Mini-Case Study on Mega-Project Management

When people think about mega-projects in Boston, the Big Dig, along with its enormous cost overruns and construction quality issues, is what comes to mind. But there’s another Boston mega-project that started at about the same time, and didn’t become an archetype for infrastructure incompetence: the Boston Harbor clean up.

In the 1980s, due to decades of pollution from poorly-treated sewage and combined sewer overflows, Boston Harbor was a stinking embarrassment. A lawsuit under the Clean Water Act resulted in the state being forced to improve stormwater and sewage treatment systems so that water quality in the harbor would recover. It’s a little surprising that there doesn’t seem to be a detailed study comparing the two projects; because both projects were constructed at about the same time in the same city, there should be less issue correcting for exogenous factors like legal precedents, quality of local contractors and engineering consultants, and political institutions.

However, a trio of articles from the fall of 2006 offers some insight. A short article in Governing cites three major factors: continuity of oversight leadership, local funding, and in-house talent at the Massachusetts Water Resources Authority (MWRA), the agency created in 1985 to oversee construction and operations of the sewer treatment system. Continuity of leadership came in the form of oversight from the same federal judge and several long-serving MWRA board members, while the use of local funds for construction created an external incentive to control costs. Inside the MWRA, a small team of talented engineers oversaw the contractors and consultants, providing strong owner representation.

In a Commonwealth Magazine expert panel on the Big Dig, Douglas McDonald, who served as Executive Director of the MWRA for nine years, cites the MWRA board of directors as the critical difference between the two projects. According to McDonald, the Executive Director had to report to the board of directors and a community advisory board every month, answering questions in real time. In contrast, leadership at the Massachusetts Turnpike Authority, which managed the Big Dig, saw more frequent turnover and political interference. McDonald says that “it’s not totally clear to whom the Bechtel corporation [which oversaw the Big Dig] ever reported.”

Lastly, in a long-form article looking at the mismanagement of the Big Dig, Boston Magazine cites the high level of in-house talent at the MWRA as the critical factor. The article quotes David Luberoff of Harvard’s JFK School of Government saying “it’s clear the state needed to have someone with Bechtel’s expertise, but the state could have done a better job of managing the managers. You have to have a small, highly skilled, highly respected group of people who could look over Bechtel’s shoulders.” In other words, a project as unique as the Big Dig is always going to be beyond the capabilities of the managing public agency, and there’s nothing inherently wrong about using outside consultants. However, strong advocacy on the owner’s part is still required.

The article goes on to quote Paul Levy, another former MWRA director, saying that “we had a 50-person project management team within the MWRA of highly paid, very experienced people… right after I hired Dick Fox, I remember [Big Dig architect and former Secretary of Transportation] Fred Salvucci calling to congratulate me, saying he wished he could do that but it was not possible under the state personnel system.” Thus, it appears that a political decision – subjecting the DOT to the state’s personnel system but exempting the MWRA – made it more difficult for the Big Dig to hire people with the skills required to oversee the project. The inability to pay wages that are competitive with the private sector is a pervasive problem for public agencies.

Readers with experience in private land development will not be surprised by any of this. As a land developer, you need to hire a team of consultants to successfully complete a large project, including legal professionals, civil engineers, architects, mechanical-electrical-plumbing consultants, structural engineers, construction contractors, and construction managers. While they are all on your payroll, they all have other interests as well, which may conflict with your priorities. Architects will select more elaborate designs and finishes, both out of professional pride and the desire to have future clients see a portfolio of high-quality work. Civil engineers don’t want to aggravate the public agencies they interact with for other projects. Construction managers don’t want the contractor community to see them as too adversarial. Contractors might be losing money on another project and looking to make up that loss on other jobs. As an owner, you must strongly advocate for your interests and priorities. If you’re asleep at the switch, you’ll end up paying too much for the job, even if the entire project team is working ethically and there are no serious issues.

The harbor cleanup project was not without issue. For example, in 1999, two workers died near the end of the project’s nine-mile long tunnel due to a failure of the improvised breathing systems that they were using. However, the project was successful in its water quality goals; today you can swim at Spectacle Island, something that would have been unthinkable in the 1980s. The MWRA seems to be one of the more respected state agencies.

Meanwhile, the problems with the Big Dig have poisoned the public debate on transportation mega-projects. People now expect that the projects will be poorly built and have massive cost overruns, which makes it much more difficult to build political support. Progressives that think cost effectiveness and public trust don’t matter, take note.

Eliminating Loop Ramps: The 10 at La Cienega and La Brea

In this introductory post on urban freeway improvements, the elimination of loop ramps and slip ramps was identified as some of the lowest-hanging fruit. It’s relatively cheap to do, and it makes things much better for bikes and pedestrians by getting rid of long, skewed crosswalks and road geometry that encourages drivers to speed. Since loops take up a lot of real estate, it also frees up a decent amount of land for development. Inspired by a recent comment, here’s a look at two interchanges on the 10: La Cienega and La Brea.

La Cienega

The westbound ramps at La Cienega are already in a tight diamond configuration, so there’s nothing to change there. The eastbound ramps are in a cloverleaf configuration, albeit a modified one, since Venice cuts through the vicinity. Thus, the northbound La Cienega on ramp to the 10 eastbound is a right on Venice, then a right on the ramp. The loops are very tight, with curve radii down in the neighborhood of 80’.

The basic idea here would be to remake the eastbound ramps in the image of the westbound ones. The interchange would become a modified tight diamond, with a new road connecting the two ramps between La Cienega and Venice. This would reconfigure the free movements to and from the ramps into normal city intersections, making them less hazardous for pedestrians and bikes. It would also yield new signalized pedestrian crossings of La Cienega and Venice, making things a little more walkable.


click to embiggen

This concept uses 10’ lanes and 6’ curbside bike lanes on La Cienega. Now, I know what you’re thinking – why is there no crosswalk on the north side of the intersection of the new ramps and La Cienega? Three-legged pedestrian crossings are horrible! The reasoning is that adding a crosswalk there makes things considerably harder for the traffic engineering, and this location is a rare exception to the rule that you should never omit a crosswalk. Because there’s nothing on either side of La Cienega on that side of the ramps, there’s no chance a pedestrian would have to use all three crosswalks in lieu of the missing crosswalk. Anybody walking here is going to either a destination north of the 10, in which case they can cross at David Ave and the onramp to the 10 westbound, or a destination south of the new ramps, in which case they can cross on the south side of the intersection.

This concept adds two lanes under the freeway bridge. It looks like this might just fit under the existing bridge, because the east side has a row of parking between the existing edge of pavement and the columns.

On the traffic side, the loop ramps are both serving over 10,000 vehicles per day. Those turning movements, which are currently free (unsignalized) right turns, will be replaced with left turn phases at the new traffic signal. Excluding the crosswalk on the north side of the intersection makes that left turn easier, reducing the green time needed for the eastbound movement. Again, this is only acceptable because it’s a special situation. Here are the traffic volumes at the new intersection.

