The African Humid Period (AHP), about 14 800 yr ago [14.8–5.5 ka (ka ≡ 1000 yr ago)], was a time of increased humidity over Africa. Paleoclimate evidence suggests that the West African summer monsoon was stronger and more extensive 6 ka than today, and that the Saharan Desert was green. Here, a regional climate model that produces an excellent simulation of today’s climate over northern Africa is used to study the dynamics of the monsoon 6 ka. Changes in insolation, atmospheric CO2, and vegetation are used to impose 6-ka conditions, and the role of each forcing is isolated. Vegetation is not interactive, and the large-scale circulation and SSTs are fixed at present-day values for the 6-ka simulations.
The regional model produces precipitation increases across the Sahel and Sahara that are in good agreement with the paleodata. However, unobserved drying is simulated over the Guinean coast region where paleodata are sparse. Precipitation increases in the Sahel are related to a northward shift of the monsoon, the elimination of the African easterly jet, and an intensification and deepening of the low-level westerly jet on the west coast. The thermal low–Saharan high system of the present-day climate is replaced by a deep thermal low. When this system becomes fully developed in midsummer, cyclonic circulations transport moisture north into the Sahara, and rainfall increases there. Surface temperatures decrease despite the increased solar forcing 6 ka because of an increase in cloudiness. A moist static energy budget analysis shows that increased low-level moisture dominates the cooling to destabilize the vertical column and enhance convection. Even though solar forcing is the ultimate cause of the AHP, the model responds more strongly to the vegetation forcing, especially early in the summer season, emphasizing the importance of vegetation in maintaining the intensified monsoon system.
Paleoclimate evidence suggests that between 8000 and 6000 years ago (8–6 ka, ka ≡ 1000 yr ago) the climate of northern Africa was considerably different from that of today. Pollen, isotopic, lake level, and charcoal records all indicate that there was much more moisture available on the surface during that time (Kohfeld and Harrison 2000), which is near the end of the African Humid Period (AHP), approximately 14.8–5.5 ka (deMenocal et al. 2000a). Markedly different vegetation types covered the land surface, and much of the Sahara was green (Hoelzmann et al. 1998).
Both spectral properties of long-term monsoon records (Rossignol-Strick 1983; Larrasoana et al. 2003) and simulations conducted with climate models (Joussaume et al. 1999; Doherty et al. 2000; Texier et al. 2000; Braconnot et al. 2000; Tuenter et al. 2003; Zhao et al. 2005) suggest that this intensified moisture regime was linked with the orbital parameters of the earth. Specifically, the fundamental cause of this intensified moisture regime was increased Northern Hemisphere summer insolation due to differences in the earth’s orbital parameters (Kutzbach 1981). Changes in the obliquity and eccentricity of the orbit had a relatively small contribution. The main parameter responsible for the increase in insolation was the longitude of perihelion, which, calculated from the vernal equinox, changed from 180.87° at 6 ka to its present-day value of 282.45°. Today, boreal summer occurs near aphelion, but 6 ka the boreal summer solstice was aligned about half way between perihelion and aphelion (Berger and Loutre 1991). This forcing was asymmetric, with the magnitude of the maximum Northern Hemisphere summer increase greater than the magnitude of the maximum Southern Hemisphere summer decrease.
In this paper, a regional climate model (RCM) that produces an excellent simulation of the present-day West African monsoon and its variability is used to study the effects of orbital and vegetation forcing on the dynamics of the monsoon. The purposes are to understand how increases in summer insolation modify the monsoon hydrodynamics and to isolate the physical mechanisms that cause the precipitation enhancements over northern Africa. Ultimately, the goal is to gain insight into the operation of today’s climate, and build confidence in our ability to simulate future climate change in this important and vulnerable region.
In the following section, a review of the paleoclimate evidence that reveals the existence of the wet climate 6 ka is presented in order to assemble validation data for the model simulation. Also, the influence of the precessional cycle on the strength of the West African monsoon is discussed and previous modeling studies of the AHP are reviewed to develop an understanding of our current state of knowledge. Section 3 provides a description of the RCM and the present day and 6-ka simulation design, and validation of the present-day simulation. Section 4 discusses the changed dynamics of the monsoon associated with radiative and vegetation forcing. Section 5 contains conclusions.
a. Proxy data evidence of the mid-Holocene
A variety of methods, including biogenic carbonate and terrigenous sediment, pollen, and diatom analyses of lake and ocean cores (Table 1), have been used to construct distributions of surface moisture, usually related to precipitation minus evaporation (P − E), for the mid-Holocene. Throughout the Sahara, all records included in the Global Lake Status Database (GLSDB: Kohfeld and Harrison 2000) and several individual findings (Table 1) indicate that P − E values were higher 6 ka as compared to present day. Farther south, in the Sahelian region, the GLSDB identifies several points between 15° and 22.5°N, 15°W and 30°E to have been wetter 6 ka. Pollen and algae analysis of an ocean core at the mouth of the Niger River (3°31′N, 5°34′E) indicates relatively wet conditions in the Niger River watershed, which includes part of the Sahel and Guinean coast (Lezine et al. 2005). Other types of proxy data agree that the Sahel was wetter (Table 1).
Along the Guinean coast, Lake Bosumtwi (6°30′N, 1°25′W) in Ghana, is a hydrologically closed crater lake that was formed one million years ago, and its lake cores are considered to be a very reliable indicator of paleoclimate conditions (Turner et al. 1996). Talbot et al. (1984) found greater P − E over Bosumtwi at 6 ka compared to present day based on lake levels and charcoal analysis. This is supported by the isotopic nitrogen analysis by Talbot and Johannessen (1992), which found that from 9.2 to 3.2 ka, the lake had a very stable water column, and the sediment analysis of Russell et al. (2003). There are no other paleoclimate sites available along the Guinean coast.
