Detecting clusters and tracing management of water distribution system using space–time scan statistics and utility network modeling

Abstract

Water-quality incidents in the water supply can occur because of errors in pressurization plant operations, leaks, and pollutant infiltration. They lead to citizen complaints about water quality. Identifying high-risk areas and contributing factors for accidents helps to prevent incidents. Here, a predictive analysis technique based on a geographic information system and complaint data was developed using SaTScan, a space–time statistical analysis program. High-risk clusters were identified using maximum log-likelihood ratios, relative risks, and Monte Carlo hypothesis testing. After a red water accident, high-risk clusters (C6–C9) were concentrated in one area. The relative risk of C7 before the accident was 4.58, double that of C2 (2.21). Using ArcGIS, a utility network dataset was created for the water-supply infrastructure in the high-risk cluster area. A utility network model enables flow simulation of the target water supply and tracing of the scope of influence. Downstream analysis of the trace showed that C8 had the greatest expected range of damage following an accident. To address this issue, four valves requiring controls were identified. C6 was predicted to suffer the most significant damage in an accident because it was closest to the water purification plant and had the largest pipe diameter.