Forecasting patterns of shifting biodiversity refuges in an increasingly flammable world

In a warming and rapidly changing world, biodiversity is increasingly threatened by more frequent, severe, and larger fires. Variation in the life history attributes and habitat preferences of species mean that they may be affected differently by fire, and hence, decision makers must account for this. Understanding how fire affects the distribution of important areas of habitat and refuges for biodiversity can help guide appropriate conservation and management actions. In 2019-20 Australia suffered widespread and devastating megafires, known as the Black Summer, and the East Gippsland region in state of Victoria was heavily affected. To estimate impacts of the fires on fauna within the region, we used a joint species distribution model fitted to data from 967 camera trap sites to estimate variation in the occurrence for 40 native and invasive animal species. We focused on the influence of the fire regime, lethal control of introduced red foxes (Vulpes vulpes), timber harvesting, precipitation, and soil gradients, on wildlife. We then predicted the spatial distribution of each native animal species in the dataset for three time periods: 1) in 2017 prior to the Black Summer megafires, 2) in 2022 following the 2019/2020 wildfires and 3) in 2030. This allowed us to estimate changes in the patterns of occurrence for each species attributable to fire. Finally, we used spatial conservation prioritisation to identify priority conservation areas (‘fire refuges) and identify where potential threats (e.g. invasive species) might co-occur with these areas. We found that fire regime variables, including time since fire and repeat short-interval fires, influence occurrence patterns for ~50% of species and this meant that the megafires influenced predicted occurrence patterns. Half of the mammal species (mostly small and medium-sized ground-dwelling mammals) were positively associated with a topographic wetness index, as were four bird species. Spatial conservation prioritisation also suggested that while there were some shifts in the location of fire havens due to the megafires, some large areas of high-conservation value persisted after fire. These areas were disproportionately likely to be long-unburnt (>= 80 years) with fewer repeat burns, have high fox baiting intensity, and low probabilities of introduced red fox and feral cat (Felis catus) occurrence. This suggests maintaining older vegetation in landscapes and managing predation pressure by invasive predators will be important to maintaining the identified high priority refuges. Our study presents a useful approach for guiding fire management before and after large disturbance events, and could be expanded to test the response of species and communities to forecasts of future fire regime scenarios. Predictive approaches such as this study will be essential for managers to understand the possible outcomes of management actions on biodiversity during a time of rapid global change.