V. Maurer et al.: Evaluation of coupled and uncoupled simulations

85:

(a)

(b)

nm TO
mean ROAM-NBS
mean obs
mean ICON-CLM
rmse ROAM-NBS
mean ICON-CLM

40

nA

?0-KM!



AS

A
$
S$

&
Na
Fe

KK

PS

N

Figure 8. Map with an overview of the evaluated station observations (a); mean station observations of hourly wind speed for 2011-2020,
mean values and root-mean squared errors (rmse) of 9 closest grid points for each station for ROAM-NBS and ICON-CLM (b); violinplot
displaying median, minimum and maximum values of hourly wind data at 100 m height at FINO1 and at the closest grid points in ROAM-NBS
 and ICON-CLM. for the vears 2004-2010 (6).

Fehmarn, which is a larger island compared to those in the
North Sea, the mean wind speed is well represented by both
ROAM-NBS and ICON-CLM, and the RMSE values are also
small compared to other stations.

3.3 Oceanographic conditions

3.3.1 Sea ice concentration and extent

The bias of ROAM-NBS and NEMO-NBS against the
Copernicus observations for the winter season (DJF) is
small (Fig. 9a), and the difference between ROAM-NBS
and NEMO-NBS is negligible (Fig. 9b). Nevertheless, in the
spring season (MAM), the amplification of sea ice extent is
prominent in the Gulf of Bothnia. While the start of the ice
growth season and the growth rate in the end of December
are well captured, the sea ice extent is overestimated during
che peak later in winter (February-March) by both simulations
 (mean annual time series of sea ice extent not shown).
Furthermore, the sea ice season is prolonged towards May

https:/doi.o0rg/10.5194/smd-19-543-2026

compared to the observations. The overestimation of sea ice
concentration as well as extent in MAM might originate from
the prevailing negative bias in SST in the Bothnian Bay and
Sea during winter and spring (Fig. 2 and the temperature profile
 for DJF at SMHISR5C4 station in Fig. 10) and an underestimation
 of salinity, especially in the Gulf of Bothnia
(see the next subsection). Further, in Nie et al. (2023), the
NEMO-SI3 model for version 4.0 was found to be most sensitive
 to the ice-ocean and atmosphere—-ice drag coefficients
and, therefore, to ice dynamics. Since only thermodynamical
 processes are modeled within ROAM-NBS and NEMO-NBS,
 further enhancements in results could be obtained by
also considering ice dynamics. First tests with ice dynamics
did not improve the overestimation of sea ice in spring. They
will need to be carried out in the future in combination with
further parameter tuning and recalibration to eliminate the
temperature and salinity bias in the Northern Baltic.

Geosci. Model Derv.. 19. 543-578, 2026