Saturday, December 5, 2015

The Coachella Valley draft Alternatives Analysis doesn't bode well

Regular intercity passenger rail from Los Angeles to the Coachella Valley has been a long studied affair, quite possibly the longest studied passenger rail service in the State. While the oft-delayed Coast Daylight has been studied since the beginning of the 1990s, studies for passenger rail service to the Coachella Valley has been studied since at least 1982 [Rail Passenger Development Plan 1984 Through 1989 Fiscal Years]. With a substantial number of studies previously conducted, Riverside County's latest one has arisen only to dash the hopes of those who might have wished for a better rail service to the Coachella Valley than the infrequent middle of the night Sunset Limited.

One would think that this rail service would be an obviously strong proposal. Such a line would connect Los Angeles not only with a large number of regular commuters, but also with an important tourism location and a developing urban region in its own right. Traffic is also regularly terrible and, as I begin to write this around 3pm on a Friday afternoon, Google Maps is projecting more than three hours from Los Angeles to Indio.  Indeed, among all of Metrolink's lines, the Riverside line has the highest occupancy rate, with an average of 264 passenger miles per train mile. But with all of these advantages going for it, the Alternatives Analysis paints a grim picture of the line's potential ridership and revenue, one so grim that it is highly unlikely to ever receive funding.

The preferred alternative identified in the analysis, Alternative 1, runs along Union Pacific's Yuma Subdivision for 72 miles before crossing onto BNSF's San Bernardino Subdivision by which it makes its way to Los Angeles by way of Fullerton. By running through Fullerton, it picks up quite a bit of additional traffic without an actual slow down compared to the other alternatives though the question is whether it's actually traffic that the Coachella Valley line is intended to address or whether it is Orange County-Los Angeles traffic that would be better taken care of by more Surfliners and Metrolink trains. I expect it's rather more of the latter.

Even though it picks up a substantial amount of additional traffic by essentially poaching from the Surfliner, the forecasted ridership is extremely low with only 189,100 forecasted for 2022. The trains themselves will be quite empty, averaging only 79 passenger miles per train mile. Only Amtrak's Piedmonts and Hoosier State average less than that (page 9) and of all the Surfliners, only 761/1761, which is the sole Los Angeles to San Luis Obispo train, has worse occupancy and the Surfliner as a whole averages nearly double this proposed train with an average of 150 passenger miles per train mile.

This low ridership is accompanied by low revenue, with only $3,245,000 projected for the line, which amounts to a farebox recovery of just 21.5% based on the Pacific Surfliner's O&M costs per train mile. Even if we were to calculate based on avoidable costs, that's less than a 50% cost recovery. There is no earthly way that this line will ever be funded based on such pitiful ridership and revenue levels. There simply isn't enough of a public benefit to justify that level of spending.

So what could be done to make it more reasonable and likely to be funded and to be more useful in general? The simplest answer is to run more trains. As the FRA noted back in 1978:

A rail passenger corridor must have certain minimum definable characteristics if it is to be seriously considered as an alternative to expanding highway and air capacity. The following characteristics are suggested:
Frequency: Minimum of 10 trains daily in each directionRationale: Most city-pairs subject to becoming corridors are receiving some level of service by long and/or short-haul trains. Such service is too often typified by infrequent and inconvenient departure times which are not amenable to the need of either the time or fare sensitive traveler. In general, frequency and scheduling should be structured to meet the demand of the traveler desiring to make a one-day round-trip with sufficient time at the point of destination. To provide an alternative to automobile transportation, and limited diversion from air travel, a minimum of five (5) a.m. and five (5) p.m. daylight departures should be scheduled. A variable in the p.m. departure schedule would be an evening train where demand demonstrates such a need. The recommended frequency considers the relative high frequency and volume of other modes of public transportation in suggested corridors. It also assumes that the fare sensitive/nonbusiness automobile traveler is secondary in importance to the higher income business trip traveler. This assumption, nevertheless should not detract from the benefits received by the fare-sensitive passenger and other rider. The suggested frequency represents a minimum level of service for an emerging corridor and is keyed to the frequency level in the Northeast Corridor at which significant diversion began to occur.

