%0 journal article
%@ 2190-4979
%A Gröger, M.,Dieterich, C.,Haapala, J.,Ho-Hagemann, H.,Hagemann, S.,Jakacki, J.,May, W.,Meier, H.,Miller, P.,Rutgersson, A.,Wu, L.

%D 2021
%J Earth System Dynamics
%N 3
%P 939-973 
%R doi:10.5194/esd-12-939-2021
%T Coupled regional Earth system modeling in the Baltic Sea region
%U https://doi.org/10.5194/esd-12-939-2021
3 
%X Nonlinear responses to externally forced climate change are known to dampen or amplify the local climate impact due to complex cross-compartmental feedback loops in the Earth system. These feedbacks are less well represented in the traditional stand-alone atmosphere and ocean models on which many of today's regional climate assessments rely (e.g., EURO-CORDEX, NOSCCA and BACC II). This has promoted the development of regional climate models for the Baltic Sea region by coupling different compartments of the Earth system into more comprehensive models. Coupled models more realistically represent feedback loops than the information imposed on the region by prescribed boundary conditions and, thus, permit more degrees of freedom. In the past, several coupled model systems have been developed for Europe and the Baltic Sea region. This article reviews recent progress on model systems that allow two-way communication between atmosphere and ocean models; models for the land surface, including the terrestrial biosphere; and wave models at the air–sea interface and hydrology models for water cycle closure. However, several processes that have mostly been realized by one-way coupling to date, such as marine biogeochemistry, nutrient cycling and atmospheric chemistry (e.g., aerosols), are not considered here.,In contrast to uncoupled stand-alone models, coupled Earth system models can modify mean near-surface air temperatures locally by up to several degrees compared with their stand-alone atmospheric counterparts using prescribed surface boundary conditions. The representation of small-scale oceanic processes, such as vertical mixing and sea-ice dynamics, appears essential to accurately resolve the air–sea heat exchange over the Baltic Sea, and these parameters can only be provided by online coupled high-resolution ocean models. In addition, the coupling of wave models at the ocean–atmosphere interface allows for a more explicit formulation of small-scale to microphysical processes with local feedbacks to water temperature and large-scale processes such as oceanic upwelling. Over land, important climate feedbacks arise from dynamical terrestrial vegetation changes as well as the implementation of land-use scenarios and afforestation/deforestation that further alter surface albedo, roughness length and evapotranspiration. Furthermore, a good representation of surface temperatures and roughness length over open sea and land areas is critical for the representation of climatic extremes such as heavy precipitation, storms, or tropical nights (defined as nights where the daily minimum temperature does not fall below 20 ∘C), and these parameters appear to be sensitive to coupling.,For the present-day climate, many coupled atmosphere–ocean and atmosphere–land surface models have demonstrated the added value of single climate variables, in particular when low-quality boundary data were used in the respective stand-alone model. This makes coupled models a prospective tool for downscaling climate change scenarios from global climate models because these models often have large biases on the regional scale. However, the coupling of hydrology models to close the water cycle remains problematic, as the accuracy of precipitation provided by atmosphere models is, in most cases, insufficient to realistically simulate the runoff to the Baltic Sea without bias adjustments.,Many regional stand-alone ocean and atmosphere models are tuned to suitably represent present-day climatologies rather than to accurately simulate climate change. Therefore, more research is required into how the regional climate sensitivity (e.g., the models' response to a given change in global mean temperature) is affected by coupling and how the spread is altered in multi-model and multi-scenario ensembles of coupled models compared with uncoupled ones.