Biodiversity indicators in semi-arid, agricultural Western Australia

The predicted future loss of native Australian species of plants and animals, in part as a result of adverse land management strategies, has led to attempts to identify areas of high biotic richness (numbers of species). Bioindicators are measures of the physical environment, or of a subset of the plants or animals, that best predict biotic richness. Ideally, bioindicators should aim at predicting as large a component of the plant or animal fauna as is possible at minimum cost. For two contrasting vegetation types, we examined remnant area, vegetation structural diversity, species richness of plants, lizards and terrestrial arthropods, and the relative abundance of individual arthropod species, as indicators of faunal richness, using correlation, principal component regression and stepwise regression analyses. The study was carried out in gimlet Eucalyptus salubris woodlands (29 sites) and shrublands (27 sites) in semi-arid, agricultural Western Australia. Sites varied considerably in grazing history (woodland) and in farming history (shrubland). Fauna sampled were lizards (woodland), scorpions (woodland), isopods (woodland), cockroaches (woodland), termites (woodland, shrubland), earwigs (woodland), hemipterans (shrubland), beetles (woodland, shrubland), butterflies (shrubland) and ants (woodland, shrubland). None of the indicator variables in any analyses effectively predicted total faunal richness for either vegetation type (<35% of variation in total richness explained). In correlation analyses for woodlands, vegetation structural diversity and plant richness, but no fauna variable, explained a high percentage of the variation in the richness of lizards (56% explained by richness of native plants, +ve), scorpions (48%, richness of native plants, +ve), termites (55%, vegetation structural diversity, +ve) and beetles (59%, litter, –ve). The richness of the shrubland fauna was poorly predicted by all indicator variables (<25% explained). When using the total richness and abundance of ant functional groups, the abundance of a subset of species within ant functional groups, and of termite and beetle species, in principal component regressions, various ant functional groups explained 42% each of the richness of scorpions and beetles, and eight beetle species explained 50% of termite richness. When remnant area, vegetation structural diversity and the richness of native plants in woodland were tested in step-wise regressions as indicators of total faunal richness, remnant area was the only significant indicator variable, explaining 33% of total richness. The richness of native plants and vegetation structural diversity explained a total of 76% of the pooled richness of lizards + scorpions + termites. No significant indicator variable was found by regression procedures for total richness, or for a subset, of the shrubland fauna. We argue that differences in the predictive qualities of vegetation structure and plant richness between the vegetation types was due, in part, to differences in the spatial heterogeneity of biotic richness, and possibly the scale at which structure was measured. The use of structural diversity or plant richness as predictors of faunal richness for different woodland types, or those with different disturbance histories, or in different geographic or climatic regions, should not be adopted without verification of their efficiency at predicting the richness of the local fauna.