LaCienega-table1 LaCienega-table2 LaCienega-sketch

Traffic volumes are from Caltrans and LADOT. This is a really rough estimate. The approach was to guess at the worst conflict group (combination of movements that can’t proceed at the same time) for each intersection, and figure out the sum of capacities needed for each movement in the group. That’s the “g/C” column (green time divided by cycle time), representing the percentage of the total intersection capacity needed for that movement. For example, the left turn from the 10 eastbound to La Cienega northbound needs 23% of the capacity at the intersection. If the total of that column is greater than 100%, or even relatively close, the intersection is close to failing.

La Brea

The existing interchange at La Brea is a full cloverleaf, with loops almost as tight as 100’ radius. However, the interchange doesn’t function like a true cloverleaf, because the outer ramps have very sharp cure radii to and from La Brea, and the offramps using the outer ramps have traffic lights instead of free-flowing turns.

The plan at La Brea would be to reconfigure the interchange as a tight diamond, using the same parameters – 10’ lanes, 6’ bike lanes. There are two options, one with the ramps tight up against the freeway, and one with the ramps intersecting La Brea near where the outer ramps do today.


click to embiggen


it’s a perfectly cromulent word

The advantage of the first option is that it lets you do the same crosswalk trick as at La Brea. However, unlike at La Cienega, there’s no extra room under the freeway at La Brea. We can steal the weaving lane to get four lanes under the bridge, but that leaves only a single lane for the left turns onto the 10. Unfortunately, that probably won’t work on the traffic side.

By pushing the ramps further away from the bridge, the second option lets you fit in a second left turn lane, though due to the lane’s short length, it might be a little optimistic to assume it could be used to its full capacity. Also, because the second option puts the new development between the ramp intersections, it’s no longer acceptable to omit a crosswalk. That makes the traffic design more challenging.

Traffic volumes on La Brea are daunting – the road serves nearly 70,000 vehicles per day here, more than many freeways that are two lanes per direction. The heaviest ramp volumes are to and from the east, all approaching 10,000 vehicles per day.

LaBrea-table1 LaBrea-table2 LaBrea-sketch

This design would require the crosswalks closest to the freeway to be concurrent with the left turns from the freeway offramps, which might be difficult given the traffic volumes.

Palmer Paradise

Now normally this is the part of the post where I’d suggest auctioning off the real estate to the highest bidder, as long as they agree to do something with it other than surface parking. That way you don’t end up with prime real estate owned by the government sitting vacant for years because it was impossible to come to a consensus on what to do with the land.

But you know what? F!@# it. These freeway-adjacent sites are right in GH Palmer’s wheelhouse. Just dial him up and let’s get us a few hundred Italianate apartments built. We can call them The Palude and The Catrame (the Italian equivalent of La Cienega and La Brea).

Traffic Troubles

While the idea of improving these interchanges for pedestrians and bikes, and freeing up space for urban development, is appealing, the worst g/C ratios approach 1 at both interchanges. More traffic study would certainly be required to see if these plans are viable.

Politically, any plan to eliminate loop ramps is going to have to win the support, or at least the grudging tolerance, of drivers. Unfortunately, these interchanges are not the best candidates for the first project, because if the first project doesn’t go well, there won’t be any more. Back to the lab again. . .

 The sausage-making behind this rough traffic analysis: I assumed the ramp volumes have the same peaking as the through movements on the 10, and 1,700 veh/hr capacity for each lane at the intersections. The critical conflict group at La Cienega was assumed to be Offramp EB – La Cienega NB – La Cienega SB left turn. At Venice, Onramp EB (from the La Cienega SB left turn) – Venice EB – Venice WB left turn. No volume was available for the last movement so it was a wild guess. The critical conflict groups at La Brea were assumed to be Offramp EB – La Brea NB – La Brea SB left turn, and Offramp WB – La Brea SB – La Brea NB left turn.

Should Streetcar Skeptics Stick a Sock In It?

Ok, to be fair, that’s not what Dave Alpert said in his Citylab piece today, but once I thought of that title, I couldn’t resist.

The article says that mixed-traffic streetcar skeptics shouldn’t be so quick to denounce the projects – “don’t let the perfect be the enemy of the good” – for five reasons: imperfect transit can still be good, limited funding makes the perfect unachievable, funding won’t get redirected towards better projects, streetcars have higher capacity than buses, and improvements can always be made in the future.

There are some larger things in play here, but first, let’s take a look at the idea of imperfect projects in general, and the five reasons offered.

An Imperfect Project Isn’t Necessarily Good

No transit project is perfect. For example, consider LA’s Expo Line. In my humble opinion, some of the stops weren’t needed – Farmdale and perhaps Expo Park/USC. All riders would agree that the Flower Street Crawl, as we call the unacceptably slow portion of the line from Jefferson/USC to the Blue Line Junction, needs improvements to make it run faster. And penny-pinching value engineers can find plenty to gripe about, like the use of low-profile catenary, the unnecessary lights mounted on OCS poles, or four-quadrant highway crossing gates with four independent pedestrian gates.

Yet on the balance, the Expo Line is still a really good project. It provides a transit service that is competitive with the freeway and arterial road alternatives, and connects several existing dense nodes of development. The Expo Line and Blue Line have some of the best new LRT ridership in the country, despite an appalling lack of upzoning.

The proposed downtown LA streetcar, on the other hand, is a very weak project, regardless of mode. It’s a one-way loop that partly duplicates existing services that are far superior. Even if it were completely grade separated, it wouldn’t be any better than underutilized downtown people movers in places like Detroit and Miami. Opposition to the streetcar isn’t just based on it being mixed-traffic, it’s that even a technically perfect project on that corridor would not be a good project from a transit planning perspective.

In general, streetcar proponents seem to discount the idea that streetcars could be bad for transit, but that possibility must be considered. A project that requires heavy operating subsidies can drain service away from other transit, like buses. Many transit advocates in Austin point to the heavily subsidized Red Line rail for causing cuts to bus service, and fear that a poorly planned Project Connect will make things worse. Even LA’s rail transit projects, which perform very well on ridership, come under fire from bus advocates like the Bus Riders’ Union, which alleges that transit dependent populations have lost bus service in order to fund rail. If you’ve ever ridden a full 204 bus down Vermont in the evening, when it’s running 20 minute headways, and transferred to a relatively uncrowded Expo Line running 10 minute headways, you can see where that perception comes from.

If you build projects that make existing transit services worse, you run the risk of losing riders, and alienating part of the political base that supports transit.

Increasing Urban Development

The Citylab post suggests an imperfect streetcar might be acceptable as a way to increase the supply of walkable, urban places, but this is not a good reason to build a transit project. If there is desire for urban neighborhoods, they will be built if zoning allows for it. Upzoning along the Expo Line would likely lead to a boom in dense residential construction on LA’s Westside, but that development would happen with upzoning even if the train wasn’t there. Where development does follow streetcars, like Portland’s Pearl District, it has been awarded large tax subsidies.

