The model would be need to be able to use either coarse x, y, z morphology data and/or high precision, large data cloud x, y, z datasets (lidar/multi-beam sonar).
The GRASS GIS (old version) has a model for channels. It solves the Saint Venant ´s equation. It takes the topography information from the GIS coverages (a DEM) and builds a basin and sub basin map using the DEM and the drainage network (from the hydrography map). Then you can simulate the rain in some basin and get an hydrogram at the end of a selected channel. You manage all the inputs as GIS coverages. Regarding dam break scenarios there were a model called MIKE 21 (I investigated long time ago a previous version called MIKE 11) from a Danish Institute. I knew it only by references, I have no experience on it. In any case, but I was digitizing some maps as input for this model. I know that he MIKE 11 version used GIS shapefiles (at that time from ARCVIEW 2 or 3) and also Landsat imagery.