[A Reexamination of the Amtrak Route Structure, page 7-26, emphasis added]

This would require, of course, that the study abandon what appears to be a frequency level based on a preordained conclusion to use the San Bernardino Subdivision. It's quite interesting how four trains per day just happens to line up quite nicely with Riverside County's four revenue slots between the Colton crossing and Riverside. Make no mistake, this would result in a significant capital expense. Union Pacific may very well want an additional track along the entire Yuma Subdivision to Indio (~$720 million) and another track may very well be needed between Colton and Fullerton (~$420 million). This would also consume almost all of the slots created by the Los Angeles-Fullerton triple tracking project, though the shared use passenger corridor currently projected by the California High Speed Rail Authority should create sufficient additional capacity for Surfliner and Metrolink expansions.

Given the low ridership per train currently projected, using a consist like the Surfliner is overkill: The current projections wouldn't even fill up a single coach car! With appropriate waivers, a DMU such as NCTD's Sprinters, but with an interior designed for intercity travel, would be much more appropriate and offer a lower cost travel method. SPRINTER's operating cost comes out to about $25 per married pair train mile (Quarterly Ridership and Financial Performance Update, page 10), about one-third that of a Surfliner's fully allocated cost. Assuming that these figures are reasonably apples to apples comparisons, this would either lower operating costs to a reasonably supportable level or allow for a substantially more frequent and useful service without breaking the bank in the process.

As it stands, this alternatives analysis is just one more in a decades long legacy of studies which never go anywhere. Coachella Valley rail service will remain just a regional aspiration until and unless planners take it seriously and put forward plans for a genuinely useful rail line. Of course, at that point, one might as well simply wait for, or pay for the acceleration of, California's Phase II high speed rail, which will include a stop in the Riverside area. This would serve the Riverside and San Bernardino region much better than the Coachella Valley rail line would while frequent bus or rail shuttle service would still provide a useful two seat ride to the Coachella Valley.

Saturday, September 5, 2015

A cost to benefit analysis of railroad electrification

The FRA recently released a technical report entitled Cost Benefit Analysis of Rail Electrification for Next Generation Freight and Passenger Rail Transportation, created as part of the recent, and terrible, Keystone West corridor study. Aside from generally being a whitewash of a preordained outcome, it contains an error so blatantly egregious that I have to question how it ever made it past any form of peer review.

4.2.2 Energy Costs
In order to determine the energy cost differences, the team estimated fuel costs per mile for both the build and no-build options. In the build option, trains would consume electricity; in the no- build option, trains would continue to consume diesel fuel. For both options, the team estimated the costs of 10 trains traveling a 75-mile round trip every day. For diesel-powered locomotives, it was assumed that the rate of consumption was 1 gallon of diesel per 3 miles traveled. For electricity-powered trains, the assumed rate of electricity consumption was 0.03 miles per kWh.

Three miles per gallon is an inversion of reality and likely stems from confusion of switching units back and forth given that they give an electrical consumption in odd form of miles per kWh. In reality, Amtrak quotes 2.3 gallons per train-mile, Metrolink gets 2.6 gallons per mile, and the Class 222 Meridian gets about 2 gallons per mile. So rather than $1.27 per train-mile, at their assumed cost of $3.80 per gallon, it would be a cost of $7.60-9.88 per mile, far above the cost of electric propulsion.

For that matter, the electricity consumption figures are on the higher end. The British Class 390 Pendolino and IC225 consume about 22 kWh per mile, though it must be admitted that the double length Class 373 consumes 66 kWh per mile and that 31-33 kWh is a good approximation for a 200m high speed train. This analysis was for a high speed alignment, which would make the high speed train's figures appropriate, but I'm not overly fond of apples to oranges comparisons.

An interesting peculiarity of this paper is that at least two of its citations, footnotes 18 and 19 from page 20 where this section is found, do not at all state what this report claims to use them for. Indeed, it's actually impossible to derive the energy consumption figures that they claim to as Amtrak's reported figures are total train-miles, total diesel consumption, and total electricity consumption: There is no distinction of diesel train-miles from electric train-miles.

So let's try our hand at creating a cost to benefit analysis of railroad electrification, specifically of existing corridors without any new high speed alignments. The costs, as provided by this analysis, are fairly reasonable and I don't have any quibble with them. These amount to $12,100 per route-mile in annual maintenance costs and $4.3 million in construction costs per route-mile or $176,515 annually over 30 years with a 1.4% discount rate for a total annual cost of $188,615.

