Methodology for Bayesian belief network development to facilitate compliance with water quality regulations

Abstract: Limited resources and drinking water quality requirements pose significant challenges to those managing small and rural drinking water distribution systems (WDSs). Real-time monitoring technologies could support regulatory compliance, if shortcomings such as false readings and data corruption could be overcome. Bayesian Belief Networks (BBNs) are proposed as a means to mitigate technological shortcomings and increase certainty about the state of a given WDS. This paper describes a methodology for the development of BBNs that integrates known system characteristics with real-time monitoring technologies to support the water quality compliance of small or rural WDSs. Expert judgement was used both in the development of the structure of the BBN and in quantifying the required probability relationships. The results of a case study application of this methodology suggest that it is useful in developing a BBN to support decision making for a WDS with limited use of real-time monitoring technology.The authors are grateful for the contributions of Dr. Bryan Karney in the development of this methodology. We would also like to thank the staff at SaskWater Inc. as well as that of the representatives from the towns of Martensville, Dalmeny, Osler and Hague, Saskatchewan for their collaboration. Special thanks go to B.S. Jong for his preliminary research and development of Figure 1. Finally, the present work would not have been possible without the financial support of the Canadian Water Network and the National Science and Engineering Research Council (NSERC)