We analyze tree growth data from Wisconsin forest inventories completed in 1968, 1983, 1996 and 2002. These show that the rate of forest growth decreased steadily over the period, in contrast to the increases predicted by CO2 fertilization models. Measured growth rate changed an average of -0.27% y-1 (95% confidence range: -0.05% to -0.49% y-1), whereas the prediction for CO2 fertilization is 0.16% y-1 (corresponding to a β of 0.36). The high statistical precision is due both to large sample sizes and positive correlations among the growth rates from different time periods within the same plot. Decreased growth occurred in stands of all ages, and so our results are not caused by age-related declines in growth (although highly significant age-related declines were also detected).

   Data allowing a direct examination of growth rates over several decades are available only for Wisconsin, but Caspersen et al. (2000) introduced an indirect method for detecting past changes in growth rate using only two sequential inventories. This method was criticized by Joos et al. (2002), who claimed that it lacked the statistical power to falsify state-of–the-art ecosystem models of CO2 fertilization. We explain both the sound points and the critical errors in Joos et al.’s argument, introduce a transparent and analytically tractable version of Caspersen et al.’s method, and check its ability to detect the decreasing growth rates in the Wisconsin data. The results show that the indirect method accurately characterizes the past changes that actually occurred, and has sufficient statistical power to falsify CO2 fertilization models, including the model in Joos et al. (2002).

     We discuss the implications of decreasing Wisconsin growth rates, together with other reasons for skepticism about the future magnitude of CO2 fertilization. In particular, the steep reductions in fossil fuel emissions required to stabilize atmospheric CO2 at 500 ppm must begin more than a decade sooner if the predictions of the CO2 fertilization models in the IPCC Third Assessment (Prentice et al. 2001) are incorrect. The difference between a terrestrial carbon sink that grows because of CO2 fertilization, and one that shrinks because it is caused by recovery from past land use, is the difference between the luxury of a substantial delay and the need to act now.