Light Rail Lines

I just spent a pleasant couple of hours looking at light rail lines. There are a surprising number coming in at greater distances than I expected. I thought the 'minimal operable segment' would be about 15 miles. I did not see that very often. Lines actually coming in between 7-20 miles, with the 18-20 range well attested to. (Not counting the bizarre Tacoma Link at 1.6 miles). Difference in initial distance seems to be the availability of an existing rail corridor. If you've got one, it's easy to go longer. Without one, using ROW from a freeway, or street right of way, the costs and complexities are higher, and the initial line tends to be shorter.

Transit Distances & Cost

Browsing planning documents, I came across one by the city of Winston-Salem, talking about the potential for a BRT or light rail between itself, and the nearby city of Greensboro. Rough distance is about 25 miles, which is far too far for a light rail line. There are certain numbers, that when I hear them, I know that the people proposing haven't done their research (or hired someone who has). Just some of my personal rules of thumb, for 'starter lines'.

Streetcar: 1-4 mile.
Light rail: 8-20 miles.

I forget where, but I once saw a graphic of all operational streetcar lines, super-imposed on one another. They were tiny, most of them between a mile and two miles. A rare few got as high as three. The Fed's funded Albequerque, which is four miles long, but they already had a 1 mile 'vintage' trolley they had been operating for the better part of a decade.

Charlotte has a plan for a 10 mile long streetcar LINE. That's not a line*--that's a system. It's a ridiculous distance to try to build at once. Streetcar networks get built one line at a time, in small segments, not enormous mega-projects. Certainly, I understand the political calculus of it. Charlotte is not going to tell the other members of its funding coalition the they aren't going to get their 'part' of the streetcar line for twenty years, during which time it may be 'delayed indefinitely' (and never built). But that's just bad transit planning.

The numbers break the bank. A 'cheap' light rail costs $25m/mile, with the average being about $35m/mile. At the $35m price, a 20 mile line runs about $700m. Assuming the Feds pick up half the price tag, the local city/MPO/transit agency still has to come up with $350m worth of NEW money.

That's not counting operations. Operations cost data is (irritatingly) mostly available in formats like "cost per passenger" or "cost per passenger mile". Cost per hour operating costs for vehicles is rare, which is irritating, because that is what is needed to determine actual operating costs. Calgary gives $163/hour for its light rail, and is considered cheap at that price.

Let us start with our hypothetical 20 mile train. Let us be (extremely) generous and assume (average) 20 mph operating speeds (actual average is 16.4), for stations a mile apart. At 20 mph, a train can go between stations in about 3 minutes. So in 15 minutes, that's five stations. Assuming 20 stations, that's 4 trains per hour, per direction (8 trains total). So that route requires 8 hours of train time every hour, at a cost of $163/hour, for 14 hours a day.  That's about $18,000 a day. Assume the train runs 6 days a week, 52 weeks a year. Annual operations cost is per mile is $5.5m.

$350m to buy the train and $5.5m to pay for the cost to run it. Assume that the initial cost doesn't have to be collected all at once, that it can be financed. UTA was paying about 15%, and has a pretty stellar credit rating. Borrowing $350m at 15% over 30 years means an annual payment of about $50,000, so such a train would cost $6m annually.

Which is a very large number, and not money most places just have lying around. The State of Utah had about $1.5b in taxable sales last quarter. Obtaining an extra $6m would mean raising the tax rate by .42%, or about .5%. So the sales tax rate would go from 7.0% to 7.5% to fund such a light rail system. That's politically difficult to achieve. For a County with an economy half the size of Utah's, (.75b in taxable revenue), double the tax raise would be necessary. Asking for a 1% rise in sales tax at the ballot box tends to fail. Half percent increases are more politically feasible. UTA got the first TRAX line built on .25%, and the expansion done on another .25%.
*A 'line' is a functionally independent unit of a transit network, that could operate on its own, cut off from the rest of the network. Most places are starting their rail networks with a single line, and building their networks from their. Each subsequent line is easier because of the network effect. By connecting by already accessible destinations

Rail Transit Distances

How far does each mode go?

Streetcar:  1-4 miles   (Sugarhouse)
Light Rail: 8 - 15 miles  (TRAX)
Commuter Rail: 20-40 miles  (FrontRunner)
Inter-City Rail: 60-300 miles (Amtrak Acela?)

