Handling incomplete and missing data in water network database using imputation methods

It is challenging to develop an extensive water mains renewal program or risk management action plan if there is incomplete, partial or missing water network data. For small and medium-sized water utilities, it may not be cost effective to invest in extensive inspection and data collection programs on existing water mains to fill data gaps. In this study, the performance of three single imputation methods (i.e., mean imputation, median imputation, and linear regression-based) and three multiple imputation methods (i.e., iterative robust model-based imputation (IRMI), multiple imputations of incomplete multivariate data (AMELIA), and sequential imputation for missing values (IMPSEQ)) are compared. The cast iron (CI) water mains data of the water distribution network (WDN) of the City of Calgary, Alberta, Canada is analyzed. Results indicate that the IMPSEQ method performed best with respect to imputing missing values in water network databases compared to the other single and multiple imputations methods used.