V. Maurer et al.: Evaluation of coupled and uncoupled simulations 
559 
bottom salinity in deeper Baltic basins can be observed. This 
ınderestimation could be mitigated in future simulations by 
employing longer spin-up periods or improved initial condi- 
:jons. Additionally, efforts should focus on refining vertical 
and horizontal mixing within the Baltic and increasing spa- 
jal resolution in the Danish Straits. For climate projections 
based on the current setup, bias correction of bottom salinity 
values in basins deeper than 40 m is necessary. 
Table 3. Statistical values of detided sea surface height model 
results vs. GESLAv3.0 observational data evaluated within Jan- 
vary 2015—-December 2019. 
Station — _— ROAM-NBS NEMO-NBS 
ımsd nn Fr ımsd nn 
Alteweser 0.93 0.11 
Bergen 0.86 0.08 67 
Borkum 0.93 0.10 85 
Bremerhaven 0.94 0.11 5 
Cuxhaven 0.94 0.12 86 
Eidersperrwerk 0.94 0.13 +7 
Emden 0.94 0.10 5 
Frederikshavn 0.91 0.09 
Helgeroa 0.90 0.08 
Helgoland 0.94 0.11 
Husum 0.95 0.12 
Landsortnorra 0.97 0.05 
Leixoes 0.50 0.09 
List 0.95 0.11 
Newlyn 0.67 0.11 
Plymouth 0.79 0.08 
Rorvik 0.90 0.0& 
Stavanger 0.855 0.08 
5tpetersburg 0.95 0.09 
Travemünde 0.92 0.07 
Fregde 0.82 0.09 
Vigo 0.61 0.09 
Viker (0.91 0.08 
0.96 0.07 9% 
0.837 0.08 69 
0.96 0.07 92 
0.97 0.07 95 
0.97 0.07 54 
0.97 0.08 94 
0.96 0.07 9% 
0.92 0.07 80 
0.91 0.07 ®%) 
0.97 0.07 93 
0.98 0.08 95 
0.97 0.04 
0.48 0.10 77 
0.97 0.07 
0.66 0.12 
0.78 0.08 
0.91 0.08 81 
0.86 0.08 €) 
0.97 0.07 93 
0.94 0.06 89 
0.833 0.08 61 
0.61 0.09 36 
093 007 86 
3.3.4 Sea surface heights 
To assess the accuracy of sea surface heights, model results 
are compared against GESLAv3.0 observational data within 
the evaluation period of January 2015 to December 2019. A 
period was selected during which all observational stations 
provided continuous hourly data, ensuring that the resulting 
statistics are fully comparable. 
Here, the sea surface height results are bias adjusted by 
subtracting the mean of SSH within the evaluation period 
from SSH results. The storm surge component of the wa- 
(er level is assessed by removing the influence of astronom- 
ical tides from the bias-adjusted sea surface height. This is 
achieved using a Demerliac filter, which separates tidal fluc- 
uations from the total signal. The remaining residual cap- 
tures the non-tidal water level variations caused by meteo- 
rological forcing, providing an estimate of the storm surge. 
Scatter plots for the detided sea surface height (represent- 
ıng the storm surge) are displayed in Fig. 13 for a station 
in the German Bight (Cuxhaven, see Fig. 8), and the Baltic 
(Landsortnorra). Statistical values such as the number of data 
points N, the correlation coefficient r, the root mean square 
deviation rmsd, and the explained variance n are calculated 
and summarized in Table 3 for the detided sea surface heights 
‚representing storm surge) at further stations covering the 
model domain. 
For detided sea surface heights, NEMO-NBS shows a 
higher correlation with the observational data than ROAM- 
NBS at nearly all stations, with minimal differences in cor- 
relation coefficients. The root mean square deviation of 
NEMO-NBS is smaller for the majority of the stations. As 
can be observed in the scatter plots, for detided sea surface 
height extreme values, especially maxima, ROAM-NBS per- 
forms better than NEMO-NBS in the German Bight, while it 
slightly overestimates maxima at Landsortnorra. The higher 
maxima at all stations obtained with the coupled ROAM- 
NBS simulation indicate a positive effect of the atmosphere- 
ocean coupling on wind speed quality. NEMO-NBS’s per- 
formance for sea surface height maxima could be improved 
5y calibrating the wind drag coefficient. Single storm events, 
which cannot be well represented spatially or temporally, 
i(ead to single outliers in the coupled model. These outliers 
are not present in the reanalysis driven NEMO-NBS results. 
Overall, the correlation of storm surge of both simulations 
with observational data is high, especially in German na- 
tional waters. At stations closer to the domain boundaries, 
5 
55 
26 
1 
storm surge correlations could be improved. Nevertheless, 
overall bias-adjusted sea surface height results strongly cor- 
relate with observational data (see Fig. A7) for the coupled 
and uncoupled models due to a high tidal component. 
4 
Variability and extreme events 
In this section, the model results of ROAM-NBS and NEMO- 
NBS are evaluated for exemplary extreme events against ob- 
servational and reanalysis data. As an adequate representa- 
tion of the inflow from the North Sea into the Baltic Sea is 
important for correctly modeling the state of the Baltic Sea 
over longer timescales, a particular focus was placed on the 
analysis of this inflow. In Sect. 4.1 and 4.2, cross-sections 
of salinity and temperature through the Baltic and sea sur- 
face heights at representative stations are evaluated around 
the Major Baltic Inflow (MBIT) in December 2014, which was 
the third-largest recorded inflow. Its influences are described 
in Mohrholz et al. (2015). In Sect. 4.3, the model simula- 
tion’s capacity to track observed marine heatwaves (MHWs) 
is examined. 
https:/doi.o0rg/10.5194/smd-19-543-20246 
Geosci. Model Dev... 19. 543-578, 2026