The development of effective fire management for biodiversity conservation is a global challenge. The highly dynamic nature of fire, the difficulty in replicating ‘real-world’ fire experiments and the need to understand population changes at large spatiotemporal scales make computer simulations particularly useful for identifying optimal fire management regimes for biodiversity conservation. We aimed to develop a flexible modelling approach with which to investigate how the spatiotemporal application of fire (i.e. management scenarios) influences savanna biodiversity. We used existing data from a landscape-scale fire experiment to develop population simulations for the common brushtail possum Trichosurus vulpecula, grassland melomys Melomys burtoni and northern brown bandicoot Isoodon macrourus across the Kapalga area of Kakadu National Park in northern Australia. We simulated how populations were expected to change between 1995 and 2015 in response to the fire patterns observed at Kapalga over this period, and under a hypothetical management scenario of extensive prescribed burning. Our models predicted a substantial decline in all three species in response to the observed fire regime at Kapalga, suggesting that the fire patterns observed at Kapalga, with the associated mechanisms and interactions with other ecological processes, were not conducive with the persistence of native mammal populations. Our prescribed burning scenario had little effect on the predicted population trajectory of the common brushtail possum and grassland melomys, but markedly improved the population trajectory of the northern brown bandicoot. These inconsistencies highlight the need for a nuanced approach to fire management across northern Australian savannas, that is tailored to local conditions and management objectives. Synthesis and applications. The modelling approach outlined here provides a basis for identifying fire patterns that are beneficial for conserving biodiversity, thereby increasing our capacity to establish clear targets for prescribed fire management. Importantly, this approach is flexible and can be easily adapted to other taxa and fire-prone ecosystems.