Wave climates in ports can determine the efficiency of port operations to a large extent, so good predictions of them are important. This thesis therefore focusses on the modelling of wave penetration in ports, using several numerical models as well as a physical model. The goal is to compare numerical wave models for the use of wave penetration in ports and to determine application boundaries. An overview of existing classes of numerical wave models and their potential for modelling wave penetration in ports is made first. This leads to the conclusion that mild-slope models, Boussinesq-type models and non-hydrostatic models are probably the most appropriate. From each of these model classes an operational model is chosen (i.e. PHAROS, TRITON and SWASH) for further study. It is demonstrated that these models are all capable of accurately reproducing analytical solutions for linear shoaling and diffraction. These two wave processes are, together with reflection, identified as crucial for wave propagation in harbours. Differences between these models are pointed out as well. Most notable are the differences in underlying equations, schematisations of boundaries and computational efficiency. To be able to study the performance of operational numerical models for wave penetration in ports, in scenarios that are more realistic than analytical benchmark test cases, measurements from a physical model are used. The physical model represents a typical seaport, is schematised to facilitate numerical modelling, uses three layouts with increasing complexity, and is subjected to a large variety of wave conditions. The physical modelling is carried out at Deltares in the framework of one of their projects and the produced dataset will become freely available. A selection of the 57 scale model tests are analysed in detail and numerically simulated using PHAROS, TRITON and SWASH. A comparison between the simulation results and the corresponding physical model measurements shows that TRITON and SWASH are capable of modelling the penetration of primary waves in ports. PHAROS simulations are less accurate and appear to overestimate wave heights in areas sheltered from direct wave attack, which is ascribed to its specific use of reflection boundaries. Wave conditions with very large wave height over depth ratios (up to 0.5), where wave breaking is significant, can still be simulated accurately with both PHAROS and SWASH. TRITON could not be used to simulate these wave conditions due to numerical instabilities. The total amount of low frequency wave energy modelled by TRITON and SWASH corresponds well with measurements from the physical model (for PHAROS this comparison is not made). Specific harbour oscillations are however only reproduced qualitatively by all three numerical models.