Battling the elements : environmental determinants of North Kimberley rainforests

Rainforests and savannas are floristically and structurally distinct vegetation types which, within an intermediate rainfall range (1000-2000 mm˜ív°year-1), can coexist in the same landscape. In those areas the relative importance of bottom-up factors, like geology and water availability, or top-down controls, such as fire and megaherbivores, as drivers of rainforest distribution is still controversial. In this thesis I aim to assess the influence of environmental factors on rainforest density and boundary change in the Wunambal Gaambera Country (north Kimberley, Western Australia), where rainforests are represented by small patches surrounded by a savanna matrix, and are potentially under threat due to climatic changes, alteration of fire regimes, and the recent introduction of cattle. I then provide management guidelines for the Indigenous land managers, taking into account western science and Aboriginal knowledge. First, I used a GIS-based approach to investigate variations in rainforest density in the north-western Australian monsoon tropics by blending existing maps, and in the north Kimberley by creating ex novo a detailed map of rainforest patches (30-m accuracy). I showed that at a regional scale climatic factors associated with water availability, such as mean annual rainfall and moisture index, were the main drivers of rainforest density. At a landscape scale geology and topography strongly affected rainforest density, supporting the hypothesis that bottom-up controls influence rainforest distribution. Nonetheless, rainforest density was higher in locations characterised by low fire frequency, implying that top-down controls may also be involved. To further investigate the role of top-down controls, I conducted a natural experiment to test the effects of disturbance on historical rainforest expansion trends and current structure of rainforests and adjacent savannas, using historical aerial photographs and vegetation transects. I selected two study locations characterised by similar climate and geology, to remove the effect of bottom-up controls, but strikingly different levels of disturbance (fire activity and cattle). Disturbance negatively affected rainforest expansion, and was associated with sharp rainforest-savanna boundaries. In disturbance-free areas savannas displayed abundance of rainforest elements, with gradual transition from rainforest to savanna. Despite the importance of fire in determining rainforest distribution, growing evidence shows that some rainforest species are able to survive a single fire event. To test this hypothesis, I compared survival rate and resprouting strategies of rainforest saplings burnt with an experimental treatment that mimicked a low-intensity savanna fire with savanna saplings burnt by an ambient early dry season fire. The results suggested that, despite low stem survival rates that restricted resprouting to basal buds, a proportion of the rainforest species found in the north Kimberley is able to survive a fire event. Basal resprouting however negatively affects growth rates, yet this does not necessarily prevent rainforest expansion under low-intensity fire regimes. Finally, I tested whether the fire management currently in place is leading to a positive change in fire regimes. I also examined the implications of vegetation management for the local fauna, with particular focus on threatened species, by using camera traps to record animal presence across the vegetation transects described above. Fire management successfully led to a shift from late dry season to early dry season fires, which are typically less intense. However, both high fire frequencies and unmanaged cattle had a negative effect on native animal distribution. Fire management should then focus on reducing fire frequency in areas exposed to prevailing wildfires through planned burning, while of in fire-protected areas a relatively low frequency of patchy prescribed fires would suffice to keep the sizes of single fuel age patches at manageable levels. In conclusion, I found that disturbance regimes have a strong influence on rainforest distribution, which is determined by the interplay of both bottom-up and top-down controls. Land management should thus be adaptive, and evolve based on changing environmental conditions between different locations and in time.