Funding is Limited

Federal funding for transit is scarce. Metropolitan regions compete with each other, and within each region, there are competing projects. This results in reductions to project scope, to try to be able to build the project for less money, or in phasing projects, to spread out costs over time as funding becomes available. For example, the Purple Line to Westwood would be better off being built as one contract, in one phase, avoiding the need to issue multiple procurement packages and the cost of mobilizing and demobilizing several times. However, Measure R funds aren’t available fast enough, so the project is split into three phases.

On the other hand, the project needs to be big enough and useful enough to make sense as a standalone job. You couldn’t build a suspension bridge with only one tower, and you probably wouldn’t build a mile of Purple Line tunnel with no stations just because that’s all you had funding for. If you can’t meet a minimum threshold of utility, you’re better off not building the project.

Funding Won’t Get Redistributed to Better Projects

This is misdirection. It’s certainly true that, due to political constraints, money can’t be shifted easily to better projects. However, that doesn’t answer the question of the usefulness of the project at hand. As Alpert points out, it’s possible that the WMATA Silver Line money could have been spent on better projects, but the Silver Line is a good project on its own. Likewise, the Westside Subway is logically the highest priority subway in LA, but the Red Line to North Hollywood got built first because of political reasons. Fortunately, the Red Line is still an incredibly useful project on its own merits.

Streetcar Capacity

Streetcar proponents often point out that streetcars have higher capacity, and therefore theoretically lower operating costs, than buses. This is only true if you’re serving a high-demand corridor, where using streetcars would allow you to save a lot of money on driver labor. Streetcar routes that are running service every 15 minutes, or even less frequently, are clearly not at the point where bus capacity is saturated. This is a guess, but I think if you have hit the point where mixed-traffic buses are inadequate to serve the demand, or where rail would offer significant operations savings, you’re probably at the point where you need exclusive lanes as well.

Future Improvements

The prospect of future improvements is a legitimate reason for accepting an imperfect project, so long as the project is set up to enable those improvements. Alpert uses single-tracking a rail line and shorter platforms as examples, and they’re good ones. LA’s Blue Line was also built with two-car platforms, later extended to three cars to accommodate high ridership.

The challenge with mixed-traffic streetcars, especially if they’re curb-running, is that they don’t easily lend themselves to future improvements. Converting curb lanes to exclusive lanes is more difficult than converting center lanes because of drainage issues, parking, and driveways. The latter, in particular, can make it difficult to extend a sidewalk platform to accommodate longer vehicles in a dense urban environment. Short downtown lines are often pitched as “starter lines”, but long lines are not workable at the speeds achieved by curb-running mixed-traffic streetcars.


There are, of course, more than enough highway boondoggles to put things in context. You could also compare streetcars to, say, Essential Air Service subsidies, which blow millions of dollars subsidizing air travel to small cities across the country. Those are good points, but public opinion is remarkably adept at compartmentalizing government waste. Rob Ford can blast city councilors for getting free zoo passes, then turn around and propose wasting billions on converting Scarborough RT to a subway. Again, projects have to be worth it on their own merits, rather than being excused by something worse.

Note that none of this should be taken to mean that streetcars are always a bad idea. The Columbia Pike project is frequently cited by streetcar proponents, and it has the potential to be a good project. For starters, it’s a straight, logical route, and they’re proposing to run 6 minute headways, which suggests existing transit demand is high enough that rail might be cost effective for operations. If it were center-running, it would offer the potential for future improvements that might lead some technically inclined observers to support it.

The Big Picture

In the big picture, the streetcar debate is part of the ongoing rift between what Alon Levy called politicals and technicals. Progressive political activists are inclined to view any expansion of rail transit services as a positive, building towards a future where there is more political support for transit expansion. Technical commenters are inclined to believe that you can only build so many bad projects before the people realize their money is being wasted.

This blog is LA-centric and written from an engineering perspective, naturally tending toward the technical side. Simply put, if the Blue Line and Red Line were running empty trains all day long, I do not think we would be building the Expo Line and Purple Line. While I understand the need to build political constituencies to support policy changes, I also think nothing succeeds like success. LA voters are demanding an expansion of rail transit services, while residents of greater Portland are pushing back against further expansions of streetcar and LRT service, putting higher priority on more frequent bus service.

Alpert’s piece concludes with a warning that “writers who think more transit is good for cities should bear in mind that not all readers necessarily agree with that basic premise”, referring to opponents who don’t want to fund transit at all. This statement is similar to Robert Cruickshank arguing that because some ideological transit opponents use efficiency as a false flag attack, progressives should actively shun the idea that efficiency matters.

Well, guess what – I don’t think more transit is necessarily good for cities! Resources are limited. Transit that is grossly inefficient, or wastes capital dollars, is not good for cities. This is a fundamental failure of allies for good transit projects to understand where their fellow advocates are coming from. But as frustrating as it may be at times, we ultimately need each other’s support. Political advocates need to learn what projects will gain long-term support by providing useful services, and technical advocates need to figure out how to improve public understanding of what makes transit useful.

And if the project is just to support condo developers, well, let them build it.

Sepulveda Pass Transit, Part 3: Mode and Alignment Through the Pass

For an overview of transit between the Westside and the Valley, see Part 1. For a close-up look at LAX, see Part 2.

The most critical part of a north-south transit line between the Westside and the Valley is Sepulveda Pass – the section that roughly parallels the 405 between Wilshire and Ventura Boulevards. Services on the Westside and in the Valley will probably end up having several branches using the pass, in order to maximize the usefulness of the pass segment. Due to the distance (about 7 miles) and engineering challenges, we’re probably only going to get one line through Sepulveda Pass in the foreseeable future. It’s critical that we get this segment right, get the most capacity for our money, and set it up to flexible enough to accommodate many services on both sides.

The two planning questions that must be answered are:

  1. What modes should the project serve? This will determine who can use the project, be it cars, buses, or trains.
  2. What should the project alignment be? This will determine what service patterns can be operated on either side of the pass and how they will relate to each other.

Question 1 comes first, because the mode choice will affect the design criteria for the project alignment, such as curvature, grades, and ventilation.

A Multi-Modal Tunnel?

The concepts that have been floated publicly are all variations on a theme. They propose building a toll auto tunnel that would also provide lanes, perhaps dedicated, for transit. The project is often pitched as a candidate for a public-private partnership.

If the alternative includes a tunnel, I don’t think auto lanes should be part of the plan, for reasons explained here. If HOT lanes are going to be part of the project, they should be converted from existing HOV lanes (or, if you insist on new lanes, new at-grade or elevated lanes, but there’s no spare capacity on the 405, the 10, and the 101 for new lanes to connect to anyway). That leaves bus and rail.

The primary trade-off between bus and rail is implementation timeline versus capacity and operating costs. If the corridor is for buses, it can be used immediately by many bus services connecting all parts of the Valley and the Westside, while a rail link from Wilshire/Westwood to Sherman Oaks would be of limited use in isolation. Choosing rail would delay the usefulness of the project until feeder lines were built on both sides. However, as passenger volumes increase, which we would expect for a useful Sepulveda Pass project, rail offers higher capacity and lower operating costs.

