In most ocean general circulation models the simulated global-scale deep-oc
ean thermohaline properties appear to be chronically colder and fresher tha
n observed. To some extent, this discrepancy has been known to be due to ex
cessive open-ocean deep convection in the Southern Ocean (SO) caused by cru
de "convective adjustment'' parameterizations on scales typically two order
s of magnitude larger than the actual convection scale. To suppress the str
ength of open-ocean convection and to thereby eventually improve the global
deep-ocean water properties, the authors first reduced convection in the S
O in an ad hoc manner by activating it every 10 days rather than every mode
l time step (20 hours). Second, a more physically based subgrid-scale conve
ction in the SO was introduced by applying the penetrative plume convection
scheme of Paluszkiewicz and Romea. With both treatments, SO convection dec
reased by about 30%, and the globally averaged deep-ocean potential tempera
ture and salinity increased substantially to within 0.2 degreesC and 0.02 p
su of observed estimates. Furthermore, the plume convection scheme led to m
ore realistic vertical temperature and salinity sections with more distinct
Circumpolar Deep Water extension toward the south and a significant improv
ement of SO sea ice in terms of its thickness and its seasonality. The resu
lts of this study confirm that in order to obtain more realistic deep-ocean
properties, open-ocean convection in the SO must be substantially weakened
and shallower. This can be achieved by adopting a more physical plume conv
ection scheme.