Implications of vegetation clearing biases for the interpretation of landscape-scale species-area relationships

Abstract

Vegetation clearing results in the loss of species from landscapes. Indeed, the area of remaining native vegetation is an important determinant of species richness in human-modified mosaics. Of interest to ecologists and landscape managers is the effect of areamthat is, how the number of species a landscape supports changes with the amount of native vegetation, as revealed by the shape and functional form of the species-area relationship. Understanding this is vital for guiding conservation interventions such as setting limits to vegetation clearing or establishing revegetation targets. Crucially though, it is not only vegetation area that affects patterns of species richness at the landscape levelmso do environmental attributes such as soil properties and topography. Complicating the matter is the fact that these attributes tend to be correlated with landscape-level vegetation area, because humans preferentially remove vegetation from landscapes suited to land uses such as agriculture. However, this interplay between vegetation area, other landscape attributes, and biased patterns of vegetation loss/retention is infrequently considered in landscape-level species-area analyses. If unaccounted for, these confounding factors may result in erroneous interpretations of the effect of area, leading to suboptimal management actions. The aim of this thesis was to examine how attributes of landscapes affect the relationship between species richness and vegetation area. Through four specific research questions, I explored in detail the hypothesis that attributes of human-modified landscapes that bias vegetation clearing also interact with vegetation area to produce landscape-specific area effects on species richness. First, I quantified correlates of vegetation clearing/retention in two regions of the southern hemisphere, and reviewed the literature to determine how often, and in what ways, biased clearing patterns are accounted for in studies relating vegetation area to an ecological response. I demonstrated that soil properties and range in elevation are reliably associated with the amount of remaining native vegetation across ~18,000 100 km2 landscapes in Australia and South Africa. Importantly though, I found that clearing biases were explicitly acknowledged in only 15 of the 118 reviewed studies. If the area of native vegetation in landscapes is a legacy of biased clearing, confounding factors like soil properties should be accounted for in analyses of area effects. Second, I explored the extent to which the effect of native vegetation area on species richness differed in 100 km2 landscapes categorised by attributes such as soil fertility, range in elevation or matrix land use. Using a case study of south-east Australian birds, I found that the shape of the species-area relationship varied substantially depending on whether landscapes were, for example, more- or less-topographically variable, or had higher or lower soil fertility. While threshold models depicting a point of sudden change in the effect of area emerged consistently, the amount of vegetation corresponding with observed thresholds differed considerably among landscape types. Therefore, aggregating and analysing species-area data from different landscape types is likely to misrepresent how species richness is affected by vegetation area. This will be exacerbated by clearing biases, because heavily cleared landscapes tend to be characterised by very different attributes to high cover landscapes. Third, I compared the effect of vegetation area on bird species richness at three scales of analysis (landscapes of 25 km2, 100 km2, 400 km2) for two regions of south-east Australia. When data for the entire study extent were analysed, a remarkable degree of scale-invariance was observedmnamely, a threshold relationship with a change-point at approximately 30% vegetation cover. However, when data were analysed for two regional subsets of the overall dataset, the effect of vegetation area, and the factors moderating this relationship, were scale-dependent. Given this finding, observed thresholds can only reliably be used to guide landscape management at the scale and in the region where the relationship was observed. Finally, I evaluated the implications of accounting for clearing biases when using species-area relationships to guide conservation, focussing on a region of Australia undergoing rapid landscape transformation. I found that using observed thresholds from species-area models that do and do not account for landscape attributes yielded different outcomes for landscape-scale species richness conservation, given a scenario of future vegetation loss. Specifically, the number and location of landscapes that could be prioritised for conservation actions varied considerably depending on the species-area model used. This research demonstrates that the effect of area on species richness differs substantially as a function of the attributes of landscapes. Crucially, clearing biases underpinned by these same attributes can confound analyses of the species-area relationship. Accounting for landscape attributes will allow for a more rigorous understanding of how species richness varies among landscapes with different amounts of native vegetation. A robust appreciation of the effect of area will provide more certainty around how much vegetation needs to be managed (i.e. protected, revegetated), and where this should occur among multiple landscapes, to avert the loss of, or enhance, landscape-scale species richness.