Disturbances by fire and harvesting are thought to regulate the carbon balance of the Canadian boreal forest over scales of several decades. However, there are few direct measurements of carbon fluxes following disturbances to provide data needed to refine mathematical models. The eddy covariance technique was used with paired towers to measure fluxes simultaneously at disturbed and undisturbed sites over periods of about one week during the growing season in 1998 and 1999. Comparisons were conducted at three sites: a 1-y-old burned jackpine stand subjected to an intense crown fire at the International Crown Fire Modelling Experiment site near Fort Providence, North-west Territories; a 1-y-old clearcut aspen area at the EMEND project near Peace River, Alberta; and a 10-y-old burned, mixed forest near Prince Albert National Park, Saskatchewan. Nearby mature forest stands of the same types were also measured as controls. The harvested site had lower net radiation (Rn), sensible (H) and latent (LE) heat fluxes, and greater ground heat fluxes (G) than the mature forest. Daytime CO2 fluxes were much reduced, but night-time CO2 fluxes were identical to that of the mature aspen forest. It is hypothesized that the aspen roots remained alive following harvesting, and dominated soil respiration. The overall effect was that the harvested site was a carbon source of about 1.6 gC m−2 day−1, while the mature site was a sink of about −3.8 gC m−2 day−1. The one-year-old burn had lower Rn, H and LE than the mature jackpine forest, and had a continuous CO2 efflux of about 0.8 gC m–2 day−1 compared to the mature forest sink of − 0.5 g C m−2 day−1. The carbon source was likely caused by decomposition of fire-killed vegetation. The 10-y-old burned site had similar H, LE, and G to the mature mixed forest site. Although the diurnal amplitude of the CO2 fluxes were slightly lower at the 10-y-old site, there was no significant difference between the daily integrals (− 1.3 gC m−2 day−1 at both sites). It appears that most of the change in carbon flux occurs within the first 10 years following disturbance, but more data are needed on other forest and disturbance types for the first 20 years following the disturbance event.