QUANTITATIVE EFFECTS OF GRAZING ON VEGETATION AND SOILS OVER A GLOBALRANGE OF ENVIRONMENTS

Multiple regression analyses were performed on a worldwide 236-site da ta set compiled from studies that compared species composition, aboveg round net primary production (ANPP), root biomass, and soil nutrients of grazed vs. protected, ungrazed sites. The objective was to quantita tively assess factors relating to differential sensitivities of ecosys tems to grazing by large herbivores. A key question in this assessment was: Do empirically based, broad-scale relationships correspond to ec ological theories of plant-animal interactions and conceptual framewor ks for management of the world's grazing lands? Changes in species com position with grazing were primarily a function of ANPP and the evolut ionary history of grazing of the site, with level of consumption third in importance. Changes in species composition increased with increasi ng productivity and with longer, more intense evolutionary histories o f grazing. These three variables explained > 50% of the variance in th e species response of grasslands or grasslands-plus-shrublands to graz ing, even though methods of measurement and grazing systems varied amo ng studies. Years of protection from grazing was a significant variabl e only in the model for shrublands. Similar variables entered models o f change in the dominant species with grazing. As with species composi tion, sensitivities of change in dominant species were greater to vary ing ecosystem-environmental variables than to varying grazing variable s, from low to high values. Increases of the dominant species under gr azing were predicted under some conditions, and decreases were more li kely among bunch grasses than other life-forms and more likely among p erennials than annuals. The response of shrublands was different from that of grasslands, both in terms of species composition and the domin ant species. Our analyses support the perception of grazing as a facto r in the conversion of grasslands to less desirable shrublands, but al so suggest that we may be inadvertently grazing shrublands more intens ively than grasslands. Percentage differences in ANPP between grazed a nd ungrazed sites decreased with increasingly long evolutionary histor ies of grazing and increased with increasing ANPP, levels of consumpti on, or years of treatment. Although most effects of grazing on ANPP we re negative, some were not, and the statistical models predicted incre ases in ANPP with grazing under conditions of long evolutionary histor y, low consumption, few years of treatment, and low ANPP for grassland s-plus-shrublands. The data and the models support the controversial h ypothesis that grazing can increase ANPP in some situations. Similar t o species variables, percentage differences in ANPP between grazed and ungrazed treatments were more sensitive to varying ecosystem-environm ental variables than to varying grazing variables. Within levels not c onsidered to be abusive ''overgrazing,'' the geographical location whe re grazing occurs may be more important than how many animals are graz ed or how intensively an area is grazed. Counter to the commonly held view that grazing negatively impacts root systems, there was no relati onship between difference in ANPP with grazing and difference in root mass; as many positive as negative differences occurred, even though m ost ANPP differences were negative. Further, there was a weak relation ship between change in species composition and change in ANPP, and no relationship with root mass, soil organic matter, or soil nitrogen. Al l three belowground variables displayed both positive and negative val ues in response to grazing. Current management of much of the world's grazing lands based on species composition criteria may lead to errone ous conclusions concerning the long-term ability of a system to sustai n productivity.