Analysis of Model Sensitivity and Uncertainty for Chlorine Transport and Decay in a Water Distribution System

There are a number of sources of uncertainty in drinking water distribution system modeling. Uncertain parameters include pipe diameters, consumer demands, hydraulic energy loss coefficients, reaction coefficients and others. Understanding the relative importance of these sources of uncertainty can improve the allocation of resources for model refinement and calibration, as well as, aid knowledge inference from monitoring data. This paper presents an analysis of uncertainty and model sensitivity for chlorine transport and decay in a water distribution system. A clustering and global variance-based sensitivity methodology is proposed to account for spatial inconsistencies found in the results of previous studies of this problem. Results are presented from small and large scalecase studies. This methodology is then used to explore the occurrence of intrusion events in a water distribution system, and the potential to detect such events through online monitoring of chlorine residual concentrations. Noise present in the chlorine monitoring signal has the potential to overwhelm the detection of an upstream intrusion and its associated chlorine demand. Results are presented from simulated intrusion events of varying magnitude and duration.