Australia has the worst mammal extinction record in the world, with medium-sized ground-dwelling mammals particularly affected. Inappropriate fire regimes and introduced predators are considered amongst the most serious threats to these species. The forests of north-eastern NSW remain the stronghold for three threatened medium-sized macropods, the Northern Long-nosed Potoroo, Parma Wallaby and Red-legged Pademelon, however the 2019-20 bushfires burnt approximately half the known records of the first two, and approximately 15% of the latter, which led to increased risk ratings and predictions of decline for all three in rapid post-fire assessments.

In this study, we tested whether the occurrence of these macropods was lower in severely burnt forest compared to unburnt or moderately burnt sites in north-eastern NSW three years post-fire, and whether this was affected by the presence of predators.

We deployed 180 camera traps for six months across four regions at sites stratified by fire severity (high-extreme = severe, moderate or unburnt) in wet and dry sclerophyll forest along 350 km latitudinal and 1000+ m altitudinal gradients. Over half a million images were classified using Artificial Intelligence models in Wildlife Insights and validated by ecologists. Multi-species Occupancy Models were used to test whether occupancy was affected by fire severity, the presence of feral cats, foxes, dingos or the native Spotted-tailed Quoll.

Potoroos were found at 46% of sites, Parma Wallabies at 34%, and Red-legged Pademelons at 19%, with occupancy significantly affected by fire severity and region. Potoroos and Parma Wallabies occurred most frequently in severely burnt sites, contrary to the prevailing expectation, while the Red-legged Pademelon was absent from severely burnt sites. In addition to fire severity, occupancy was influenced by predator presence, although the effects varied depending on understorey density and vegetation type.

Fire is a key ecological process in the development and maintenance of temperate coastal heathlands, where lack of fire can lead to a transition to scrub or woodland. Inappropriate fire regimes, along with other threatening processes, have contributed to an ongoing decline in the extent and condition of coastal heathlands. Prescribed burning is a widely-used management strategy to maintain heathland vegetation structure and flora and fauna diversity. Understanding the responses of native and invasive mammal species to prescribed burning is important to enable adaptive management of heathlands for biodiversity conservation.
This study investigates flora and fauna responses to prescribed burning of patches of long-unburnt (40 years plus) coastal heathland at a Bush Heritage reserve in Tasmania. Six transects were sampled annually in Spring, following Winter burning, for up to 3 years post-burn. Three additional long-term transects were surveyed four times since 2008. We recorded vegetation structure and shrub species cover before and after prescribed burning and at control (unburnt) sites, plus soil surface cover and frequency counts of herbaceous species in quadrats. Camera traps with lures to attract animals deployed before and after burning and in long-unburnt vegetation provided a measure of relative activity of mammal species.
Shrub cover and mean height was lower in recently burnt sites, particularly for obligate seeder species. Mammal species displayed varying responses to burning. Most of the 14 mammal taxa detected were infrequent in both burnt and unburnt vegetation. Small mammals, including introduced rodents, showed a preference for unburnt vegetation, except for pygmy possums. Wallabies and brush-tailed possums were more frequent in burnt areas.
Our results suggest that patch burning creates habitat heterogeneity which supports flora and fauna species with differing life history traits. Further work is needed to understand the role of fire intensity in maintaining heathland biodiversity and preventing woody thickening.

Many Australian native plant species are perceived as inherently flammable, leading to resistance toward their use in revegetation of urban and peri-urban areas. However, a subset of native species exhibit fire-resistant, fire-retardant, and/or fire-tolerant traits. The Fire Wise project, led by the Foundation for National Parks & Wildlife, aimed to identify and promote local provenance native species that support hazard reduction while maintaining biodiversity and ecological function.
We partnered with 14 community-based organisations across three states, prioritising areas affected by the 2019–2020 bushfires. Collaborators contributed field-based knowledge of species responses to fire, particularly regarding flammability, ember resistance, and post-fire regeneration. Through community workshops, site observations, and literature synthesis, we identified 110+ Fire Wise species. Selection criteria included structural traits (e.g. moisture content, growth habit) influencing fire behaviour (e.g. windspeed), and ecological value (e.g. habitat provision).
Project outcomes include the development of region-specific plant lists, demonstration gardens, increased propagation and seed collection of selected species via community-based nurseries, and the creation of a publicly accessible online portal. These resources promote informed plant selection that reduces fire risk and recovery costs, while supporting native biodiversity in peri urban landscapes. By integrating local ecological knowledge with fire-response traits, this project contributes to foundational understanding of plant-fire interactions and offers a scalable model for community-driven fire resilience.

