Evaluation of non-uniqueness in contaminant source characterization based on sensors with event detection methods

Due to the present state of sensor technology, during a water distribution contamination event, sensors may be able to detect only the presence of a contaminant and not necessarily the complete concentration profile. Some sensors trigger a detection based on a specified threshold concentration of observation, yielding a binary detection/no-detection signal. Event detection can also be based on observed concentrations of water quality parameters, such as pH and chlorine, which are routinely measured. These concentration observations are then processed through event detection algorithms to yield a detection/no-detection signal. These event detection techniques filter the measured concentrations at sensors to produce a discrete signal. When using this filtered information to characterize the contamination source, the certainty of identifying a unique solution is likely reduced, i.e., a set of widely different source characteristics may provide a match for the sensor observations. The authors previously presented an evolutionary algorithm-based procedure for source characterization and for assessing non-uniqueness by generating a set of maximally different alternatives. The procedure is extended here to characterize a contaminant source and any non-uniqueness arising by using sensor information processed through different event detection methods.