The public drinking water supply of southern Germany is characterized by a rather decentralized network. Due to the hydrogeological setting in these parts of Germany many of the small water works with an average capacity of 50 m3/h have to treat raw water extracted from karstic or cliffy aquifers. These raw waters tend to be contaminated with particles and pathogens acquired during snowmelt or after strong rainfalls. In the last decade ultrafiltration has become the technology of choice for the removal of the aforementioned contaminants. Flux decline caused by unanticipated membrane fouling is the main limitation for the application of ultrafiltration membranes. This paper describes how membrane fouling phenomena can be predicted by using a statistical approach based on data from large scale filtration systems in combination with field and lab experiments on raw water quality and membrane performance. The data defines water quality and respective fouling phenomena both in technical scale filtration plants and in lab experiments of eleven different raw waters. The method described here is more economically feasible for small water works when compared to typical pilot experiments that are used for high capacity water works.
Keywords: drinking water treatment, fouling prediction, modified fouling index, ultrafiltration