Deutsche Hydrographische Zeitschrift - German Journal of Hydrography 
474 
In a next step we determined the high (HW) and 
low water (LW) times for both curves and compared 
the values. At the WLR the HWs appeared 
16.3± 10.7 minutes earlier than the model HWs, 
and the LWs appeared 8.3 ± 10.8 minutes earlier. 
This difference may be partly due to the different 
sampling intervals (10 and 15 Minutes). However, 
the delay of the model is significant. 
7 Discussion 
It has been demonstrated that the BSH model 
represents tidal conditions quite well. Tidal analysis 
of currents at different depths showed good agree 
ment in direction and speed. Differences were found 
with respect to some higher harmonics (M 4 , MS 4 , 
M 6 , 2MS 6 ). This is in accordance with the compari 
son of water level data. The differences for high and 
low water times between both data sets may also be 
caused by a spurious representation of higher har 
monics. As higher harmonics are mostly generated 
within the model area by non-linear effects, the error 
could probably be reduced by improving the repre 
sentation of topography or bottom friction. However, 
the low standard deviation of water level differences 
(± 0.08 m), reflecting not only tidal but also surge ef 
fects, shows that the model’s tidal error at UFS DB 
is very small. 
Looking at the representation of residual cur 
rents, the current structure is found to agree in gen 
eral. Differences in the surface layer are mainly 
caused by missing ADCP data near the surface 
(side-lobes). In general, current magnitude, vector 
speed, and kinetic energy agree quite well. How 
ever, some differences are observed in the current 
direction in the layer between 20 and 30 m. While 
measurements show a sharp shear at about 21 m 
depth, the gradient in the model predictions is much 
smoother. A reason for this is the coarse vertical 
grid spacing of 4 respectively 6 m which prevents 
the formation of sharp gradients. 
On the other hand, density distribution also 
affects the baroclinic residual currents and their ver 
tical structure. As a wrong representation of the 
density structure would cause wrong residual cur 
rents, the salinity and temperature profiles mea 
sured at UFS DB will be compared with model re 
sults in the following. Figure 8 shows the temporal 
evolution of measured and computed temperature 
profiles at UFS DB for the period of current mea 
surements. Both figures demonstrate the warming 
of the whole water column in summer and show a 
succession of periods with stratification and mixing. 
In general, model temperatures are somewhat 
above the measured values. The mean deviation is 
approximately 0.6 °C at the surface and in the bot 
tom layer and 0.8 °C in the middle of the water 
column. Larger differences at mid-depths are attrib 
utable to a weaker stratification in the model than in 
nature. As has been mentioned above, the model is 
not capable of simulating sharp vertical gradients. 
However, as temperature stratification and current 
shear occur at different water depths, there was no 
direct link between both phenomena. Looking at the 
salinity profiles, there was only weak haline stratifi 
cation in the summer of 1999. The difference 
between surface and bottom salinities was 0.3 in 
nature and 0.1 in the model, with a mean deviation 
of 0.5 between model data and measurements. 
Therefore, the density field at UFS DB - and hence 
the baroclinic currents - was influenced much more 
by temperature than by salinity. 
We may conclude that in summer 1999 the op 
erational circulation model of the BSH in principle 
predicted a realistic description of hydrodynamics at 
UFS DB. Earlier comparisons in the close vicinity of 
Helgoland exhibited significant deviations between 
ADCP and model currents resulting from strong to 
pographical gradients which could not be resolved 
by the model grid. At UFS DB, which is located in an 
area without significant topographical gradients, dif 
ferences between ADCP and model data were 
found at depths where strong current shear or 
strong stratification occurred. One reason for this is 
that gradients cannot be simulated with a vertical 
resolution of more then 4 metres. Another possible 
cause could be an error in the parameterization of 
vertical eddy diffusion. This will be investigated in a 
future experiment.