Gold Line Foothill 2B on the Cheap

In my post on the virtues of LA’s LRT system, I noted that from a network perspective, you want to avoid building anticipatory lines and start with where you already have demand. The Gold Line Foothill Extension Phase 2B is a little weak in that regard; the northern edges of the San Gabriel Valley are lower density (by LA standards, though still higher than most US suburbs). In theory you could allow a lot of new development there by upzoning, but who knows if it will happen? Indeed, the project has many features that should give you pause: extension for the sake of extension, lower density areas, shared ROW and expensive infrastructure (long flyovers at Lone Hill Av and Towne Av) to accommodate low-volume freight without upsetting FRA bureaucrats.

As a result of this, the project has been taking some licks on Twitter recently from folks like Erik Griswold and Market Urbanism. The momentum behind Phase 2B has been compared unfavorably to the lack of double-tracking and other improvements to Metrolink’s San Bernardino Line. However, between LA Union Station and La Verne, these routes don’t serve the same places. The obvious advantage of Metrolink is that it will (should) be a much faster ride from Pomona to LA Union Station, serving the middle of the San Gabriel Valley (Covina, Baldwin Park, and El Monte). The Gold Line serves Pasadena, a regional node (and regional nodes are more important in LA than other cities), along with Northeast LA and the northern fringe of the San Gabriel Valley (Arcadia, Monrovia, Duarte, Irwindale, Azusa, Glendora,  and San Dimas).

So it seems like there is still value in a transit service along the corridor. However, since the nature and density of development aren’t right for full-blown LRT, we should consider a much more cost effective option. This would be sort of like the old “interurbans”, borrowing from shared infrastructure best practices in places like Karlsruhe.

I’d split the project into three sections:

  • Azusa/Citrus to Glendora (1.1 miles): this section would remain typical double-track LRT, and run on the same headways as the existing Gold Line.
  • Glendora to La Verne (5.8 miles): this section would be single-track with a single passing siding at San Dimas, and run on 15 minute peak headways.
  • La Verne to Montclair (5.3 miles): this section would operate over the existing Metrolink San Bernardino Line, which is presently all double-track, and run on 15 minute peak headways.

Note that throughout, I’m assuming the existing freight track and grade crossings between Azusa and La Verne are unsuitable for conversion to LRT operations, and must be replaced entirely. This may or may not be true.

Azusa/Citrus to Glendora

I’m a little more optimistic about the Gold Line to Azusa than some people. A huge new master-planned community, Rosedale Azusa, is under construction on the north side of Azusa/Citrus Station, and it will have a ton of apartments and townhomes on the part of the site closest to the station. Across the city line, Glendora just annexed some adjacent vacant land, suggesting that someone may have development plans in store there as well.

In terms of the development along the corridor, Glendora is pretty similar to Azusa. Therefore, I’d keep this section double-track, all at grade except for the diagonal crossing of the intersection of Grand and Foothill. The Glendora Station should be located at this intersection, which means it would have to be an elevated station. The reason that the station should be here, instead of at Vermont, is that this location facilitates a transfer to transit on Grand, one of the main north-south arterials in this part of the San Gabriel Valley. (I should add a rule to my network design principles: never build infrastructure that screws up an obvious future line.)

The freight track that’s been maintained from Azusa/Citrus west as part of Phase 2A would tie in to the LRT track just east of Azusa/Citrus. An at-grade runaround track would be provided at Glendora Station, since the viaduct grades would be too steep for freight.


This section would require track, signals, OCS, and traction power infrastructure similar to what’s proposed under the current Phase 2B plan.


Ok, so far all I’ve done is increase the cost of the project. Fortunately, we’ll more than make up for it on the rest of the corridor.

Glendora to La Verne

At almost 6 miles but with only one station, this portion runs through less intense development in Glendora, San Dimas, and La Verne.

