Coastal wetlands are major global carbon sinks; however, they are heterogeneous and dynamic ecosystems. To characterize spatial and temporal variability in a New England salt marsh, greenhouse gas (GHG) fluxes were compared among major plant-defined zones during growing seasons. Carbon dioxide (CO2) and methane (CH4) fluxes were compared in two mensurative experiments during summer months (2012–2014) that included low marsh (Spartina alterniflora), high marsh (Distichlis spicata and Juncus gerardii-dominated), invasive Phragmites australis zones, and unvegetated ponds. Day- and nighttime fluxes were also contrasted in the native marsh zones. N2O fluxes were measured in parallel with CO2 and CH4 fluxes, but were not found to be significant. To test the relationships of CO2 and CH4 fluxes with several native plant metrics, a multivariate nonlinear model was used. Invasive P. australis zones (?7 to ?15 ?mol CO2·m?2·s?1) and S. alterniflora low marsh zones (up to ?14 ?mol CO2·m?2·s?1) displayed highest average CO2 uptake rates, while those in the native high marsh zone (less than ?2 ?mol CO2·m?2·s?1) were much lower. Unvegetated ponds were typically small sources of CO2 to the atmosphere (<0.5 ?mol CO2·m?2·s?1). Nighttime emissions of CO2 averaged only 35% of daytime uptake in the low marsh zone, but they exceeded daytime CO2 uptake by up to threefold in the native high marsh zone. Based on modeling, belowground biomass was the plant metric most strongly correlated with CO2 fluxes in native marsh zones, while none of the plant variables correlated significantly with CH4 fluxes. Methane fluxes did not vary between day and night and did not significantly offset CO2 uptake in any vegetated marsh zones based on sustained global warming potential calculations. These findings suggest that attention to spatial zonation as well as expanded measurements and modeling of GHG emissions across greater temporal scales will help to improve accuracy of carbon accounting in coastal marshes.