The proxy data are presented as a basic means for validating the 6-ka simulations, but this comparison is not intended to be the focus of this paper, which is to describe changes in the dynamics of the monsoon due to orbital forcing. Also note that, while the simulations are summer averages, P − E estimates from lake cores represent an annual signal. However, since it is a reasonable assumption that the main precipitation signal over the Sahel at 6 ka is captured during the summer months, lake cores may still be used for comparison to the model simulations.
b. Orbital forcing of the West African monsoon
Analysis of Saharan dust in a Mediterranean Sea core dating back to three million years before present (3 × 106 yr BP) indicates that changes in orbital parameters, especially precession, strongly influence the West African monsoon (Larrasoana et al. 2003). Hematite concentrations in the dust particles suggest an origin between 21° and 30°N, so the amount of dust is used as a measure of the northward extent of the monsoon, although its concentration is also influenced by changes in wind speed. When precipitation reaches northward into the dust origination region of the Sahara, there is less dust available for transport to the Mediterranean. Throughout the core, minima in dust amounts correspond to minima in precession, indicating that a stronger and northward extended monsoon is associated with low precession. A minimum in precession (modulated by eccentricity) occurs when the summer solstice is aligned with perihelion. The 41 000-yr obliquity cycle is also evident in the core, with a phased relationship linking minimum dust and stronger monsoons to maximum obliquity.
East Mediterranean sapropels dating back to 464 ka also support the idea that orbital forcing influences the strength and northward extent of the West African monsoon (Rossignol-Strick 1983). Sapropels are present in the core always and only when a threshold value of the northern summer monsoon index is reached. The index is high when the value of maximum insolation is large and the maximum occurs in the Northern Hemisphere.
Sapropel evidence demonstrates that strengthened monsoons can occur despite the presence of large northern ice sheets, for example at 176 ka (Rossignol-Strick 1983). Also, Masson et al. (2000) simulate with an AGCM strengthened African and Indian monsoons at 175 ka, a time when glacial conditions existed at mid and high latitudes. Both geological evidence and climate modeling study suggest that the global cooling associated with ice sheets is not sufficient for aridity in the Tropics and supports the idea that the main monsoon forcing is tropical insolation.
Simulations with ECBilt, a global fully coupled atmosphere–ocean–sea ice climate model of intermediate complexity, also support the idea that African monsoon precipitation is increased and extends farther north when precession, modulated by eccentricity, is at a minimum and obliquity is at a maximum (Tuenter et al. 2003). The effect of precession dominates and influences the amplitude of the response to obliquity, while obliquity differences have no effect on the precession-induced changes. In response to precession changes from minimum values to maximum, the timing of maximum precipitation shifts about one month further into the summer and low-level westerly winds between 10° and 15°N strengthen by over 2 m s−1. While vegetation changes are not included in this model, Tuenter et al. (2003) speculate that they may provide a positive feedback mechanism.
c. Previous model simulations of the AHP
Although the most recent maximum in Northern Hemisphere summer solar insolation occurred around 11 ka, and 6 ka is thought to be a period of climate instability in Africa (Gasse 2000), several individual modeling studies (e.g., Doherty et al. 2000; Texier et al. 2000) and model comparison projects (Zhao et al. 2005), including the Paleoclimate Modeling Intercomparison Project (PMIP: Joussaume et al. 1999; Braconnot et al. 2000) focus on 6 ka to avoid uncertainties in the extent of ice sheets, and that time is adopted here for consistency and to enhance comparison. The period from 8 to 6 ka was generally marked by high West African precipitation rates despite fluctuations on the time scale of centuries that may have been related to climate instabilities arising from nonlinear interactions between the atmosphere and vegetation. In a GCM asynchronously coupled with a vegetation model, Claussen (1998) found both vegetated and nonvegetated stable solutions over the Sahara under prescribed present-day orbital parameters, indicating the potential for rapid climate change between two stable states. However, Zeng and Neelin (2000) demonstrated that, when their intermediate level model is forced with sufficient SST variability, the multiple equilibrium solution fails, indicating the existence of a single equilibrium state.
In the PMIP group (Joussaume et al. 1999), 18 different atmospheric GCMs were run with the same present-day and 6-ka boundary conditions. Only radiative forcing, from changed orbital parameters, and reduced CO2 was used in the 6-ka simulations. Present-day land surface conditions and SSTs were prescribed. Each GCM simulates a strengthened monsoon for 6 ka. However, each model underestimates the summer northward migration and the intensity of the rainfall expected from the geological data, likely due to the use of present-day vegetation. A second phase of the PMIP project, known as PMIP2 (Crucifix et al. 2005), has begun to include interactions with the ocean and vegetation more fully in GCM simulations of 6 ka.
A compilation of the present-day and 6-ka simulations of seven AOGCMs reveals that including an interactive ocean produces more West African summer precipitation than PMIP GCMs with a comparable atmospheric component due to increased moisture advection from the Atlantic Ocean related to enhanced land–sea contrast (Zhao et al. 2005). Present-day vegetation is prescribed for both present-day and the 6-ka simulations. The simulated precipitation increases for 6 ka are in better agreement with the paleodata, but are still too weak.
When prescribed land surface attributes in GCMs are changed to reflect estimates of the conditions at 6 ka from proxy data, the models tend to produce additional precipitation in response to the radiative forcing of 6 ka (e.g., Kutzbach et al. 1996). Also, a GCM fully coupled to a dynamical vegetation model simulated about a 50% reduction in North African desert area as calculated from present day and an increase in summer precipitation (Doherty et al. 2000). However, both of these simulations still underestimate the northward extension of the monsoon as evidenced by the geological data.
The 6-ka simulations, which include coupling between atmosphere, vegetation, and ocean with CLIMBER, a coarse-resolution climate system model of intermediate complexity, find Eurasian summer warming of up to 4°C, Saharan summer precipitation increases of around 1.5 mm day−1, and a significant reduction in the Saharan Desert (Ganopolski et al. 1998). Vegetation feedbacks were shown to play an important role in the intensity of the West African monsoon. Claussen et al. (1999) find that in order to simulate the abrupt change in Saharan vegetation that is suggested by geological evidence, CLIMBER-2 must be coupled with vegetation. Additional ocean coupling influences the timing of the abrupt change. From these studies, it is evident that the vegetation 6 ka must be included to simulate the proper response. While CLIMBER-2 simulates the present-day climate of Africa reasonably (Petoukhov et al. 2000), the coarse resolution is a disadvantage for comparison of the 6-ka simulations with the paleodata.
