After the tragedy that's right over the border in Lac Megantic, it was only a matter of time before someone squawked about the lack of safety and the horrible environmental harm involved in transporting crude oil by rail. In fact, the Bangor Daily News already covered an aspect of this question last Summer in reaction to the increase in crude traveling by rail across Maine to a refinery in New Brunswick. But just a few hours ago, I spotted the first news article questioning the relative safety of pipelines vs. trains and more are sure to follow as the horror of the Lac Megantic disaster sinks in (see

Of course, the various professional associations that represent the rail and pipeline businesses have had their say as to which mode of transport is safer. This isn't really a big deal at all for transporting North Dakota crude across Maine to New Brunswick because there aren't any petroleum pipelines in central or northern Maine to compete with the two local rail carriers. Now if you're someone deeply invested for or against the proposed Keystone XL pipeline project, then looking at the relative safety of shipping oil sands dilbit by rail vs. pipeline is a big deal, especially since the controversial US State Department Environmental Impact Statement ("EIS") uses some of those arguments in favor of pipelines.

It will be interesting to see if the State Department reissues the Keystone EIS to include the Lac Megantic train accident as further support for the pipeline.

It is worth taking a look at the safety of shipping crude in pipelines vs. railroad tank cars. This is made difficult since the federal database on train accidents at the Federal Railroad Administration is even worse and more crypic than the old USGS surface and ground water database - and that's saying something given the notoriety of the infamous USGS water data interface. In fact, after fighting with the database on the Federal Railroad Administration website, I gave up and decided to use some data from the Association of American Railroads ("AAR").

The AAR data is from a two-page info piece that presents the AAR's assertion that shipping crude by rail is safer than shipping via pipeline. To do so, the AAR calculated an average spill rate on a per barrel and per mile basis. They dug up numbers for miles of track and miles of pipeline for this analysis. They also confined their analysis to exclusively crude oil. This approach has some major problems. Due the to differences between railroads and pipelines, their numbers are based on the no-no of false comparison. You can look at their two-page analysis for yourself at

The first problem with the AAR approach is the focus on crude. It's not really the right thing to focus on. The safety of tanker cars is the important factor in spills, not the contents. The gig with rail transport is the variety of rail cars used. There are many different tank cars types, though the major categories are pressurized or unpressurized, and insulated or uninsulated. Petroleum oils are commonly shipped in insulated non-pressurized tank cars. To evaluate tanker car safety for shipping crude, one should look at all hazardous liquid spills involving insulated non-pressurized tank cars. The spill rate for insulated non-pressurized tanker cars will be underestimated if one looks at just crude oil spills. So long as the liquid inside is appropriate for the tank car design, then the actual contents by themselves are unlikely to cause a spill. It's the external factors which are important in spills, things like tanker car design, the car's condition, train speed, weather and rail damage.

Any tanker car spill rate which is based solely on crude oil will be be lower than the actual spill rate

Using a spill rate based on the amount of track traveled is also problematic because one can ship multiple liquids over differing route. Selecting an appropriate amount of track mileage for a crude spill rate is therefore slippery and subjective. Also, consider that there is more track mileage than pipeline mileage. While oil pipeline mileage is essentially a constant with fixed-location end points, rail mileage can vary depending on the route taken. The AAR tried to compare a shipping mode with fixed mileage and and one known product with one that can ship any liquid commodity over a wide choice of routes. These differences add up making any comparison inadvisable. Comparing apples and gorillas is a false comparison.

Comparing spill rates on a per mile basis is questionable since total rail mileage is variable.

Even if spill-per-mile rates are problematic, there is still a way to compare pipelines vs. trains. The trick is to drop per mile measurements entirely and to convert spill rate to a dimensionless quantity. In this case, an appropriate dimensionless spill rate is the volume spilled vs. the total volume shipped. The subjective amount of mileage traveled is now no longer a problem and concerns of disparate infrastructure can no longer bias the results. Gross volume shipped is unaffected by other variables.

Let's look at the numbers. Data are from the AAR, the National Transportation Safety Board, the Association of Oil Pipe Lines and the Pipeline & Hazardous Materials Safety Administration ("PHMSA").