Per the EPA, Tier IV locomotives have the following emission factors in terms of grams per brake horsepower per hour:
PM10
0.015
HC
0.04
NOx
1.00
CO
1.28

With a conversion factor of 20.8 to grams per gallon for large line-haul and passenger trains we have the following emissions per gallon:
PM10
0.312
HC
0.832
NOx
20.80
CO
26.624
Per route-mile, each bidirectional frequency (one train in each direction) has this consumption:

Class 222
1,460
Amtrak
1,679
Metrolink
1,898
Using the median 2014 South Coast Emission Reduction Offset transaction costs as a proxy for the social costs of diesel emissions, and therefore the social benefit of eliminating them with electrification (assuming 100% clean energy), gives us this:


Class 222 Amtrak Metrolink Value per ton (Median 2014 South Coast) Value per gram
PM10
455.520
523.848
592.176
$520,548
$0.574
HC
1,214.720
1,396.928
1,579.136
$24,658
$0.027
NOx
30,368
34,923.2
39,478.4
$63,014
$0.069
CO
38,871.040
44,701.696
50,532.352
$5,479
$0.006






Cost per frequency
$2,638.51
$3,034.29
$3,430.06


Considering just the social benefits, maintenance is paid for with 3.5-4.5 daily frequencies. This alone does not justify electrification in most corridors, however, as it would require 55 frequencies, five times the current level of the Pacific Surfliner, to justify solely on environmental grounds. As we established earlier, however, electric propulsion is quite a bit cheaper than diesel propulsion; for a corridor intercity or commuter train with consumption of 22 kWh per train-mile, we can expect a cost of $2.42 per train-mile, $5.18-7.46 cheaper than diesel.

Again, maintenance is very quickly paid for: 2-3 frequencies will pay for the maintenance, but construction costs require rather more: 32-47 frequencies. Combined, we see that it takes 21-29 bidirectional frequencies for benefits to match the costs of railroad electrification.


Frequencies to break even Social benefit per frequency Cost savings per frequency
Class 222
29
$2,638.51
$3,781.40
Amtrak
25
$3,034.29
$4,613.60
Metrolink
21
$3,430.06
$5,445.80
In California, this would indicate that it would be justified to electrify Caltrain between San Jose and San Francisco. With increased service, electrification would also be justified on Metrolink's San Bernardino Line as well as LOSSAN between Burbank and Irvine (Metrolink and Pacific Surfliner) and Oceanside and San Diego (Coaster and Pacific Surfliner).

For freight trains, the decreased fuel costs play a much larger role, and more importantly, the only one that the board of directors actually care about, resulting in break even at fewer frequencies. From the 2014 STB R-1 reports, we see that, for the Class I railroads, there is an average consumption of 6.92 gallons per train-mile; a comparable figure for electric traction would be 86.5 kWh per train-mile. Because of the significantly greater fuel consumption, the pay off is much quicker: Only 9 trains per day are needed in each direction with social benefits included or 15.4 when only considering fuel costs.

Of course, private companies aren't going to be using Federal discount rates and will likely be seeking money on the open market. While this will be more expensive, it won't be enormously so. Union Pacific recently sold 40 year bonds at 3.875%; if I've done the math correctly, this would come out to $212,374 per mile of track, pushing the break even points to 10 and 17.3 frequencies. In Southern California, this would justify the electrification of the Alameda Corridor, Sunset Corridor, and Southern Transcon (slide 9).

I've ignored a few costs so far. One is that of locomotive maintenance, which is, unsurprisingly given the rather shoddy nature of the study, completely unmentioned in the original study. According to EMD, maintenance costs for electric locomotives are 30% cheaper than those for diesel-electric locomotives.

