High-resolution ensemble projections of ocean warming and sector-specific indices in the North and Baltic Seas

Abstract. This study investigates future temperature changes in the surface and benthic layers of the North Sea and the Baltic Sea using a high-resolution, three-dimensional regional hydrodynamic ensemble. To simulate future ocean states, we forced our regional ocean model with a five-member atmospheric ensemble from EURO-CORDEX, using their dynamically downscaled variables under the high-emission RCP8.5 pathway across three 30-year climatological time slices: historical (1971–2000), near-future (2031–2060), and far-future (2071–2100). Rigorous validation against long-term satellite observations reveals systematic, seasonally dependent model biases, marked by winter overestimation and summer underestimation of sea surface temperatures (SST). To mitigate these anomalies, a monthly linear scaling (MLS) bias correction scheme is implemented. Projections indicate an accelerated three-dimensional warming trend toward the end of the century, with pronounced regional hotspots exceeding 4–5 °C in semi-enclosed arctic-amplification zones like the Bothnian Bay and the Gulf of Finland. To demonstrate why bias correction is vital and to illustrate the socioeconomic and ecological value of bias-corrected projections, the model results are converted into two sector-specific indicators. The surface-based "bathing suitability index" reveals an expansion of water temperatures above 20 °C along the western Baltic coast, indicating prolonged tourist seasons alongside heightened risks of pathogenic Vibrio proliferation. Moreover, the bottom-based "spawning survival index" suggests that spring sea bottom temperatures (SBT) in the North Sea will consistently transcend the optimal 7 °C threshold for Atlantic cod (Gadus morhua), driving an inevitable northward habitat migration. These results highlight the critical necessity of surface and vertical bias correction when deploying hydrodynamic models to inform regional climate adaptation frameworks.