Next: PALACE Floats
Up: Density Structure
Previous: WOCE Hydrographic Cross-sections
The Bravo mooring at (
) kept temperature and
salinity measurements throughout the winter of 1996-97. The model produces a
slow deepening of the mixed layer and a rapid restratification in spring as
seen in observation (see Fig. 12). The timing of the deepest convection,
mid-March, is consistent within 10 days between the model and real
observations. The depth of convection is almost 1500 meters in the data but
approximately 1000 meters in the model. Further investigation of the model
shows that convection as deep as 1500 meters occurred in the region directly
west of Bravo, but not at that site. Also, the surface layers of the model
react to the atmospheric forcing too quickly and strongly. A pool of fresh,
cold water is present during the onset of convection. A closer look at the
salinity restoring field indicates that virtual salt flux is freshening the
ocean at this time. In fact, the spring restoring field is very fresh because
it is weighted by a summer field with salinity of only 34.2 at Bravo. Even with
the very fresh surface properties, this patch of water is statically unstable
for 20 days at the height of winter. Although such a cold, fresh patch of water
is not observed in data, it has no effect on the continuing increase of mixed
layer depth. Previous model runs which completely mixed unstable profiles
instantaneously always produced homogeneous patches of convected water
(C. Herbaut, personal communication). The convective scheme in this model
partially mixes statically unstable profiles every 30 minutes. For the unstable
profile found a Bravo in mid-March, the convective scheme would take
approximately 120 iterations (2.5 days) to remove the instability in the
absence of other forces. The convective scheme can lower the density of surface
water by a maximum of
every 30 minutes. However, a heat loss
of
directly taken from the upper 15 meters would raise density
by roughly
every 30 minutes. In the absence of strong vertical
diffusivity in the model, supercooled surface layers of only 15-30 meters deep
are present. Observed convection homogenizes on timescales of 12-24 hours, so
an increased convective adjustment frequency is suggested. Secondly, increased
vertical diffusivity is recommended for the immediate surface layers. By the
end of March, the model water column is homogeneous to 1000 meter depth,
although this process takes much longer than observed. Use of a more frequent
convective adjustment is advised in order to homogenize this upper layer. By
late spring, the model's surface temperature increases rapidly in the upper 50
meters. The temperature is over 6 degrees Celsius while only 3.5 degrees was
observed. These problems do not extend to any great depth and are a result of
the mixed layer physics again. The KPP model should be able to deepen the
seasonal mixed layer in spring and summer by parameterizing the effect of wind
stirring. On the other hand, KPP would probably not help homogenize the water
column during winter, because it has a very crude convective scheme. In the
case of a multi-year model run, the compacted warm layer after restratification
may affect convection in subsequent years, but for our purposes here the timing
and nature of convection is consistent with data .
Next: PALACE Floats
Up: Density Structure
Previous: WOCE Hydrographic Cross-sections
Jake Gebbie
2003-04-10