To explore this transformation requires a shift in focus from the primary consumers to the primary producers and associated landscape processes. Few detailed studies at this level have been reported specifically for desertification in the Sahel. Under the assumption that these processes are not geographically unique, a generalized account has been produced based on Warren and Maizels (1977) and Reining (1978), and supplemented by research from elsewhere.

   The direct effects of grazing are on the yield (standing crop) of the herbage (grass and forb) and browse (woody perennial) layer. Depending upon life form, grazing affects population dynamics of the species. Overgrazing can completely eliminate perennial--but may have little effect on annual--plants (see Noy-Meir, 1975). The spread of the grazing impact is determined by the landscape heterogeneity, the herbivore type, and the time of year. Separation of, or competition among, herbivore species is determined by these three dimensions--location, forage type, and time. Diet preferences are characteristically different among herbivore species. Diets change with time as either the relative abundance or quality, or both, of the grazed plant species changes with time; see for example the diet studies published by Coppock et al. (1986a) and Moore (1987).

   A reciprocal interaction exists between the vegetation and the herbivore. The quality of the diet in terms of energy and nutrient content is determined by the selective capability of the herbivore, but is set by the base level offered by the forage (Coppock et al., 1986b; Sinclair, 1975; Skarpe and Bergstrom, 1986). In these environments highly opportunistic consumers appear to enjoy an advantage (Coppock et al., 1986a; Moore, 1987).

   How much of the standing crop of vegetation can be harvested without seriously affecting the future productive capacity of the grass sward is the central question of range management. In a seasonal or stochastic rainfall environment the productivity can be expected in pulses. A nomadic system as described above would harvest intensively for a short time, usually close to the peak of productivity, and depart before the pulse of soil water was depleted. Thus grass growth would continue, root reserves be replenished and reproduction take place without the stress of grazing.

   In variable and unpredictable ecosystems the level of harvesting by long-lived herbivores must be low for populations to be sustained. The more variable the rainfall--pasture growth cycle, the lower the overall level of offtake (Caughley et al. ,1987). A harvesting pattern under nomadic grazing is intermittent and varies according to seasonal conditions. The feedback controlling the herbivore usage is primarily the availability of drinking water and secondarily the biomass levels. The two will be positively correlated and act in concert; high-rainfall years have abundant casual drinking water, whereas in dry years animals are forced to migrate elsewhere, presumably before the level of offtake becomes deleterious. This model is simplistic and ignores the interactions that occur at plant level between defoliation and subsequent growth, which will be discussed later.

   The most influential long-term consequence of sustained and intense grazing on the productivity of the herbage layer is indirect in action. It is affected through altering the patterns of soil water and nutrient flow in the vicinity of plants and results from the cumulative physical impact of the herbivores on the partitioning and redistribution of rainfall. Both nutrient cycling and water flow are patterned in a scale of size of individual plants (Noy-Meir, 1975; Noble and Tongway,1986). The removal of fallen dead plant material (litter) by grazing and (mostly) trampling changes the partitioning of rainfall at the soil surface. As the surface is bared, infiltration is decreased and runoff and redistribution increased. This change is first catalysed by the removal of litter which has served to reduce the kinetic energy of raindrop impact and facilitate infiltration close to the perennial plants (Braunack and Walker, 1985; Noble and Tongway,1986).

   The removal of litter and exposure of the soil surface initiates two positive feedback loops. The first is that baring the soil enhances the formation of soil crusts by direct impact of raindrops. These crusted soils immediately begin to reduce infiltration and encourage redistribution of rainfall. In the longer term the reduced soil water regimes will reduce plant growth, and therefore cover, over the soil and the positive feedback loop is established.

   The loss of litter also facilitates the erosion and redistribution of soil by wind (Warren and Maizels,1977; Noble and Tongway,1986). This is just as effective as water erosion in dramatically altering the spatial patterns of soil water and nutrient flow. Like water erosion it can also be a positive feedback process because of the non-linear relationships, the erosive power of either agent and the relative proportion of protective plant cover. Very small changes in plant cover at or about the threshold, catalyse disproportionately larger changes in erosion (Fig. 4.4).

   The second exacerbating loop is the change in microclimate that comes with reducing plant cover. The greater the reduction in plant cover, the more hostile to plant growth and establishment the physical environment becomes. In particular the temperature and evaporative regimes experienced by plants, especially seedlings, can be dramatically amplified by increasing the area of bare soil in the sward. The outcome of this feedback has often been observed in fenceline contrasts where ungrazed grasslands remain greener longer than do grazed ones. It has been observed at many scales, from the small exclosures of plant research (e.g. Belsky, 1987) to the large areas that are directly observable by satellite (e.g. Wade, 1974; Otterman, 1981; Otterman and Tucker, 1985).

   The end-result of these processes is desertification. The positive feedback loops continue to enhance runoff and redistribution of water at the expense of infiltration. This runoff carries with it both soil and nutrients, and seals the surface over which it runs, further adding to the positive feedback loop. Where the nutrients and water are redistributed depends upon the characteristics of both landscape and rainfall. Redistribution may exaggerate or rescale existing spatial pattern. For example, where water was mostly redistributed over the scale of inter-plant distance (0.1 m) in stable communities, it may now be redistributed over the larger scales of erosion cells (Pickup, 1985) or the clumping or banding of vegetation (Foran,1987  Mabbutt and Fanning, 1987). Ultimately the redistribution of water and nutrients from the most frequently experienced rainfalls may routinely reach the scales of landscape drainage representing an export of water and nutrients, both of which critically control the productivity of arid ecosystems.

   These positive feedback processes result in a transformed landscape. An appealing metaphor is an unravelling of a fabric, the fabric of ecological relationships, and the pattern of the landscape frays (Wolf, 1986). As the scale of patterning of the system increases, the remnant pockets of vegetation begin to function independently of one another and the level of connectedness and interaction within the system declines.