This section should be built as single track, with the track located so that constructing a future second track would be easy. Since vehicle malfunctions most often occur at stations, the logical place for the sole siding is at San Dimas station. This station would be a center platform with pedestrian at-grade access, located just east of San Dimas Av. With a critical single-track section of just less than 5 miles, if we assume an average operating speed of 50 mph between Glendora and San Dimas, it will take a train 6 minutes to traverse this section. With 6 minutes eastbound, 6 minutes westbound, and some schedule pad, the target 15 minute headway is achievable.

We should also eliminate the dedicated freight track on this section. Freight can run at night or during midday on the same track as the LRT trains. The current design for this section has two long, costly flyovers, one each at Lone Hill Av and Towne Av, so that the exclusive freight track can run on the north side of the ROW and service customers on the north side of the track between those streets. The freight track runs on the south side in the rest of the project. It is illogical to spend so much money on exclusive track and grade separations for a freight branch currently serving one train per day, with little prospect for growth.

Significant cost savings can be achieved just through building only one shared track instead of two LRT tracks and one freight track, but there are additional savings that could be had in systems (a nebulous term that refers to signals, OCS, and traction power).

LA’s existing LRT network runs on fixed-block cab signals. Current LACMTA design criteria specify that the LRT branch lines (i.e. everything except the future Regional Connector and the existing Blue/Expo between 7th/Flower and the Flower/Washington junction) must be able to support 3 minute headways. Minimum headways are a major driver of costs in signal systems, because shorter headways require shorter blocks, which means more signal locations and all their expensive equipment. If we’re going for 15 minute headways, designing the signals for 3 minutes is overkill, and we could go with much longer blocks. Really, there’s no need to have any signal blocks between Glendora and La Verne, other than at the San Dimas siding.

We could take this a step further and say that we don’t really need cab signals – at all. Even the FRA (more on them later) allows you to run 59 mph with no signals and 79 mph without cab signals, speeds well above what would be necessary for this service. For the purposes of cost estimating, let’s say we have cab signals. I’m not willing to accept the risk of a train overrunning a signal at San Dimas, resulting in two trains heading towards each other on single track, with neither train being warned of the impending collision. Since all trains will presumably be stopping at San Dimas, we can put out a fixed approach signal heading into the station in each direction. On the five mile section between Glendora and San Dimas, we’ll need a few cut sections, which should be placed such that they don’t interfere with logical future infill stations, say at Loraine and Lone Hill.



The final cost savings we can squeeze out of this section is using spring switches instead of power machines. The switch is set up so that a spring holds the switch in the correct position for facing moves (trains that are heading towards the switch points, from the single-track to the double-track). For trailing moves (trains heading from the double-track to the single-track), the weight of the train pushes the switch points into the correct position, compressing the spring. Thanks to some railfans in Virginia, we have confirmation that spring switches are used on mainline LRT operations.

For traction power, we could save money by building fewer or smaller substations, since we won’t need as much juice to support 15 minute headways. The next step would be to consider eliminating electrification altogether, which I’ll address separately. For cost estimating, assume the line is electrified.


La Verne to Montclair

Between La Verne and Montclair, we already have two tracks. Headways on the San Bernardino Line are constrained by a long single-track section that’s stranded in the middle of the 10 between the 710 and El Monte. Fixing that is probably a billion dollar project, which won’t be happening anytime soon. So let’s make the most this double-track capacity and just run the Gold Line trains on these tracks.

Since we’re saving money elsewhere, why not make this project benefit Metrolink too? Today, the SB Line bumps up against Arrow Hwy just east of San Dimas Canyon, just across the street from the Gold Line corridor. The SB Line then curves south before meeting up with the Gold Line ROW east of White Av, after a 40 mph reverse curve that crosses Arrow Hwy at grade. The fix would be to swing the SB Line north onto the Gold Line ROW east of San Dimas Canyon. Arrow Hwy could fly over the SB Line on a new highway bridge; there are no intersections or driveways to preclude the bridge, and there’s plenty of ROW. I’m showing this as an optional cost. The Gold Line would merge with the SB Line here at a new interlocking.

