The 
O/
O ratio of atmospheric O
is higher than that of
average seawater H
O by 23.5
[ Kroopnick and Craig, 1972].
This observation was first made independently by Dole [1935] and
Morita [1935] and has become known as the Dole effect. It is caused
mainly by discrimination against 
O during respiration, as was
realized in early investigations [ Lane and Dole, 1956]. By this
reasoning, photosynthesis produces O
from H
O with the same
O/
O ratio as the H
O, while respiration preferentially
removes 
O
from the air. A steady-state balance is achieved
when the 
O/
O ratio of atmospheric O
is enriched relative
to photosynthetic O
by the discrimination factor associated with
respiration.
It is now recognized that additional processes also contribute to the Dole
effect. While careful investigations have confirmed that photosynthesis
produces O
without fractionation [ Stevens et al., 1975; Guy
et al., 1993], the isotopic composition of photosynthetic water can vary,
and these variations will be passed on to the O
produced by
photosynthesis. On average, this leads to an increased Dole effect
because of evaporative enrichment of 
O in leaf water [ Dongmann,
1974]. Additional processes which influence the Dole effect are the
equilibrium fractionation of 
O/
O between dissolved and
gaseous O
, which is relevant because O
consumed by respiration is
derived from dissolved O
, and photochemical processes in the
stratosphere which lead to a slight decrease in 
O of O
through exchanges of oxygen atoms between O
and CO
[ Bender et
al., 1994a].
Estimating the effective average fractionation factor for global
respiration is complicated because O
consumption can occur via several
distinct biochemical pathways [ Guy et al., 1989; Guy et al.,
1993; Bender et al., 1994a] including the light reactions, such as
the Mehler reactions and photorespiration reactions, and the dark
reactions, such as the cytochrome pathway and the alternative
cyanide-resistant pathway. To compute the global respiratory contribution
to the Dole effect it is necessary to know the fractionation factors and
the relative O
consumption for each pathway at the global scale [
Berry, 1992; Bender et al., 1994a].
Respiration in the deep sea requires special consideration because here
respiratory O
utilization depletes a significant fraction of the O
originally present in the water. If total depletion occurred, then the
effective fractionation for respiration in the deep sea would be zero
because the 
O/
O ratio of the removed O
would be equal to
the 
O/
O ratio of the O
originally dissolved in the
water. In the case where O
is only partially depleted, the effective
respiratory fractionation factor can be calculated based on the percentage
O
depletion that actually occurs [ Bender et al., 1994a].
A recent budget of the global contributions to the Dole effect by
Bender et al. [1994a] is presented in Figure 3.
This budget adopts the value of 4.4
[ Farquhar et al., 1993]
for the average enrichment of terrestrial chloroplast water relative to
SMOW. The budget takes account of respiratory fractionation using
fractionation factors from Guy et al. [1989], Guy et al.
[1993], Kiddon et al. [1993], and Bender [1990], and using
estimates of the global O
uptake on land and in the ocean from
Farquhar et al. [1980], Guy et al. [1993], and Keeling and
Shertz [1992].
Interestingly, this budget yields an estimate for the global Dole effect
of 20.8
which is significantly smaller than the observed value of
23.5
. The difference may either reflect errors in the values adopted
or unknown additional processes. A possible problem is the value of
4.4
adopted from Farquhar et al. [1993] for average chloroplast
water. A value of 8.7
would bring the budget into balance, and
Bender et al. [1994a], argue that a higher value is plausible given the
uncertainties involved. In this case, however, the Farquhar et al.
[1993] budget for 
O/
O of CO
would be out of balance.
One possible way of reconciling both the CO
and O
isotope budgets
might be by increasing the 
O of chloroplast water and
decreasing the 
O of CO
leaving soils relative to the
Farquhar et al. [1993] budget (M. Bender, personal communication). Some
additional flexibility may be provided by the fact that the O
and
CO
budgets depend on different weighted averages of chloroplast water.
