Statistics of catastrophic hazardous liquid pipeline accidents

Highlights: • The existence of Power-law behaviors in the upper tail of the actual distribution has been verified with the combination of two graphic tools. • Power law model is capable of modeling the upper tail of the empirical distribution of hazardous liquid pipeline accidents. • The occurrence of a catastrophic pipeline accident is relatively common from a statistical standpoint. • A new method has been proposed to estimate the probability of catastrophic hazardous liquid pipeline accidents and evaluate the pipeline safety management. The sparse data of catastrophic pipeline accidents implies large fluctuations in the empirical distribution's upper tail, which makes it hard to estimate the true probability. To address this problem, the block-maxima(BM) method and the peaks over threshold (POT) method from extreme value theory have been applied to describe the tail behaviors of the hazardous liquid pipeline accidents occurred in the United States, 1986–2019. The regularity of the scaling in the tails of the estimated power-law distribution indicates that the most catastrophic events are not abnormal values, but consistent with the global pattern of hazardous liquid pipeline accidents. Moreover, there are similar regularities in the severity of pipeline accidents with different measurement scales, which are consisted with the scale-free property. GEV model and GPD model have been built to estimate the probability of catastrophic hazardous liquid pipeline accidents. Compared with classical statistical models, these two models are more accurate. Another potential practical application of the fitted models is that they can be used to directly evaluate the pipeline safety management regardless of the details of pipeline accidents. The results provide a new insight to calculating the environmental risk triggered by hazardous liquid pipelines and deliver pivotal information to pipeline managers so that potential pipeline accidents could be immigrated or prevented scientifically.