Tropical savanna ecosystems are a major contributor to global CO2, CH4 and N2O greenhouse gas exchange. Savanna fire events represent large, discrete C emissions but the importance of ongoing soil-atmosphere gas exchange is less well understood. Seasonal rainfall and fire events are likely to impact upon savanna soil microbial processes involved in N2O and CH4 exchange. We measured soil CO2, CH4 and N2O fluxes in savanna woodland (Eucalyptus tetrodonta/Eucalyptus miniata trees above sorghum grass) at Howard Springs, Australia over a 16 month period from October 2007 to January 2009 using manual chambers and a field-based gas chromatograph connected to automated chambers. The effect of fire on soil gas exchange was investigated through two controlled burns and protected unburnt areas. Fire is a frequent natural and management action in these savanna (every 1-2 years). There was no seasonal change and no fire effect upon soil N2O exchange. Soil N2O fluxes were very low, generally between -1.0 and 1.0μg Nm-2h-1, and often below the minimum detection limit. There was an increase in soil NH4 + in the months after the 2008 fire event, but no change in soil NO3 -. There was considerable nitrification in the early wet season but minimal nitrification at all other times. Savanna soil was generally a net CH4 sink that equated to between -2.0 and -1.6kg CH4ha-1y-1 with no clear seasonal pattern in response to changing soil moisture conditions. Irrigation in the dry season significantly reduced soil gas diffusion and as a consequence soil CH4 uptake. There were short periods of soil CH4 emission, up to 20μg Cm-2h-1, likely to have been caused by termite activity in, or beneath, automated chambers. Soil CO2 fluxes showed a strong bimodal seasonal pattern, increasing fivefold from the dry into the wet season. Soil moisture showed a weak relationship with soil CH4 fluxes, but a much stronger relationship with soil CO2 fluxes, explaining up to 70% of the variation in unburnt treatments. Australian savanna soils are a small N2O source, and possibly even a sink. Annual soil CH4 flux measurements suggest that the 1.9million km2 of Australian savanna soils may provide a C sink of between -7.7 and -9.4 Tg CO2-e per year. This sink estimate would offset potentially 10% of Australian transport related CO2-e emissions. This CH4 sink estimate does not include concurrent CH4 emissions from termite mounds or ephemeral wetlands in Australian savannas.