Reservoirs are a significant source of atmospheric carbon (C), but their emission rates vary in space and time. Here, we compared C emissions via diffusion, ebullition, and degassing pathways for six large hydropower reservoirs in the southeastern US that were previously sampled in summer 2012. We revisited the same stations and used a similar methodology to assess which emissions pathways were dominant during the two times. While recent models have suggested that degassing could be a dominant pathway, in these reservoirs, the contribution of degassing to total emissions was low, typically < 5%. Instead, we found that CO2 diffusion was the dominant emissions pathway in 2022. All six reservoirs were CO2 sources in the summer of 2012, but sinks in the summer of 2022. Next, we explored drivers of spatial variation and found relationships associating indicators of greater algal production with lower CO2 but higher CH4 emissions. Finally, to explore drivers of temporal variability, we sampled one reservoir during three drawdown phases (full summer pool, mid-drawdown, and winter pool). Temporal variation in CO2 diffusion rates paired with seasonal productivity patterns, and we found that lower water levels associated with lower hydrostatic pressure and reduced distance for oxidation resulted in the highest CH4 emissions rates. Our results demonstrate that important regional differences are not yet reflected in efforts to produce C emissions estimates for reservoirs globally. Measuring C emissions from multiple pathways and understanding their spatial and temporal responses and variability is vital to reducing uncertainties in global upscaling efforts.