Fire and biodiversity: responses of grass-layer beetles to experimental fire regimes in an Australian tropical savanna.
Up to 50% or more of the savanna landscapes of northern Australia are burnt each year, but the effects of these fires on savanna faunas are poorly known, especially for arthropods. We investigated the responses of grass-layer beetles to three fire regimes (unburnt - fires excluded; early - annual fires lit early during the dry season; late - annual fires lit late during the dry season) as part of a landscape-scale fire experiment at Kapalga in Australia's Northern Territory, where replicated fire regimes were applied to 15-20-km2 landscape units over a 5-year period. We also sought to identify beetle species that might act as indicators of particular fire regimes, and examined the extent to which analysis at the family level is sufficient to reveal patterns evident from species-level analysis. Beetles were sampled using sweep nets during February (mid-wet season) and May (end of wet season) each year, from 1989 (18 months prior to the imposition of fire treatments) to 1995 (after 5 years of treatments). A total of 3865 beetles from 233 species was recorded, with the dominant families Chrysomelidae (leaf beetles) and Curculionidae (weevils) contributing 91% of all beetles and 57% of total species. Total beetle abundance, species richness, family richness and the abundances of one of the 10 most common species were significantly affected by fire treatment according to ANOVA. In all cases, beetle abundance and richness were similar in the two burning treatments, but greater than in the unburnt treatment. These differences were only apparent during the second half of the experiment. The responses were mirrored by changes in composition at species, family and guild levels, according to multivariate analysis. According to indicator analysis, five beetle species were significant indicators of late fires, and one of early fires. All these species were infrequently recorded and therefore of limited use for management. However, two common species were indicative of burnt habitats if early and late treatments are considered together. Family richness was highly correlated (r2=0.615) with species richness, and multivariate analysis of family-level abundances revealed the compositional changes in relation to fire that were evident in species-level analysis. Family was therefore a reliable surrogate of species for detecting beetle responses. Although we found significant effects of our experimental treatments, in the broader context of overall beetle dynamics the assemblages appeared to be remarkably resilient to fire. This reflects a long history of association with frequent fire, and is consistent with many other components of the northern Australian fauna, where fire appears to be of secondary importance to variation in rainfall and soil type. The lack of effect of fire timing on grass-layer beetles calls into question a management paradigm that focuses largely on fire intensity, and suggests that fire management needs to be more mindful of fire frequency.