Newland Head Conservation Park lies on the Fleurieu Peninsula and contains one of the Region’s few remaining sandy heath vegetation communities. This remnant is critical habitat for a range of threatened flora and fauna and requires careful management to maintain populations of those species. Large areas of the reserve have no recorded fire history and observations showed senescing vegetation and declining diversity.

In 2021 the National Parks and Wildlife Service conducted an ecological burn aimed at stimulating the germination of obligate seeders, including several threatened species, before their seedbanks were exhausted. This intervention was also required to restore habitat structure for species such as Western Beautiful Firetail (Stagonopleura bella samueli) and Chestnut-rumped Heathwren (Hylacola pyrrhopygia parkeri). Monitoring at Newland Head Conservation Park included pre and post fire shrub cover transects as well as pre and post fire fauna occupancy. We will highlight the results of this monitoring and demonstrate how monitoring is used to guide ecological burning and why this adaptive approach is critical to the management of threatened species in the Mount Lofty Ranges.

Contemporary climate change, land clearing, fire suppression, and inappropriate management policies are resulting in rapid fire regime changes and managing these changes usually requires knowledge of historical fire regimes. One commonly used source of historical fire data is derived from satellite imagery, which has wide geographical coverage and semi-automated processes for burn scar identification. However, satellites are not well suited to detecting small fires at scales relevant to management, or low intensity fires which do not burn overstory vegetation or significantly change surface reflectance. This gap in historical fire data limits our ability to manage fire for conservation and human safety. We aimed to improve estimates of fire frequency from satellite data by incorporating mapped fire history data from public land and environmental co-variation. We produced a generalisable modelling workflow, applying boosted regression trees, generalised linear, and generalised additive models to predict fire frequency in an eastern Australian case study. Generalised linear models estimated low fire frequencies well and were appropriate for mapping land as either burnt or unburnt in recent decades. This has useful applications for situations where researchers need to separate whether particular sites had experienced fire or not. Generalised additive models most accurately estimated higher fire frequencies and were most appropriate for mapping the total number of fires over recent decades (i.e., fire frequency) for most vegetation types. Our models were useful for eucalypt vegetation, but less accurate for rainforest vegetation, likely due to its unique fire regime. Thus, model selection is dependent on the application and vegetation type, but our approach can improve the accuracy of satellite derived fire history estimates, assisting fire management and conservation.

Australia’s diverse ecosystems give rise to, and are shaped by, a wide range of fire regimes. Most Australian ecosystems burn with some frequency, with fire intervals ranging from a few years to several decades, except in areas that are too dry, infertile, or wet to support fire. At seasonal to decadal scales, fire occurrence is controlled by (seasonal) patterns of vegetation growth and desiccation, causing fires to follow rains in the arid grasslands of the interior and droughts in the forests of southern Australia. These climate-fire patterns are routinely, though informally, used by AFAC (Australasian Fire and Emergency Service Authorities Council) to issue seasonal bushfire outlooks. Released quarterly, AFAC bushfire outlooks highlight increased ‘bushfire risk potential’ for grassy regions in response to above-average rainfall and in forests in response to relatively dry and warm conditions. Being based on expert analysis of fuels, recent rains and temperatures plus Bureau of Meteorology climate outlooks, the seasonal bushfire outlooks are difficult to reproduce or formally evaluate.

We present a novel data-driven approach to seasonal bushfire forecasting using Deep Learning. Our model learns spatiotemporal patterns linking remotely sensed fire activity with time series of remotely sensed vegetation greenness, soil water balance, and climate variables, as well as location factors related to ignition probabilities such as distance to roads, powerlines and railways. The model ingests 3 months of input data to predict fire probability at 4 km resolution, 1–3 months into the future, across Australia. Model evaluation against independent burned area data shows forecast accuracies of 80–90%, consistent across both tropical and temperate regions, and across typical and extreme fire seasons. These seasonal forecasts offer objective, reproducible guidance for fire management and serve as a valuable resource for exploring Australia’s unique pyrodiversity.