The correct prediction of flooding in urban areas is an important challenge to secure the values and fulfil public regulations. Traditional sewer simulations deliver the basic information for a rudimental flood protection, but the interaction between sewer and surface runoff can only be considered by a bi-directional modelling. Therefore detailed information about the relevant structures on the surface is necessary, which can partially be delivered by airborne laser scan data. This data have to be refined to get as detailed information about the endangered areas as possible. But the plenitude of information leads to high requirements on computer capacity and performance. This paper shows different approaches to predict the sewer caused flooding in urban areas. The approaches have been checked on two testing areas in Germany and the developed tool will be implemented in a commercial software system soon. This approaches, which partially base on each other, make a stepwise refinement of the model and narrowing of the affected areas possible. The developed algorithms to thin the digital terrain model and the well proven method to parallelize the calculation on more than one processing units secure an effective calculating process.