For the O
budget, the average needs to be weighted by GPP plus
photorespiration, while for CO
the average needs to be weighted by the
flux of CO
out of stomata, which is equal to the gross flux of CO
into stomata minus GPP. In any case more work is needed to construct
mutually consistent budgets for 
O in both atmospheric O
and CO
.
Bender et al. [1994a] estimate that the Dole effect which would
result from exchanges with the oceans alone is around 2 to 3
lower
than that which would result from terrestrial exchanges alone (see
Figure 3). This difference would be even larger if a 
O value
higher than 4.4
is adopted for globally averaged chloroplast water.
Either way, the overall magnitude of the Dole effect is sensitive to the
ratio of gross primary production on land to gross primary production in
the oceans. This suggests that measurements of the Dole effect and its
variation over time may be used to constrain relative variations in
terrestrial and marine productivities [ Bender et al., 1994a]. To
succeed, this application requires accounting for changes in the isotopic
enrichment in leaf water and any other influences on the Dole effect using
independent methods.
Variations in atmospheric 
O/
O of O
over the past 130
thousand years have been reconstructed from measurements on ancient air
samples extracted from bubbles in polar glaciers [ Bender et al.,
1985; Sowers et al., 1991; Bender et al., 1994c]. The
O/
O ratio of atmospheric O
has closely followed the
O/
O ratio of surface seawater as established from sediment
records [ Shackleton and Pisias, 1985], which in turn has varied due
to the expansion and contraction of the continental ice sheets. The Dole
effect, i.e., difference in 
O/
O ratio between atmospheric
O
and surface seawater, has been constant to around 
0.5
over
this period, with possible small cyclic variations with a period of 23
thousand years corresponding to the precession period of the earth's
orbital axis [ Bender et al., 1994a]. The high degree of constancy
can probably be explained only if some of the factors controlling the Dole
effect changed in ways that compensated for each other. This could occur,
for example, if reductions in terrestrial productivity during glacial
conditions were accompanied by reductions in marine productivity [
Bender et al., 1994a]. The similarity in the patterns of 
O
variations in ice core O
and sediment records has made it possible to
establish more firmly the age of the air extracted from ice cores relative
to the sediment chronologies [ Sowers et al., 1991].
Variations in 
O of atmospheric O
must lag behind
variations of 
O in surface seawater by the turnover time of
atmospheric O
with respect to gross photosynthesis and respiration.
If the sediment and ice core chronologies were improved sufficiently, this
turnover time, currently estimated at 1500 years, could be directly
determined [ Bender et al., 1985; Bender et al., 1994a].
How variable is 
O of atmospheric O
on shorter time
scales? Temporal and spatial surveys [ Dole et al., 1954;
Kroopnick and Craig, 1972] showed that 
O in the present
atmosphere is constant to at least 0.25
. More recent tropospheric
measurements indicate that 
O is constant to least 0.03
(M. Theimens, personal communication). Known sources of variability are
expected produce changes about an a order of magnitude smaller than this.
For example, we can expect 
O to be lower in summer than in
winter in both northern and southern hemispheres by about 0.002
. This
estimate is based on assuming that the 0.01% seasonal increase in the
atmospheric O
/N
ratio (see next section) is driven by the input of
photosynthetic O
that is 20
lower in 
O than atmospheric
O
. Seasonal variability might also be caused by seasonal phase
differences in gross photosynthesis in the oceans or on land, or by
seasonality in leaf water 
O. Detecting such small changes
may eventually be feasible with very precise mass spectrometric
measurements.
In summary, our knowledge of variations in 
O of atmospheric
O
is limited to variations over recent glacial cycles and these
variations are largely consistent with a constant Dole effect over this
period. The Dole effect places constraints on the globally averaged
composition of metabolic water which, in turn, constrains the relative
magnitude of gross photosynthesis on land and in the ocean.