Detecting Spatial Patterns in Water Distribution Pipe Failures during Extreme Weather Events

As infrastructure assets throughout the U.S. approach the end of their service life, many water distribution systems are also increasingly operating outside of the climate conditions for which they were designed. For instance, in February 2021 Winter Storm Uri crippled water distribution systems across Texas due to extensive pipe and equipment failures. Understanding if spatial patterns of pipe failures exist can aid utilities in directing limited funds when replacing aging pipes and increasing resiliency to extreme weather. Spatial autocorrelation analysis has previously identified spatial clusters in pipe failure data in distribution networks to support asset management and pipe failure prediction modeling. However, previous work has not emphasized the discrete events that frequently account for a greater portion of failures. Extreme events (e.g., severe hot or cold weather) place an intensified strain on water infrastructure and frequently cause an increase in pipe failures, stretching utility resources during critical response times. This study examines 20 years of pipe failure data from a southern U.S. city to determine if there are distinct spatial patterns present during extreme events. Here, we use control charts to isolate extreme events in pipe failure data and demonstrate that sharp increases in failures align with the occurrence of extreme weather. We then apply spatial autocorrelation analysis to determine if statistically significant clusters of failures exist and if clusters differ during extreme events compared with “normal” operating conditions. The results offer additional insights to utilities seeking to both improve asset management practices and better prepare for extreme weather events.