Spatio-temporal fire regimes are likely to shift with changes in land use and climate. Such a shift in the disturbance regime has been suggested from recent reconstructions of the regional fire history in the Mediterranean-type woodlands and shrublands of Western Australia which suggest that fire was much more frequent before 1930 (fire intervals of 3-5 years) than it is today (fire intervals of 8-15 years). To investigate the potential biodiversity consequences of such changes in fire regime for fire-killed woody species in high diversity Mediterranean-type shrublands we developed a spatial model for the serotinous shrub, Banksia hookeriana, that grows on sandy dunes of the Eneabba sandplain, Western Australia. We sought to identify the conditions of fire size and temporal distribution of ignition events under which the spatially-separated populations in this species are able to persist, and whether this encompasses the fire regimes proposed by recent fire history reconstructions. We tested two fire frequency-size distribution scenarios based on alternative models of fire spread: 1) fire spread as a function of interpatch distance, leading to mostly fires of the same size, and 2) fire spread as a function of interpatch distance and the age of the vegetation (time since last fire) as a measure of available fuel, leading to a broader range of fire sizes with more small fires.

Results/Conclusions

In the first case, population persistence was only possible for mean ignition intervals at the landscape scale of >5 years. In the second case, persistence was possible for the whole range of fire interval distributions suggested by the empirical data. However, persistence is hardly possible if the mean return fire interval at the local scale (i.e. for individual dunes) is <7 years. Our results show that the spatially-separated plant populations can potentially persist over a wide range of temporal fire regimes at the landscape scale, so long as there are buffering mechanisms at work (e.g. feedback between flammability and vegetation age) which reduces the probability of large fires at short intervals on the landscape scale. Our findings do not agree with the empirical fire reconstructions for the early and pre-European period and suggest that this fire history record is incompatible with the conservation management needs of many fire-killed woody species of this highly biodiverse region.