There is also some concern with using GCM simulations to compare the West African monsoon of the present day and 6 ka. In addition to difficulties in simulating the full strength of the monsoon 6 ka, some of the models do not verify well for present-day simulations over northern Africa. A model is taken to verify well if the precipitation is within an order of magnitude of the observed and has a summer maximum that falls over the continent instead of over the ocean and if the circulation reproduces that described in section 3. While GCMs are useful for global studies, there are disadvantages when using a GCM to focus on one region. The GCMs may have inadequate resolution for regional studies and may use parameterizations that are not suitable for a given region. For example, in an analysis of 18 state-of-the-art coupled ocean–atmosphere GCM simulations, Cook and Vizy (2006a) found that one-third of the models did not even capture the summer migration of the ITCZ onto the African continent. Such poor present-day simulations make comparison of present-day dynamics with those of the simulated humid period problematic. Additional complications arise due to differences in land surface albedo initialization for paleoclimate simulations, which can significantly impact the West African climate (Bonfils et al. 2001).
However, while the present-day simulation of the RCM validates very well and has high resolution, the assignment of boundary conditions for paleoclimate simulations is troublesome. Paleoclimate boundary conditions derived from GCMs do not have the appropriate resolution needed for use with the RCM and may introduce errors if the boundaries of the RCM are over regions where the GCM does not validate well, while use of present-day boundary conditions eliminates any influences from outside of the domain. While Differnbaugh and Sloan (2004) use paleoclimate boundary conditions derived from a GCM, this study utilizes present-day boundary conditions with the implications discussed in the following section.
3. Model description, simulation design, and model validation
The Pennsylvania State University–National Center for Atmospheric Research (PSU–NCAR) Mesoscale Model (MM5) version 3.6 (Grell et al. 1994) is utilized for all simulations. MM5 is a nonhydrostatic model with 24 vertical sigma levels. The top of the atmosphere is set to 50 hPa for this tropical application.
Physical options for the simulations include the Blackadar planetary boundary layer scheme (Blackadar 1979; Zang and Anthes 1982), the Kain–Fritsch convective scheme with shallow convection turned on (Kain and Fritsch 1993), the Dudhia simple ice explicit moisture scheme (Dudhia 1989), and the Rapid Radiative Transfer Model (RRTM) longwave radiation scheme (Mlawer et al. 1997). These choices were made based on previous experience in modeling the northern African climate with this model (Vizy and Cook 2002, 2003; Hsieh and Cook 2007). Land surface characteristics are specified using the 10′ resolution 24-category U.S. Geological Survey (USGS) dataset (Fig. 1a), and SSTs are prescribed from Shea et al. (1992), as represented in the reanalysis. Soil moisture, which heavily influences evaporation rates, and surface albedo are prescribed by vegetation type and held constant throughout the simulation. Evapotranspiration and runoff processes are not included in the model. Note that, whereas vegetation in the real world is interactive, thus a feedback, the vegetation is held constant throughout the simulations, so is a forcing in this study.
Each simulation uses 90-km horizontal resolution and a 90-s time step, and runs for 137 days from 15 May to 30 September with the first 16 days disregarded for model spinup. The 3-hourly model output is averaged to form monthly climatologies. The classical calendar approach is used for all simulations, as differences between surface air temperature in simulations with the classical and angular calendar definition are small over northern Africa for 6 ka (Joussaume and Braconnot 1997).
To simulate climate, MM5 is modified to a regional climate model, as detailed in Vizy and Cook (2002). Previous testing, which continued RCM simulations for a second year, produced a monthly mean climate that was very similar to the first year, suggesting that the model is, indeed, capturing the climatology.
To test the influence of boundaries on the interior of the domain, present-day simulations are run on two domains with the same physical options and resolution. The smaller domain covers 26.4°S–35°N, 48.6°W–48.6°E and is represented in Fig. 1b, with the northern and southern borders drawn in. The larger domain covers the entire region shown in Fig. 1b and extends from 29°S to 45°N, 48.6°W to 48.6°E. The two integrations yield similar results, confirming that the boundaries chosen do not influence the results in the interior of the domain. Results presented below are from the simulation with the larger domain.
For simulations of the climate at 6 ka, the same resolution, integration length, domains, and physical parameterizations are used as for the present-day simulations. The obliquity, eccentricity, and CO2 concentrations are set to 23.5°, 0.016 710 22, and 330 ppmv for the present day, and 24.1°, 0.018 682, and 280 ppmv (Raynaud et al. 1993) for 6 ka. Also, since the position in the precessional cycle 6 ka was about one-fourth cycle different than the present-day position, the longitude of perihelion calculated from the vernal equinox is changed from 282.45° during the present day to 180.87° for 6 ka. These differences in the eccentricity, obliquity, and longitude of perihelion cause a difference in the incoming solar radiation at the top of the atmosphere (Fig. 2). The largest increase in solar radiation occurs in late July near 45°N.
Land surface types for 6 ka (Fig. 1b) are specified according to Hoelzmann et al. (1998), which is based on interpolated biome reconstructions. The northern boundary of this dataset falls a few degrees of latitude short of the northern boundary of Africa. Realizing that pollen data in this region is scarce (Prentice et al. 2000), we assume that steppe replaced desert to 35°N based on lake status of wetter conditions (Kohfeld and Harrison 2000). All other land surfaces are assumed to be of present-day type (Prentice et al. 1998).
