Ecological models can act as interesting tools to support decision-making in river restoration management. In particular models which are capable of predicting the habitat requirements of species are of considerable importance to ensure that the planned actions have the desired effects on the aquatic ecosystem. To this end, Artificial Neural Network (ANN) models were tested and optimized for the prediction of the habitat suitability for Gammarus pulex, a relevant indicator species in water quality assessment. Although ANN models are in general quite robust with a rather high predictive reliability, the model performance had to be increased with regard to simulations for river restoration management. In particular, it has been shown that spatial and temporal expert-rules could possibly be included. Migration dynamics of downstream drift and upstream migration of the organisms and migration barriers along the river (weirs, culverted river sections, ...) might indeed deliver important additional information on the effectiveness of the restoration plans, and also on the timing of the expected effects. In this context, an additional in-stream migration model for Gammarus pulex was developed. This migration model, implemented in a Geographical Information System (GIS), has been used to simulate a practical river restoration scenario for a river in Flanders, Belgium. The case study illustrated that the removal of a weir, at a particular site, resulted in the improvement of the habitat suitability for Gammarus pulex. The ANN models predicted that after restoration the habitat was suitable again for Gammarus pulex. The migration model indicated that the restored parts of the river would be recolonized within about 2 months. In this way, decision makers can have an idea whether and when a restoration option will have a desired effect.