Estimates of net primary (NPP) and ecosystem production (NEP) are needed for tropical savanna, which is structurally diverse but understudied compared to tropical rainforest. Estimates of NPP and NEP are available from eddy covariance and inventory methods, but both approaches have errors and uncertainties. We used both methods to estimate carbon (C) fluxes for an upland mixed grassland and a seasonally flooded forest to determine the correspondence in C cycling components derived from these methods and assess the contribution of the various C cycling components to the overall NEP. Both techniques provided similar estimates of NPP, NEP, and gross primary production (GPP). Belowground NPP accounted for 49-53% of the total NPP for both ecosystems, followed by aboveground litter (26-27%) and wood (16-17%) production. Increases in water availability increased the potential for C storage, but the mechanism was different in the savanna types with an increase in soil moisture causing higher NPP in the mixed grassland but lower ecosystem respiration (R_eco) in the Cerrado forest. Compared to other savanna ecosystems, the mixed grassland had a similar rate of R_eco but lower productivity and C use efficiency (CUE = NPP/GPP = 0.28). The Cerrado forest had a high CUE (0.58) and similar C flux rates to other tropical savanna forests and woodlands. While our measurements are spatially and temporally limited, the agreement in C fluxes estimated using inventory and eddy covariance methods suggest that the C cycle estimates for these savanna ecosystems are robust.