Small headwater dams and their resulting reservoirs are a ubiquitous feature of stream networks, with approximately 2 million small reservoirs in the continental United States alone. By impeding the downstream movement of water, sediment, and organic matter, dams can increase greenhouse gas emissions from inland waters, but the factors that influence the spatial and temporal variability in small reservoir emissions are poorly studied, complicating inferences about the factors controlling net emissions across reservoirs. We characterized in-reservoir spatial and diel variability in emissions from four small reservoirs (0.001-0.008 km2) in Clarke County, GA, USA over 24-hour periods during Aug-Sep, when annual temperatures are high. Every three hours, we measured (1) the diffusive flux of CO2 and CH4 at 10 locations in each reservoir using a static flux chamber method and (2) the ebullitive flux of CH4 at 25 locations using a stationary bubble trap method. We used sequential Gaussian simulation with simple kriging to interpolate our flux estimates at unmeasured locations in the ponds. Total estimated emissions ranged from 1.92 to 16.5 g CO2-eq m-2 day-1, which is within the range of previous observations in small reservoirs. The dominant emissions pathway varied by pond: CO2 diffusion contributed the most CO2 equivalents in two ponds, CH4 diffusion in one pond, and CH4 ebullition in one pond. Ebullition was spatially variable, but when detected was greater at shallower depths. The diffusive flux of CO2 exhibited distinct diel patterns: three ponds had lower and one pond had higher CO2 fluxes at night compared to day. Only one pond exhibited diel patterns in the diffusive flux of CH4, with lower fluxes at night, and no ponds exhibited diel patterns in ebullition. This information can be used to optimize sampling effort in future studies and to incorporate climate impacts into long-term dam management decisions.