A time-dependent climate-change experiment with a coupled ocean-atmosp
here general circulation model has been used to study changes in the o
ccurrence of drought in summer in southern Europe and central North Am
erica, In both regions, precipitation and soil moisture are reduced in
a climate of greater atmospheric carbon dioxide. A detailed investiga
tion of the hydrology of the model shows that the drying of the soil c
omes about through an increase in evaporation in winter and spring, ca
used by higher temperatures and reduced snow cover, and a decrease in
the net input of water in summer. Evaporation is reduced in summer bec
ause of the drier soil, but the reduction in precipitation is larger.
Three extreme statistics are used to define drought, namely the freque
ncy of low summer precipitation, the occurrence of long dry spells. an
d the probability of dry soil. The last of these is arguably of the gr
eatest practical importance, but since it is based on soil moisture, o
f which there are very few observations, the authors' simulation of it
has the least confidence. Furthermore, long time series for daily obs
erved precipitation are not readily available from a sufficient number
of stations to enable a thorough evaluation of the model simulation,
especially for the frequency of long dry spells, and this increases th
e systematic uncertainty of the model predictions. All three drought s
tatistics show marked increases owing to the sensitivity of extreme st
atistics to changes in their distributions. However, the greater likel
ihood of long dry spells is caused by a tendency in the character of d
aily rainfall toward fewer events, rather than by the reduction in mea
n precipitation. The results should not be taken as firm predictions b
ecause extreme statistics for small regions cannot be calculated relia
bly from the output of the current generation of GCMs, but they point
to the possibility of large increases in the severity of drought condi
tions as a consequence of climate change caused by increased CO2.