1. Eddy covariance was used to measure the net CO2 flux of Alaskan moist-tussock and wet-sedge tundra ecosystems between 1 June and 31 August 1995. The sites were located within 2·5 km of each other and, depending on wind speed and thermal stability, the eddy-covariance measurements integrated surface fluxes over 0·8–26·5 ha (mean 3·8 ha) at the moist-tussock site and 0·1–4·2 ha (mean 0·6) at the wet-sedge site.
2. Both sites were net sinks for atmospheric CO2 during the 92-day measurement period but wet-sedge tundra accumulated 1·5 times more CO2 than moist-tussock tundra. Wet-sedge tundra was a net CO2 sink of −6·4 mol m−2 and moist-tussock tundra was a net sink of −4·6 mol m−2 over the June–August measurement period.
3. Estimates of whole-ecosystem respiration (R) were made using multiple non-linear regression by quantifying the response of measured nocturnal CO2 efflux (Fn) to fluctuations in air temperature and water-table depth (r2 = 0·68). Partial regression analysis revealed that water-table depth explained relatively more of the variance in Fn (45%) than temperature (11%). R was estimated to be 14·9 mol m−2 for moist-tussock tundra and 5·4 mol m−2 for wet-sedge tundra over the 92-day measurement period. The large difference in R was apparently owing to a substantially higher water table in wet-sedge tundra.
4. Estimated rates of gross primary production (GPP), calculated from measured net CO2 flux and estimated R, were 40% lower for wet-sedge tundra over the 92-day measurement period. Rates of GPP integrated over diel (24 h) periods were substantially larger for moist-tussock tundra ecosystems throughout June and July but during the month of August, total diel rates of GPP for moist-tussock and wet-sedge tundra were comparable.
5. Our results demonstrate that R and GPP of moist-tussock and wet-sedge tundra ecosystems can be reliably estimated from eddy-covariance measurements using functional relationships developed from plot-scale studies. Given the large spatial differences in hydrology, the larger sink observed for wet-sedge tundra was probably the result of relatively lower rates of R.