Maximising the spatiotemporal variability of prescribed fire (i.e. pyrodiversity) is often thought to benefit biodiversity. However, given mixed empirical support, the generality of the pyrodiversity hypothesis remains questionable. Here, we use a simulation experiment to explore the effects of spatiotemporal fire patterns on the population trajectories of four mammal species in a northern Australian savanna: northern brown bandicoot (Isoodon macrourus), northern brushtail possum (Trichosurus vulpecula arnhemensis), grassland melomys (Melomys burtoni), and northern quoll (Dasyurus hallucatus). Underpinned by data from a landscape-scale fire experiment, we simulated mammal population trajectories under three scenarios of fire size (ambient, small/dispersed fires, large/clumped fires) and three levels of dispersal ability (low, moderate, high) over a 21-year period across the Kapalga area of Kakadu National Park. The simulated population size of all four species declined markedly, regardless of fire spatial pattern and dispersal ability. However, the predicted final population size (i.e. number of individuals in the final timestep of the simulation) for the northern brown bandicoot, northern brushtail possum and grassland melomys were significantly influenced by fire size, with declines most severe under the small/dispersed fire scenario. Our results suggest that maximising the dispersion of small fires at the expense of disturbance refugia (such as less-frequently burnt areas) may exacerbate the severity of mammal decline. This highlights the importance of considering trade-offs between spatial (i.e. fire dispersion) and temporal (i.e. fire frequency) aspects of pyrodiversity, and the potential risks when applying fire management for biodiversity conservation without a firm understanding of the requirements of the target species.