The conventional drinking water treatment plant at Lake Constance consisting of microstraining, ozonation, rapid sand filtration and chlorination, produces drinking watts for almost 3,500,000 people. Although compliance with the drinking water regulation was guaranteed at every time, discussions about a possible breakthrough of parasites and their insufficient inactivation during disinfection resulted in thinking about a further improvement of particle removal of the existing water treatment process. Therefore a pilot-scale ultrafiltration plant was operated for two years with sand filtered and microstrained surface water, While the conventional water treatment process achieved a removal of particles (1-100 mu m) around 90%, LTF improved particle removal by more than 99% with total particle counts <1 particle per mt in the ultrafiltrate. The most important operational parameter achieved during the pilot experiment was permeability. Its mean value of 1.25 L/hm(2)kPa at 5 to 6 degrees C was marginally influenced by the concentration of chlorine for the periodic back-washing. Although particle removal was very effective, calculations concerning investment costs prove UF as very expensive compared to costs of alternative techniques like direct filtration. However, considerations are made at BWV to use membrane filtration for recycling filter backwash water.