Siemens, meanwhile, quotes a range of 30-70% less maintenance cost for electrics compared to diesels.
Referring once again to the R-1 reports, we see that there are 2.63 locomotive unit miles for every train-mile, or to put it another way, the average train has 2.63 locomotives, and that it costs $1.65 per locomotive unit mile, which comes out to $4.34 per train-mile. At the low end, electrification saves $1.30 per train-mile for every freight train that passes through and that's without any consideration of the potential need for fewer locomotives with electric trains. For commuter and intercity trains, assuming that they have similar maintenance costs to freight trains, the savings would amount to $1-2 per frequency depending on whether they have a single locomotive or are double ended, such as the British IC125 or All Aboard Florida's new trains will be. This reduced maintenance cost doesn't amount to overly much, only $730-1,898 per frequency per year for every electrified mile, but on busy corridors it can quickly add up. Just over six freight trains per day in each direction will offset the maintenance costs of the electrification for instance and sixteen would account for the maintenance costs on commuter and intercity lines.

But what of the price of the locomotives? The original study quotes a price of $7 million for Amtrak's new ACS-64, a price of $2.6 million for diesel locomotives and proclaims itself happy: The comedy of errors continues. In reality, Amtrak is paying $6.66 million per locomotive and $2.6 million for a diesel passenger locomotive is simply laughable. Amtrak may have paid $2.4 million for the P-42s back in 1996, but inflation's been a touch more than $200,000 over the last 20 years and Metrolink is currently purchasing EMD's F125 for $6.84 million each, including 20 bought at an option price of $6.295 million.  For intercity trains, Siemen's Charger, purchased by several states and All Aboard Florida, is coming in at $7 million each, the same price the analysis was quoting for an electric locomotive. If anything, based on actual prices paid, it would be cheaper to choose electric locomotives. It must be admitted, however, that SEPTA is paying significantly more, $8.6 million per locomotive, for their ACS-64s. This may be the result of only a single bidder (thanks to Buy American requirements) and purchasing of spare parts that were not purchased by Amtrak as part of their ACS-64 contract.

For Tier 4 freight locomotives, I have yet to come across anything giving an actual number. California's Air Resource Board estimates $3 million per locomotive, which is a fairly significant increase over the $2.3 million per unit that Florida East Coast spent acquiring its ES44C4s. While $3 million per unit is cheaper than $4.3 million for Bombardier TRAXX or $5.125 million for Siemens Vectron, it must be kept in mind that the electrics also have significantly more power: 7,500 horsepower for the TRAXX and 8,600 horsepower for the Vectron against the 4,400 horsepower of the ES44C4 or EF44AC (the Tier 4 replacement). Additionally, these are significantly faster locomotives: American diesel-electric locomotives are typically limited to about 75mph while the TRAXX is rated for 87mph and the Vectron, in its standard configuration, is built for 100mph; if the speed is not considered worthwhile, it's likely that they could be built for a lesser price. Be that as it may, it's worth considering that, for horsepower limited trains, a Vectron can replace diesels on a 2:1 basis and a TRAXX can replace them on a 3:2 basis. In such case, the electrics are cheaper to purchase than the corresponding number of Tier 4 diesels, by about $500,000-1,000,000 dollars and will save more than a hundred thousand dollars a year in reduced maintenance costs simply from requiring fewer locomotives, before even consideration of the lowered maintenance costs of the electrics.

There are three final issues which negatively impact the costs and benefits of electrification. The first is that of clearing sufficient overhead to install railroad electrification: Not every place is the wide open plains and tunnels, overhead passes, or other environmental issues may result in substantially higher costs for particular stretches of track. This must be evaluated, however, on a case by case basis, rather than simply trying to throw in a blank figure for every bit of electrification that may be done.

Second is the fact that oil prices, and with them the cost of diesel, have plummeted since the original study was created. While I have used the original numbers, and I fully expect the price of oil to return to its highs within a decade, the financial benefit from lowered fuel costs is currently significantly lower than what I've gone through above using the original numbers.

Lastly, there is the issue of operational flexibility, to which I cannot attach a number. Electric locomotives must stay on electrified track and cannot be detached for duty elsewhere unless, as some versions have, they possess an auxiliary diesel engine, at the cost of reduced range and speed. The increasing use of distributed power by freight trains means that switching between diesel and electric power can be a costly and time consuming affair. Again, this is not something I can quantify financially and would need to be evaluated, by professionals, on a line by line basis.

 On the other hand, this does not also evaluate the increased speed, acceleration, and frequencies which electrification may make possible and which would make train service more attractive than other modes. Frequently these are the overriding benefits which decide the case for commuter and intercity rail electrification but, at the risk of growing repetitious, these would need evaluation on an individual basis.