AGCM and many observational studies clearly demonstrate the importance of regional and global SSTAs on interannual and interdecadal variability of rainfall in the Sahel for today’s climate (Folland et al. 1986; Rowell et al. 1992, 1995; Vizy and Cook 2002). Unfortunately, SST reconstructions for 6 ka cover only a small region. The faunal record from an ocean core off the coast of West Africa (20°45′N, 18°35′W) suggests that 6-ka STs at this one location may have been up to 4.5°C cooler (deMenocal et al. 2000a). However, it is not certain how widespread the cooling may have been. If this core is recording changed upwelling patterns, for example, there may not be a large influence on African precipitation because the forcing region would be small. The Uk37 analysis at this same site and a nearby site (19°00′N, 20°10′W), is in disagreement with the faunal record, and suggests SSTs were about 1°C warmer during the humid period (Zhao et al. 1995). For lack of information about the larger-scale SST climatology, we use present-day SSTs and cannot evaluate any impact of a climatological change in SSTs. When reliable SST distributions for 6 ka are available from coupled GCMs and observations, we will evaluate their influence.
The 1949–2002 (National Centers for Environmental Prediction) NCEP–NCAR reanalysis climatology is used to specify initial and lateral boundary conditions for 6 ka as well as for the present day. Lateral boundary conditions from GCMs are not used because of problems with validation in this region, as discussed in the previous section and in Cook and Vizy (2006a).
To test the influence of this specification of lateral boundaries on the interior of the domain, the 6-ka simulations are run on both the larger and smaller domains (Fig. 1b). Differences are small, confirming that the boundaries chosen do not influence the results in the interior of the domain and providing additional support for the use of present-day boundary conditions for the 6-ka simulations. Vizy and Cook (2005) and Cook and Vizy (2006b) report similar results in their regional modeling study of South America at the Last Glacial Maximum.
There are several implications of using present-day lateral boundary conditions for 6 ka. Any remotely forced influences, such as Mediterranean SSTs (Rowell 2003), obliquity forced moisture transport from higher latitudes (Tuenter et al. 2003), and teleconnections from the Antarctic or ENSO, are not included in the simulations. Also, the effects of changing the large-scale temperature gradient between Eurasia and the tropical Atlantic on the tropical easterly jet and the monsoon are not included. However, by including the primary forcing functions at 6 ka, we capture the primary physics responsible for the changed precipitation distribution.
Four model integrations were performed, and these are referred to as the “present-day,” “6ka_RV,” “6ka_R,” and “6ka_V” simulations. The present-day simulation is run with present-day solar parameters, CO2 concentration, and land surface types, and the 6ka_RV simulation has those of 6 ka. The 6ka_R uses the solar parameters and CO2 concentration at 6 ka but present-day land surface types, while 6ka_V, uses present-day solar parameters and CO2 and land surface types of 6 ka.
Before investigating changes in the monsoon circulation at 6 ka, the degree of realism in the simulation of the present-day summer monsoon dynamics is evaluated.
Figures 3a and 3b show 925-hPa winds and geopotential heights from the NCEP–NCAR reanalysis (Kalnay et al. 1996) and RCM climatologies, respectively, averaged over June–September (JJAS). For this comparison, the RCM output is interpolated onto the coarser grid (2.5° × 2.5°) of the reanalysis, and mass-weighted vertical averaging is used to place the RCM output onto the reanalysis levels. In both the reanalysis and the model, the thermal low centered near 15°N introduces the exceptionally strong meridional geopotential height gradients that are characteristic of the West African monsoon. Southerly winds transport moisture onto the continent across the Guinean coast and into the thermal low. Near 10°N on the west coast, the RCM resolves the low-level westerly jet identified by Grodsky et al. (2003). The jet is only weakly represented in the NCEP–NCAR analysis, and this may be a result of the lower resolution and the scarcity of data for the assimilation in this region.
Many of the prominent features of the continental circulation near the surface (Figs. 3a and 3b) are quite shallow. At 850 hPa (Figs. 3c and 3d), the thermal low structure is much less pronounced, and the southerly flow across the Guinean coast is not present. In the RCM, the low-level westerly jet on the west coast near 10°N is much weaker than at 925 hPa.
The Saharan high, which marks the divergent center that overlays the convergence into the near-surface thermal low, is prominent at 600 hPa (Figs. 3e and 3f). The African easterly jet, which is generated when Coriolis accelerations act on the northerly outflow from the Saharan high, is centered near 15°N in both the model and reanalysis. The Saharan high is a little weaker in the RCM compared with the reanalysis and does not extend as far north. As a consequence, westerly flow onto the continent occurs north of about 23°N in the model, while in the reanalysis this flow is confined to far northern Africa.
Given uncertainties in observed precipitation climatologies and the scale dependence of these observations, two datasets are used to validate the RCM’s precipitation climatology. The Climate Research Unit (CRU) dataset is a 0.5° by 0.5° monthly rainfall climatology derived from rain gauges covering the years 1961–90 (New et al. 1999). The JJAS-averaged CRU precipitation (Fig. 4a) shows a very strong meridional gradient between 15° and 20°N. Three precipitation maxima, in order of decreasing strength, are centered on the west coast near 7.5°N, 12.5°W; the coast of Cameroon at 5°N, 10°E; and in Ethiopia. There is also a smaller maximum at 7.5°N, 25°E associated with the maximum over the coast of Cameroon and a small maximum over northern Algeria at about 35°N, 5°E.
For comparison with the CRU precipitation dataset, the RCM output is presented on the model’s own grid (Fig. 4b). The RCM captures the meridional gradient between 15° and 20°N and, as with the CRU precipitation, there is a maximum over northern Algeria. The west coast maximum is weaker and located farther inland in the RCM than in the CRU data, while the Ethiopian maximum is stronger. The maxima over the Cameroon highlands and 7.5°N, 25°E are split into four separate maxima, and these are also located slightly inland.
We also compare the RCM precipitation with the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) dataset (Fig. 4c) to have information over the oceans. This is a 2.5° by 2.5° monthly rainfall climatology covering the years 1979–2001, compiled from rain gauge estimates and satellite values (Xie and Arkin 1997). The 0.5 mm day−1 line is a few degrees farther north over western Africa and the 0.5 mm day−1 line in Algeria extends farther southward in the CMAP climatology than in the CRU dataset. The JJAS RCM precipitation interpolated onto the CMAP grid (Fig. 4d) shows good agreement with the CMAP rainfall. The largest difference is a lower precipitation rate and northward displacement of the ITCZ by about 2.5°.
