**1. Introduction**

Unprecedented changes in climate, urbanization, and economic development are increasing the pressures that societies are enforcing on ecosystems [1]. Developing sustainable ecosystem services is subsequently a priority for conservation management, with savanna ecosystems a landscape of primary concern. Savannas are mixed plant communities comprised of grasses and woody vegetation that cover approximately a quarter of the Earth's land surface, including roughly half of the African continent [2]. Savannas are an extremely important socio-economic landscape in Africa, with over 80% of savanna land used to raise livestock [3], underpinning the economic stability of many countries [4,5]. The dynamic nature of savannas means they are susceptible to changes, particularly shifts in plant community composition associated with an increase in woody vegetation [6,7]. A particularly concerning aspect of this increased density of woody vegetation is the reduction of grasses and herbs by encroaching woody species. These negative impacts are occurring at an increasingly frequent rate worldwide [8–10], which is a major threat to the ecosystem stewardship of these economically important landscapes.

The transition of savanna ecosystems to open shrubland across Botswana, and in particular the western part of the Kalahari, presents a considerable threat to the conservation of the economically important ranching industry. In order to develop adaptive managemen<sup>t</sup> strategies, the underlying environmental drivers of woody vegetation species need to be better understood. By understanding the environmental drivers responsible for the diversity and abundance of woody vegetation, we can develop predictive models to identify 'high-risk' areas, and provide managers, farmers, and governments with decision support across savanna landscapes. Previous research addressing the ecological processes responsible for the observed vegetation patterns have often found conflicting results regarding the importance and significance of these environmental drivers [11–15], thus limiting the use of this knowledge as the basis for decision-making at a landscape scale. These differences will be discussed below in the context of savanna ecosystems.
