Fire drives animal population dynamics across many ecosystems. Yet, we still lack an understanding of how most species recover from fire and the effects of fire severity and patchiness on recovery processes. This information is crucial for fire-mediated biodiversity conservation, particularly as fire regimes change globally. We conducted an experiment to test whether post-fire recovery is driven by in situ survival or recolonisation, and to determine whether this varies with fires of increasing percentage area burnt (burn cover) and severity. We used the pale field rat (Rattus tunneyi) as a model, because it represents the extinction process for a suite of mammal species suffering population collapse across Australia’s northern savannas. Our treatments spanned a gradient from patchy, low severity fires (simulating early-dry season management burns) to thorough, high severity fires (simulating wildfires). We performed capture-mark-recapture, vegetation and aerial surveys before, six weeks after and one year after fire. Six weeks after fire, pale field rats were only captured in unburnt patches of vegetation, and capture rates were proportional to the amount of unburnt habitat. One year later, both vegetation and pale field rat populations recovered across all sites. However, population recovery after low severity fires was likely achieved through in situ survival and reproduction in unburnt micro-refuges, compared to recolonisation driving recovery after high severity fires. Synthesis and applications. Pale field rat persistence is strongly dependent on the retention of unburnt habitat patches within fire affected areas. Management strategies that increase micro-refugia within burnt areas may facilitate pale field rat population recovery. Globally, building recovery mechanisms into fire management will be vital for supporting the long-term persistence of fire-affected species.

R.tunneyi.FireExperiment_READ_ME.txt Read me document detailing datasets. Prelim_Supp_R_code.docx Full R code and documentation for preliminary analyses. Full_R_code.docx Full R code and documentation for all analyses in manuscript. SECR Recapture data used for supplementary SECR analysis, to determine if capture rate approximates abundance. Patchiness_Index.xlsx Data/calculation of patchiness index. Recaptures_per_Site.csv Capture-mark-recapture data. Trap_Autocorrelation.csv Data for preliminary trap spatial autocorrelation analysis. Trap_Data.csv Capture data and trap/site meta-data over three trapping sessions. Trap_Data_Sex_Age.csv Capture data and trap/site meta-data over three trapping sessions, split into sex and age. Veg_Point_Intercept_Data.csv Vegetation survey data.

All data, analyses and R code are documented in the READ_ME and word documents provided.