Data from Columbia Glacier are used to identify processes that control
calving from a temperate tidewater glacier and to re-evaluate models
that have been proposed to describe iceberg calving. Since 1981, Colum
bia Glacier has been retreating rapidly, with an almost seven-fold inc
rease in calving rate fr om the mid-1970s to 1993. At the same time, t
he speed of the glacier increased almost as much, so that the actual r
ate of retreat increased more slowly. According to the commonly accept
ed model, the calving rate is linearly related to the water depth at t
he terminus, with retreat of the glacier snout into deeper water, lead
ing to larger calving rates and accelerated retreat. The Columbia Glac
ier data show that the calving rate is not simply linked to observed q
uantities such as water depth or stretching rate near the terminus. Du
ring the retreat, the thickness at the terminus appears to be linearly
correlated with the water depth; at the terminus, the thickness in ex
cess of flotation remained at about 50 m. This suggests that retreat m
ay be initiated when the terminal thickness becomes too small, with th
e rate of retreat controlled by the rate at which the snout is thinnin
g and by the basal slope. The implication is that the rapid retreat of
Columbia Glacier (and other comparable tidewater glaciers) is not the
result of an increase in calving as the glacier retreated into deeper
water. Instead, the retreat was initiated and maintained by thinning
of the glacier. For Columbia Glacier, the continued thinning is probab
ly associated with the increase in glacier speed and retreat may be ex
pected to continue as long as these large speeds are maintained. It is
not clear what mechanism may be responsible for the speed-up but the
most likely candidate is a change in basal conditions or subglacial dr
ainage. Consequently, the behavior of tidewater glaciers may be contro
lled by processes acting al the glacier bed rather than by what happen
s at the glacier terminus.