2001. Std dev for JAS for the period 1990–2001 from (c) the gauge analysis, (d) CFS forecasts, and (e) CFS (corrected). Contour interval is 0.3 mm day−1. Values greater than 1.2 (1.5) mm day−1 are shaded light (dark).

  • Fig. 3.

    (a) Mean zonal wind at 600 hPa from the CDAS, averaged from 1990 to 2001. Contour interval 2 m s−1. Values less than −8 m s−1 are shaded. (b) Same as in (a), but for the mean zonal wind at 700 hPa from the CDAS. (c) Same as in (b), but from the ensemble CFS forecasts from 1990–2001. (d) Same as in (a), but ensemble mean from the CFS (corrected) forecasts. (e) Same as in (a) but for SIMs. (f) Same as in (a) but for the AMIP.

  • Fig. 4.

    SST difference between the ensemble SST forecasts for JAS with the initial conditions in Jun averaged from 1981 to 2001 and the corresponding SSTs from the CDAS. Contour interval is 1°C. Positive values are shaded.

  • Fig. 5.

    (a) Time–longitude plot of the CDAS SST JAS mean difference averaged from 10°S to 10°N and the CDAS mean for the period 1981–2001. Contour interval is 0.5°C. Positive values greater than 1°C are shaded. (b) Same as in (a), but for the difference between the CFS ensemble forecasts and the CDAS mean. (c) Same as in (a), but for the CFS (corrected) ensemble forecasts. (d) Same as in (a), but for the SST difference averaged from 35° to 55°N. (e) Same as in (d), but for the difference between the CFS forecasts and the CDAS mean. (f) Same as in (d), but for the CFS (corrected) ensemble forecasts.

  • Fig. 6.

    (a) REOF 1, (b) REOF 2, and (c) REOF 3. Contour interval is 0.5 nondimensional units. Negative values are shaded. (d) RPC 1, (e) RPC 2, and (f) RPC 3 for the observed SSTs (open circles) and for the SST anomalies from the CFS forecasts after removing the observed SST mean (dark circles).

  • Fig. 7.

    Correlation between the observed rainfall and RPC 1 from observed SSTs for (a) RPC 1, (b) RPC 2, and (c) RPC 3 for the period 1950–2001. Contour interval is 0.1. Values statistically significant at the 5% level are shaded.

  • Fig. 8.

    (a) Mean volumetric soil moisture from 0 to 10 cm for Jun averaged from 1990 to 2001 from the R2. Contour interval is 0.05. Values greater than 0.3 are shaded. (b) Same as in (a), but for Jul. (c) Same as in (a), but for the CFS ensemble forecasts in Jul. (d) Same as in (a), but for AMIP. (e) Same as in (b), but for the AMIP. (f) Same as in (c), but for SIMs in Jul.

  • Fig. 9.

    Vertical profile of the zonal wind averaged over 0°–10°E from the (a) CDAS, (b) AMIP, (c) SIMs. Contour interval is 2 m s−1. Values less than 6 m s−1 are shaded. (d), (e), (f) Same as in (a)–(c), but for dT day−1 averaged over 5°W–15°E. Contour interval is 2 × 10−6 K m−1. Values greater than 8 × 10−6 K m−1 are shaded.

  • Fig. 10.

    (a) Difference in soil moisture from 0 to 200 cm between the SIMs and AMIP for JAS averaged from 1990 to 2001. Contour interval is 0.02. Values greater than 0.06 are shaded. (b) Same as in (a), but for evaporation difference. Contour interval is 0.5 mm day−1. Values greater than 0.5 mm day−1 are shaded. (c) Same as in (b), but for P. (d) Same as in (b), but for EP. (e) Ensemble JAS mean temperature at 2 m from the SIMs for 1990–2001. Contour interval is 2°C. Values greater than 32°C are shaded. (f) Same as in (e), but for AMIP.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 251 64 2
PDF Downloads 169 45 2