Quantitative integration of the literature on the effect of elevated C
O2 on woody plants is important to aid our understanding of forest hea
lth in coming decades and to better predict terrestrial feedbacks on t
he global carbon cycle. We used meta-analytic methods to summarize and
interpret more than 500 reports of effects of elevated CO2 on woody p
lant biomass accumulation and partitioning, gas exchange, and leaf nit
rogen and starch content. The CO2 effect size metric we used was the l
og-transformed ratio of elevated compared to ambient response means we
ighted by the inverse of the variance of the log ratio. Variation in e
ffect size among studies was partitioned according to the presence of
interacting stress factors, length of CO2 exposure, functional group s
tatus, pot size, and type of CO2 exposure facility. Both total biomass
(WT) and net CO2 assimilation (A) increased significantly at about tw
ice ambient CO2, regardless of growth conditions. Low soil nutrient av
ailability reduced the CO2 stimulation of WT by half, from + 31 % unde
r optimal conditions to + 16 %, while low light increased the response
to + 52 %. We found no significant shifts in biomass allocation under
high CO2. Interacting stress factors had no effect on the magnitude o
f responses of A to CO2, although plants grown in growth chambers had
significantly lower responses (+ 19 %) than those grown in greenhouses
or in open-top chambers (+ 54 %). We found no consistent evidence for
photosynthetic acclimation to CO2 enrichment except in trees grown in
pots < 0.51 (- 36 %) and no significant CO2 effect on stomatal conduc
tance. Both leaf dark respiration and leaf nitrogen were significantly
reduced under elevated CO2 (- 18 % and - 16 % respectively, data expr
essed on a leaf mass basis), while leaf starch content increased signi
ficantly except in low nutrient grown gymnosperms. Our results provide
robust, statistically defensible estimates of elevated CO2 effect siz
es against which new results may be compared or for use in forest and
climate model parameterization.