Numerical models are essential tools to understand coastal seas, but long-term simulations are expensive and slow. To reduce this effort, researchers often simulate only a few “representative years” - usually chosen by experience rather than by clear criteria. In this talk, we present a simple and objective method to select representative years for the Baltic Sea. Using long-term atmospheric data, we identify a small set of years that captures typical conditions as well as extremes. We test these years with high-resolution ocean model simulations. We show that this approach reproduces the key features of long-term simulations while reducing computational costs substantially. The method offers a transparent way to design efficient model experiments and can be applied to other coastal regions as well.