Biodiversity loss is often attributable to multiple interacting pressures that are moderated across environmental gradients. These processes contribute to complex responses that are challenging to interpret and understand. Well-designed and well-implemented monitoring can play a vital role in this but is rarely undertaken.

Mounting evidence suggests that current fire regimes across savanna ecosystems have contributed to the decline of a range of biota. Hence, contemporary fire regimes are at odds with conservation goals.

Using an extensive spatio-temporal monitoring dataset from three large National Parks in northern Australia, we applied generalised linear mixed models to examine: 1) trends in mammal richness and abundance through time and how these vary across environmental gradients such as productivity or landscape position (e.g., terrain ruggedness); and 2) how fire, a potential driver, is moderated by environmental gradients.

Across 24-years, major declines in mammal richness and abundance were observed with greater reductions in less rugged and less mesic habitats. Patterns of decline were related to multiple aspects of fire regimes, but not changes in vegetation structure and composition, or the arrival of Cane Toads Rhinella marina and concomitant poisoning of predatory animals during consumption. More pronounced declines occurred in sites that were exposed to larger fires, had less long unburnt (5 years without fire) vegetation, and were more distant from large, long unburnt patches.

Relationships between mammal persistence and fire varied among vegetation communities, with the strongest fire effects observed in lowland woodland – relatively drier with few barriers to fire spread – where the availability of long unburnt areas moderated declines more than in other vegetation communities.

Synthesis and applications. Current fire regimes are contributing to mammal declines in northern Australian savanna. Collectively, our results highlight: 1) the value of long term monitoring, and importance of considering landscape position when assessing faunal responses to landscape-level perturbations like fire; 2) that significant improvements in fire regimes are required to ameliorate mammal declines; and 3) the need to shift management focus from retaining small, short-lived unburnt patches towards preserving relatively large contiguous areas of long unburnt habitat, particularly in less rugged lowland and sandstone woodlands.

We monitored the mammal community using standardised trapping and nocturnal searches.

During the 1995 – 2015 monitoring period, sites accessible by vehicle were sampled for 72 hrs, and sites accessible only by helicopter were sampled for 48 hrs. Sampling occurred from April to November, with each site visited in the same month and sampled with the same effort between sessions. For the 48 hr surveys, 30 Elliott traps (33 x 10 x 9 cm) were placed equidistantly around the perimeter of a 50 x 50 m trapping quadrat, six cage traps (56 x 20 x 20 cm) were placed at each corner and mid-way along the perimeter, and three 20 L pitfall traps (with 10 m of drift-line fence) were placed within the quadrat. For the 72 hr surveys, the same number of trap nights were achieved using 20 Elliott traps, four cage traps, and two pitfall traps. Between 2016 and 2019 sampling occurred from May – August, and the trapping period was increased to 96 hours, using 16 Elliott traps, eight cage traps, and three pitfall traps. Prior to analysis, the 96 hr sampling data was truncated to either 48 or 72 hrs by omitting captures from the third and/or fourth night to match the previous sampling effort for each site. Cage and Elliott traps were open at night and baited with a mixture of peanut butter, honey, and oats. All traps were checked within 1 h of dawn on each sampling day. All captured mammals were identified to species level and released at the point of capture without being marked. Additionally, spotlight searches were performed on each night of sampling within a 1 ha square overlaying the trapping quadrat. All mammals seen during a 10-minute meander search of the site were recorded.

This study focused on small to medium sized native non-flying mammals (<5 kg body mass, herein referred to as mammals), the group for which marked declines have previously been reported (Woinarski et al., 2010): i.e., we did not consider bats, the dingo Canis familiaris and a small set of macropod species. An abundance value for each mammal species was tallied as the sum of all captures and all individuals recorded during spotlight searches (herein referred to as captures) at each site in each sampling session.

Microsoft Excel