Four options come to mind:

  • A guideway exclusively for buses
  • A guideway exclusively for rail
  • A hybrid guideway running both buses and trains (not as crazy as it sounds; Seattle is running a tunnel like this right now)
  • A larger guideway with four lanes, two for rail and two for bus (or hybrid)

The first two options just seem underwhelming for the context. We’re not talking about the Gold Line from Azusa to Claremont or an improvement to an arterial corridor that’s got parallel arterials to be upgraded a mile away on either side. This is it – the one big project between the Westside and the Valley that we need to facilitate more growth between Sylmar and Long Beach. You don’t want it to end up like the MBTA Green Line, right?

Capacity Counts

Some more serious numbers: in the post on capacity, we estimated about 5,000 pax/hr per direction for bus (standing load, 60 second headways) and 15,000-20,000 pax/hr per direction for LRT (standing load, 2 minute headways, 3 or 4 car trains). For comparison, the five lanes of the 405 (we’re ignoring the climbing lane and auxiliary lanes) have a capacity of about 12,000 veh/hr per direction. Obviously, the passenger capacity depends on how many people are in each car; assuming 1.2 pax/veh (not unreasonable for commuting), that’s 14,400 pax/hr per direction.

That gives you an idea of the magnitudes of how many people can be moved by each mode. You can vary the assumptions as you like (double articulated buses, longer trains, higher occupancy in cars). Bus headways below 60 seconds are probably beyond the point where rail offers higher reliability and lower operating costs. The inclusion of bus would be mainly motivated by the desire to put the facility to use immediately, without waiting for long branch rail lines to be built.

That puts a transit option with one lane in each direction in the same league as the existing 405, so maybe that’s enough. On the other hand, the relentless congestion on the 405 suggests there’s a crap ton of latent demand – in other words, a lot more people would be traveling through Sepulveda Pass if it were easier to do. We want this project to relieve the 405, but also to facilitate economic growth on the Westside and in the Valley. With that in mind, a large diameter tunnel with four tracks may be the way to go.

To see why we might want a tunnel with two lanes or tracks in each direction, consider the effect of branching. Since Sepulveda Pass is a natural bottleneck, we should be serving several parallel north-south transit lines, bringing them together for a trunk through the pass and allowing transfers. In the opening post, we identified up to four corridors on each side to be served. With an operational headway of 2 minutes and one track in each direction, that’s 8 minute headways on the branches. This is short of Metro’s design criteria, which calls for operational headways of 5 minutes on LRT branches. With a large diameter tunnel and two tracks in each direction, operational headways of 4 minutes would be achievable on the branches.


In the introductory post, I defaulted to the assumption of a tunnel the whole way from Westwood to Sherman Oaks. Alon Levy rightly called that assumption out in the comments, prompting a look at some elevated and hybrid options.


An elevated option is self-evidently going to follow the 405. This is both the best horizontal alignment and the best vertical alignment that does not involve a tunnel.


From a technical standpoint, the critical section of the alignment is the approximately 1.5-mile long 5.5% grade on the north side of the pass. Light rail vehicles (LRVs) can handle short 5%-6% grades without issue; in fact, there are 5%-6% grades in many places on the new Expo Line for grade separations. However, I’m not sure if vehicle braking performance would suffer on such a long downgrade, and it might be difficult to support the required headways.

Let’s assume 2 minute operational headway and 90 second design headway (Metro’s current design criteria for a trunk LRT line is 2.5 minutes operational and 100 seconds design). Safe braking distance, for signal design, must include (a) distance traveled during reaction time, (b) braking distance, and (c) a buffer between vehicles. If you’re using fixed signal blocks, the buffer might be the vehicle overhang; for Communications Based Train Control (CBTC), let’s use an assumed imprecision in the system’s knowledge of where the vehicle is located.

Metro’s current design criteria specifies 9.8 seconds of reaction time. This might seem like a lot, but it has to cover equipment reaction time, operator reaction time, and brake build up. This value isn’t atypical in US practice. For braking, Metro specifies a distance of 0.733*S2/(B+0.2G), where S is speed, B is the braking rate (assumed to be 2.0 mphps), and G is the profile grade. Let’s assume 200’ for vehicle location imprecision (more precisely, 100’ for each train, with the worst possible combination of errors.

For a design speed, let’s assume 60 mph. For safe braking, you need to assume the entry speed when braking starts is higher due to a combination of speedometer error and equipment tolerance. To keep things simple, let’s assume 65 mph. That yields a reaction distance of 934’ and a braking distance of 3441’, for a total of 4575’ (including the 200’ CBTC buffer). Using 0.2G underestimates the effect of gravity a little; if you calculate the braking distance based on a 2.0 mphps braking rate adjusted by the laws of motion, you’ll get 5029’.

Okay, so that’s the separation you need from the rear of one train to the front of the train behind it. If you want the theoretical headway, you need the distance from the front of the train to the front of the train behind it. In other words, you have to add the length of the train. In this case, that’s four 90’ LRVs for 360’. If you have fixed signal blocks, you also need to add the length of one clear block of track, as shown below, but since we’re assuming CBTC, we’ll ignore that distance.


That gives a total distance, based on Metro criteria, of 5389’. At 60 mph, that’s 61 seconds of travel time, essentially a 1 minute theoretical headway. Even if you assumed fixed signal blocks and added a clear signal block distance, it would seem that a 2 minute operational headway is within the realm of possibility.

Note that this is still a simplification; the headway impact of having a station, presumably at Ventura Blvd, at the bottom of the grade would have to be determined by simulation. This analysis also ignores other potential physical constraints, for example the ability of the LRV to continually put out maximum braking force for that long or the impact of wet rails, that wouldn’t be an issue on shorter grades. Premature rail wear, such as rail corrugation, might occur. These issues are well beyond my experience. (Hint, hint, technically inclined commenters.)

From a route planning perspective, the elevated alignment is not ideal at either end. At the south end, you end up at the 405 and Wilshire, west of the proposed Wilshire/Westwood station on the Westside Subway. It wouldn’t be too hard to deviate west to the Veterans Hospital; however, this is bound to be a low demand station. Wilshire/Westwood is a much better location for the transfer, because it will eliminate the need for many people on the north-south transit lines to transfer in the first place. It wouldn’t be too hard to get over to Veteran Av by crossing the cemetery (they’re the abutters least likely to complain). That makes the transfer reasonable, but still puts the stop at the very margin of UCLA and Westwood. From there, the line would probably head back towards Sepulveda, but more on that another time.


At the north end, the first stop would naturally fall at Sepulveda/Ventura. North of there, the line could hop over to Sepulveda Blvd at the 101 or at Burbank, and follow Sepulveda north through the Valley. Sepulveda is good corridor, and deserves a high quality transit service, but most of the interest in the Valley seems to prioritize Van Nuys over Sepulveda. Getting from Sepulveda to Van Nuys would require a one mile jog to the east, and the resulting zigzag would be bad route planning. However, Sepulveda/Ventura is a decent node in its own right.