Tuesday, August 11, 2015

New auto plant in Moreno Valley grows the case for reverse commute and Coachella Valley service

With the announcement that Fisker Automotive will be building a factory for a new line of electric vehicles in Moreno Valley, it's time to acknowledge that the Inland Empire is becoming a worthy urban area in its own right rather than merely a suburban home for workers in Los Angeles and Orange Counties.

This is hardly an unexpected development of course: It's not merely residential land prices that have been rising in the more coastal areas of Southern California, but industrial and commercial as well. Meanwhile, commutes on the 91 best described as a hellish abomination, dropping to an average speed of 14-22 mph. It's only natural then, that, that companies are increasingly taking advantage of the lower land prices and and lower wages that need to be paid to employees (on account of the lowered housing prices) and turning the Inland Empire into a jobs center.

As the Inland Empire becomes more and more important, it behooves California, Riverside County, and LOSSAN to recognize this with appropriate rail services. Frequencies upon the Riverside and 91 Lines need to be both increased and have reverse commute service added and added urgency should be given to make the Coachella Valley service, which has been languishing in study hell for the past 30 years, an actual reality. This particular plant isn't a justification in and of itself for such service: It's not ideally located for train service, either commuter or intercity, and I don't expect Fisker to become a major player. It is, however, an important symbol of the increasing importance of the Inland Empire which deserves such service.

Thursday, July 9, 2015

Pacific Surfliner Ridership Update

At the upcoming 7/9/2015 Technical Advisory Committee meeting, LOSSAN is providing an update on the Pacific Surfliner's ridership and for those of us with an interest in some of the nuts and bolts of the data, it's like Christmas in July. Unfortunately it doesn't have all the information one might want; the station-pair information is vague rather than specific as LOSSAN used to receive a decade ago and as the Capitol Corridor is currently releasing, but it does continue great information nonetheless; in particular, the average monthly ridership per train, broken down into coach and business, is quite possibly unprecedented.

It's a bit unfortunate that the ridership doesn't show the values, but luckily it's possible to extract it. I've also included a daily ridership figure in order to better show off 761/1761 and 790/1790 which are the only two frequencies which receive separate numbers for running on weekdays and weekends.  I'm a bit mystified, honestly, by that, since there are other trains which have separate weekend timings, but don't receive a different number for it. For the purpose of this, I've assumed an average month has 30 days, 8 of which are weekends.

TrainTotal ridershipBusiness classCoachBusiness %Daily ridership
562
6,474
433
6,041
6.7%
216
564
5,491
642
4,849
11.7%
183
565
6,439
749
5,690
11.6%
215
566
5,807
784
5,023
13.5%
194
567
7,177
749
6,428
10.4%
239
572
6,088
853
5,235
14.0%
203
573
6,826
783
6,043
11.5%
228
579
5,666
819
4,847
14.5%
189
580
9,423
1,346
8,077
14.3%
314
582
10,268
1,065
9,202
10.4%
342
583
6,896
1,065
5,831
15.4%
230
591
8,547
995
7,552
11.6%
285
595
4,718
328
4,390
6.9%
157
761
3,383
256
3,128
7.6%
154
763
14,342
1,908
12,434
13.3%
478
768
12,797
1,768
11,029
13.8%
427
769
12,761
2,049
10,713
16.1%
425
774
14,483
2,505
11,977
17.3%
483
777
13,218
2,154
11,064
16.3%
441
784
19,751
2,189
17,562
11.1%
658
785
16,274
2,152
14,121
13.2%
542
790
9,987
1,276
8,711
12.8%
454
796
8,090
714
7,376
8.8%
270
1761
1,452
152
1,300
10.5%
181
1790
4,894
609
4,285
12.4%
612
761/1761
4,835
408
4,427
8.4%
161
790/1790
14,880
1,885
12,996
12.7%
496

I've also redone the chart to join the weekday and weekend trains, making it quite apparent that the 790/1790 pair performs significantly better than the original chart makes it appear.

From the passenger miles per train, we can also determine how far the typical journey is per train and how productive each train is in terms of passenger miles per train mile.


