Interactive effects of increases in atmospheric CO2 and reductions in plant
species diversity were investigated in planted calcareous grassland commun
ities in northwestern Switzerland. The experimental communities were compos
ed of 5, 12, and 31 species assembled from the native species pool. The stu
dy aimed at testing whether the CO2 responses of ecosystems change when spe
cific sets of species are lost from plant communities. Species were selecte
d so that the proportion of grasses, legumes, and non-legume forb individua
ls remained constant across levels of diversity. The most diverse plant com
munity had approximately the same diversity as the surrounding grassland, a
nd species occurring in less diverse communities were subsets of the specie
s in the more diverse communities. The factorial atmospheric-CO2 treatment
was applied using 50-cm-tall, open-bottom, open-top wind screens. Plant com
munity-level responses and the responses of the individual species were ass
essed over a period of five years. A significant positive correlation betwe
en plant community diversity and biomass was detected, but this effect was
not present on all dates. Significant effects of elevated CO2 on community
biomass were only found in the first years of treatment. CO2 effects were l
argest in the communities with the highest number of plant species and were
primarily due to the presence of responsive species not present in the les
s diverse communities. The time dependency of community responses to elevat
ed CO2 and species diversity was related to shifts in community structure o
f the experimental plots. Community responses at the beginning of the exper
iment were dominated by the response of species with a less competitive/str
ess-tolerant life history, These species were successively lost from experi
mental plots as the experiment proceeded, and the observed community-level
effects became smaller. Changes in species composition over the experimenta
l duration were affected by elevated CO2 in the way that species loss was r
educed (i.e., coexistence of species performing well at the beginning and a
t the end of the five-year period increased) and the way that community eve
nness was increased (i.e., dominance was reduced). Based on these results o
ur main conclusions are that (1) community-level responses to CO2 enrichmen
t depend on the species present; (2) the positive correlation between produ
ctivity and species numbers was caused by different species at the beginnin
g and at the end of the experiment; (3) therefore, a large, redundant speci
es pool is important in assuring high productivity under altering environme
ntal conditions; (4) elevated CO2 has the potential to substantially alter
the structure of grassland communities, even if community productivity does
not increase; and (5) a short-term effect of elevated CO2 may be misleadin
g when attempting to predict longer-term effects.