A hybrid alignment would follow the same route as the elevated alignment from Wilshire to the 405 just north of the Sepulveda Blvd ramps. This would require about 3.5 miles of tunneling, just a little more than half of what the full tunnel would require.


This alternative would save some money over the full tunnel alignment, because elevated construction is usually cheaper than tunneling. It would also allow the northern approach to be constructed at a much gentler grade, around 1.0%, than the 5.5% grade required by the elevated option, and greatly reduce the length of the 3.0% grade on the southern approach.

From a route planning perspective, this alternative is also somewhere in between the elevated option and the full tunnel option. The southern end would suffer the same drawbacks as the elevated option, but the northern end would be in a better location, as described under the full tunnel option.


The tunnel alignment would follow the approximate route of the tunnel that has been proposed publicly, from Wilshire/Westwood to Ventura/Van Nuys. This route would be in tunnel the whole way. It might be possible to build some of the route at-grade through UCLA’s campus, but it’s probably not worth the effort to bring the line to the surface for such a short distance.

This alternative would cost the most, but it would have the best track geometry, with a ruling grade of 1.0%.


From a route planning perspective, it’s also the best option at both ends of the alignment. At the south end, it puts the Wilshire/Westwood stop in the right place for both transfers to the Purple Line and for local destinations at UCLA and Westwood.


At the north end, it lines up perfectly with Van Nuys, the highest priority north-south corridor in the Valley, and yields reasonable geometry for additional branches to the west towards Sepulveda, Reseda, and Balboa.


Boring Questions

Assuming a tunnel is going to be part of the selected alternative, the cross section of the tunnel is the next question. With the exception of the Blue/Expo Line tunnel on Flower Street, all of the transit tunnels in LA were constructed with the same cross section, consisting of two single-track tunnel bores, connected every so often by emergency cross passages. The stations are center platforms located between the two bores.

For Sepulveda Pass, you’d have a few options:

  • Four single-track bores, built in pairs either simultaneously or sequentially. In this option, you would probably build two tracks at the outset, leaving the next two tracks as a future project.
  • Two two-track bores, again likely leaving the second set of tracks as a future project.
  • One four-track bore.

Alon Levy pitched large diameter tunnel boring machines (TBMs) as money-savers because the station platforms can be located inside the bore; I’m not sure how much they’d save for an LA-type station, relative to the costs of the additional excavation.

However, I think a large diameter TBM might make sense for the Sepulveda Pass project for different reasons. For one thing, when you do two single-track tunnels, you have to make a decision about how many TBMs to buy. Do you buy two TBMs, at considerable up-front capital expense, and allow both bores to proceed simultaneously? Or do you buy one TBM, and bore each tunnel sequentially, paying the price of a longer construction schedule? Using a larger diameter tunnel means buying fewer TBMs and a shorter construction process.

Personally, I like the idea of one four-track bore with two tracks on each level. One level could be used for rail right from the outset, with the other level used for express bus services between the Valley and the Westside. In the future, the bus level could be converted to rail if needed for capacity. The advantages in time and cost are many: construction of launching pits is only needed once, the full capacity is available after completing one bore, working near an active transit line is avoided, and labor costs are reduced by minimizing complexity and shortening the duration of construction. This approach also avoids the tendency of future capacity improvements to remain forever in the future.

Some recent examples of large diameter tunnels include the M30 freeway in Madrid (inner diameter 44.13’), Line 9 in Barcelona (inner diameter 35.8’), and the Alaskan Way tunnel in Seattle (diameter 56’). The TBM in Seattle is, of course, currently broken down, but don’t let their crummy execution sour you on the concept of a TBM that large. Barcelona Line 9 was apparently built to be just large enough for a four-track section, to allow crossovers between stations, but that seems like a really tight section for four tracks. On the other hand, 56’ would probably overdo it and result in high costs for the launching pits and excavation.

A 45’ diameter tunnel would allow four tracks, along with space for breathing room to fit in mechanical and electrical equipment. In particular, with a long tunnel like Sepulveda Pass, it might make more sense to set up the ventilation like a freeway tunnel, with continuous clean air and polluted air levels below and above the travel ways, respectively. In contrast with most transit tunnels, which depend on the piston effect, this design would hopefully allow the ventilation system to meet the requirements of NFPA 130 without restricting the system to one train per direction in the tunnel between stations. Such a restriction would cripple a long tunnel’s capacity to the point that building it would be almost pointless. (The NFPA 130 requirement is actually one train per tunnel vent zone; relying on the piston effect means that each length of tunnel between consecutive stations is operated as one vent zone.)


The space to the sides of the tracks would accommodate electrical and mechanical systems, emergency egress, and ventilation as needed.

For an overview of large diameter tunnel costs, see this post on long freeway tunnels.


There are several feasible alignments and mode alternatives through Sepulveda Pass. While an elevated facility following the 405 is theoretically cheaper, it may be less so in this case because it would have to be constructed over and around an active freeway. The hybrid and full tunnel options offer better routes, and might be worth the trouble, especially if a high capacity tunnel can be built in one bore (and we can reign in US tunneling costs a little). An option that has provisions for both bus and rail will allow higher utilization of the tunnel from the beginning, without needing to wait for all the branch rail lines to be finished.

Components of Highway Funding Shortfalls

I forget where, but I recently heard another story about highway funding that specifically mentioned declining vehicle miles traveled (VMT) and increasing fuel efficiency as causing reductions in gas tax revenues. The impact of inflation on the purchasing power of gas tax revenues was mentioned only in passing (to note that the federal gas tax has not been increased since 1993).

The VMT and fuel efficiency trends are real, but are they the main show? Well, we have data! So here’s a look at trying to tease out the impacts of decreasing VMT, increasing fuel efficiency, and inflation on transportation funding at the federal level.

Warning: all of the calculations here are very rough. There are many variables involved that are beyond the scope of this post. A detailed study would probably be a good project for a grad student somewhere. As it is, take the results here as indicative of the order of magnitude of impacts.

Detailed methodology is explained at the bottom of the post.

The federal gas tax was last raised in 1993, to 18.4 cents per gallon. Since that time, inflation has eroded the purchasing power of the gas tax (i.e. the same amount of money now buys less roads). In addition, in recent years there has been a notable increase in fuel efficiency, and a stagnation of VMT. Because people are driving more fuel efficient vehicles, and driving them fewer miles, the amount of gas used is going down, and so are fuel revenues.

To figure out the impact of each factor, we need to look at data trends over the last 20 years and pose some reasonable counterfactuals. For data, I pulled VMT from the St Louis Fed’s FRED service, inflation from the BLS Inflation Calculator, and fuel efficiency from the EPA’s Fuel Economy Trends Report. For counterfactuals, I looked at the following scenarios:

  • VMT growing at a constant 2.33%/year, the approximate rate at which it grew from 1993 to 2005.
  • Gas tax being adjusted annually to account for inflation.
  • Vehicle fuel efficiency held constant at 1993 level.
  • Combination of all three factors.

