At a deciduous forest in the southeast United States (Walker Branch Watershed, Oak Ridge, Tennessee), as at other sites with tall vegetation and/or complex terrain, it is difficult to temporally integrate eddy covariance data to obtain long-term estimates of net ecosystem exchange of carbon dioxide (NEE), primarily because of suspected systematic nocturnal errors. Therefore, although eddy covariance data can be invaluable, additional tools such as empirical gap-filling methods, independent measurements of CO2 flux using chambers, and simulations using canopy process models are often necessary to obtain reliable annual carbon uptake estimates. Two independent approaches for estimating annual NEE using these tools at the Walker Branch site are discussed. One approach is to cumulatively sum the full set of eddy covariance measurements over time. The second approach is to sum the output of NEE from a biophysical canopy exchange model (CANOAK). CANOAK incorporates independent chamber measurements on leaves, soil, and stems and is driven using the observed canopy architecture, meteorology, soil water content, and soil temperatures to predict NEE. Both methods estimate similar trends and magnitudes of daytime (daylight hours) soil respiration and NEE over 5 years. Both methods also suggest similar differences among years (interannual variability). These two estimates of NEE are used to address possible measurement bias errors at this site and to provide plausible estimates of annual NEE. The estimated mean annual NEE at this site is -574 g C m-2 yr-1 between 1995 and 1999, ranging from -470 g C m-2 y-1 (1995) to -629 g C m-2 yr-1 (1999) (negative NEE indicates uptake by forest). ¿ 2001 American Geophysical Union