V. Maurer et al.: Evaluation of coupled and uncoupled simulations 
565 
Overall, the evaluation of variability and extreme events 
shows that both NEMO-NBS and ROAM-NBS can generally 
reproduce but underestimate the Major Baltic inflow event, 
chat they are able to represent storm surge events, and cap- 
ture MHWs. 
duction of wind speed compared to ICON-CLM. Future im- 
provements in NEMO-NBS could include a time-dependent 
chlorophyll field that leads to a season-dependent absorp- 
tion of radiation by the ocean. Improved radiative forcing 
at the ocean surface could reduce the SST bias in all sea- 
sons. Further, a calibration of the lateral and vertical diffu- 
sion parameters could enhance the transport over steep ridges 
in the bathymetry and therefore weaken the salinity bias. In 
t'he coupled system, it could be an option to send the ocean 
albedo over water to the atmospheric part, but then an adap- 
tation in the NEMO coupling interface would be necessary. 
The comparison of seasonal mean sea ice concentration 
between the NEMO-NBS and ROAM-NBS simulations and 
observational datasets reveals that both simulations tend to 
overestimate sea ice concentration, particularly during the 
spring season. While the simulations show good agreement 
with observations in winter, they significantly amplify sea 
ice extent in the Gulf of Bothnia during spring. This discrep- 
ancy is likely linked to the cold bias and an underestimated 
salinity in the region. Additionally, the lack of ice dynamics 
in the current model configurations may contribute to these 
inaccuracies. Incorporating dynamic ice processes alongside 
thermodynamic ones and parameter tuning of the thermody- 
namic ice model may improve the models’ performance and 
alignment with observed sea ice behavior. 
The validation of modeled sea surface salinity (SSS) 
against observational data from December 1993 to Novem- 
ber 2020 reveals consistent spatial patterns and systematic 
biases in both ROAM-NBS and NEMO-NBS simulations. 
While surface salinity is generally well captured in the open 
Baltic Sea and certain coastal regions, persistent underes- 
timations are observed near the Norwegian and German 
coasts, as well as within the Baltic Sea. Conversely, SSS is 
overestimated in the transition zone between the Baltic Sea 
and the North Sea. These biases may be attributed to overly 
strong prescribed freshwater runoff, which is not exclusively 
observation-based, and insufficient representation of saline 
inflow from the North Sea. Prescribing runoff throughout the 
water column and enhancing vertical mixing could improve 
sea surface salinity in NEMO-NBS and ROAM-NBS simula- 
tions. For the generation of the historical simulations, an on- 
line coupled runoff model will be used as in Ho-Hagemann 
et al. (2024), which is already available in the setup but 
was not used for better comparability between the coupled 
and the ocean-only simulation. In deeper layers, both models 
consistently underestimate salinity, particularly in the Baltic 
hasins, although the surface layers show good agreement 
with observations. The major inflow events are qualitatively 
reproduced but quantitatively underestimated. Overall, while 
the models effectively capture large-scale salinity patterns 
and seasonal behavior, further refinement of boundary con- 
ditions and freshwater forcing is necessary to improve deep 
water salinity representation and coastal accuracy. 
The comparison of mean temperature and salinity profiles 
from the ROAM-NBS and NEMO-NBS simulations against 
5 Conclusions 
Evaluation results from the ERA5S/ORAS5-driven evaluation 
simulation of the coupled regional ocean-atmosphere model 
ROAM-NBS were presented. ROAM-NBS will be used to 
produce regional climate projections, which will contribute 
to the EURO-CORDEX ensemble. Therefore, ROAM-NBS 
and the simulations with the individual stand-alone versions 
of the ocean (NEMO-NBS) and the atmosphere (ICON- 
CLM) were assessed with respect to different observations 
and reanalyses. NEMO-NBS as well as ROAM-NBS exhibit 
a small SST bias, which is on area-average about +0.5K. 
For individual seasons and regions, it can also reach larger 
values. Especially over the Atlantic Ocean, a cold bias pre- 
vails for all seasons except summer. In the Baltic Sea, a cold 
Dias prevails. The SST bias is slightly increased in summer 
in ROAM-NBS compared to NEMO-NBS by about 0.3 to 
0.4K. For both ROAM-NBS and NEMO-NBS, there is no 
increase in the bias with time throughout the evaluated pe- 
:iod of 1979-2020. Warming trends in the North and Baltic 
Sea are well reproduced by the simulations. 
The surface temperature difference against ERAS5 exhibits 
clearly larger values over land than over the ocean. The near- 
surface air temperature bias over land is overall negative for 
both ICON-CLM and ROAM-NBS, with a small overesti- 
nation of the diurnal minimum temperature and a more pro- 
nounced underestimation of the diurnal maximum temper- 
ature in all seasons. The temporal evolution of mean tem- 
peratures over land is generally in agreement with observa- 
tional data and reanalyses, with the largest cold bias in Spain 
ınd Portugal especially after 1995 and a warm bias in the 
drier, more continental region of south-east Europe. Differ- 
ences between ROAM-NBS and ICON-CLM are very small 
in all land regions and years. Over the ocean, the SST differ- 
ences between ROAM-NBS and ICON-CLM reflect the bias 
of NEMO-NBS. The sign of the SST difference coincides 
with the sign of the differences for sensible and latent heat 
Aux, wind speed, and precipitation, 1.e. in regions and sea- 
sons where the SST is higher in ROAM-NBS than in ICON- 
CLM, also the heat fluxes, wind speed, and precipitation are 
higher, and vice versa. This relationship means that the SST 
5ias introduced into the coupled system by NEMO-NBS in- 
fluences the atmospheric fields over water, but as shown be- 
fore, this does not have a systematic influence on the land ar- 
eas. Compared to station observations, wind speed over the 
ocean is underestimated, which is slightly more pronounced 
in ROAM-NBS than in ICON-CLM due to the SST cold bias 
along the German coast in ROAM-NBS, which causes a re- 
https:/doi.o0rg/10.5194/smd-19-543-20246 
Geosci. Model Dev... 19. 543-578. 2026