Trends since 1993 are shown in the following graph:

yearly revenue loss

Logically, the loss of purchasing power due to inflation has an immediate and increasing effect, tempered by slower inflation in recent years. The emergence of VMT and fuel efficiency as significant factors in revenue losses is relatively recent, within the last 5 years.

The next graph shows cumulative losses to transportation funding since 1993:

cumulative revenue loss

The reason to look at cumulative losses is that transportation funding is usually authorized as multi-year bills, and surpluses (or deficits) to the highway trust fund roll over from year to year. The loss of purchasing power due to inflation dominates; change in VMT trends only recently emerges as a relatively minor factor. Changes in fuel efficiency have had basically zero effect on the long-term solvency of the fund, as recent gains in fuel efficiency have just barely offset losses in fuel efficiency in the late 1990s and early 2000s.

While these are only order of magnitude results, they show pretty clearly that inflation has been the most important factor. It is beyond the scope of this post to argue if additional federal transportation funding is needed, but if it is, raising the gas tax is the easiest and most effective way to do it. (Note that theoretically, declining revenues due to declining VMT isn’t even a problem in the first place, since less driving means less roads are needed. And due to climate change, trends that result in higher fuel efficiency and lower VMT are net positives for society.)

So why so much talk about declining VMT and fuel efficiency? Wild guesses: first, it’s a convenient narrative if you want to replace the gas tax with a VMT tax and/or tolls. Second, it makes the problem seem more complicated, which creates the opportunity for Serious People™ to opine. And lastly, it provides a scrap of cover to incompetent politicians who would look like dunces if everyone realized that the problem could be solved by a 25-word piece of legislation raising the tax and indexing it to inflation.


A detailed assessment of the impact of inflation, VMT, and fuel efficiency trends on gas tax revenues is beyond the scope of this post, which is intended to determine orders of magnitude. Data was processed as follows.

VMT data was pulled from the St Louis Federal Reserve Bank’s FRED service. I used the January data point for the moving 12-month average for each year from 1993 to 2014. VMT trends started changing in 2005, so I calculated the annual rate of change from 1993 to 2005 as about 2.33%, then projected a VMT trend from 1993 to 2014 at that rate. The impact of VMT trends was calculated as a function of the difference between actual VMT and “implied VMT” calculated at 2.33% growth per year.

Inflation data was pulled from the BLS’s CPI calculator. I calculated the tax rate for each year from 1994 to 2014 that would yield the same purchasing power as 18.4 cents in 1993. The impact of inflation was calculated as the difference between revenue that would have been collected if the tax had been adjusted for inflation every year and revenue that was collected at 18.4 cents per gallon. This methodology probably systematically underestimates the impact of inflation, since costs of construction materials increased more quickly than general CPI between 2000 and 2007.

Fuel efficiency data was pulled from the EPA’s Fuel Economy Trends Report. This data provides fuel efficiency for vehicles by model year, not for the actual composition of the US vehicle fleet. Therefore, it can’t be used directly because that would overestimate the impact – model year 2014 cars are more efficient, but most cars on the road are older. I generated a crude fleet MPG by summing the product of model year fuel efficiencies and an approximation of vehicle fleet composition by age.

I then calculated assumed revenue amounts for each year as follows (where T = tax rate):

  • Actual revenue for each year: R = T(VMT/MPG).
  • Implied revenue assuming VMT continued to grow at 2.33%, and the gas tax was adjusted yearly for inflation, and no gains in fuel efficiency: S =T’ (VMT’/MPG’).
  • Implied revenue assuming actual VMT trends, but gas tax adjusted yearly for inflation and no gains in fuel efficiency: W = T’(VMT/MPG’)
  • Implied revenue assuming the actual gas tax (i.e. unchanged), but VMT continued to grow at 2.33% and no gains in fuel efficiency: X = T(VMT’/MPG’)
  • Implied revenue assuming actual gains in fuel efficiency, but VMT continued to grow at 2.33% and gas tax adjusted yearly for inflation: Y = T’(VMT’/MPG)

As a point of reference, the world’s finest information source says that the federal gas tax raised $25b in 2006. The methodology here generates an estimate of $27.4b. Good enough.

The revenue losses due to each factor were then calculated as the difference between the theoretical revenue with all historic trends (S) and the theoretical revenue with actual trends for each factor (W, X, Y):

  • Total revenue loss = S – R
  • Revenue loss from VMT trends = S – W
  • Revenue loss from inflation = S – X
  • Revenue loss from fuel efficiency = S – Y

Note that the total revenue loss does not equal the sum of the components, i.e. (S-R) ≠(S-W) + (S-X) + (S-Y). This is because the factors are not independent. For example, if fuel efficiency goes up, the revenue lost from a decline in VMT will go down, because the missing VMT represents a smaller amount of gas. This is why the three factor lines on the graphs do not sum to the total.

Downtown LA is Responsible for 20% of Housing Built Since 1999, and That’s Terrible News

Shane Phillips has a post over at Better Institutions looking at the proportion of housing built in LA since 1999 that’s located downtown. He calculates it to be about 20%, based on state data and a Downtown Center Business Improvement District Report. The report is generous in its definition of downtown, including Skid Row and the Fashion, Arts, & Industrial Districts, and stretching well into Westlake and Chinatown. Nevertheless, by any standard the amount of development in downtown is impressive. About 20,000 units have been built in the last 15 years, with another 20,000 in the pipeline for the next 5-10 years.

A pro-growth stance from the city has resulted in mid-rise buildings and towers popping up all over the place on top of former parking lots, putting the land to much more productive use. Meanwhile, the adaptive reuse ordinance (ARO) has allowed once-vacant historic office buildings to find new live as apartments, condos, and hotels. Michael Manville writes in UCTC Access that the ARO alone was responsible for 6,500 units of housing in the historic core between 1999 and 2008.

All of this is good. Turning parking lots into higher value land uses is good; putting abandoned buildings back to use is good. The neighborhoods around downtown are in danger of being victims of its success when it comes to gentrification, but more on that later.

So what’s the problem? The problem is that percentages have numerators and denominators. And in this case, the downtown boom is making the numerator bigger, but a severe lack of housing production citywide has made the denominator much smaller. In fact, based on the same state data, all of LA County added about 215,000 housing units between 1999 and 2014. In other words, in a county of 10 million people, a neighborhood of just 50,000 has been responsible for over 9% of new residential construction.

In short, the problem is that other neighborhoods across LA have not seen nearly as much growth. As Shane correctly points out, one neighborhood can do only so much. Read the USC Casden Multifamily Forecast and you’ll see neighborhood after neighborhood with almost no new inventory added from 2009 to 2013. East LA, Alhambra, Montebello, & Pico Rivera, zero. El Segundo, Hermosa Beach, & Redondo Beach, zero. Granada Hills, Northridge, & Reseda, zero. Paramount, Downey, Bellflower, & Norwalk, zero. The list goes on and on.

