Background Tropical savannas represent a large proportion of the area burnt each year globally, with growing evidence that management to curtail fire frequency and intensity in some of these regions can contribute to mitigation of climate change. Approximately 25% of Australia’s fire-prone tropical savanna region is currently managed for carbon projects, contributing significantly to Australia’s National Greenhouse Gas Inventory. Aims To improve the accuracy of Australia’s national carbon accounting model (FullCAM) for reporting of fire emissions and sequestration of carbon in savanna ecosystems. Methods Field data from Australian savannas were collated and used to calibrate FullCAM parameters for the prediction of living biomass, standing dead biomass and debris within seven broad vegetation types. Key results Revised parameter sets and improved predictions of carbon stocks and fluxes across Australia’s savanna ecosystems in response to wildfire and planned fire were obtained. Conclusions The FullCAM model was successfully calibrated to include fire impacts and post-fire recovery in savanna ecosystems. Implications This study has expanded the capability of FullCAM to simulate both reduced emissions and increased sequestration of carbon in response to management of fire in tropical savanna regions of Australia, with implications for carbon accounting at national and project scales.