Right off the bat, it's obvious that pipelines do a better job in terms of injuries and death. Using 2011 data, fatalities for five modern transportation modes are:

  • 32367 highway deaths
  • 800 marine deaths
  • 759 railroad deaths
  • 494 aviation deaths
  • 14 pipeline deaths

Pipeline fatalities are lower by an order of magnitude compared to other shipping modes. This is likely due to the fixed nature of pipelines, where the pipe and pumping equipment are stationary. All other shipping modes involve the motion of both heavy machinery and product through space, a complex system of many moving parts. It's a no-brainer that a lot that can go wrong in any complex system. In comparison, pipelines are much simpler since only the product moves, not the pipeline.

US pipelines transport approximately 11.3 billion barrels of petroleum per year. Approximately 52% of this is crude so a total of ~5.9 billion barrels of crude oil travel in pipelines.

The rail transport data from AAR was decadal so I converted it to an annual basis instead. Here's the converted AAR data:

  • Average annual barrels shipping by rail: 20,632,788.
  • Average annual barrels spilled by rail: 206.

As already mentioned, that spill rate probably underestimates the risk of spills from tanker cars.

The AAR also reported a spill volume for pipelines:

  • Average annual barrels spilled by pipelines: 43,131.

We can test the validity of the AAR spill amount for pipelines by comparing it to other reported estimates for pipeline spills. The database from the PHMSA website is set-up to report on all petroleum products in pipelines, including liquified gasses. Here's the average spilled by pipelines based on that data:

An average of 105,370 barrels of all pipeline products are spilled every year.

Assuming that 52% of those pipeline products are crude oil, then an average of 54,792 barrels of crude in pipelines is spilled every year.

Now we can calculate annual average spill rates on a dimensionless volume-only basis:


20 632 788 barrels/year shipped on trains     
206 barrels/year spilled
9.98 e -06 = fraction spilled
= 0.000 998 %


5 900 000 000 barrels/year in pipelines (52% of PHMSA volume)
43 131 barrels/year spilled from pipelines (AAR data)
7.31 e -06 = fraction spilled
= 0.000 731 %


5 900 000 000 barrels/year in pipelines (52% of PHMSA volume)
54 792 barrels/year spilled (52% of PHMSA volume)
9.29 e -06 = fraction spilled
= 0.000 929 %

There's very little difference between these numbers.

If the data were actually available for the miles that every barrel of crude traveled from oil field to refinery, the most rigorous way to calculate spill rates would be on a basis of per mile of actual travel, which is not the per mile basis from before which was calculated using the total fixed track mileage or pipeline mileage (both constants). But that data might be impossible to generate since you would need access to the shipping records of every railroad and pipeline. One can make a strong argument, however, that the number of miles traveled by crude on rail vs pipeline will not be significantly different. Why? Because the distances between oil fields and refineries will be about the same regardless of shipping mode. Those miles are not going to vary much.

The bottom line here is that the dimensionless volume-basis spill rate is about the same for both rail and pipeline. If this is really true, then we must look at other variables before one can say that pipelines are preferable to rail for moving crude. We already know that shipping by rail is around ten times more lethal than shipping by pipeline. What we don't know, and would like to know, is impact on the environment. We've looked at quantity of spills, but we haven't considered the "quality" of those spills. This is not a stupid question. A leaking pipeline is going to impact the environment differently than a spill during a train accident. Location for oil spills matters! A spill off a rail car into a river is a different beast than a leak from a buried pipeline that may dribble down into a potable aquifer. So given that the volume-basis rate of spills is approximately the same for pipelines and rail, the real question should be which transport mode impacts the environment the least.

I suspect the folks in Lac Megantic already have an opinion on that.

This blog post was editted on 30 July 2013 to improve readability.


Association of American Railroads, "Just the Facts – Railroads Safely Move Hazardous Materials, Including Crude Oil,", accessed 7 July 2013.

Association of Oil Pipe Lines (2013), "About Pipelines,", accessed 7 July 2013.

Christou, P. A. (2010), "Silhouettes of Rail Cars, Tank Trucks and Chemical Tanks," ttp://, accessed 7 July 2013.

Federal Railroad Administration Office of Safety Analysis (2013, July 5),, accessed 7 July 2013.

National Transportation Safety Board, "34,434 Transportation Fatalities In 2011, ", accessed 7 July 2013

Pipeline & Hazardous Materials Safety Administration, "All Reported Pipeline Incidents,", accessed 7 July 2013.