Reading US Pipeline Incident Data

01

What the incident records actually contain

US pipeline incident data is compiled from mandatory operator reports and published by system: gas transmission, gas distribution, hazardous liquids, and LNG. Each system has its own reporting thresholds and mileage base, so a count of incidents is only meaningful when it is read against the mileage and the year it covers. The cause fields in those records come from official filings, and their wording follows regulatory categories rather than plain description, which is why the same event can appear under different labels across systems. The 49 CFR parts 190 to 199 set the reporting and integrity rules that produce the records in the first place.

The records are not a single table. They are a set of annual submissions from operators, each tied to a system and a reporting year. Gas transmission and gas distribution are counted separately, even when the same company runs both, because the reporting thresholds differ. Hazardous liquids, which covers crude oil, refined products, and other dangerous fluids, sits in its own series. LNG facilities form a fourth group with its own characteristics.

For each record, the fields typically include the location, the system, the commodity, the volume or quantity released where relevant, the consequences, and a set of cause codes. The cause codes are the part most often misread. They are assigned according to definitions in the reporting instructions, not according to an investigator’s narrative. A single incident can carry a primary cause and contributing factors, and the primary cause is the one that drives the aggregate statistics.

Mileage matters as much as counts. An operator with more pipeline miles will, all else equal, appear more often in absolute counts. Annual mileage reports exist precisely so that rates can be calculated against exposure. Without the mileage denominator, a state or a system with a large network looks worse than a small one for reasons that have nothing to do with safety performance.

A useful reference point for anyone starting from scratch is the documentation collected at pipeline incident records, which organises the material by system and keeps the annual mileage reports alongside the incident counts. That pairing is what makes the numbers comparable across years.

02

Which systems are reported separately?

Gas transmission lines move gas at higher pressure over longer distances, often between states. Gas distribution lines are the lower pressure network that delivers gas to homes and businesses. The two are reported under different thresholds, and the consequences of a failure differ: a transmission line failure can affect a wide area, while a distribution failure is usually local but more frequent per mile.

Hazardous liquids pipelines carry crude oil, gasoline, diesel, and other products. Their incident records include release volumes, which gas records do not, and this changes how severity is expressed. LNG facilities are fewer in number and are reported as fixed sites rather than as linear networks, so their records do not fit a per mile logic at all.

Because of these differences, a total across all four systems is rarely informative on its own. The more defensible approach is to read each system against its own mileage and its own reporting history, then compare trends within a system over time. Cross system comparisons are possible but require stating which normalisation was used.

03

What do the declared causes mean?

The cause fields come from the operator’s report, filed under the definitions in the applicable part of 49 CFR. Common categories include corrosion, material and weld failures, excavation damage, equipment failure, and incorrect operation. These labels are administrative. They tell you how the event was classified for reporting, not how it was investigated in a forensic sense.

Corrosion, for example, covers both internal and external mechanisms, and the subcategories matter if you are trying to understand whether integrity management or coating practice is implicated. Material and weld failures group together manufacturing defects, construction defects, and seam issues, which have different origins. Excavation damage is usually a third party event, but the record may not distinguish between a contractor working near the line and a homeowner digging without a locate request.

Reading the labels cautiously means checking the reporting year, because definitions and code sets have changed over time. A category that exists today may not have existed twenty years ago, and a category that was split into two will show an artificial drop in the old series. Anyone building a time series from these fields has to reconcile the code sets before plotting anything.

04

How does 49 CFR frame the data?

49 CFR parts 190 to 199 are the regulations that govern pipeline safety reporting, integrity management, and operator qualification. Part 191 covers incident and annual reports for gas pipelines. Part 195 does the equivalent for hazardous liquids. Part 192 contains the design, construction, and maintenance rules for gas pipelines, and Part 195 for liquids. Parts 198 and 199 deal with damage prevention and drug and alcohol testing programmes respectively.

The reporting thresholds in Parts 191 and 195 define what must be filed. A release below the threshold does not appear in the national dataset, which means the dataset is a record of reportable events, not of all events. This is the single most important interpretive limit. Any statement about frequency that ignores the threshold is comparing a filtered set to an unfiltered reality.

Integrity management requirements, introduced through amendments to Parts 192 and 195, require operators of certain pipelines to identify high consequence areas and to assess the condition of the pipe in those areas. In line inspection, often called smart pigging, is one of the permitted assessment methods. The results of those assessments are not published as incident records, but they shape the maintenance decisions that precede or prevent incidents. The regulatory framework therefore produces two different kinds of public information: event records and compliance context.

05

How should the numbers be compared?

Comparisons are legitimate when the systems, the years, the thresholds, and the mileage base are held constant. A rate per thousand miles of pipeline per year is the standard normalisation for transmission and distribution. For hazardous liquids, release volume per incident is often added because the consequence profile differs. For LNG, incident counts are small enough that a single event can dominate a year, and rates are not meaningful.

Methodology notes matter more than the headline figures. If a dataset states that it uses reportable incidents only, that is a different population from one that includes all reported releases. If it uses a rolling five year average, the trend line will lag. If it reclassifies causes according to a modern code set, the historical series will not match the original filings.

A glossary is not a formality here. Terms such as significant incident, high consequence area, and maximum allowable operating pressure have specific regulatory definitions, and using them loosely produces conclusions that the underlying records do not support. The same applies to cause labels: reading them as plain English rather than as defined categories is the most common source of error in secondary summaries.

06

What the records can and cannot show

The records can show how many reportable events occurred in a given system in a given year, what the declared cause was, and what the consequences were in terms of injuries, fatalities, and property damage. They can show trends when the code sets and thresholds are held constant. They can show which operators appear more often, though not why without further context.

The records cannot show how many near misses occurred, how many releases fell below the reporting threshold, or how many potential failures were caught by inspection and never became incidents. They cannot show the condition of the network as a whole, because inspection results are not published in this form. They cannot support a claim that one system is safer than another without a stated normalisation and a stated period.

For readers who need the source behind a specific claim, the practical route is to go back to the annual filings and the mileage reports, confirm the reporting year and the threshold, and check the cause code against the definition in force at the time. That is slower than reading a summary, but it is the only way to know what a number actually represents.