The δ13C values of atmospheric carbon dioxide (CO2) can be used to partition global patterns of CO2 source/sink relationships among terrestrial and oceanic ecosystems using the inversion technique. This approach is very sensitive to estimates of photosynthetic 13C discrimination by terrestrial vegetation (ΔA), and depends on δ13C values of respired CO2 fluxes (δ13CR). Here we show that by combining two independent data streams – the stable isotope ratios of atmospheric CO2 and eddy-covariance CO2 flux measurements – canopy scale estimates of ΔA can be successfully derived in terrestrial ecosystems. We also present the first weekly dataset of seasonal variations in δ13CR from dominant forest ecosystems in the United States between 2001 and 2003. Our observations indicate considerable summer-time variation in the weekly value of δ13CR within coniferous forests (4.0‰ and 5.4‰ at Wind River Canopy Crane Research Facility and Howland Forest, respectively, between May and September). The monthly mean values of δ13CR showed a smaller range (2–3‰), which appeared to significantly correlate with soil water availability. Values of δ13CR were less variable during the growing season at the deciduous forest (Harvard Forest). We suggest that the negative correlation between δ13CR and soil moisture content observed in the two coniferous forests should represent a general ecosystem response to the changes in the distribution of water resources because of climate change. Shifts in δ13CR and ΔA could be of sufficient magnitude globally to impact partitioning calculations of CO2 sinks between oceanic and terrestrial compartments.