Abstract. The accuracy of morphological predictions is generally measured by an overall point-wise metric, such as the mean-squared difference between pairs of predicted and observed bed levels. Unfortunately, point-wise accuracy metrics tend to favour featureless predictions over predictions whose features are (slightly) misplaced. From the perspective of a coastal morphologist, this may lead to wrong decisions as to which of two predictions is better. In order to overcome this inherent limitation of point-wise metrics, we propose a new diagnostic tool for 2-D morphological predictions, which explicitly takes (dis)agreement in spatial patterns into account. Our approach is to formulate errors based on a smooth displacement field between predictions and observations that minimizes the point-wise error. We illustrate the advantages of this approach using a variety of morphological fields, generated with Delft3D, for an idealized case of a tidal inlet developing from an initially very schematized geometry. The quantification of model performance by the new diagnostic tool is found to better reflect the qualitative judgement of experts than traditional point-wise metrics do.