The quality of a morphological prediction is often expressed by an overall grid-point based skill score based on the Mean Squared Error (MSE) between the predicted and observed bed levels (Sutherland et al., 2004). Although the MSE is a good measure of the overall error between model and observations, it tends to penalize rather than reward the model’s capability to provide information on features of interest such as scour holes, accumulation zones and migrating tidal channels; a feature that is predicted correctly in terms of timing and size but is (slightly) misplaced leads to a relatively large MSE as compared to a smoother forecast. This makes it difficult to demonstrate the skill of a high variability prediction (Anthes, 1983). Our aim is to overcome this inherent limitation of the MSE and other grid-point based error metrics. To that end, we introduce a new distance measure for 2D morphological change that explicitly takes (dis)agreement in spatial patterns into account.