The combined single-track railroad would cross Wheeler, A, D, E, and White at-grade, merging in with the existing double track section of the SB Line east of White. La Verne Station would be a center platform just west of E.


At Pomona, Claremont, and Montclair, new high level platforms would be needed for the Gold Line. At Montclair, a new siding track would be built on the north side of the ROW so that Gold Line trains could layover at the station without interfering with Metrolink operations. Fortunately, there’s an existing interlocking just across Monte Vista, so the siding track could be added to that interlocking rather than being an entirely new interlocking. The westbound home signal on the northernmost track would need to be relocated.


Again, for the purposes of cost estimating, assume that the line is electrified and uses cab signaling. I’ll explain electrification and signaling concerns like PTC in more detail below.



Avoiding the cost of electrification infrastructure is another way to save capital costs. The service could be operated with DMUs from Montclair to Glendora, with passengers transferring to the Gold Line there. The disadvantages are significant: forced transfers, introduction of new type of rolling stock, captive fleet, inability to interline with the Gold Line, and inability to change service patterns without incurring major capital costs.

There are also disadvantages to electrification: freight limitations due to wire height, and the introduction of traction return current on the SB Line between La Verne and Montclair.

To me, the advantages of electrification outweigh the drawbacks. It is undesirable to have a captive fleet of a new type of rolling stock, especially such a small fleet, and the transfer is a major inconvenience to passengers. Finally, it is bad form to lock in service patterns with hard infrastructure constraints.

Meanwhile, the disadvantages of electrification are relatively easy to address. It is possible to construct the OCS to be high enough to accommodate double stack freight clearances (22’-6”). For example, UTA’s Trax LRT system is set with a wire height of approximately 22’, which would allow for double-stack freight. I’m not sure if LACMTA’s existing LRT vehicles have pantographs capable of operating on wires that high, but it would be a reasonable assumption.

At any rate, between Glendora and La Verne, we’re talking about a minor dead-end freight branch line with low volume, all of which is local traffic. BNSF isn’t about to start running double-stack unit intermodal trains here like they do on the 91 Line. Between La Verne and Montclair, there’s probably a little more freight, but UP is running the majority of its traffic on the Los Angeles Sub and the Alhambra Sub. There aren’t any active customers between La Verne and Montclair, so any oversize items could be brought in from either the west via the Alhambra Sub or the east via the San Gabriel Sub. For the freight service that’s going to be operating here for the foreseeable future, 19’ of clearance would be just fine.

The existing signal system on the SB Line between La Verne and Montclair is probably not compatible with traction return current in the rails. The existing system most likely uses electronic track circuits, which use coded pulses of electricity in the rails to communicate information about track occupancy and signal aspects to adjacent signal locations. This system can be modified by adding equipment that uses a modulating frequency to distinguish the signal information from traction return and cab signal indications for trains. Cross bonds at signal block boundaries would also be required.

Platform Height, Length, and Width

Current LACMTA design standards call for LRT platforms to be built for level boarding (3.25’ above top of rail), long enough to accommodate three-car trains. Level boarding is absolutely the way to go, for legal reasons (ADA), moral reasons (equality of access), and practical reasons (boarding/alighting efficiency & vehicle interoperability).

However, three-car trains are probably overkill for service on this line, so I’m going with one-car trains. This does save a little money on construction costs. It should go without saying that space should be left to extend the platforms to three-car length in the future. If you want to get really chintzy, you could build wooden platforms, but for the sake of argument, I’m going to assume typical concrete platforms.

Level boarding platforms that comply with ADA requirements on the gap between an LRT vehicle would have a platform edge 54.77” (4’ 6.77”) off track center per LACMTA standards. These platforms are therefore not compatible with normal mainline passenger or freight equipment, which is up to 5’ 4” off track center. For Glendora and Montclair, which would be exclusive Gold Line use, that’s not an issue. For San Dimas Station, where freight would likely be run at night, the platform could be built with flip-up edges to allow passage of freight.

