Passive acoustic monitoring (PAM) is an important technique to assess the presence of marine mammals and, if necessary, mitigate the effects of anthropogenic noise sources upon them. The complexity of the ocean acoustic environment makes accurate localisation of marine mammal vocalisations difficult. It is, therefore, important to be able to predict the detection and localisation performance of PAM to ensure sensors are optimally placed and any resulting actions are suitably informed. A variety of acoustic models can be used to predict the performance of a sensor field ranging from simple 1D models up to full 3D models. Each type of model has advantages and disadvantages. Simple 1D models offer the most advantages in terms of solution time and minimal inputs but their predictions can have reduced accuracy. This is especially true in areas of complex ocean floor topography, such as on a shelf break, where complex 3D acoustic effects can occur. This study investigates the variation in the predicted acoustic field from 1D up to 3D models on a shelf break and how with increasing model fidelity the picture of the acoustic environment changes and its impact on acoustic monitoring advice. © Crown copyright (2022), Dstl.