Aquaculture ponds represent a biogeochemical hotspot of the global carbon cycle. However, accurate estimations of their carbon budgets are hindered by a limited understanding of the temporal variability of carbon fluxes across time scales. In this study, the eddy covariance (EC) approach was applied to quantify net ecosystem CO2 exchange (NEE) over algae-shellfish aquaculture ponds (razor clam cultivation) in Zhangjiang Estuary of Southeast China from January 2020 to June 2022, aiming to assess the diurnal variability of NEE during various management periods. The EC-based NEE over the ponds showed strong diurnal variations with daytime sink and nighttime source, mainly controlled by photosynthetically active radiation and air temperature, respectively. The ponds acted as a net sink system during razor clam (daily mean flux of −0.42 μmol m−2 s−1) and shrimp-crab (−0.50 μmol m−2 s−1) farming periods with a stronger daytime sink than a nighttime source, while it was the opposite during the drainage period acting as a net source (0.40 μmol m−2 s−1). The strength of sink/source also differed between the early and late stages of the razor clam farming period, with much larger (∼ 16 times) net carbon uptake in the late stage as a result of the strong daytime sink and near-neutral nighttime flux. The diurnal variability of NEE over the ponds was overall larger than other aquatic ecosystems and tended to increase with air temperature. Previous estimates of daily NEE from discrete daytime-only samplings might have biased the actual carbon budgets if no diurnal correction was applied in the temporal aggregation. The confirmed strong temporal variability of NEE across time scales highlighted the importance and necessity of continuous and high-frequency flux measurements in assessing the carbon budgets of algae-shellfish aquaculture ponds.