Statistical Modeling of the Probability and Duration of Hazardous Liquid Pipeline Shutdowns: A Hurdle Regression Approach

Operational shutdowns following hazardous liquid pipeline incidents are critical but poorly understood events that impact the U.S. energy supply. Although prior research has investigated the causes and outcomes of pipeline failures, limited work has explained what drives both the likelihood of a shutdown and the duration once it begins. The goal of this study is to address this gap by developing a hurdle regression model to examine the two-stage shutdown mechanism in pipeline incidents, using the Pipeline and Hazardous Materials Safety Administration (PHMSA) incident dataset from 2010 to 2025. The hurdle model consists of a logistic regression restricted to pre-decision predictors to model the probability of shutdown, and a lognormal regression to model the duration of those leading to shutdown. The results revealed that distinct factors are associated with each outcome. Shutdown probability is associated with pre-decision operational and contextual indicators, including operating pressure at the time of incident, accident type, location, monitoring presence, and response delay. In contrast, shutdown duration is associated with logistical complexity and post-incident severity, including incidents at pipeline crossings, pressures exceeding 110% of the maximum operating pressure, and reported property damage. These findings, while exploratory in nature given the use of public incident data, offer practical reference points for operators and regulators who aim to shorten recovery time and strengthen the resilience of energy infrastructure.