Although biomass burning of savannas is recognised as a major global source of greenhouse gas emissions, quantification remains problematic with resulting regional emissions estimates often differing markedly. Here we undertake a critical assessment of Australia's National Greenhouse Gas Inventory (NGGI) savanna burning emissions methodology. We describe the methodology developed for, and results and associated uncertainties derived from, a landscape-scale emissions abatement project in fire-prone western Arnhem Land, northern Australia. The methodology incorporates (i) detailed fire history and vegetation structure and fuels type mapping derived from satellite imagery; (ii) field-based assessments of fuel load accumulation, burning efficiencies (patchiness, combustion efficiency, ash retention) and N : C composition; and (iii) application of standard, regionally derived emission factors. Importantly, this refined methodology differs from the NGGI by incorporation of fire seasonality and severity components, and substantial improvements in baseline data. We consider how the application of a fire management program aimed at shifting the seasonality of burning (from one currently dominated by extensive late dry season wildfires to one where strategic fire management is undertaken earlier in the year) can provide significant project-based emissions abatement. The approach has wider application to fire-prone savanna systems dominated by anthropogenic sources of ignition.