Ocean Dynamics 
7er 
BE 
or 
L 
535°N 
35° 
Z4°N 
Canuto et al. (2002) 
Canuto et al. (2010) + realizab. 
Fig. 5 Drift of an object over 72 h from 5 December 2013 10 UTC to 8 
December 2013 10 UTC (the starting point is marked by a blue star, the 
end points are marked by blue circles) forced by the two different current 
data sets 
Canuto et al. (2002) 
Canuto et al. (2010) + realizab. 
2:ig. 7 Oil drift simulation results of’ the artificial oil spill after 72 h on 8 
December 2013 at 10 UTC (the position of the oil outlet is marked by a 
’lue star, the black and red crosses represent oil positions) forced by the 
two different current data sets 
„Ar 
6 Conclusion 
5.2 Oil drift 
Also in the case of the artificial oil spill, the results of the 
drift simulation, which was forced by the currents generat- 
ed without explicit realizability and stability checks, differ 
significantly from those of the drift simulation, which was 
forced by currents generated with explicit realizability 
checks. In particular, after a 72-h simulation, the area con- 
taminated by the oil is significantly larger in the case with- 
out realizability and stability checks (Fig. 7). In an emer- 
gency, the lack of accuracy of the drift forecast caused by 
the missing realizability checks could lead to very high 
additional costs, in any case, it means a poorer basis for 
resource planning. 
10 
[N 
87 
—_ —— np —7 - # 
S S S S S S 
y S V S N S 
V a V a V a 
S S S S S 
SS SS S SP F F SF 
S S S S S S S 
Y Y Y Y Y Y Y 
Fig. 6 Distance of the calculated object positions of the two drift 
calculations 
In this study, an extension of realizability and stability criteria 
:o turbulence closure schemes with double diffusion was pre- 
zsented, which were previously only known or at least docu- 
mented for those without double diffusion. The lack of explicit 
cealizability and stability checks led to spatially and temporal- 
ly limited instabilities with nonphysical model (especially cur- 
rent) forecasts during storm or strong wind events, which were 
even found in operational products such as the CMEMS NRT 
product of the Baltic Sea. 
Unfortunately, this problem is not limited to a short- 
term obviously wrong current product. Although the insta- 
»ilities found in the circulation model were only of short 
duration and spatially strongly limited, it could be shown 
hat the instabilities can be of considerable relevance for a 
downstream drift product, since they can significantly re- 
duce the quality of this service. For operational services in 
particular, this is therefore a serious issue to be taken into 
account in the usually very extensive validation proce- 
Jdures. In order to guarantee the prediction quality at a con- 
stant good level, the validation of special individual events 
should always be performed in addition to the statistical 
validation. Special events are not only storms, which are 
accompanied by extreme water levels and strong currents. 
For example, sudden cold or heat bursts have the potential 
(o cause short-term instabilities in the models too. As 
shown in this study, for the validation of special events, 
which are only of short duration, technical validation such 
as the e-tests is suitable in addition to the physical valida- 
jon. In any case, it should by no means be neglected in 
aperational use but should rather increase in scope in the 
course of automation. 
2 ‚pringer