During the past 20 years there have been a large number of GCM sensitivity studies (Mintz, 1984; Dirmeyer and Shukla, 1993) that show that increases in albedo and decreases in vegetation cover and soil moisture tend to further reduce rainfall over the Sahel. We do not know of any GCM sensitivity study challenging this basic result. We present here the results of a recent GCM sensitivity study (Xue and Shukla, 1993 and an unpublished manuscript) in which the Center for Ocean-Land-Atmosphere Interactions (COLA) model was integrated with two sets of land-surface conditions over the region south of the Sahara desert. In one set of integrations, referred to as the "desertification" experiment, it was assumed that the area approximately between 10° and 20°N and 15° and 40°E was covered by shrubs above bare soil. In the other set of integrations, referred to as the "reforestation" experiment, the area approximately between 15° and 40°E was covered by broadleaf trees above ground cover. The first set of surface conditions (desertification) was considered to represent the exaggerated state of current conditions. The second set of surface conditions (reforestation) represented a hypothetical situation in which vegetation was maintained in the margins of the Sahara Desert. It was an exaggeration of the conditions during the 1950s, when rainfall over the Sahel was above normal, and the latitude of the 200 mm isoline of seasonal rainfall in the western Sahel was at about 18°N. The model was integrated with identical global SST patterns for each change in the land-surface conditions.
Figure 2a shows the difference between the observed rainfall averaged for 1950 plus 1958 and 1983 plus 1984. Figure 2b shows the difference between the desertification and reforestation experiments for seasonal mean rainfall. The similarity between Figures 2a and 2b is remarkable. Both show negative rainfall departures between 10° and 20°N and a positive rainfall departure to the south of the negative departures. The dipole nature of both the observed and the model rainfall anomalies suggests a southward displacement of the mean rainfall pattern.
Since the only changes in these integrations were those in land-surface conditions, it is reasonable to conjecture that if the natural variability of the global climate system were to produce an initial drought, the strong atmosphere-land interaction over the Sahel region could contribute toward the persistence of that drought.
We further propose that since natural changes in the Sahel rainfall have led to large-scale changes in human activities in the region, it is quite likely that the degradation of the land surface was exacerbated by human activities (overgrazing, deforestation, soil desiccation, etc.), thereby producing further reduction in the Sahel rainfall. The exacerbating effects of human activities could have been especially large during the past 30 years because the population density has been much higher than at the beginning of the century.
The various conjectures put forward in this paper so far have been synthesized in a schematic diagram shown in Figure 3. It is seen that the atmosphere-land interaction has a positive feedback mechanism, so that initial drought caused by the natural variability of the climate system is perpetuated by increases in albedo and decreases in vegetation, soil moisture, and surface roughness. In addition, the human-induced effects also produce an increase in albedo and a decrease in vegetation, soil moisture, and surface roughness. Thus, both the planetary-scale effects and the human-induced effects are mutually reinforcing the dry conditions, which has led to an unprecedentedly long and severe drought over the Sahel.
If we accept the premise presented in the beginning of this paper, that the Sahel drought was initiated by natural