Fire is probably the greatest natural and anthropogenic environmental disturbance in Australian tropical savannas with large tracts of savanna burnt annually.  These landscape scale fires are likely to have massive impacts on the regional water, energy and carbon dioxide exchanges and as a result may have important feedbacks to the atmosphere and regional to global climate.  We initiated the Savanna Fire Experiment (SaFE) at Howard Springs, Darwin, N.T., to examine the impacts of burning on carbon dioxide, heat and moisture budgets. Eddy covariance measurements of surface fluxes were commenced in August 2001.  This paper examines measurements of the initial fluxes during the fire and post burn effects as well as the longer term (>2 year) vegetation recovery. Fire and the subsequent scars radically alter the surface energy budgets of tropical savanna lands through reduced surface albedo (0.12 to 0.06), increased available energy for partitioning, increased sensible heat fluxes (40%) and decreased latent heat fluxes (75%). There was a relatively rapid recovery of water but not carbon exchanges driven by canopy flush around six weeks following fire.  In addition, biomass burning on this scale in Australian savannas is responsible for releasing large amounts of carbon dioxide into the atmosphere and important in determining the overall net biome productivity.  Fire is an important disturbance factor that will reduce Net Ecosystem Production through initial burning emissions and recovery.  Net Ecosystem Production minus fire losses was +0.8 (2001-2002) & +0.3 (2002-2003) and Net Biome Production was estimated at -0.6 tC ha^-1 y^-1.  There are possibilities for reduced carbon emissions and carbon sequestration in this environment.