This post explains my approach to modeling pipeline accidents along the Dragonpipe (Mariner East pipeline system). I will be doing a number of posts about the consequences of accidents at various sites, and it didn’t make sense to repeat this background information in each post.

In addition to the basic information described here, site-specific information is required to model each accident location. That information is described in the individual posts for the location involved.

A list of sites I have already analyzed, with links to the corresponding posts, can be found at the end of this post. It will be updated as more locations are analyzed.

Modeling the worst case. In modeling a pipeline accident, many factors need to be taken into account as inputs to the process. Some of the critical ones are assumptions about the size of the pipeline rupture or leak, the weather conditions, and the particular liquefied gas being transported. I make the assumptions that result in the greatest potential for casualties and destruction. For a pipeline like Mariner East carrying “natural gas liquids” (NGLs), the main worst-case assumptions are:

  • A complete rupture of the pipeline
  • Very light winds
  • The contents of the pipeline is propane
  • An assumed delay (typically 10 minutes, but any time beyond 5 minutes gives similar results) before the cloud of propane vapor finds an ignition source

This worst-case focus is somewhat different from that taken by the recent risk assessments, which tend to focus more on a variety of cases, including the most likely cases, not the worst ones. Both types of analysis have their role, but I think considering the worst cases is essential for emergency preparedness. If first responders were only prepared for the common cases, they would be left unprepared to deal with the really disastrous ones. On the other hand, if they are prepared for the worst case, they are ready for anything.

It is also worth noting that Australia, which requires pipeline builders to take worst cases into account in their plans, has a far better pipeline safety record than we do in the US.

Which pipeline? When considering the consequences of a rupture at a given location, the first question to ask is which pipeline to focus on. The Mariner East system actually consists of four pipelines:

  • Mariner East 1 (an 8-inch line from the 1930s that is currently in operation)
  • Mariner East 2 (a 20-inch pipeline, currently being constructed)
  • Mariner East 2X (a 16-inch pipeline, currently being constructed)
  • A 12-inch pipeline (the “bypass” pipeline) built in 1949 that is to be used to bypass parts of ME2 and ME2X that haven’t been completed. The section that is being incorporated into the Mariner East system runs from Wallace Township in northern Chester County to Middletown Township in central Delaware County.

Because of variations in the size and pressure of these pipelines, the consequences of a rupture are different for each. I have generally used the 20-inch pipeline as the basis for my analysis, except as noted below.

Which risk assessment? Two different risk assessments have been done of the Mariner East system. Both of them focused on the 20-inch pipeline (because it represents the greatest risk). The first to be published was the Citizens’ Risk Assessment, which was initiated by State Senator Andy Diniman and funded by a combination of crowdsourcing and funds from local municipalities. It was performed by Quest Consultants of Norman, OK. It used a software modeling package called “Canary” to calculate the consequences of a leak.

The second risk assessment was commissioned by Delaware County and performed by G2 Integrated Solutions of Houston, TX. It used a software modeling package called “Phast” to calculate the consequences of a leak.

Because the two organizations used different software, they came up with different consequences for a pipeline rupture. The area subject to a lethal “flash fire” from a rupture was about 8 times greater in the Delaware County report. That report also indicated a large area that would experience a lethal “overpressure” (the shock wave from the explosion). The Cititzens’ assessment did not analyze that source of fatality.

Once the Delaware County report became available, I began using it as the basis for most of my accident modeling. (It was not yet available for the first few locations that I modeled.) I used it because it describes a more serious risk than that described by the Citizens’ report, and emergency responders need to be ready for the worst case.

A technical note on the difference between the two assessments: in both of them, the boundary of the propane cloud is defined by what engineers call the “lower flammable limit”. As the gas expands from the leak, it mixes with air. At some point, although it continues to expand, it becomes too diluted to burn (i.e. it is no longer “flammable”). That point is the lower flammable limit. The difference between the vapor cloud sizes calculated in the two risk assessments may be related to different ways of calculating the lower flammable limit.

