So there I was, quietly minding somebody else’s business, when I received notification that Railway Age had released a new Rail Group on Air podcast, “TRB Long Freight Train Consensus Study.”
The podcast featured a virtual panel of six members of the TRB special committee that produced the “Long Freight Trains: Ensuring Ensuring Safe Operations. Mitigating Adverse Impacts,” TRB Special Report 353. Also present was TRB’s Senior Program Officer. The panel was moderated by Bill Vantuono, Editor-in-Chief of Railway Age.
Not wasting a moment, Bill opened the discussion quoting the words of a Contributing Editor to RA:
I was disappointed in the committee’s report, which was not able to establish–per ton-mile, car-mile, train-mile, or other denominator common to all manifest (mixed as opposed to unit) freight trains—a causal connection between the frequency, the rate or the severity, of derailments and train length.
Imagine my surprise when I realized after repeated listening, that the Contributing Editor was me? It dawned on me, finally, that there was a reason the words sounded familiar. At the same time, I finally located my lost eyeglasses, which were exactly where I left them—on top of my head.
The virtual panel, however, was none too pleased. Peter Swan quickly picked up the gauntlet I didn’t realize I had thrown down:
Yeah, I think that was a very carefully crafted statement. First off, we did a statistical analysis and as such we don’t really ever prove causation in a statistical analysis. We show correlation or relationships between variables or events. So, he’s right in saying that we didn’t establish any causation. However, we did show a very significant link between the length of through trains, which includes manifest trains in intermodal trains and the increased rate of derailment of trains associated with train handling train makeup in FRA derailment causes that would be linked to that. And in that we showed that for all four of the largest U.S. railroads, there was a significant linkage between the size of their through trains and the rate of derailments. And by the way, the rate of derailments says how many derailments do you have? And we divide it by gross ton-miles. So, it’s also untrue to say that we didn’t link it to gross ton-miles or car-miles or train-miles; we actually did.
So, I would wonder how carefully he read the report to have missed that. I would also like to go on to say that the other two main critics of reports—the AAR, and an obscure journal, the American Journal of Transportation, that had blurbs that were very, very similar, both of which missed or ignored the major points of our paper and also the recommendations everybody’s talked about. They said you shouldn’t run trains over a certain length. We said nothing about that. What we said was, as you increase the length of trains, it’s very important that you put them together correctly or you risk increasing the in-train forces to such an extent. You’re going to get a derailment.
So there, Schanoes. And tell us, how closely did you read that report?
Well, I read it closely enough to assign the source of my disappointment not to the committee, but rather to AAR’s apparent failure to satisfy the committee’s request for additional details. Those additional details regarding train lengths are essential to determining if there existed specific train lengths where derailment frequency exceeds the portion of traffic assigned to trains of those particular lengths.
We all know that, particularly but not exclusively in the era of jointed rail and 40-foot boxcars, there was a critical range of speed, generally between 12 and 22 mph, that was likely to produce derailments due to “harmonic rock off.” Is there a parallel critical train length?
I assumed that if the committee’s brief was to investigate the impact of trains greater in length than 7,500 feet, then actual train lengths would be of such critical importance that no proxy for length would be allowed. I further assumed that those actual train lengths were the details that AAR was reluctant to provide the committee.
But now, I’m not so sure. Did the committee attempt to establish train lengths? Did it make that request to AAR or FRA or specific railroads? Were the requests, if made, rejected?
I am certain though, that “assume” still makes an ass out of you and me.
I read it closely enough to remember reading …
As a general matter, manifest trains create operational challenges due to the mixture of rail car types, designs, sizes, and weights. All railcars in a train are subject to longitudinal forces that create draft and buff load conditions and to lateral forces, especially at curves. These in-train forces can lead to broken equipment, including drawbars and couplers, and cause the wheels of a car to leave the rail when negotiating curves. The magnitude of these forces will differ among cars that vary in size and weight, and the movement and mitigation of the forces will differ among cars having different drawbars and coupling devices with or without cushioning units (pp. 1-2).
And:
In this regard, it is important to emphasize that current policy interests pertaining to long trains stem largely from recent trends within the rail industry to build and operate increasingly longer manifest trains, which haul a mix of freight in many different types of railcars (p.14).
I’m wondering how, given these variations, the committee could possibly accept a proxy so blind, insensitive to variation for actual train lengths as number of cars?
