Abstract. Reservoirs are important sources of greenhouse gases (GHG) to the atmosphere and their number is rapidly increasing, especially in tropical regions. Accurately predicting their current and future emissions is essential but hindered by fragmented data on the subject, which often fail to include all emission pathways (diffusion, ebullition, degassing, and downstream emissions) and the high spatial and temporal flux variability. Here we conducted a comprehensive sampling of Batang Ai reservoir (Malaysia), and compared field-based versus modeled estimates of its annual carbon footprint for each emission pathway. Carbon dioxide (CO2) and methane (CH4) diffusive fluxes were higher in upstream reaches. Reducing spatial and temporal sampling resolution resulted in up to 64 and 28 % change in flux estimate respectively. Most GHGs present in discharged water were degassed at the turbines, and the remainder were gradually emitted along the outflow river, leaving time for CH4 to be partly oxidized to CO2. Overall, the reservoir emitted 2639 g CO2 eq m−2 yr−1, with 90 % occurring downstream of the dam, mostly in the form of CH4. These emissions, largely underestimated by predictions, are mitigated by CH4 oxidation upstream and downstream of the dam, but could have been drastically reduced by slightly raising the water intake elevation depth. CO2 diffusion and CH4 ebullition were lower than predicted, whereas modeled CH4 diffusion was accurate. Investigating latter discrepancies, we conclude that exploring morphometry, soil type, and stratification patterns as predictors can improve modeling of reservoir GHG emissions at local and global scales.