A Simulation Model for Shrub Ecosystem in the Semiarid Karoo, South Africa

Abstract

Plant community dynamics in semiarid regions appear to be "event—driven." The aim of our model is to attain an understanding of the main processes determining the spatial and temporal dynamics of a shrub community in the semiarid Karoo on a large temporal scale and to identify the significant events that drive this shrub community. Vegetation at the study site covers 15—20% of the soil surface and is dominated by five shrubs: Brownanthus ciliatus (Mesembryanthemoideae), Ruschia spinosa (Ruschioideae), Galenia fruticosa (Aizoaceae), Pteronia pallens (Asteraceae), and Osteospermum sinuatum (Asteraceae). Grasses and annuals play little part in the dynamics of this vegetation. The model is based on detailed life history data for the five dominant species and on monthly long—term rainfall data for this region. The method of "dynamic automata" is employed to model individual plants. Growth, death, seed production, germination, and seedling establishment are modelled over long time scales in annual time steps under the influence of the stochastic ad unpredictable rainfall in ungrazed rangeland. In the absence of grazing, survival of seedlings depends on their competitive ability during the seedling stage and their ability to compete with established plants in neighboring cells. The model shows that the dynamics of this shrub community are typified by episodic and discontinuous changes in species composition with intervening quasistable phases lasting some decades. The reason for this episodic behavior is that both recruitment and mortality of plants depend on particular conditions. Using 93 complete data sets with monthly rainfall data taken at the weather station in Prince Albert we show that rainfall is only sufficient for seedling recruitment in 44% of all years for B. ciliatus and in < 30% of all years for the four other species. We identify two types of abrupt and discontinuous changes in species composition: (1) big recruitment events that can only occur if plant density is low and if rainfall conditions are extraordinarily favorable and (2) big mortality events that can only occur if cohorts, originating from big recruitment events, die within a short period of time and if little further recruitment has taken place. However, this behavior is not a property of the biota but is generated by rainfall input to the model. By using a different rainfall input, the model can also display regular cyclic succession. Therefore there appeared to be no contradiction between sudden, discontinuous changes and gradual, continuous and reversible changes in vegetation composition.