As global temperatures increase, the potential for longer growing seasons to enhance the terrestrial carbon sink has been proposed as a mechanism to reduce the rate of further warming. At the Niwot Ridge AmeriFlux site, a subalpine forest in the Colorado Rocky Mountains, we used a 9-year record (1999–2007) of continuous eddy flux observations to show that longer growing season length (GSL) actually resulted in less annual CO2 uptake. Years with a longer GSL were correlated with a shallower snow pack, as measured using snow water equivalent (SWE). Furthermore, years with a lower SWE correlated with an earlier start of spring. For three years, 2005, 2006, and 2007, we used observations of stable hydrogen isotopes (δD) of snow vs. rain, and extracted xylem water from the three dominant tree species, lodgepole pine, Engelmann spruce, and subalpine fir, to show that the trees relied heavily on snow melt water even late into the growing season. By mid-August, 57% to 68% of xylem water reflected the isotopic signature of snow melt. By coupling the isotopic water measurements with an ecosystem model, SIPNET, we found that annual forest carbon uptake was highly dependent on snow water, which decreases in abundance during years with longer growing seasons. Once again, for the 3 years 2005, 2006, and 2007, annual gross primary productivity, which was derived as an optimized parameter from the SIPNET model was estimated to be 67% 77%, and 71% dependent on snow melt water, respectively. Past studies have shown that the mean winter snow pack in mountain ecosystems of the Western US has been declining for decades and is correlated with positive winter temperature anomalies. Since climate change models predict continuation of winter warming and reduced snow in mountains of the Western US, the strength of the forest carbon sink is likely to decline further.