For La Verne, Pomona, and Claremont, that wouldn’t be acceptable, because Metrolink equipment needs to run at the same time as Gold Line equipment. These platforms could be designed with 5’ 8” clearance and platform extenders, similar to those that NYC MTA had at the old South Ferry station. Mechanical components are undesirable, especially at open stations, because they require diligent maintenance, but I don’t see a way around this that wouldn’t have high capital costs, like gauntlet tracks. (And really, it’s a symptom of how spoiled we are in the US that people would argue for millions of dollars in capital costs to avoid having to do regular maintenance.) I’ve included an allowance for platform extenders in the costs.

Service Pattern

By going with one-car trains, I’ve tipped off that I don’t plan on making the service a straight extension of some Gold Line trains. I propose that the new service run one-car trains between Montclair and Pasadena. An offsetting number of three-car Gold Line trains would short turn at Pasadena so that the number of trains on the line between Pasadena and Montclair would stay the same.

service pattern

The short turn three-car trains would turn at Sierra Madre Villa, where there’s already an interlocking just west of the station. To expedite the short turns, home signals would be added at the east end of the platform. The one-car trains would turn just south of Fillmore, where an interlocking and short new siding would be constructed at what appears to be a previous siding location. The overlap complicates operations, but it’s logical given that it results in all short turn trains serving all Pasadena stations. I’m assigning a capital cost here of $3m ($2m for signals, $500k for track, $500k for OCS & miscellaneous crap).


A more ideal setup would be as I’ve sketched below, because it would allow the short turn trains to diverge and merge without interfering with opposing traffic. There may be space for this if you do strip ROW takings to the north and south, but I’m not going to make that assumption.


A rapid transit service that has some vehicles turn back before the end of the route is more operationally challenging than one where all trains run to the end of the line. You’re trying to insert the short turn trains back into traffic quickly enough that they don’t delay following outbound trains, but without screwing up the inbound headways either. A scheme with all three-car trains, and some turning back at Glendora, would have one such location (Glendora). The scheme proposed above has two such locations (Glendora and Pasadena) and is therefore a little trickier.

However, it offers considerable savings on the capital cost side because it allows shorter platforms and reduces the size of the vehicle fleet needed to operate the service. It also reduces traction power loads, which reduces the costs of substations. There are operations cost savings as well, since a one-car service will use less electricity. (The operations costs won’t be 1/3 of a three-car train, since much of the cost is the labor of the driver.)

Not to short change the operational challenge of running this type of service pattern, but if the MBTA can make the Green Line work, with its 90-second headways, multiple turnback locations, and merges from branch lines that aren’t even under central control, we can make this work.

Regulation Protestation

You didn’t think I would just come out and propose some crazy Euro scheme without at least considering the US-specific regulations that have precluded this type of service, did you?

Freight Clearances

First of all, on the freight side, if UP has retained final say over freight clearances on the SB Line, that is a potential fatal flaw, because it would be virtually impossible to overrule them. Getting CPUC to approve the clearances might be challenging enough. The cost of building gauntlet tracks at La Verne, Pomona, and Claremont would probably exceed what could be justified by expected ridership. For the sake of argument, let’s assume that flip-up platform edges are acceptable.

PTC and Cab Signals

We’re getting to some genuine weirdness here. But let’s think everything through.

  • LRT trains won’t interface with the PTC system. But they’ll be forced to stop by the cab signals, so they won’t rear end each other or Metrolink trains.
  • Metrolink trains won’t interface with the cab signals. But they’ll be forced to stop by the PTC system.
  • If a Metrolink train is accidentally dispatched onto the Gold Line, it will be stopped by a PTC wayside unit fixed at stop. (Freight running at night would be stopped by the PTC and call dispatch for permission to proceed.)
  • If a Gold Line train is accidentally dispatched to the west of La Verne or east of Montclair, it will be stopped by getting no cab signal. (Electrification infrastructure should extend just far enough west and east to allow the train to remain on the wire in the event of such a mishap.)

That leaves the issue of the Gold Line interface with the “kitchen sink” functionalities of PTC – the miscellaneous goodies that have nothing to do with the Chatsworth crash that supposedly moved Congress to pass the Rail Safety Improvement Act of 2008. In this case, that means overspeed derailments and work zone incursions.

