This paper reports the results of the analysis of eddy covariance CO2 data obtained at a successional forest of maple and aspen at Camp Borden in southern Ontario, Canada, between July 1995 and December 1997. Main findings are (1) The Michaelis-Menton model explains >50--65% of the observed variance of the daytime net ecosystem carbon exchange (NEE) during the growing season; leaf wetness appears to be an important variable contributing to the remaining variance. (2) The whole-ecosystem respiration rate as a function of the 5-cm soil temperature shows a seasonal hysteresis (higher rate in the later part of the year), suggesting a nonnegligible contribution by deep soil/roots and the influence of litter age. (3) There is evidence of photosynthetic activities immediately after the spring snowmelt/soil warming, but the daily NEE did not switch sign till about 40 days later; our best estimates of the annual net carbon uptake by the ecosystem (net ecosystem production (NEP)) are -1.0, -1.2, and -2.8 t C ha-1 yr-1 for the periods July 19, 1995, to July 18, 1996, January 1 to December 31, 1996, and January 1 to December 31, 1997, respectively, with an uncertainty of ¿0.4 t C ha-1 yr-1. (4) The higher NEP value in 1997 than in 1996 was caused by lower growing season soil temperature, cooler spring and fall transitional periods, and higher photon flux in 1997; possible enhancement in canopy photosynthetic capacity may also have played a role. In addition, three main sources of uncertainties, data gap, fetch, and mass flow, are discussed. It is suggested that collective use of the methods available for assessing the whole-ecosystem respiration (friction velocity threshold, mass flow theory, and dark respiration from the forest light response) may increase the confidence level of NEP estimates. ¿ 1999 American Geophysical Union