Incorporating spatial correlation in a markov chain Monte Carlo approach for network model calibration

To utilize a drinking water distribution system network model in any decision making process requires a calibrated network model. Typical calibration methods assume known consumer demands and adjust pipe roughness coefficients to fit pressure measurements and storage levels. However, these data contain little explicit information related to hydraulic residence time and travel path, which are necessary to improve water quality representations. Recent field-scale tracer tests have been shown capable of collecting data related to hydraulic residence time and flow path that can be used to adjust demand pattern multipliers to fit the observed tracer data. Both problem types (estimating pipe roughness coefficients or demand pattern multipliers) can have a spatially distributed component. This research extends an existing Markov chain Monte Carlo calibration algorithm by incorporating spatial correlation into the parameter estimation framework. Results will be generated using synthetic test data to evaluate the ability of the calibration algorithm to regenerate the known roughness coefficients or demand pattern multipliers. Copyright ASCE 2005.