Context -Bush encroachment has been observed in southern African savannas for several decades. This form of land degradation reduces the abundance of (palatable) grass biomass on farms, which in turn can support less livestock sustainably. A reduction in livestock usually results in lower income or less food for farmers and can therefore threaten livelihoods. The vegetation dynami cs of savannas are complex and multifactorially interconnected. Spatially explicit, process-oriented rangeland models are ideal tools to study these complex vegetation dynamics and the phenomenon of bush encroachment. These models can aid in optimizing management to reduce bush encroachment and give guidance to focus future research on the most relevant factors influencing vegetation processes and bush encroachment in savannas.Objectives -This dissertation aims to identify and examine the key factors that can cause and influence the emergence and expansion of bush encroachment in southern African savannas. To this end, we incorporated potentially important factors and processes in a simulation model to study their impact on bush encroachment and to be able to test different environmental and management scenarios with a focus on realistic vegetation dynamics and livestock management.Methods -To identify the potentially most important environmental and management factors that influence the vegetation dynamics of bush encroachment, we interviewed farmers and stakeholders, held workshops and had discussions with scientists and leading experts in the field.Through programming a landscape generator (PioLaG) and a process-oriented rangeland simulation model (Midessa), we gained the ability to set these influential factors and processes in relation to another and to observe and test their cause-and-effect relationship in terms of the impact on vegetation type distribution and livestock condition.Results -The most important factors that determine the current state of savanna rangelands are heterogeneous precipitation patterns, soil moisture, vegetation competition and succession, fire events, seed availability, grazing and browsing, wood reduction by firewood remo val or through arboricides, as well as other management factors such as livestock type and stocking density, type and characteristics of the grazing system, and farmer back-up plans for drought periods. Simulation results showed that introducing browsers and/or mixed feeders to a farm will greatly slow down bush encroachment. A grazer/browser mix thus seems to be a promising approach for a nonchemical management tool against bush encroachment. Careful and adaptive farming strategies including a sophisticated rotational grazing system, will improve vegetation conditions for livestock.Reducing the livestock density on a farm has also been found to slow down bush encroachment and delay the decline of livestock condition. From a modelling perspective , an interesting observation was that spatially heterogeneous distributed precipitation had a major impact on Abstract II simulation results ...