Why did/do I regard such “granularity” as essential? One reason is regulation. Congress wastes a lot of money and time but there’s a chance Congress wants to demonstrate its responsiveness to public complaints and use the study to prompt FRA to exercise rule-making authority in this matter. An intelligent regulation requires something more precise than an application based on average train lengths measured in numbers of cars rather than number of feet.
The committee produced the chart below consolidating a portion of its research into train length and the trend in derailments:
When I read the report the first time, I was struck by two things: 1) the near “flat-lining” of the rate of derailments at BNSF; 2) the accelerated rate of increase for these derailments on Norfolk Southern, a rate sharply higher than the rates attributed to CSX and Union Pacific.
The second time I read the report, I was struck by the same two things, but I had developed at least part of an explanation. Regarding NS, clearly what was driving the derailment rate could not be attributed to increased average train size, as the rates on NS were so much higher even when the average cars per train were just 60% of the current average.
And BNSF? I thought, “this is what happens when data isn’t filtered to remove the frequency of intermodal trains from the calculation.” The committee had justified the inclusion of intermodal trains:
For reasons explained in discussing that figure, the through-train data include intermodal trains in addition to manifest trains, but this aggregation should not be problematic because the inclusion of intermodal trains is likely to depress the calculated average number of cars per train (p. 28).
My thinking is a little bit different. If I’m taking data from four railroads and two of the four exhibit rates so extraordinarily different from the other two, I would be extremely reluctant to include those rates as evidence. I might even reject all the calculated rates until I could establish some better filters.
Then there’s the committee’s consideration of the relationship of train length to severity of derailments:
Research on train derailments caused by equipment and mechanical failures shows that the number of cars derailed is correlated to the length of the derailed train.Furthermore, the literature shows that the number of cars derailed is highly correlated with the likelihood of hazardous materials being released and other severe outcomes(pp. 25, 26).
Makes sense, right? More cars, more cars derailed. More cars, more likelihood of hazardous material containing cars, more likelihood of hazardous material cars derailing, more likelihood of derailed hazardous materials cars releasing their contents—except “making sense” isn’t the issue. Measurement of the actual damages from train derailments is the issue, and to make actual sense of that we use the (almost) universal standard of dollars. Money isn’t everything, but it’s the measure of cost. Measurement of derailment costs in relation to train lengths is essential.
The committee did produce this:
The UIUC (University of Illinois Urbana-Champaign) researchers examined the relative severity of derailments involving DP (Distributed Power) and non-DP trains. They found that the median train length and weight of DP trains was greater than that of non-DP trains. They also found that DP trains derail at a higher speed than non-DP trains. Based on these differences in train length and weight and speed at the time of derailment, one would expect that derailments involving DP trains will be more severe on average (i.e., derailing more cars). However, when the researchers compared the median number of cars derailed per derailment, they did not find a statistically significant difference in derailment severity among accidents involving DP and non-DP trains. These results suggest that, even though DP trains tend to be heavier and longer than non-DP trains, they may be helping to reduce derailment severity by helping to manage in-train forces and handling challenges of longer trains (pp. 45, 46).
The committee did not specify how the UIUC measured train length and/or if the method of, or resulting data, is proprietary.
Regarding the report’s sections on long-train impacts on Amtrak performance and grade-crossing obstruction, I read them but didn’t comment on them. The report relies upon impressions, opinions, the old anecdotal evidence. It is unclear if the committee examined Amtrak’s records of train delays or examined details from FRA’s database of complaints regarding blocked crossings. However, these two issues, freight train interference and blocked crossings, have plagued Amtrak performance and local motorists, respectively, since … as long as I can remember, and I’m old.
That the committee did not address these issues using direct records is not a failure of the committee, but emphasizes that without the actual measures, “quantities,” of the problems, producing and managing a program for improvement is simply not possible.
My thanks to Bill Vantuono for providing me with the transcript of the podcast, which I read as closely as I did the committee’s report.
David Schanoes is Principal of Ten90 Solutions LLC, a consulting firm he established upon retiring from MTA Metro-North Railroad in 2008. David began his railroad career in 1972 with the Chicago & North Western, as a brakeman in Chicago. He came to New York in 1977, working for Conrail’s New Jersey Division. David joined Metro-North in 1985. He has spent his entire career in operations, working his way up from brakeman to conductor, block operator, dispatcher, supervisor of train operations, trainmaster, superintendent, and deputy chief of field operations. “Better railroading is 10% planning plus 90% execution,” he says. “It’s simple math. Yet, we also know, or should know, that technology is no substitute for supervision, and supervision that doesn’t utilize technology isn’t going to do the job. That’s not so simple.”