And, of course, here's the information for that in spreadsheet form:

Train Total ridership Passenger miles (000s) Average trip length Passenger miles/train mile
562
6,474
363.1
56.1
95
564
5,491
386.2
70.3
101
565
6,439
461.2
71.6
120
566
5,807
438.1
75.4
114
567
7,177
458.3
63.9
119
572
6,088
472.8
77.7
123
573
6,826
481.4
70.5
125
579
5,666
424.9
75.0
111
580
9,423
694.9
73.7
181
582
10,268
657.4
64.0
171
583
6,896
539.1
78.2
140
591
8,547
720.9
84.3
188
595
4,718
317.0
67.2
83
761
3,383
293.9
86.9
60
763
14,342
1,176.7
82.0
163
768
12,797
1,119.0
87.4
155
769
12,761
1,292.1
101.3
179
774
14,483
1,759.5
121.5
168
777
13,218
1,574.9
119.1
150
784
19,751
1,603.7
81.2
222
785
16,274
1,537.2
94.5
213
790
9,987
1,009.1
101.0
131
796
8,090
665.9
82.3
92
1761
1,452
152.5
105.1
86
1790
4,894
536.2
109.6
192
761/1761
4,835
446
92.3
67
790/1790
14,880
1,545
103.8
147

So what are some conclusions we can draw from this?

First and foremost, 761 (and the weekend 1761) are absolutely abysmal. This is the current northbound Coast Daylight slot and one hopes that there would be a significant increase in traffic when it is actually extended to San Jose or San Francisco as the Coast Daylight (eventually, one day, quite possibly just a few minutes before the universe dies of heat death). That's not necessarily an unreasonable proposition, 790/1790 is the southbound counterpart and is an above average performer. It's probably a case of leaving Los Angeles too early in the morning without a correspondingly significant destination to arrive at.

Amtrak has been suggesting the past year that the Coast Daylight would be better off running from San Diego to San Jose rather than Los Angeles to San Francisco. At the same time, LOSSAN has been looking at an early morning from San Diego using existing equipment. It seems pretty obvious that 761/1761 should be extended to be an early morning (approximately 5:00 AM) departure from San Diego, picking up commuter traffic and possibly some early morning business and theme park travelers, before continuing on its current low performing route. I honestly don't expect it to pick up too much additional Amtrak traffic, the Surfliner's primary market being leisure travel, though Metrolink monthly pass holders would likely be quite pleased with it, but it's worth the experiment: As it is, with the current Surfliner average yield of 28¢ per passenger-mile, and an estimated $35 per train-mile in marginal operating expenses, 761/1761 currently covers about half of its running costs, possibly less as it is significantly below the average for business class use as well. The first three departures from San Diego are above or close to break even and the first departure, which also heads continues north of Los Angeles, is the best performing of them. If it averages the same as the current 500-series trains (which run only between San Diego and Los Angeles), which is just over break-even under the above assumptions, it would mean an additional 82,900 annual riders and $1.67 million in revenue for the Surfliner.

There's also good reason to believe that continuing to extend the current 500-series trains north of Los Angeles will result in improved ridership and revenue. These full (or "fuller," depending on how you wish to consider San Luis Obispo as part of the Surfliner) length runs are 29% more productive, in terms of passenger-miles per train-mile, than are the 500-series trains, and have average trip lengths 25% longer, something that shouldn't be too surprising since the corridor is nearly doubled by the extension to Santa Barbara. LOSSAN should continue to improve the locally owned portions of the line north of Los Angeles, terminating at Chatsworth with a new layover facility for as many trains as possible if need be, while pushing for improvements by Union Pacific to the Santa Barbara Subdivision.

Lastly, I have a chart created on a bit of a whimsy that I'm not sure actually shows anything terribly significant except that the first and last trains of a day underperform (which may undercut the rationale for a 5:00 AM San Diego departure) and possibly that people want to take a train around meal times. With Los Angeles as the top station for the Surfliner, and being the origin or destination for 21% of passengers, I looked at the productivity (in terms of passenger-miles per train-mile) for trains according to when the train arrived or departed Los Angeles. It would be better, of course, to have the information as to how many passengers are actually on board a train during any given hour of the day, but I'm not sure even Amtrak has that information, at least not without a rather laborious process of collection.