Evaluation results from the reanalysis-driven evaluation simulation for the years 1979–2021 with a regional coupled ocean–atmosphere model (ROAM) are presented. The coupled setup portrayed here is one of the first regional climate modeling systems to couple the ICON atmosphere model in climate limited-area mode (CLM) with the ocean model NEMO for the North and Baltic Sea (NBS), using a flux-based OASIS coupling approach. Along with the simulation with the coupled model configuration ROAM-NBS, the simulations with the uncoupled components (ICON-CLM and NEMO-NBS, respectively) are analyzed and compared with various observational datasets. ROAM-NBS complements atmosphere-only climate projections with the same atmospheric model and setup, which will all be published in accordance with EURO-CORDEX specifications. Climate projections by ROAM-NBS will enrich the data available to support the German Strategy for Adaptation to Climate Change (DAS), especially for our target region, which are the German national waters.</p> <p>In general, the mean model climate is well represented by all setups. The sea surface temperature bias is, on average, below ±0.5 K, with an improved representation of extreme sea surface height events in the uncoupled setup. SST biases of ROAM-NBS directly lead to biases of surface heat fluxes, wind speed, and precipitation over land. However, the mean influence on the land areas is negligible. The evaluations of ocean variables indicate a strong agreement of ROAM-NBS with NEMO-NBS. Compared to observations, both simulations overestimate sea ice concentration and extent in spring, especially in the Gulf of Bothnia. Mean temperature profiles in the Baltic Sea show that both simulations generally reproduce observed profiles. ROAM-NBS exhibits a cold bias in deeper layers, especially in the Gotland Deep, while NEMO-NBS shows better agreement at the sea surface and bottom. Stratification analyses confirm a fresh bias in Baltic bottom layers, especially in the Gulf of Bothnia, with NEMO-NBS performing slightly better in this region. Major inflow events are captured but underestimated. Sea surface height and wind surge highly coincide with observational data, with NEMO-NBS slightly outperforming ROAM-NBS in correlation. The marine heat wave (MHW) evaluation against observations in the North and Baltic Sea demonstrates that the simulations capture the inter-annual variability of MHW characteristics.</p> <p>Overall, the coupled simulation demonstrates good performance for both the atmosphere and the ocean, and the setup is now ready to be used for producing coupled regional climate projections for Europe.