Deutsche Hydrographische Zeitschrift - German Journal of Hydrography 
466 
culation model "BSHcmod". The model output in 
cludes currents, temperature, salinity, and sea level 
data. The meteorological input is generated by an 
atmospheric model of the DWD in Offenbach. The 
accuracy of the meteorological input, especially the 
local wind field which acts as an external force onto 
the sea surface, is essential for the currents within 
the surface model layer. The quality of the wind in 
put can be estimated by comparing it to local DWD 
measurements recorded on board UFS DB. The val 
idation of the model results is essential for several 
BSH tasks, e.g. water level prediction and storm 
surge warnings. As the results of BSHcmod consti 
tute basic data which are used to drive Lagrangian 
and Eulerian drift and dispersion models, a realistic 
computation of currents is of vital importance. 
2 Instrument setups 
The ADCP was a 300 kHz BroadBand Work 
horse Sentinel manufactured by RD Instruments. It 
was mounted in a bottom frame with the upward 
looking transducer 0.5 mab. The ADCP divided the 
water column into equally spaced depth cells (bins) 
whose length was 2 m for this application. The sam 
pling interval (ensemble time) was 15 minutes. Us 
ing 45 pings per ensemble the standard deviation 
amounts to 0.9 cm/s. With a beam angle of 20°, the 
ADCP measurements of the upper 6 % (about 3 m) 
of the water column are contaminated due to side 
lobe effects (Gordon [1996]). ADCP current data 
are means over the whole depth range of each bin. 
The Aanderaa water level recorder (WLR7) 
was attached to the ADCP bottom frame. It had a 
sampling rate of 10 minutes. The sensor is based on 
a pressure controlled oscillator with an integration 
time of 40 seconds and an accuracy of about 1 cm. 
After recovery the pressure data were corrected for 
atmospheric pressure variations by means of the 
meteorological data recorded at UFS DB. 
3 Model data 
The BSH circulation model (BSHcmod) is a 
three-dimensional baroclinic numerical model which 
in nightly routine runs produces forecasts up to 
48 hours ahead (Kleine [1994]). Currents, water 
levels, water temperatures, salinity, and ice cover in 
the North and Baltic Sea are computed on two nest 
ed and interactively coupled grids. The new model 
version (Dick et al. [in prep.]) which is in operation 
since January 1,1999, was improved with respect to 
horizontal and vertical resolution. Horizontal grid 
spacing in the German Bight and western Baltic Sea 
is 1.8 km, and 10 km in the other North and Baltic 
Sea areas. The model also simulates the falling dry 
and flooding of tidal flats, allowing complex process 
es in the highly structured coastal waters (tidal flats, 
sandbanks, tidal channels, barrier islands) and wa 
ter exchange with the open sea to be simulated re 
alistically. 
The model is driven by meteorological fore 
casts of the DWD’s atmospheric models, by tides 
and external surges entering the North Sea from the 
Atlantic as well as by river runoff from the major riv 
ers (Fig. 1). Heat fluxes between air and water are 
computed in the BSH model using air temperature, 
cloudiness, and specific humidity above the sea 
(Müller-Navarra and Ladwig [1997]). The tidal 
predictions at the model’s open boundaries are cal 
culated from the harmonic constants of 14 tidal con 
stituents. External surges entering the North Sea 
are computed by a two-dimensional hydrodynamic 
model of the Northeast Atlantic. 
The circulation model BSHcmod simulates 
density driven (baroclinic) currents which depend on 
the prevailing temperature and salinity distributions. 
As hydrodynamics are also influenced by ice condi 
tions in the North Sea and Baltic, the circulation 
model is additionally coupled to an ice model simu 
lating the formation, melting, and drift of sea ice. 
The time step of the model is 1.5 minutes. How 
ever, water level and current forecasts are only 
stored with a temporal resolution of 15 minutes. All 
other model results are stored with a temporal reso 
lution of 60 minutes. 
Table 1 shows the vertical spacing of the model 
layers at UFS DB and the location of the ADCP bins 
which are used for the comparison. The model out 
put is a vertical mean over the respective layer. For 
the comparison we selected those bins which were