Development and comparison of SAVANNA ecosystem models to explore the concept of carrying capacity

This research used integrated model and database systems to explore savanna ecosystem responses to selected land use policies based on different carrying capacity definitions. Two models were tested inside the savanna ecosystem of the Kruger National Park (KNP), Republic of South Africa. The ACRU-Veld model, intended to be a simplified representation of ecosystem processes, was designed and constructed for this research effort. The SAVANNA model, developed outside South Africa, was also used to represent a complex, spatially sensitive ecosystem model.;After the construction of input databases such as climate, soils, vegetation, herbivores and fire, two testing sites were chosen in the KNP, the Sabie River Basin (SRB) and the Northern Plains Region (NPR). The ACRU-Veld and SAVANNA models were tested over a six year period (1989-1995) and compared with monitored herbaceous layer and herbivore population data. At the test site scale, the ACRU-Veld herbaceous biomass predictions closely followed monitored totals while the SAVANNA model under-predicted most biomass totals. In addition, herbaceous biomass predictions were analyzed in the more detailed and spatially homogenous Land Type scale. Both models showed success at representing biomass in some Land Types and incorrectly predicted biomass in others. The SAVANNA model was more successful than ACRU-Veld in simulating herbivore responses to ecosystem dynamics.;After the model testing phase, several grazing-related policies were simulated over a twenty year period (1971-1991). These policy scenarios included wildlife, livestock and mixed wildlife/livestock grazing with intense and low-level management options. The policy scenarios were analyzed with respect to water, herbaceous vegetation and herbivore population dynamics. Both models showed a water-limited environment with different estimates of herbaceous layer and herbivore production potentials. The ACRU-Veld model simulated water-limited biomass with little impact of herbivory in most scenarios while the SAVANNA model showed large differences in herbaceous biomass and herbivore condition with respect to the different scenarios.;While both models were capable of simulating the differences between various grazing strategies, these initial results are not to be taken without some qualifications. More testing is required to assess the role of the woody vegetation in the policy scenarios. This modelling research constitutes a first step in the iterative cycle of gaining confidence in the use of databases and models to explore complex ecosystem processes.