Ecosystem process models were used to estimate seasonal and annual landscape carbon fluxes in the boreal forests near Fairbanks, Alaska. The forest landscape is a successional mosaic of recently burned to old-growth stands and productive to unproductive stands growing on a variety of soil types. A model of forest dynamics simulated annual production and decomposition for a successional mosaic of 250 stands growing on five representative sites. A model of daily ecosystem photosynthesis and respiration decomposed these annual carbon fluxes into a seasonal cycle. Simulated annual aboveground tree production, annual moss production, and annual forest floor decomposition were not significantly different from observed data. Monthly estimates of the normalized difference vegetation index derived from NOAA's AVHRR sensor, which is thought to represent canopy photosynthesis, accounted for 85% of the seasonal variation in landscape-averaged tree carbon flux. On average the landscape absorbed 74 g C m-2 yr-1. Carbon uptake by trees was the dominant flux (225 g C m-2 yr-1). Carbon loss during microbial respiration was of secondary importance (144 g C m-2 yr-1). Carbon uptake during moss growth and carbon loss during fire were less important (23 and 30 g C m-2 yr-1, respectively). Potential model errors indicate that the net landscape carbon uptake represents the lower bounds of the possible uptake. If valid for the circumpolar taiga, this net flux extrapolates to an uptake of 0.85 to 1.11 Gt C per year. ¿American Geophysical Union 1991