Understanding the nature of risks associated with onshore natural gas gathering pipelines

Abstract

Onshore gathering pipelines for transport of natural gas are a key part of the growing production industry in the United States. The safety of gathering pipelines has become an increasing public concern. However analysis of the data available in the database available from the federal agency responsible for regulation, leads to the conclusion that the risk of failure of gathering pipelines is equal to or possibly lower than that of transmission lines. Over the past two decades, incidents in natural gas onshore gathering pipelines have included no fatalities and 16 serious injuries. Unfortunately these statistics are based on the 10% of gathering pipelines that are currently federally regulated. A substantial increase in the failure rate (rate of significant incidents in the DOT database) of gathering pipelines is evident over the last two decades. It is shown that this does not correspond with a decline in pipeline safety. The variation in the rate of significant incidents can be largely explained by the increased natural gas prices over the corresponding time period. Onshore gathering pipelines are typically younger in construction age (43% are <30 years old; whereas only 25% of onshore transmission pipeline segments are 30 years or younger). This may explain the lower failure rates for onshore gathering pipelines. The 95% confidence interval of failure rate for onshore gathering pipelines is 3.5 × 10−5 to 7.0 × 10−5 per km yr, whereas it is 8.4 × 10−5 to 9.7 × 10−5 per km yr for onshore transmission pipelines. Injury rate of onshore gathering pipelines is estimated as 38%–78% of that of onshore transmission pipelines at 95% confidence level. Difference in the fatality/injury rate between onshore transmission pipelines and onshore gathering pipelines, however, is not statistically significant. The available evidence does not support the contention that gathering pipelines have high risks or that the risk has recently increased. However, these conclusions must be tempered by recognizing the limited sample size of our data set.