Spurious thresholds in the relationship between species richness and vegetation cover

Aim Thresholds often exist in the relationship between species richness and the area of remaining habitat in human-modified landscapes, prompting debate about the mechanisms responsible. We hypothesize that if speciesarea relationships differ with underlying factors such as landscape productivity, and such factors correlate with patterns of habitat clearance, then spurious thresholds can arise where the separate speciesarea relationships intersect. We assessed whether this phenomenon could explain landscape-level speciesarea relationships for birds occupying 31 landscapes of 100 km2 in eastern Australia. Location Eastern Australia. Methods Landscape-level species richness estimates were modelled as a function of the percentage of native vegetation remaining in the study landscapes. The performance of traditional speciesarea curves and continuous and discontinuous piecewise models was compared using an information theoretic approach. Separate models for high- and low-productivity and high- and low-fragmentation landscapes were examined to determine whether they implied different speciesarea relationships. Results The speciesarea relationship exhibited a rapid change-point at approximately 40% vegetation cover, but this was most parsimoniously explained by two disjunct slopes rather than a continuous threshold model or a classic speciesarea curve. Exploration of models fitted separately to high- and low-productivity landscapes suggested that such landscapes may differ in their characteristic speciesarea relationships. Main conclusions The observed pattern is consistent with the spurious threshold hypothesis, and opens a new avenue of enquiry into the processes behind apparent ecological thresholds. This hypothesis may be valid in other regions where clearing history is confounded by underlying factors such as landscape productivity, and demands further research. In such systems, real thresholds for different landscape types may occur at different levels of cover, or might not exist at all. If so, a simple space-for-time substitution may not be valid, and management prescriptions based on threshold values (e.g. 40%) will be flawed