Although surface temperature is not assimilated in the reanalysis and, therefore, is a model product, differences in surface temperature are important in the analysis of the climate 6 ka, so a comparison between the NCEP–NCAR reanalysis and RCM is included. The reanalysis (Fig. 5a) places maxima over 306 K over northwestern and northeastern Africa and a strong meridional temperature gradient at 15°N. These features are well represented by the RCM (Fig. 5b). While the pattern of surface temperature of the European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA: not shown, Gibson et al. 1997) is comparable to the RCM and NCEP reanalysis, the ECMWF is up to 6 K warmer than the RCM and NCEP reanalysis over the Sahara.
The RCM simulates the magnitude and structure of the summer monsoon precipitation very well. Differences between the data and RCM are not as drastic as in most coupled ocean–atmosphere GCMs (Cook and Vizy 2006a). In GCMs with prescribed SSTs, 5 of the 14 models either fail to simulate the summer precipitation maximum over Guinea or simulate rainfall too far north into the Sahara (Gadgil and Sajani 1998). The accuracy of the RCM simulated precipitation is comparable to that of the more accurate GCMs for this region. Equally important, the model also reproduces the circulation features including the southerly low-level flow from the Gulf of Guinea and the thermal low and Saharan high system. High spatial and temporal resolution and carefully chosen parameterizations give the RCM an advantage for simulations of northern Africa, as compared to GCMs, which are not designed for regional studies and sometimes do not reproduce the summer monsoon well. The disadvantage to the RCM is the problem of paleoclimate boundary conditions, as previously discussed.
a. Comparison of the present-day and mid-Holocene monsoon dynamics
Under radiative and vegetation forcing of 6 ka, the RCM produces a wetter climate over the Sahel and the Sahara, in agreement with the paleodata, with high rainfall rates (>4 mm day−1) extending to 25°N over the central Sahel (Fig. 6a). Differences between the 6ka_RV and present-day simulations (Fig. 6b) indicate that rainfall rates are up to 12 mm day−1 higher over the Sahel and Sahara 6 ka. Significance of these results is suggested by the fact that summer precipitation differences 6 ka are much higher than the maximum JJAS variability in the CPC dataset spanning from 1979 to 2004, which shows a maximum deviation of ±1 mm day−1 from the average precipitation in this region.
In 6ka_RV, rainfall decreases of up to 6 mm day−1 are simulated over the Guinean coast region (5°–10°N, Fig. 6b). These differences are on the order of the variability shown in the CPC dataset, suggesting the results in this region may be significantly influenced by a mechanism not included in the simulations, for example, the SSTs 6 ka. The simulated decrease is a major discrepancy between the simulation and the paleodata since the very reliable lake core at Bosumtwi indicates wetter conditions 6 ka (e.g., Talbot and Delibrius 1980; Talbot et al. 1984; Russell et al. 2003). According to the CRU climatology, about 67% of the annual precipitation in this region occurs during JJAS, so it is also possible that during the months not included in the present study, there is a precipitation increase over the Guinean coast 6 ka that would, in the annual average, cause that region to be wetter. Increased precipitation in this region may also be dependent on land surface interactions. Future annual simulations will better address these issues. Here, we focus on understanding the dynamics that brings higher precipitation rates to the Sahel and Sahara.
In June (Fig. 7a), the precipitation enhancement 6 ka is evident across the Sahel, with differences up to 12 mm day−1 in the central Sahel. These anomalies strengthen up to 16 mm day−1 east of the Greenwich meridian in July (Fig. 7b), but do not spread farther north. In August and September, the precipitation enhancements develop over the central Sahel and extend northward into the Sahara (Figs. 7c,d).
Since local recycling (through evaporation) contributes about 10%–30% to the moisture budget for both the present-day and 6-ka simulations equally, examining the mass-weighted vertical integral of the moisture transport vectors (the product of the specific humidity and the horizontal wind) provides a first-order understanding of the overall moisture sources for the summer rainy season. In the present-day simulation (Fig. 8), the primary net moisture source west of the Greenwich meridian is flow across the Guinean coast near 5°N, and moisture is transported away from the continent to the west over the Sahel near 15°N. East of the Greenwich meridian, there is an additional source of moisture from the north. Rowell (2003) demonstrated the importance of this moisture transport from the Mediterranean for sustaining eastern Sahel rainfall. There is also some moisture transport into the eastern Sahel from the Indian Ocean, and from the North Atlantic in the west. Each of the four months of the present-day simulation shows this pattern of moisture transport.
In the 6ka_RV simulation, in June (Fig. 9a), moisture transport across the Guinean coast is similar to the present day (Fig. 8) but the moisture crossing the Guinean coast penetrates deeper into the Sahel. Also, easterly moisture transport off the west coast near 15°N is absent, replaced by weak westerly transport onto the continent near 8°N. In July (Fig. 9b), this westerly transport intensifies and spreads northward to nearly 15°N. A distinct cyclonic circulation develops over the central and eastern Sahel, and moisture is transported into the northern Sahara from the south and southeast between 5°W and 10°E. This transport is strongest in August and September (Figs. 9c and 9d), consistent with the rainfall anomaly (Figs. 7c and 7d).
This overview of the moisture transport indicates that three main factors contribute to the precipitation anomalies 6 ka: 1) the continuation of the moist flow from the Gulf of Guinea into the Sahel, 2) the lack of moisture transport off the west coast, and 3) the northward moisture transport into the Sahara later in the season. Circulation anomalies dominate over specific humidity differences (not shown) in controlling the moisture transport.
The anomalous moisture transport away from the Guinean coast countries and into the central Sahel occurs during all four summer months (Figs. 9a–d) and is associated with wind anomalies between 825 (Figs. 10a,b) and 725 hPa. The southwesterly flow just west of the Greenwich meridian between 8° and 10°N in the 6ka_RV simulation is associated with a deepening of the continental thermal low centered near 20°N. The flow has a pronounced divergent component down the height gradient near 10°N that transports moisture away from the Guinean coast region and into the central Sahel.
