Desertification was first named forty years ago to describe the destructive transformation of savannah ecosystems in Africa. 'These are real deserts that are being born today, under our eyes, in the regions where the annual rainfall is from 700-1500 mm' (Aubreville,1949, translated in Glantz and Orlovsky,1983). Twenty years later the process became a large-scale reality; it was swift, extensive, and severe.

   Within the catalogue of ecosystem perturbations desertification could be classified as a 'pulse' experiment; a rapid and severe alteration to the system (Bender et al.,1984). Such experiments theoretically can yield information about the direct interactions between ecosystem components as the system 'relaxes' back to equilibrium. However, the actual difficulties in achieving this goal are formidable (Bender et al., 1984; Yodzis, 1988).

   An equilibrium-centred view is not appropriate for arid ecosystems. Desertification appears to have been an irreversible perturbation. The landscapes of the Sahel now operate within a new domain of stability with attributes and behaviour more closely akin to ecosystems typical of more arid climates.

   Within the limitations of the evidence available, and admittedly with much inference and generalization, it has been possible to construct a feasible explanation of the ecological processes involved in desertification using a very simple model of grazed arid ecosystems (Fig. 4.5). The model has just four components (soils, herbaceous vegetation, herbivores, and man), and two external influences (external society and climate). They are connected by ten interactions, each of which may comprise several processes, e.g. herbivory, infiltration. The pairs of interactions may function as loops and be either strong or weak, positive or negative. In terms of dynamic food chain links this system would be ranked as 2 (Fretwell,1987).

   Given the recorded sequences of events in the Sahel, this model was adequate to explain the observed consequences of those events.

   The nomadic pastoral system has persisted for a considerable time experiencing droughts known to have been as severe as those of 1969--75. Therefore we may conclude that this system had evolved as a stable, resilient strategy. This conclusion is supported by the similarity in grazing strategy between the nomadic pastoralism and the large migratory herbivores, the wildebeest and the kob.

   In its undisturbed state the system was driven by climate (rainfall), interaction (9), and controlled by negative feedbacks between vegetation and herbivores, interaction (6), and to a lesser extent between herbivores and man, interaction (4). In the former, the negative nature of the feedback was determined by the nutritional limits to herbivore health, growth, and reproduction offered by the low-quality but reliable dry-season grasslands. The grasslands acted as a 'passively harsh' environment for herbivores (Oksanen,1988). In the latter case the obligate but hazardous dependence of the pastoralist on the herbivores must also be seen as a limit to human populations through similar controls of disease, growth, and reproduction.

   The patterns of nomadic movements of the livestock can be interpreted as a strategy to harvest the higher-quality, but less predictable, arid grasslands. This nomadic strategy incorporated an additional control in the form of availability of water for livestock. Its availability determined the distribution and density of herbivores sympathetically with the productivity of the grass-lands.

   Consistent with this model the modern epoch of desertification can be explained by one small but catalytic change. It was that the isolation, and therefore the integrity, of the nomadic pastoral society was breached by the interaction with alien economic systems starting about 1950. This produced two responses. First a pulse of population growth of pastoralists and (then) livestock passed from the higher levels to the lower. Secondly a controlling negative feedback between vegetation and herbivores, lack of drinking water, was overridden by the supply of permanent water. This allowed the pulse of disturbance passing down the system to be transmitted to the vegetation and soils where before it would have been suppresssed.

   Under additional stress of the first drought (1969--75), the interaction exceeded some threshold whereby the interaction (8), previously weak and positive, was transformed into strong negative feedback on plant growth and establishment by soil erosion by wind and water. This continued in a self-reinforcing way until a new domain of resilience had been reached. Most interactions had been spatially uncoupled, productivity at all levels had declined, become episodic, and the landscape looked like a desert.

   The collapse of the primary productive system, i.e. diminished production, was transmitted back up the system (1975--?) and interaction (2) became a strong positive feedback on the further aid intervention (1). The cycle repeated itself by 1980--? Given the continual input of ecologically inappropriate intervention, the cycle will continue at an ever-increasing frequency.

   In summary, desertification is a tale of two feedbacks. A nomadic pastoral system was stabilized by the negative feed-backs between the grazed, the grazer, and the grazier. The ecological processes that comprised these feedbacks, e.g. energy and nutrient flow, were determined by the characteristics of arid ecosystems wherein the influence of space and time have exaggerated importance.

   The nomadic system of the Sahel was destabilized by the effective removal of two negative controlling feedbacks. The resultant pulse of disturbance and the drought precipitated a negative feedback of soil erosion which desertified the productive base of the system. This pulse of response, a new and severe negative feedback, passed back up the system only to solicit a further positive feedback of destabilizing aid to renew the cycle again.

   Desertification as an ecological phenomenon reinforces the view that, for arid ecosystems, abiotic factors are dominant. Simple unidirectional representations of ecosystems are consistent, adequate, and intuitively appealing.