The loss of utilisable rangeland in semi-arid areas results in huge economic and social costs worldwide. Only adaptive management strategies are able to cope with these systems, which are largely driven by unpredictable and stochastic rainfall. Additionally they are characterized by strong feedback mechanisms between economic and ecological factors. This study aims to contribute to the identification of basic principles for sustainable grazing management. The approach emphasizes learning from existing management systems through the use of ecological-economic modelling. Two apparently successful management systems in Namibia are used as a starting point for a broader analysis: the Gamis Karakul sheep farm and the land use system of the semi-nomadic Ova-Himba. Although the economic systems differ strongly, their management seems to have similarities: the importance of pasture resting and of adapting livestock numbers to available forage. This PhD thesis contributes substantial insights about the relevance and functioning of pasture resting for sustainable grazing management in semi-arid regions. Assessment of the two case studies leads to the hypothesis that resting in the rainy season, particularly during wet years, is fundamental for ensuring pasture productivity under low regeneration potential of the vegetation. The thesis highlights that resting during wet years acts as a risk reducing strategy. Additionally, the study reveals that access to economic risk management strategies, such as rain-index-insurance, may change farmer´s behaviour towards less conservative strategies. The used approach - learning from existing apparently successful grazing strategies by ecological-economic modelling - offers a powerful tool for tackling new questions related to global change. The scope and the limits for generalizing the key factors discovered for sustainable grazing management can be easily detected under changing ecological, climatic and economic conditions.