Abstract. Quantifying system-wide biogeochemical dynamics and ecosystem metabolism in estuaries is often attempted using a long-term continuous record at a single site or short-term records at multiple sites due to sampling limitations that preclude long-term monitoring. However, differences in the dominant primary producer at a given location (e.g., phytoplankton versus benthic producers) control diel variations in dissolved oxygen and associated ecosystem metabolism, and they may confound metabolic estimates that do not account for this variability. We hypothesize that even in shallow, well-mixed estuaries there is strong spatiotemporal variability in ecosystem metabolism due to benthic and water-column properties, as well as ensuing feedbacks to sediment resuspension, light attenuation, and primary production. We tested this hypothesis by measuring hydrodynamic properties, biogeochemical variables (fluorescent dissolved organic matter – fDOM, turbidity, chlorophyll a fluorescence, dissolved oxygen), and photosynthetically active radiation (PAR) over 1 year at 15 min intervals at paired channel (unvegetated) and shoal (vegetated by eelgrass) sites in Chincoteague Bay, Maryland–Virginia, USA, a shallow back-barrier estuary. Light attenuation (KdPAR) at all sites was dominated by turbidity from suspended sediment, with lower contributions from fDOM and chlorophyll a. However, there was significant seasonal variability in the resuspension–shear stress relationship on the vegetated shoals, but not in adjacent unvegetated channels. This indicated that KdPAR on the shoals was mediated by submerged aquatic vegetation (SAV) and possibly microphytobenthos presence in the summer, which reduced resuspension and therefore KdPAR. We also found that gross primary production (Pg) and KdPAR were significantly negatively correlated on the shoals and uncorrelated in the channels, indicating that Pg over the vegetated shoals is controlled by a feedback loop between benthic stabilization by SAV and/or microphytobenthos, sediment resuspension, and light availability. Metabolic estimates indicated substantial differences in net ecosystem metabolism between vegetated and unvegetated sites, with the former tending towards net autotrophy in the summer. Ongoing trends of SAV loss in this and other back-barrier estuaries suggest that these systems may also shift towards net heterotrophy, reducing their effectiveness as long-term carbon sinks. With regards to temporal variability, we found that varying sampling frequency between 15 min and 1 d resulted in comparable mean values of biogeochemical variables, but extreme values were missed by daily sampling. In fact, daily resampling minimized the variability between sites and falsely suggested spatial homogeneity in biogeochemistry, emphasizing the need for high-frequency sampling. This study confirms that properly quantifying ecosystem metabolism and associated biogeochemical variability requires characterization of the diverse estuarine environments, even in well-mixed systems, and demonstrates the deficiencies introduced by infrequent sampling to the interpretation of spatial variability.