Skip to main content

Contamination in drinking water distribution systems: Consumer exposure risks and source identification

Final Activity Report Summary - COCERSI (Contamination in drinking water distribution systems: Consumer exposure risks and source identification)

Vulnerability of drinking water distribution systems (DWDS) to contamination events has been always a major concern of water utilities and regulatory agencies. This research project developed along three main interacting topics. One is the development of methods to identify optimal sensor location for early warning should a contamination event occur. Second topic is the contamination source identification should a contamination event occur and be detected in the system. Third line of research has been the study of the efficacy against a contamination event of disinfectant residual maintenance in the distributed water and the trade-offs of these strategies to reduce consumer exposure risks. In addition, some work has been carried out the improve water quality modelling trough sampling design and disinfectant residual monitoring within the distribution system. This part is critical for all other topics, since operational research in this area has to rely on good models to mimic real dynamical conditions.

The most recent approaches that have been proposed for sensor location design in DWDS have different mathematical formulations and require different computational efforts; both depending on design objectives, simplifying assumptions and solution methods adopted. The main goal of the work developed has been instead to propose a model to identify optimal water quality monitoring locations in DWDS that somehow unifies the different approaches in a unique coherent general framework, applicable to steady or unsteady flow conditions, and that may accommodate different design objectives. Various design objectives may be described as identical mixed-integer linear programs whose formulation differences are only the cost function coefficients and not the decision variables nor the constraints. The model proposed may reproduce equal solutions of most recent approaches when design characteristics -- in particular design objective, network dynamical properties and contamination ensembles -- are set equal.

Recently, there is a grown interest in investigating methodologies that may allow using the gathered information obtained by the sensors to identify the locations or the network areas from which the contamination might have started. Such capability could be very important both for emergency response strategies, for example to isolate only part of the network rather than affecting the entire DWDS; but more importantly to better investigate, and possibly eliminate once for all, the cause of contamination. The main challenges of this problem are the inherent non-uniqueness of the solution due to limited data availability compared to the typical size of a real DWDS; and the computational effort required to solve the problem due to the significant number of space and time variables. Research activity has developed along two main topics: (1) sampling design to optimal identify sensor locations that may improve inversion; (2) the implementation and comparison of different approaches (Least-Squares, Bayesian and Entropic) for solving ill-posed inverse problems for assigned sensor networks.

Finally, early detection technologies will become available for installation in the coming years. Therefore, it is necessary to address several issues associated with the use of warning systems and residual maintenance design. A holistic simulation framework for designing protection strategies has been proposed. A focal point of the work is how residual maintenance can be tailored keeping into account of the presence of a warning system and vice-versa. A multi-species network water quality model that includes disinfectant decay and disinfection kinetics quantitatively assesses the risk of delivering contaminated water to consumers. For an assumed set of intrusion scenarios, the system vulnerability is quantified as the cumulative distribution function of the number of consumers that receive contaminated water before the warning system raises an alarm. Design and effectiveness of these consumer protection measures are discussed and illustrated by applying the framework on example networks