The disabling of the moisture transport off the west coast of the continent 6 ka is associated with two differences between the present-day and 6-ka circulation. One is an intensification and deepening of the low-level westerly jet near 10°N (see Fig. 3b), which enhances moisture transport onto the continent from the Atlantic Ocean. Throughout the season, the westerly jet in the 6ka_RV simulation is much stronger and deeper than in the present-day simulation (Figs. 3b and 3d), and extends farther north. (The figures in section 2 were interpolated to the NCEP–NCAR reanalysis standard levels for the model validation, but the flow is very similar at 850 and 825 hPa.) These differences in the westerly jet are consistent with the strengthening and deepening of the low over northern Africa, and the related intensification of the already impressive meridional height gradients across West Africa. This westerly flow is essentially geostrophic.
The change in the low-level westerly jet can be understood from a vorticity budget perspective. The vorticity equation, with the tilting term and vertical advections neglected and ∂ζ/∂t set to zero for the climatology, is
where u is the zonal wind, υ is the meridional wind, V is the horizontal wind vector, f is the Coriolis parameter, β is ∂f /∂y, and ζ is relative vorticity. Applying this equation to the lower troposphere, the right-hand side of the equation can be thought of as being imposed by the condensational heat. In the present-day monsoon, the maximum condensational heating—and the maximum in the divergence term—occurs between 5° and 10°N. A balance for the climatology is achieved when the southerly flow from the Gulf of Guinea advects low planetary vorticity to balance the column stretching due to the condensational heating (Cook 1997). The maximum condensational heating 6 ka occurs between 15° and 20°N, and the Sverdrup balance is less effective due to stronger Coriolis accelerations. Instead, the vorticity balance is achieved by a strengthening of the low-level westerlies, which advect low relative vorticity from the Atlantic high onto the continent.
The association between strengthened low-level westerly winds and increased precipitation over the Sahel 6 ka is similar to an important present-day mode of variability (Grist and Nicholson 2001). Also, a stronger low-level westerly jet may act to cool SSTs through entrainment and latent heat loss (Grodsky et al. 2003), so would support the ocean core analysis that suggests that SSTs off the west coast were cooler 6 ka due to increased upwelling (deMenocal et al. 2000a).
The other circulation difference that leads to a disabling of moisture transport off the west coast is the destruction of the African easterly jet (AEJ) in the 6ka_RV simulation. Low pressure over northern Africa extends into the middle troposphere in the climate of 6 ka (Figs. 10c and 10d). In sharp contrast, the low is confined below about 750 hPa and is overlaid by the Saharan high (Fig. 3f) in the present day. The Saharan high does not form in the 6ka_RV simulation, so the AEJ, being a consequence of outflow from the Saharan high, is not a feature of the 6-ka climate.
While the precipitation differences between the present day and 6 ka south of about 20°N are in place throughout the summer season, increased precipitation 6 ka in northern Africa occurs primarily during August and September. Its development is related to the seasonal evolution of the thermal low, which becomes stronger as the summer season progresses (cf. Figs. 10a and 10b). The flow into northern Africa shifts from easterly in June and July to southeasterly in August and September, transporting moisture into the region and resulting in the large increase in rainfall from June (Fig. 7a) to August (Fig. 7c). Note that, while we are able to assess changes over this region during the summer, the northern coast of Africa receives significant rainfall during the nonsummer months, and the precipitation maximum over this region is not considered to be part of the monsoon.
b. Forcing processes of the mid-Holocene climate
As detailed in section 3, the mid-Holocene forcing involves differences in insolation, atmospheric CO2, and vegetation. We combine solar and CO2 forcing of 6 ka into one radiative-forcing simulation (6ka_R) and contrast that simulation with one in which anomalies are forced only by vegetation differences (6ka_V).
We begin by examining the 6ka_R simulation, imagining that the changes induced in the African climate by the insolation differences triggered the changes in vegetation, and then the full response developed. This is a convenient construct used to gain more insight into physical processes, but in nature the vegetation changed slowly as the insolation evolved, and the “radiative-forcing-only” simulation was never a reality.
Figure 11 shows precipitation differences between the 6ka_R and present-day simulations averaged over JJAS. A comparison with Fig. 6b indicates that the radiation-induced rainfall anomalies are not as intense and do not agree as well with the paleodata as those with 6ka_RV forcing, but the structure is similar. The RCM 6ka_R simulation is in better agreement with the paleodata over the Sahel and Sahara than the 18 GCMs of the PMIP program (Braconnot et al. 2000). As discussed in section 2, the PMIP models produce rainfall increases that are confined within a few degrees of 10°N. The RCM maximum precipitation increase is located at 17°N, where the GLSDB (Kohfeld and Harrison 2000) indicates much wetter conditions.
Most of the precipitation differences that contribute to the summer average occur in August and September with radiative forcing of 6 ka. In June and July, when differences in insolation are small (Fig. 2), precipitation anomalies are also small (Figs. 12a and 12b). Coincident with the maximum solar forcing in August and September, precipitation differences are much larger and extensive (Figs. 12c and 12d). The ITCZ is shifted to the north, and positive perturbations spread far north across the central Sahel and Sahara.
Comparison of the 6ka_R (Figs. 12a–d) with the 6ka_RV simulation (Figs. 7a–d) suggests that, when the radiative (solar) forcing is weaker in the first half of the summer, precipitation perturbations are forced by the vegetation changes and, when strong solar forcing is in place late in the season, vegetation enhances the precipitation increase induced by radiative forcing alone. This result is confirmed by the 6ka_V simulation (not shown).
Figures 13a–d display the vertically integrated moisture transport for each month of the 6ka_R simulation. During June and July (Figs. 13a and 13b), moisture transport patterns are similar to those of the present day (Fig. 8), including westward moisture transport off the west coast near 15°N. An exception is the weak anomalous moisture transport away from the Guinean coast countries and into the central Sahel. The small drying along the Guinean coast region in the first part of the summer (Figs. 12a and 12b) is related to this change in moisture transport.