The normal speed profile of this section of track would be enforced by the cab signals. But any slow orders would not; nor would temporary speed restrictions for work zones. Fitting the LRT fleet with PTC equipment would cost a lot of money. Can’t we just admit that PTC was a huge boondoggle, and. . .

(muffled sounds of struggling as FRA goons drag me out into the streets of Palms and beat me with spare Colorado Railcar parts)

Ok, where was I? If we can’t get an exemption from PTC functionality for the LRT trains, it should theoretically be possible to interface the PTC system with the cab signals, and use software to drop the cab signal when there’s a slow order or work zone. I’m not a huge fan of unique installations like that but as a last ditch solution, it might be acceptable.

Current Wars

There is one other issue that I’ve ignored so far: any future Metrolink electrification would likely use that punk Tesla’s AC distribution system, 2x 25 kV, while LRT uses DC at 750 V. You could write volumes on it, but for now, note that DC systems have proven to be the most common choice for LRT, while 2x 25 kV AC is the standard for commuter rail and intercity trains. Ultimately, this problem will have to be solved by having a small dedicated fleet of trains for this LRT service capable of running on both 750 V DC and 2x 25 kV AC. Dual mode trains are nothing new.

Metrolink electrification is not on the horizon, it doesn’t make sense to introduce 2x 25 kV AC on such a short segment of track, and it doesn’t make sense to procure a new, small, unique vehicle fleet if it can be avoided. The cost of a few dual mode vehicles for this service would practically be a rounding error in the overall cost of Metrolink electrification, but would be a major expense for this service as a standalone project. Therefore, I see no issue with going with 750 V DC for the time being.

What About Wheels?

Some light rail systems that are streetcar legacy systems, like the MBTA Green Line, use vehicles with wheels that have a flange depth less than mainline rail. These systems also often use flange-bearing frogs through turnouts. Using that type of turnout is not an option if you’re sharing tracks with mainline commuter and freight trains. To be honest, I’m not sure what type of wheels LACMTA LRT vehicles use, or if it’s an issue to run shallow flange wheels through mainline rail special trackwork – perhaps a reader could provide some insight? The answers to these questions could have some bearing on what turns out to be the most practical choice.

Total Cost

Throwing some money at design and contingencies, we come out at about $130m, or just over $10m/mile. Not bad for 12.2 miles of new rail transit, right?


Rough Schedule and Cycle Time

Based on typical speeds, I’d expect end to end running time from Fillmore to Montclair to be about 47 minutes. Allowing 5 minutes for turn at Fillmore and 8 minutes for turn at Montclair, that’s a cycle time of 105 minutes. To run 15 minute headways, you’d only need 7 vehicles. Meanwhile, short-turning an equivalent number of three-car trains at Sierra Madre Villa instead of Azusa/Citrus reduces peak period requirements by 6 vehicles. (See that? See what I did there?)

Moar Cheaper Pleez

The obvious ways to drive down costs even further are (a) start the single-track concept at Azusa/Citrus and (b) avoid electrification and the changes to signaling that result from it. The problem with (b) is that you then have to procure new vehicles, either dual modes or DMUs (which would never be allowed in the downtown tunnels because of fire hazards) or new EMUs that have battery backup for running between the SB Line merge and Montclair. With the latter, I’m assuming that you’d still electrify the line from Azusa/Citrus to the SB Line merge, and provide recharge capability at Montclair so that trains wouldn’t get stranded.

I estimate the cost of option (a) as $115m. For option (b) with EMUs, I estimate the cost at about $107m. For option (b) with DMUs, I estimate the cost at about $84m. For both scenarios in option (b), I included a vehicle cost of $10m. That’s much less than it would cost to procure a new fleet of 10 vehicles for the service, but I’m crediting these alternatives for reducing LRT fleet requirements. Ignoring that credit would bring the costs closer to $150m for EMUs and $125m for DMUs. This assumes that the new vehicles could be maintained at existing yards and shops.

