Plant Ecology 150: 77–96, 2000.© 2000 Kluwer Academic Publishers. Printed in the Netherlands.77Simulated long-term vegetation response to alternative stocking strategiesin savanna rangelandsGerhard E. Weber1, Kirk Moloney2& Florian Jeltsch31Department of Ecological Modelling, UFZ Centre for Environmental Research, Leipzig, Germany (E-mail: [email protected]);2353 Bessey Hall, Department of Botany, Iowa State University, Ames, IA 50011-1020, USA,3Universität Potsdam, Vegetationsökologie und Naturschutz, Potsdam, GermanyKey words: Grazing heterogeneity, Grazing management, Livestock grazing, Rangeland management,Semiarid rangeland, Vegetation changeAbstractIncreasing cover by woody vegetation, prevalent in semiarid savanna rangelands throughout the world, is a de-grading process attributed to the grazing impact as a major causal factor. We studied grazing effects on savannavegetation dynamics under alternative stocking strategies with a spatially explicit grid-based simulation modelgrounded in Kalahari (southern Africa) ecology. Plant life histories were modeled for the three major life forms:perennial grasses, shrubs, annuals. We conducted simulation experiments over a range of livestock utilizationintensities for three alternative scenarios of small scale grazing heterogeneity, and two alternative strategies: fixedstocking versus adaptive stocking tracking herbage production. Additionally, the impact of the duration of themanagement planning horizon was studied, by comparing community response and mean stocking rates after20 and 50 years. Results confirmed a threshold behavior of shrub cover increase: at low, subcritical utilizationintensity little change occurred; when utilization intensity exceeded a threshold, shrub cover increased drastically.For both stocking strategies, thresholds were highly sensitive to grazing heterogeneity. At a given critical utilizationintensity, the long term effect of grazing dependedon the level of grazingheterogeneity: whereas under low hetero-geneity, shrub cover remained unchanged, a large increase occurred under highly heterogeneous grazing. Hence,information on spatial grazing heterogeneity is crucial for correct assessment of the impact of livestock grazingon vegetation dynamics, and thus for the assessment of management strategies. Except for the least heterogeneousgrazing scenario, adaptive stocking allowed a more intensive utilization of the range without inflating the risk ofshrub cover increase. A destabilizing feedback between rainfall and herbage utilization was identified as the majorcause for the worse performance of fixed compared to adaptive stocking, which lacks this feedback. Given theusually high grazing heterogeneity in semiarid rangelands, adaptive stocking provides a management option forincreasing herbage utilization and thus returns of livestock produce without increasing degradation risks.IntroductionSemiarid savanna rangelands are characterized bylarge fluctuations in amount and distribution of annualrainfall and, as a consequence, exhibit a high degreeof variability in primary production (Friedel 1990).Traditionally, livestock management has attempted tocope with this variability by stocking at supposedlylow or conservative fixed rates, in an effort to pro-mote long term productivity of the range (Mace 1991).The validity of this management paradigm has beenchallenged, not least due to a trend towards increasingwoody cover reported from many semiarid rangelands(Archer et al. 1988; Dean & MacDonald 1994; Hacker1984; Van Vegten 1982). For this trend, livestockgrazing itself is reckoned a major cause (Buffington& Herbel 1965; Fleischner 1994; Skarpe 1990b; VanVegten 1983). However, if livestock is kept at stockingrates considered to be low by managers why then iswoody cover increasing? The key may be found in78the underlying conceptual model used to understandthe processes drivingvegetationdynamicsin semi-aridecosystems.According to the Clementsian (1916) successionalparadigm, rangeland community dynamics are com-posed of a continuous sequence of reversible statesproceeding towards a single persistent climax in theabsence of grazing or other disturbances. Grazingcounteracts the successional tendency, and a given uti-lization intensity results in a specific equilibrium stateof the vegetation lying somewhere on the continuumfrom an ungrazed or lightly grazed climax state to aheavily grazed early-successional state. Thus, man-agement would ideally select a stocking rate (i.e.,carrying capacity) that keeps the rangeland commu-nity at a desirable successional state, producing amaximum sustainable yield under grazing. However,Westoby et al. (1989) suggest that the equilibriumconcept of continuous and reversible compositionalchanges does not apply to semiarid rangelands. Theyargue that a more appropriate paradigm would be thestate-and-transition concept, which views the systemas being composed of a limited number of states con-nected by a more complex pattern of transitions thanallowed under the Clementsian notion of succession.For example, an adverse climatic event under condi-tions of heavy grazing could trigger a transition from arange dominated by perennial grasses towards a shrubdominated state; however, a reverse transition to thegrassdominated state would be extremelyunlikelydueto low mortality rates and high drought resistance ofwoody vegetation (Westoby et al. 1989). The state-and-transition concept thus offers an explanation forthe observation that degraded rangelands often do notrecover even under drastically reduced utilization in-tensities (Milton & Dean 1995). As a consequence,proponents of the state-and-transition concept advo-cate adaptive stocking as an alternative managementstrategy for avoiding the possibly hazardous conse-quences of fixed stocking based on the range succes-sion approach (Abel & Blaikie 1989; O’Reagain &Turner 1992; Walker 1993). Thus, the two concep-tual models of rangeland dynamics leave us with twofundamentally different management strategies: fixedversus adaptive stocking.In rangeland management, like in any other fielddealing with complex dynamical systems, modelingapproachesare a common means of studying the avail-able strategies. However, most models of rangelandsystems are of limited utility for long-term assess-ments of the impact of different management strate-gies. For example, one group of studies lacked anyfeedback of herbivory on primary production (Riech-ers et al. 1989;