Moisture transport in August and September (Figs. 13c and 13d) in the 6ka_R simulation is considerably different from that of the present day (Fig. 8), and more similar to that of the 6ka_RV simulation (Figs. 9c and 9d). In particular, the westward moisture transport off the west coast is reversed, and pronounced southerly transport supports precipitation increases in the central Sahara.
The anomalous moisture transport away from the Guinean coast countries and into the central Sahel in the 6ka_R case is related to a relatively small strengthening of the thermal low in June and July at 825 hPa (Fig. 14a). Later in the summer, as seen in the August–September average (Fig. 14b), the thermal low develops further and meridional height gradients strengthen. The low-level westerly jet on the coast intensifies, and a fully developed cyclonic flow over the central Sahel brings moisture into the northern Sahel and central Sahara, similar to the 6ka_RV simulation.
In the 6ka_RV simulation at 625 hPa, the AEJ forms in June, but the flow shifts to westerly in July. In the 6ka_R case, the AEJ persists throughout the summer. It is similar to the present-day jet early in the summer (Fig. 14c), but weaker during August and September (Fig. 14d). The effect of 6-ka radiative forcing alone, then, is to weaken the AEJ but not eliminate it, while the inclusion of the 6-ka vegetation eliminates the jet. This association between the AEJ and the meridional soil moisture gradient is consistent with GCM simulations (Cook 1999).
Closely linked to the AEJ is the Saharan high. Under radiative forcing 6 ka, the high is slightly weaker than in the present-day simulation early in the season (Fig. 14c) and much weaker late in the season (Fig. 14d).
The sum of the precipitation generated in the 6ka_R and 6ka_V simulations (Fig. 15) has a similar distribution but larger magnitude than the precipitation of the 6ka_RV simulation (Fig. 6a). This indicates that radiative forcing and land surface changes generate nonlinear interactions that decrease the magnitude of the rainfall perturbation. The importance of nonlinearities was also recognized by deMenocal et al. (2000b). This issue will be explored in future simulations with an interactive land surface model.
c. Tropical climate change: Warm and wet versus cool and wet
Proxy data reconstructions of surface temperature 6 ka for northern Africa are severely lacking, but we examine the modeled surface temperature changes here because a better understanding of the relationship between surface temperature and precipitation for tropical climate dynamics is needed.
If all other factors remain the same, anomalously wet surface conditions will be associated with surface cooling in a climatology. Over a wetter surface, evaporation carries more of the heat transfer from the surface to the atmosphere than the sensible heat flux. These latent heat fluxes operate at lower surface temperatures than sensible heat fluxes; that is, transferring a given amount of heat from the surface to the atmosphere by sensible heating requires a greater difference between the surface and surface air temperature (generally, a higher surface temperature) than that required by evaporative cooling. But, of course, all other factors do not remain the same, and it is not clear what relationship should be expected between precipitation and surface temperature when clouds, vegetation type, and many other factors change as well.
South America during the Last Glacial Maximum (LGM ∼ 21 ka) presents an interesting example. Although there is controversy, much of the geological proxy data for this time is interpreted as indicating that conditions in the Amazon basin were drier and cooler (Farrera et al. 1999). But it is difficult to understand how a drier surface in the Tropics can be significantly cooler than a wetter surface. Cook and Vizy (2006b), for example, find that the LGM drying over South America occurs in austral spring when the inflow of moisture to the Amazon basin is reduced in association with a cooler equatorial Atlantic in the RCM. The resulting precipitation reductions are accompanied by surface drying and reduced cloud amounts, and both of these factors act to increase surface temperature in austral spring.
In the 6ka_RV case, with a wetter surface, we find that much of the surface of northern Africa is cooler than in the present-day simulation despite the increase in solar radiation at the top of the atmosphere (Fig. 16a). The RCM simulates surface temperature anomalies as cool as 5 K in the regions with the largest positive precipitation anomalies. While the noble gas temperature analysis of groundwater in southwestern Niger (Beyerle et al. 2003) supports near-surface cooling 6 ka, we realize that the magnitude simulated by the RCM may be linked to model parameterizations.
In the 6ka_R simulation (Fig. 16b), surface temperature decreases of about 1 K are confined near 15°N. There is surface warming to the north, despite mild precipitation increases over the Sahara (Fig. 11). Thus, the RCM simulations indicate that including vegetation differences between the 6 ka and present day enhances the surface cooling 6 ka. This is similar to the surface temperature response GCM simulations with orbital forcing of 6 ka and present-day vegetation (Braconnot et al. 2000), and is consistent with the results of Doherty et al. (2000), who associate significant decreases in surface temperature with vegetation feedbacks in a GCM fully coupled to a dynamical vegetation model.
An analysis of the seasonally averaged surface heat budget reveals why the surface is cooler 6 ka over the central Sahel (14°–25°N, 0°–30°E). As detailed in Table 2, despite increased incoming solar radiation at the top of the atmosphere and lower surface albedo due to the vegetation differences 6 ka, surface temperatures are cooler due to a higher atmospheric albedo from increased cloud cover from present day. Decreases in the sensible heat flux approximately balance increases in latent heat flux, as expected for a wetter surface. There is some additional heating of the surface in the 6ka_RV simulation due to increases in longwave back radiation from an atmosphere with more clouds and higher water vapor amounts, but the cooling effects in the shortwave response dominate.
The surface heat beat budget over this same region for the 6ka_R simulation provides an explanation for the surface temperature anomalies in Fig. 16b. Regions of decreased surface temperature are associated with increased cloud cover. Surface heat fluxes increase because in the northern part of the region, where precipitation increases are smaller, surface temperatures increase.
The moist static energy (MSE) budget shows how the intensified monsoon of 6 ka is supported despite the surface temperature decrease. MSE is the sum of the sensible, latent, and (geo)potential energy according to
where cp is the specific heat of air at constant pressure, T is air temperature, L is the latent heat of vaporization of water, q is specific humidity, g is the acceleration due to gravity, and z is height. MSE increasing with altitude denotes a stable atmosphere, so increases in low-level MSE destabilize the vertical column and promote convection.