Think About Your Future

There’s one final thing to consider when building infrastructure, and it’s especially true of rail infrastructure: it’s never going to be cheaper or easier to build it than it is right now, when nothing’s in the way. A surefire way to look foolish in front of the public is to undersize something and have to expand it in 5 years.

Adding a second track is more costly when you have to work around an active railroad instead of on an almost vacant ROW. That means if you are going to build a “startup line” or single-track line, you need to be pretty certain that the single-track will be sufficient for long enough to make the future capital premium worthwhile in today’s money. I wouldn’t have gone to the trouble of writing this post if I didn’t think a single track would work for a while, but this is still something to take seriously.

Fire Away

Alright, I’m pretty sure there are a lot of questions and issues I missed, and this isn’t a slam dunk. So, fire away: tell me what I got right, what I’m overlooking, what I could improve, and your thoughts on the overall viability.


2 thoughts on “Gold Line Foothill 2B on the Cheap

  1. Henry Fung (@calwatch)

    It’s better to just try to settle with the FRA using some sort of flagging or time-delay offset than build the wacky flyover in Pomona. The idea is sound but at some point in the next 20-30 years an additional track will need to be built at great expense and disruption. Better to do it right now than do it half now and half later. Interim solutions, like the Orange Line, never work out well.

    In my mind the Gold Line eastern extension will be no less productive than, say, MAX to Hillsboro or the San Diego Trolley to Santee. The lack of a station at Lone Hill is a big issue, one that I’ve brought up all the way at the early stages, but Glendora wants the car dealership and big box sales tax revenue (and note that, at the time, the Glendora Marketplace was still farmland). For events such as the Rose Parade, Rose Bowl, and Claremont Colleges graduation (note the paucity of quality hotels in the Pomona Valley, compared to Pasadena or Arcadia), you’ll need at least two car trains, even if they run every 15 minutes. Operationally I would agree that the Gold Line east of Monrovia will probably be every other train during the peak, running every 12-20 minutes, similar to the end of the Blue Line to Long Beach.

    In particular the substations will need to be sized to the final condition to avoid overloads during a Rose Parade situation, when MTA may want to run multiple car trains for operational efficiency. The single car platforms will save a bit of money, and you could have the operator or security lock the rear cars (the power saved would not be significant compared to the operational issues posed at Fillmore). While you could save money with single tracking, in the future this line will get popular and need to be double tracked. This ended up being an operational constraint in San Diego and Sacramento, where existing single track line needed to be double tracked just for operational effectiveness, since headways remained at 15 minutes. Since bridges are the hardest to expand after they are built I would leave the Lone Hill and Foothill/Compromise Line bridges as double track bridges and use those as pullout tracks in the event of an operational disruption. East of San Dimas to Montclair they are less necessary, in the event of a service issue, you could have buses down Arrow Highway or Bonita in a pinch, but west of there the 57 freeway acts as a natural bottleneck where traffic congests. The sharing of tracks from La Verne to Montclair is novel and given proper PTC may actually be feasible in the medium term. The biggest issue is the platforms and delays in loading, in which case pull-out stations for the LRT may be a better option.

    1. letsgola Post author

      Thanks for the detailed reply; especially appreciated from someone who knows the San Gabriel Valley really well.

      Many of the issues come down to how far in the future the second track will be needed. If it’s far enough away, single track would make sense despite the capital cost premium for working on an active railroad. The operational issues for things like turning back at Fillmore, or pull-out stations on the shared section, are certainly something to consider, and they could take up a lot of dispatchers time. I’m not sure how LACMTA does it now, but something the Automatic Vehicle Identification system on the MBTA Green Line, where the trains themselves make route requests to the signal system, might work.

      As for special events, IMHO they are a relatively minor consideration in the overall scheme of transportation planning. I wouldn’t want to provide any extra freeway or parking capacity just for special events. Still, there should be some ability to provide extra trains (and a station & siding at Lone Hill would help with that).


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s