What about pipelines besides the 20-inch one? The two risk assessments provide the information necessary to model a rupture of the 20-inch pipeline, but there are some locations where ME1 and the 12-inch pipeline deviate from the route of the 20-inch pipeline, or where the 12-inch pipeline will be operating for a long time (perhaps years) before the 20-inch pipeline is completed and in operation. In those locations, a different approach to modeling is needed.

Fortunately, the Citizens’ Risk Assessment opened up an opportunity for modeling the other pipelines. In connection with the risk assessment itself, the citizens group that sponsored it also leased the “Canary” software package that Quest Consultants uses to model consequences. The package is widely used in the petrochemical industry to model pipeline accidents and accidents at various kinds of petrochemical plants. It has been field-tested to make sure its results match actual results of experimental pipeline releases. A group of local people involved with pipeline analysis and emergency response was given a two-day training course on using the software, and I was able to be one of those trained.

Using the Canary software, I have been able to analyze the consequences of pipeline ruptures of the other pipelines in the Mariner East system, and those analyses form the basis for some of my blog posts.

Vapor cloud, flash fire, shock wave, jet fire. In each case modeled, it is assumed that the escaping liquid would vaporize instantly and form a large cloud of flammable propane during the initial 10 minutes. The exact dimensions of the cloud would depend on the pipe diameter and (to a limited extent) on the pipeline pressure, but in most cases it would range from about 1500 feet in length and width to as much as 1.3 miles in length and 0.75 mile in width. (The details on the flammable cloud dimensions are in the posts for specific sites.)

In 10 minutes, the cloud was assumed to find an ignition source. The entire cloud would burn within a few seconds, killing anyone in it who was outdoors. This is called a “flash fire”, and it is the most dangerous part of the process. Being indoors would provide some protection, depending on the degree to which the vapor had penetrated the building (potentially drawn inside by the ventilation system). The flash fire would also set on fire any flammable materials, including wood-frame houses, in the vapor cloud.

At the same time, the flash fire would trigger an “overpressure event” (a shock wave) that would kill anyone within a specific radius of the center of the cloud, whether indoors or outdoors. In the case of the 20-inch pipeline, the width of the fatal overpressure area given in the Delaware County report would be about 0.8 mile. That largely overlaps with the flash fire area, adding perhaps 10% additional outdoor fatalities and a large number indoors. We don’t know the corresponding area for the smaller pipelines, because the Canary software (which I use to analyze the smaller pipelines) does not include a comparable shock wave calculation.

Following the flash fire and shock wave, the propane would continue to flow from the pipe for hours, even if the company immediately closed the valves to isolate the damaged segment. The rupture location would still be extremely hot, and the remaining gas exiting the pipeline would ignite instantly. That gas, igniting on exit, would create a “jet fire”—a much smaller but hotter fire similar to a blowtorch, emanating from the pipe. Depending on exactly where the rupture was relative to nearby buildings and their access roads, the jet fire could impede rescuers from getting to those who survived the flash fire. The jet fire would burn for hours, gradually subsiding, until the gas had entirely emptied out from the ruptured section of pipeline.

In general, each of my analyses has focused on the vapor cloud, flash fire, and shock wave, but I do mention the jet fire in cases where it could affect rescue efforts.

Effects not modeled. There are limits to the two available reports and the Canary software that have prevented me from modeling some aspects in more detail. For example, in hilly areas the vapor will travel downhill, and will collect in low-lying areas, changing the shape of the vapor cloud and the area of the resulting flash fire. That has been an important factor in some of the worst NGL accidents, but the software used for these reports makes the assumption that the ground is more-or-less level.

Burns and other injuries would extend beyond the area of the vapor cloud. Some of the additional effects (not modeled in my analyses) would include potentially toxic fumes from burning materials in the area, shock wave injury and death beyond the 0.8 mile area mentioned above, damage to buildings from both fire and the shock wave (creating yet more deaths and injuries), and numerous other potential causes of death and injury, such as flying debris. None of these are modeled in my work. But I think the flash fire is the really critical outcome to focus on, and that is what I have done.

Locations and pipelines modeled. I have already published blog posts describing the consequences of a pipeline rupture at the locations in the list below. Each location name is a link that you can click to see the corresponding post. This list will be updated as new locations are modeled.