The present-day simulated MSE budget at 18.5°N, 17°E averaged over the summer is shown in Fig. 17a. An unstable profile is never present in monthly means because the processes that stabilize convection operate on much smaller time scales. The neutral MSE profile is the result of convection. The temperature term is the largest contributor to the MSE throughout the troposphere, and the geopotential energy term becomes increasingly large above 700 hPa. The latent heat term is relatively small and is nearly zero above 500 hPa.
Differences in the MSE budget between the 6ka_RV and present-day simulations at the same location (Fig. 17b) demonstrate that the increased instability of the atmosphere over the Sahel and Sahara is not related to low-level temperature increases, but to moisture increases. Below 700 hPa, anomalies in the temperature term are negative, which contribute to stabilizing the column. Although the moisture term is small in the full field, the increases below 700 hPa dominate over the temperature decreases, resulting in a positive MSE anomaly. Differences in the geopotential energy term are negligible. The MSE profiles of the 6ka_V (not shown) and 6ka_RV simulations are very similar, supporting that vegetation forcing dominates the MSE differences.
A regional climate model is used to investigate differences in the dynamics of the West African monsoon during the late African Humid Period, approximately 6 ka (6000 yr ago). At this time, summer insolation was about 5% greater than today in the Northern Hemisphere, mainly due to changes in the longitude of perihelion of the earth’s orbit, and the Sahara and Sahel were much wetter. GCM simulations applied to this problem underestimate the increase in precipitation suggested by the paleodata.
The regional model has a large domain—including all of Africa and extending to 48°W over the Atlantic. The horizontal resolution is 90 km, and all simulations are run with a 90-s time step for 137 days, from 15 May to 30 September, with the first 16 days disregarded for spinup. Climatological, but seasonally varying, surface and lateral boundary conditions are prescribed. The present-day control climatology validates well and is compared with simulations that include both full and partial 6-ka forcing.
Three forcing functions are used to impose 6-ka conditions, namely, changes in the earth’s orbital parameters, a reduction in atmospheric CO2 concentrations from 330 to 280 ppmv, and a change in land surface types to reflect the “greening” of the Sahel and the Sahara 6 ka (Hoelzmann et al. 1998).
In the simulation with all three forcing functions applied, summer precipitation rates increase across the Sahel and the Sahara, with the largest precipitation enhancements of over 8 mm day−1 in the central Sahel. These values agree well with the paleodata presented in section 2. Farther south, however, between 5° and 10°N, the model simulates drying. This is in disagreement with reliable evidence from the crater lake Bosumtwi at 6°N in Ghana. This may be due to the use of present-day SSTs in 6-ka simulations. The analysis here concentrates on the atmospheric dynamics related to the higher precipitation rates in the Sahel and Sahara.
Positive precipitation anomalies of the 6-ka climate are attributed to three differences in the circulation.
In the 6-ka monsoon, the moist flow from the Gulf of Guinea into the Sahel continues farther northward. This is essentially a northward shift of the present-day monsoon dynamics, in which (to first order) the Sverdrup balance maintains the climatology. That is, low-level stretching associated with midtropospheric condensational heating is balanced by the advection of low planetary vorticity air by the southerly flow across the Guinean coast.
The net loss of moisture off the west coast that occurs in the present-day climate is reversed in the 6-ka climate. This occurs for two reasons. One is that the African easterly jet does not form in the 6-ka simulation because the strong meridional surface temperature gradients that characterize today’s climate in association with surface wetness contrasts between the ITCZ and the Sahara Desert are eliminated. The other reason is that the onshore westerly jet at low levels near 10°–15°N is intensified and deepened in the 6-ka climate. This intensified westerly jet contributes to the climatological vorticity balance by advecting low relative vorticity air from the tropical Atlantic high to balance the low-level stretching associated with midtropospheric condensational heating over the Sahel. Both of these dynamical responses are similar to modes of interannual variability seen in present-day climate.
Higher precipitation rates are delayed until later in the summer (July and August) in the Sahara. In contrast to the present-day climate, a deep thermal low develops over the central Sahel in the 6-ka climate in June and July and the Saharan high does not form. The resulting cyclonic circulation transports moisture into the Sahara in July and August.
Despite the increased insolation of 6 ka, surface temperatures are cooler, with simulated maximum cooling of about 5 K across the Sahel from 15° to 20°N. Surface temperatures are cooler primarily due to a reduction in solar radiation reaching the surface due to increases in cloud cover. Limited temperature proxy data show agreement, but reconstructions are severely lacking. An analysis of the moist static energy budget is used to show how rainfall can increase in the Tropics despite such surface cooling. The vertical column is destabilized, and convection is enhanced by increased low-level specific humidity. This increase is related to the large-scale circulation and moisture transport by the processes listed above, as well as increases in evaporation.
A simulation in which only 6-ka radiative forcing is used, with no changes in the vegetation, indicates that forcing by insolation (and CO2) alone is not enough to eliminate the African easterly jet, so the response in the precipitation field is significantly weaker in this case. This could partially explain the weak response in GCM simulations that do not account for changed land surface characteristics.
A simulation to isolate the influence of 6-ka vegetation shows that vegetation forcing alone produces precipitation and circulation changes similar to those of the simulation with the 6-ka radiative and vegetation forcing, especially in the early part of the summer. However, the ultimate forcing for the Africa Humid Period is insolation, that, in the real climate, causes these vegetation changes.
Given the prominent role of vegetation forcing, future simulations will take into account land surface–atmosphere interactions more completely, and simulations with a potential vegetation model are planned as well to eliminate the dependence of the modeling on an external specification of vegetation at 6 ka.
Dr. E. K. Vizy provided invaluable support and advice in running the simulations, and his help is gratefully acknowledged. Financial support was provided by NSF Awards 0123797 and 0415481. We also wish to thank V. Masson-Delmotte, two anonymous reviewers, and Journal of Climate editor, D. M. Straus, for their constructive suggestions during the review process.
Corresponding author address: Christina M. Patricola, Dept. of Earth and Atmospheric Sciences, Cornell University, 3152 Snee Hall, Ithaca, NY 14853. Email: firstname.lastname@example.org