There is much uncertainty about the net carbon (C) exchange of boreal forest ecosystems, although this exchange may be an important part of global C dynamics. To resolve this uncertainty, net C exchange has been measured at several sites in the boreal forest of Canada as part of the Boreal Ecosystem-Atmosphere Study (BOREAS). One of these sites is the Southern Old Aspen site at which diurnal CO2 and energy (radiation, latent, and sensible heat) fluxes were measured during 1994 using eddy correlation techniques at different positions within a mixed 70 year old aspen-hazelnut forest. These measurements were used to test a complex ecosystem model “ecosys” in which mass and energy exchanges between terrestrial ecosystems and the atmosphere are simulated hourly under diverse conditions of soil, management, and climate. These simulations explained between 70% and 80% of diurnal variation in ecosystem CO2 and energy fluxes measured during three 1 week intervals in late April, early June, and mid-July. Total annual CO2 fluxes indicated that during 1994, aspen was a net sink of 540 (modeled) versus 670 (measured) g C m−2 yr−1, while hazelnut plus soil were a net source of 472 (modeled) versus 540 (measured) g C m−2 yr−1. The aspen-hazelnut forest at the BOREAS site was therefore estimated to be a net sink of about 68 (modeled) versus 130 (measured) g C m−2 yr−1 during 1994. Long-term simulations indicated that this sink may be larger during cooler years and smaller during warmer years because C fixation in the model was less sensitive to temperature than respiration. These simulations also indicated that the magnitude of this sink declines with forest age because respiration increases with respect to fixation as standing phytomass grows. Confidence in the predictive capabilities of ecosystem models at decadal or centennial timescales is improved by well-constrained tests of these models at hourly timescales.