El 
, 
V. Maurer et al.: Evaluation of coupled and uncoupled simulations 
dalone ocean models are ideally forced by the output of 
‚egional atmospheric models, the ocean simulations can 
only be delivered with a considerable delay compared 
to the global climate simulations due to the downscaling 
chain. Thus, one advantage of using a regional coupled 
ocean-atmosphere model for climate projections, compared 
co the stand-alone ocean component, is the independence 
of regional climate model (RCM) projections from CMIP6. 
Moreover, the regional coupled model allows us to deliver 
consistent information on climate and climate change for the 
atmosphere and the ocean in the North and Baltic Sea (NBS) 
‚egion, with a particular focus on the German coasts. For the 
Baltic Sea region, a number of investigations using regional 
coupled models were conducted within the framework of 
Baltic Earth (https://baltic.earth, last access: 9 July 2025). 
Gröger et al. (2021) review progress on coupled modeling 
in that context. The investigations considered in the review 
outline different aspects of the added value of regional 
coupled models. Gröger et al. (2021) summarize that only 
online coupled high-resolution ocean models can represent 
small-scale ocean processes accurately. They also conclude 
‘hat the demonstration of the added value of coupled models 
over their uncoupled counterparts is often influenced by 
biases in datasets, such as runoff, used for the forcing of 
the uncoupled versions. Christensen et al. (2022) analyzed 
RCM projections with and without ocean coupling, forced 
by global climate model (GCM) simulations provided 
with the fifth phase of the coupled model intercomparison 
project (CMIP5S). Their focus was on climate change in the 
Baltic Sea region. They showed that the coupled simulations 
can exhibit differences in future sea surface temperatures 
and sea ice conditions compared to the respective uncoupled 
versions, which can locally modify the climate change 
signal. 
As shown by Gröger et al. (2021), different regional cou- 
pled ocean-atmosphere models have been in use for the NBS 
region. These are coupled versions of CCLM and NEMO 
(e.g. Pham et al., 2014; Primo et al., 2019; Ho-Hagemann 
et al., 2020), RCA4 and NEMO (Gröger et al., 2015; Di- 
eterich et al., 2019), REMO and MPIOM (Sein et al., 2015), 
or Hirham and HBM (Tian et al., 2013). Karsten et al. (2024) 
present a recent development of a coupled ocean-atmosphere 
model, which couples the atmosphere and the ocean com- 
ponent (CCLM and MOM5, respectively) via an exchange 
grid. Bauer et al. (2021) were coupling ICON and GETM 
via an ESMF exchange grid. However, the ocean domains 
of their coupled models only encompass the Baltic Sea, or 
an even smaller domain in the case of Bauer et al. (2021), 
which is merely a small part of the whole EURO-CORDEX 
domain used for the atmosphere. A first version of a cou- 
pled regional ocean-atmosphere model incorporating ICON 
in climate limited-area mode (ICON-CLM) and NEMO was 
presented by Ho-Hagemann et al. (2024). The modeling sys- 
tem is called GCOAS’T-AHOL, just as its earlier version (Ho- 
Hagemann et al... 2020). Ho-Hagemann et al. (2024) found 
that the new version of GCOAST-AHOI could well capture 
near-surface air temperature, precipitation, mean sea level 
pressure, and wind speed at a height of 10 m. However, there 
was a prevailing negative sea surface temperature (SST) bias 
of 1-2 K, which they attributed to an underestimation of the 
downward shortwave radiation at the surface. 
Here, we introduce ROAM-NBS, a new version of a re- 
gional coupled ocean-ice-atmosphere modeling system cov- 
ering the full EURO-CORDEX domain for the atmosphere 
and the North and Baltic Sea for the ocean. ROAM-NBS 
combines the ICON-CLM atmosphere model (version icon- 
2024.07) with the NEMOv4.2.0 ocean model and the Sea 
Ice modelling Integrated Initiative (SI3) thermodynamic sea 
ice model, coupled via OASIS3-MCT using a flux-based ex- 
change approach. Compared to the version by Ho-Hagemann 
et al. (2024), ROAM-NBS is based on a later NEMO version 
and includes a new ocean bathymetry, which is specifically 
designed for a good representation of the German coastline. 
Moreover, our setup also integrates a refined treatment of 
radiation in NEMO based on prescribed chlorophyll distri- 
butions, supporting a realistic representation of shallow and 
stratified shelf seas. With the use of a later ICON release, 
a more recent NEMO version with major updates, higher- 
resolution coastal bathymetry, and enhanced representation 
of surface fluxes and radiation, ROAM-NBS represents a 
methodological advance over previous GCOAST configura- 
tions. 
Using ROAM-NBS and different configurations of 
GCOAST-AHOI described by Ho-Hagemann et al. 
(2020, 2024), which all employ an online coupled ocean 
for the NBS region, CMIP6 climate projections will 
be downscaled for the EURO-CORDEX region (Jacob 
et al., 2014). These coupled regional climate projections 
will complement the RCM simulations of CORDEX- 
CMIP6 (https://github.com/WCRP-CORDEX; last access: 
23 May 2025). The simulation status is updated regu- 
larly and can be viewed at https://werp-cordex.github.i0/ 
simulation-status/CORDEX_CMIP6_status.html#EUR-12 
(last access: 23 May 2025). The evaluation simulation 
analyzed in this article defines the setup of ROAM-NBS 
that will be used for downscaling. In addition to publishing 
the data on the ESGF nodes within the EURO-CORDEX 
community, ROAM-NBS will be applied to generate an 
ensemble of climate projections that can be used for climate 
adaptation measures in German national waters. Related 
evaluations will be published on https://das.bsh.de (last 
access: 9 July 2025). 
The comparison of coupled simulations against their un- 
coupled counterparts, which are forced by high-quality re- 
analyses like ERA5S (Hersbach et al., 2020) at the ocean- 
atmosphere interface, can never be a fair one. Thus, the most 
important added value of using a regional coupled model 
for climate projections is not shown when evaluating the 
reanalyses-driven evaluation simulation, where we can pro- 
vide good forcing data for both components for uncoupled 
Geosci. Model Dev... 19. 543578, 2026 
https://doi.ore/10.5194/smd-19-543-2026