Housing prices are largely determined regionally, which makes it impossible for one neighborhood to upzone its way out of price increases. If you’re near desirable neighborhood XYZ that has very little new construction, it doesn’t matter what you do, eventually you’ll be “XYZ-adjacent” and it’s game over. On the Westside, you have to wonder how long places like Palms and Pico-Robertson can last with demand radiating east and south from Santa Monica and Venice, despite Palms being relatively friendly to new construction.

Even in cities with a strong traditional form like NYC, with a huge CBD dominating regional employment, concentrating all housing development near the core is a mistake. New York YIMBY recently chronicled the woes of NYC’s small builders, who have been driven out of business by downzoning in the outer boroughs. That has resulted in a decrease in the amount of market-rate housing being built for middle income earners, making the city’s affordability problems worse.

In a city like LA, with highly decentralized employment, concentrating housing development in the core makes no sense at all. The hottest office markets in LA are on the Westside, where the tech industry is concentrated in Santa Monica and Venice. Growth in that market has spread south to Playa Vista and the Howard Hughes Center. Century City office developers hope to capitalize on it as well, while others in commercial real estate expect growth to continue moving south to El Segundo. Whatever the reasons, the office market in Downtown LA remains weak, with plenty of vacancy and virtually no new construction.

The lack of a corresponding residential boom on the Westside exacerbates existing imbalances. The pull of Westside employment long ago made the “reverse” commute direction on the 10 freeway the peak direction (traffic is worse going away from downtown in the morning, and towards it in the afternoon). It would not be surprising at all if the peak travel direction on the Expo Line and Westside Subway ends up following a similar pattern.

Beyond the local issues of the Westside, there are job centers scattered all over LA County. Employment growth is not going to be concentrated in downtown, so why should housing growth? Distributed housing growth spreads out the impacts as well as the benefits, and helps prevent gentrification and development from flooding into a localized area.

Why Is Downtown Booming?

To be sure, Downtown LA has become a desirable place to live. It’s walkable, has good access to freeways and transit, and offers an increasingly diverse mix of restaurants, bars, and retail. It’s centrally located, making it (relatively) easy to live there and commute to the Westside, Hollywood, and parts of the San Fernando and San Gabriel Valleys. The architecture, especially the historic office and hotel buildings, is unparalleled in the region. That explains the demand side.

The supply side is explained by the factors mentioned before – the adaptive reuse ordinance and a strong (sometimes, maybe a little too strong) pro-growth stance from the city. As Manville writes, the conversions of historic buildings would have been impossible without the ARO, so it’s worth recapping the significant relaxation of land use regulations that the ARO provides:

  • No restriction on density based on lot size (though minimum apartment sizes apply)
  • Existing non-conforming FAR, setbacks, and heights do not require a variance
  • No new parking spaces required (existing parking must be maintained, but is not required to be bundled with dwelling units)
  • Automatic “by-right” entitlement for rental units in commercial or R5 zoning in buildings constructed before 1974
  • No environmental clearance for projects constructed “by-right”

This allows adaptive reuse projects to avoid almost all the NIMBY bugaboos, and deprives opponents of the leverage provided by the need to obtain discretionary approvals. It also allows projects to avoid the need to build expensive parking; as Manville writes, many developers have chosen to provide none or to offer it off-site.

The city has also facilitated growth downtown by other means, for example, selling the air rights above the convention center.

Why Are Other Neighborhoods Not Growing?

For most of the city, though, development doesn’t come so easy. Increasing demand has not been met by a boom in supply. Most neighborhoods don’t have a large supply of parking lots or vacant buildings to be redeveloped, and the city has been very reluctant to try to buck NIMBYism in the R1 zoned single-family residential (SFR) neighborhoods.

As a case study, consider the draft rezoning plans being developed for the five Expo Line Phase 2 stations that are within the City of LA (Culver City, Palms, Expo/Westwood, Expo/Sepulveda, and Expo/Bundy).

At Expo/Bundy and Expo/Sepulveda, there are significant amounts of land currently zoned M2 (light industrial). The plans propose maintaining some of that zoning, while converting other areas to new industrial zones including “New Industry”, “Hybrid Industrial (Max 30% Residential)”, and “Hybrid Industrial (Min 30% Job-Generating)”. The “industrial” classification is a little deceiving, since it allows office, R&D, media, and technology developments. Nevertheless, the New Industry zone precludes residential development entirely and only permits retail and restaurants as ancillary uses, and this is the most prevalent new zone. At Sepulveda, only two blocks are zoned Hybrid Industrial (Max 30% Residential), while at Bundy, four blocks are given that designation and three are given Hybrid Industrial (Min 30% Job-Generating). At Expo/Sepulveda, R1 zoning less than 0.25 miles from the station will remain. To the city’s credit, at Expo/Bundy planners did at least propose upzoning the R1 properties between the Expo Line and Pico, as potential options on the base plan.

At Expo/Westwood, almost the entire 0.25-mile radius around the station is currently zoned R1, even on the arterials (Overland and Westwood). The plans goal is to “preserve character of existing SFR neighborhoods”  and that’s what we’ll get, because all the R1 zoning is proposed to remain. The plan calls for upzoning a few R2 properties to R3, a largely symbolic gesture because that only increases density from 2 du/lot to 6 du/lot (assuming 5,000 SF lots). The lone bright spot for development is an upzoning of Pico between Sepulveda and Westwood to RAS4 (12 units per 5,000 SF lot with ground floor retail), but this amounts to only small portions of nine blocks fronting Pico.

The Palms plan might appear to be better, because it rezones Venice Blvd and Motor Av for a new “Mixed-Use (Min 20% Job-Generating)” zone with FAR of 2.0-3.6. However, Venice and Motor are currently zoned C2, which under the current zoning scheme already allows purely residential projects at R4 density. The Mixed-Use (Min 20% Job-Generating) zone therefore reduces some flexibility by requiring a commercial component. The small-scale residential and commercial developments that line Motor today couldn’t be built under that zone.

At Culver City, it’s more of the same industrial zoning, with three large blocks directly across Venice zoned New Industry and one further west, currently the site of a commercial plaza, for Hybrid Industrial (Max 30% Residential).

The plan also calls for current parking requirements to apply, except in “limited circumstances”.

The limited zoning changes produce the results you’d expect. The Spring 2014 outreach presentation projects that the plan will allow the construction of 4,422 new housing units by 2035, satisfying market demand of 3,800 to 6,400 units. So while downtown booms, under this plan, the Expo Line corridor won’t, because you can’t build a ton of housing if your zoning doesn’t allow for it. On the demand side, I submit that it is simply beyond belief that there will only be demand for 6,400 housing units within walking distance of those five transit stops in the next 20 years.


The downtown boom is great for LA, and it shows that when we want to, we can be pro-growth and get a lot of development built. But when growth is restricted across so much of the rest of the city, there will still be pressure on regional housing prices, and gentrification will continue. Downtown’s growth is remarkable, but we still need to figure out how to increase housing production elsewhere, so that the city can make space for all Angelenos, current and future.

