Wetland restoration is proposed as a way to mitigate climate change through carbon sequestration, but methane (CH4) emissions from these systems complicate actual climate benefits. Wetlands are the largest natural source of CH4, and climate feedbacks are expected to increase future emissions. Land-use change alters soils, vegetation, and hydrology, yet the extent to which these legacies influence CH4 emissions from restored wetlands is unclear. To better understand CH4 dynamics across a gradient of human disturbance, we analyzed four years of continuous CH4 flux data from a least disturbed ("natural") wetland, a restored wetland, and a cultivated former wetland on the mid-Atlantic Coastal Plain (Maryland, USA). We quantified temporal relationships between CH4 flux and key biophysical drivers, including soil temperature, gross primary productivity (GPP), and hydrologic variables, and examined how relationships varied across time scales. Fluxes at the natural and restored wetlands were comparable in magnitude, had similar temporal patterns, and responded similarly to biophysical drivers. At the diel scale, wetland CH4 fluxes were primarily driven by GPP, but average diel cycles were the inverse of GPP, with highest fluxes at night. At the seasonal scale, wetland fluxes were driven by the hysteretic effects of soil temperature and water level. In contrast, CH4 fluxes at the drained cultivated site were an order of magnitude lower and showed different interactions with biophysical drivers. Our results indicate the recovery of processes underlying CH4 emissions in a restored wetland, and suggest that the projected rise in CH4 emissions from natural wetlands may be mirrored in restored wetlands. As wetland restoration targets many ecosystem services beyond climate mitigation, accurately predicting CH4 emissions from these systems is essential for assessing their net climate impact.