A Probabilistic Approach for the Design of an Early Warning Source Water Monitoring Station

Abstract

This thesis involves the design of an early warning source water monitoring station for a riverine source of drinking water. These stations provide downstream water utilities with advanced notification of contamination events so they have time in which to implement a response.

Many threats facing riverine water supplies, such as accidental spills, are uncertain in nature. Therefore, designing a monitoring station for the detection of these events requires a probabilistic modelling approach. Sources of uncertainty considered in this research include the location, mass and duration of a spill event as well as the flow at the time of the spill and the water quality model parameters. Probability distributions for each of these uncertainties were defined and a Monte Carlo experiment was conducted.

The design objectives include maximizing the probability of detection and maximizing the probability of having a threshold amount of warning time. These objectives are in conflict with each other because the probability of detection improves as the station moves closer to the intake and the amount of warning time increases as the station is located further upstream. Values for the competing objectives were calculated for a number of potential monitoring station locations at multiple sample intervals and the tradeoff solutions were analyzed.

This methodology was applied to the Hidden Valley Intake which services the Regional Municipality of Waterloo’s Mannheim Water Treatment Plant. The Hidden Valley Intake is located in Kitchener, Ontario and withdraws up to 72 ML of water per day from the Grand River.

Based on an analysis of the Monte Carlo simulation results for the case study application, it was found that locating the monitoring station near the Victoria Street Bridge, approximately 11 km upstream of the intake, represents the best tradeoff in the design objectives. Sampling at least once per hour is recommended to increase the amount of warning time.

The impact of various sources of uncertainty was also explored in this thesis. It was found that the flow at the time of a spill and the spill location are the only sources of uncertainty that significantly impact the probability distributions of relevant model results.