<p>As with most aquatic ecosystems, reservoirs play an important role in the global carbon (C) cycle and emit greenhouse gases (GHG) as carbon dioxide (CO₂) and methane (CH₄). However, GHG emissions from reservoirs are poorly quantified, especially in temperate systems, resulting in high uncertainty. We compared reservoir C emission estimates and uncertainty of diffusive, ebullitive, and degassing pathways in six hydropower reservoirs in the southeastern United States among four data sources: two field-based surveys and two models (including the GHG Reservoir “G-res” Tool). We found that CH₄ diffusion was most similar across data sources (modeled minus observed, bias = −&#xa0;21&#xa0;g CO_2-eq m⁻² y⁻¹) and had low relative uncertainty (coefficient of variation, CV = 0.98). On the other hand, CO₂ diffusion was least consistent across data sources (bias = −&#xa0;518&#xa0;g CO_2-eq m⁻² y⁻¹). Both field surveys indicated strong negative CO₂ diffusion (i.e., CO₂ uptake) at all reservoirs, while G-res estimated positive CO₂ diffusion. By extension, total C emissions showed similar discrepancies, leading to high uncertainty in upscaling and interpreting reservoir source-sink dynamics. Finally, CH₄ ebullition had the highest relative uncertainty (CV = 2.77) due to high variability across sites. We discuss limitations of field surveys and these models, including temperature-based annualization methods, varying definitions of ebullition zones, low sampling resolution, and lack of dynamism. Future field efforts focused on capturing variability in CO₂ diffusion and CH₄ ebullition will be especially valuable in reducing uncertainty and improving models to advance our understanding reservoir GHG emissions.</p>