Transit Costs: OCS Edition

Last weekend, I tweeted a few photos of Expo Line Phase 2 overhead contact system (OCS) construction. The OCS is part of the traction electrification system (TES) that provides electricity to trains, and includes the poles, cantilevers, wires, and associated hardware that you see along the track. The other part of the TES is the traction power system (TPS) which consists of electrical substations along the tracks, the cables that connect the substations to the OCS, and the rails, which serve as the negative return.

In this post, we’ll explore some OCS choices that can impact the cost of a light rail project. The OCS is something you have to build, and it’s not going to be the thing that breaks the bank. Nevertheless, a million here, a million there is still real money, right? In the same way that Jarrett Walker describes the choices between frequency and coverage, I’m going to take the soft approach and tell you that none of these options are really wrong, but you need to have an honest and open discussion about the cost implications. A good engineer should provide the owner with an understanding of the options available and the implications of each, and faithfully execute the design as efficiently as possible, but the ultimate design direction is made by the owner.

The discussion here is geared towards DC electrification, which is used for most light rail systems. Intercity rail is electrified with AC; the concepts are the same, but the greater working clearances required for higher voltages will result in some different decisions (e.g. side poles instead of center poles).

System Type and Height: Normal Profile Simple Catenary, Low Profile Simple Catenary, or Single Wire?

Simple catenary refers to the system you see on most of the Expo Line, Blue Line, Gold Line, and Green Lines. This system has two wires, called the messenger wire and contact wire. The messenger wire is the upper wire and is more visibly parabola-shaped than the contact wire, which, as the name suggests, is the wire the train’s pantograph touches. The system height refers to the vertical distance between the messenger wire and the contact wire at the poles.

A normal profile simple catenary system might have a system height of 4’. This lets you maximize the distance between poles (the span length) based on other factors like wind loading and track curvature, without worrying about maintaining separation between the messenger wire and the contact wire. So why aren’t all systems normal profile simple catenary? Because people have decided they’re visually unappealing.

A normal profile simple catenary system takes up a lot of your field of vision and has long hangers connecting the messenger wire to the contact wire. To reduce the visual impact, you can use low profile simple catenary. This has a lower system height – say 2’. This reduces the visual impact of the wires, but because the messenger wire and contact wire start out closer at the poles, they’ll get close to each other sooner, and you can’t space the poles as far apart.

Thank goodness the OCS doesn’t block too much of the view of the LADWP power lines on the other side of the tracks, right?

As an example of pole spacing, here’s a straightaway on normal profile OCS on the Blue Line with poles at 180’, and one on the Green Line with poles at 220’. Meanwhile, the straightaways on the Expo Line’s low profile OCS max out at about 140’.

The main impact here as far as cost goes is that you need more poles, and pole foundations. As you can see, the comparison between the Green Line and Expo Line suggests a low profile system will need three poles for every two poles on normal profile system. That’s too high, because curves will still need closer pole spacing, as will special track work like crossovers. Nevertheless, using a low profile system can result in a considerable increase in OCS costs. The engineer should be able to give a rough idea of the impact of changing the system height for the project. The owner must decide if it’s worth the cost.

Finally, there is the single wire system. This is often the most preferred system by politicians and city boosters because it results in the fewest wires in the sky.

Naturally, single wire is also often the most expensive. The pole spacing is reduced because without the messenger wire, you need more supports to keep the contact wire on an acceptable profile. However, the hidden costs of single wire systems are worse. With only one wire in the air carrying current, there is not enough ampacity (ability to carry electricity) in the system. To avoid the need for additional substations, power must be supplied to the contact wire at shorter intervals, and this is done with electrical feeders in underground duct banks. This can add significant costs.

Single wire systems are most suitable for low speed services in touristy areas, which is why you often see them for streetcars. They’re also suitable for tunnels and other places with constrained vertical clearances, where the cost of feeder cables and additional supports is less than the cost of increasing the vertical clearance to accommodate a simple catenary system.

Poles: Wide-Flange, Tubular, or Ornamental?

This is again strictly an aesthetic decision. Wide-flange poles are, as the name suggests, wide-flange steel beams, the kind you can practically order off the shelf. That makes them cheap, because you just say you need so many W this by that members of lengths X, Y, and Z. They’re also very utilitarian and you rarely see them in urban contexts. Here they are in action on the Northeast Corridor, where it would be hard for anything to outdo the New Haven’s old rusty trusses with outside utility overbuild.

Tubular poles (or tapered tubular poles) are a little bit more pleasing to the eye, probably because they tend to look like the poles used on old systems. They’re a little more expensive, but not terribly: this Central Corridor document suggests that the costs to go from wide-flange to tubular poles for 5 miles of ROW was just $1m. Of course, if you have chosen a low-profile simple catenary, you’ll have more poles, and the cost will be greater.

Ornamental poles will cost you quite a bit more, like any custom design. Here are a couple examples, one in Australia and one in St Petersburg (the Leningrad one, not the one that had a Madame Tussaud Wax Museum from 1963 to 1989).

A couple of final points: wood poles are always an option; I’m not aware of any modern system that’s using them but they are cheap. Of course, they’ll probably wear out faster than steel poles. The built-up truss style poles you see on older systems are labor-intensive to manufacture, and standard wide-flange steel beams are widely available. If material costs for steel go up, maybe pre-fabricated truss poles like these ones in Australia will be more cost-effective. (The source page for that photo has other good photos of wide-flange poles too.)

Mixed Metals

Another potential costs savings in OCS design is using aluminum for the messenger wire instead of copper. Aluminum has a lower ampacity, which means you have to use a slightly larger gauge wire, but this is more than offset by the cheaper cost of aluminum. Aluminum conductors are sometimes used in building construction to save money; for example, the Staples Center used aluminum feeder cables. On the other hand, this paper (registration required) recommends the use of copper to avoid the need to connect different metals electrically, which can create a galvanic couple. This page from the Indian Railways Fan Club says that Indian Rail tried aluminum contact wires but this was unsuccessful due to oxidation and mechanical failures. I’m not sure why they tried aluminum for the contact wire; copper is definitely the way to go there.

My knowledge here is very limited, but I’ve heard that there are places outside the US that use aluminum for the messenger wire and copper for the contact wire. Certainly, this is an option worth considering if the price of copper stays as high as it has since it surged before the 2008 financial crisis.


The OCS is just a cost of doing business, and it’s usually not a large portion of the total costs. The big overruns come from unnecessary project elements, labor inefficiencies, differing site conditions and associated delays, change orders, and so on. But, while the cost of a house is mostly determined by the size of the house, a lot of little extravagances with the finishes can noticeably drive up the costs too. Do you really need marble floors in the bathroom?

As an engineer, I usually come down on the side of cost efficiency. If it was up to me, you’d get nothing but standard wide-flange poles and normal profile simple catenary. No one’s life is going to be measurably worse from having to look at those poles with two wires, and no one’s life is going to be measurably better by getting to look at an ornamental pole with one wire – certainly not in comparison to having or lacking access to good transit. At least, that’s my opinion. If your city decides it’s better to have fancier looking things, that’s your collective choice. Just have an open and honest discussion about the costs.