Hydraulic model calibration for the Girdwood, Alaska water distribution system

Abstract

The possible EPA promulgation of regulations requiring flushing programs for water distribution systems to regulate water age is encouraging many utilities to develop a better understanding of their systems. This usually involves the development of a hydraulic model. For the hydraulic model to be of use it must be calibrated using collected field data. Few established guidelines exist for utilities to perform such data collection and calibration. There are many different types of data that can be collected using many different methods. There are also various model calibration methods that can be used. This study sought to develop an optimized sampling plan and calibration method for a small utility that defines the best practice for data type, location, quantity, and collection conditions. A large quantity of many data types were collected at various locations in the system. Different methods were used to calibrate the model with various data sets and the model accuracy was evaluated using a second independent data set. The effect of model input parameter accuracy and pipe grouping during calibration on overall accuracy was also investigated. Data type, location, quantity, and collection conditions had an impact on calibration accuracy. High headloss data from fire flow tests provided better calibrations than low loss data from static pressures. Carefully selected test locations resulted in more efficient calibration than evenly distributed test locations. There was found to be a point of diminishing returns when investigating the amount of data used in calibration versus calibration accuracy. Uncalibrated model input parameters such as elevations and pump characteristics also had a significant impact on model calibration accuracy.