So Winston-Salem to Greenboro is really talking about Commuter Rail, not light rail or BRT. Funny to think that UTA has now stretch FrontRunner from Ogden to Provo. 88 miles, give or take. Stations every ten miles. While TRAX has stations every mile or so. Ergo Inter-City rail has stations every 100 miles or so? Acela stopping every 30 miles or so.

Average Speed & Travel Speed

This should be obvious: Average speed is not the same as maximum speed.

If I drive between my house and the University along the freeway, I'm reaching 80 mph (and traveling at that speed for a goodly portion of the distance). I think of my 'travel speed' as 80 mph, regardless of the time I spend waiting at lights, or driving cross-town to access the freeway. But when I use the distance traveled (~17 miles) and actual travel time (45 minutes) to calculate* my rate, I find out my average speed is much, much lower--about 23mph.


When transit planners talk about transit, they habitually talk about the AVERAGE speed, rather than the maximum speed, but fail to make a distinction between the two. For a bus, an AVERAGE speed of 8-10 mph is normal. 12 mph is really really good, while 4 mph sucks. A back of the envelope calculation on the AVERAGE speed on TRAX between Sandy, Utah and Downtown SLC is about 30 mph.

*Distance = Rate * Time, thus Time/Distance = Rate

EmX goes BATS

Normal buses also have to deal with the consequences of running within a stream of automobile traffic. 
To avoid snarling traffic, buses must pull off the right of way, in a 'bus pullout'. It prevents them from slowing automobile traffic, but has significant consequences for the bus.  First, the bus must slow down and turn out of the travel lane rapidly, jolting passengers on board. Second, the bus must wait for a gap in traffic to re-enter the lane, which substantially slows the bus.

The EmX suffers from none of these flaws. The EmX doesn't quite have it's own lane, but it has a 'BAT', a 'Bus and Turn' Lane. No cars permitted, unless they are turning into either a curb-cut, or at an intersection. Not quite as good as a full lane, but certainly better than a shared lane.

It also seems to have been cheap and easy to create. It looks like the EmX's BAT lane used to be the 'safety area' at the edge of a high-speed arterial. The safety area is a 8' wide stretch of pavement between the outmost travel lane and the curb, and is designed to let drivers veer several feet without either running off the road or running into anything. More than wide enough for a lane, and many safety areas vanish as they are re-striped as lanes as traffic volumes grow.

EmX - When curb quality counts

Normal bus stops are just bad. To pick up passengers, a bus must leave its lane, lumber into into a slanted, debris filled gutter, jolting and tilting everyone on board. And they must approach slowly so as not to spray waiting passengers with the mud and slime in the puddle.

The EmX stations are a strange hybrid between normal bus stops and transit stations. They are a contiguous part of the sidewalk, (so no land or ROW costs), but with very different curbs.The edges of roadways are typically described as 'curb and gutter', reflecting their duel function in preventing cars from running off the road and in channeling water. (Roadway engineers design a 'hump' into the road it aid the latter function).

The EmX curbs are 12" high (compared to a normal 6" or 8" curb), and covered gutters. For normal buses, entering the debris filled and slanting gutter means jolting everyone on board, and splashing anyone at the stop with the water in the gutter. It also leaves a 'gap' between curb and bus floor that's impossible to cross in a wheelchair. In contrast, the EmX stations make it possible for the bus to roll up right next to the curb, with minimal gap. The taller curb matches the hight of the bus floor almost exactly. Wheelchairs can roll on and roll off without the need for a lift.

Why Favelas Exist

Favelas are generated when rapid urbanization causes population to outstrips the existing housing supply. The rapid growth in population drives up the value of housing. In a normal market, that triggers the development of additional housing until the two equalize. But  the essential input of housing is urban land. As the value of housing is rising, so is the value of urban land. Because of the durability of buildings, urban land can be developed only once per generation, so it makes sense to wait to develop until the value increase in urban land has stopped/slowed. While the present  value of urban land is high, the expect value of urban land in the future is higher still. As a result, rather than developing a property to its present ‘highest and best use’, landowners instead hold out for the future, when the highest and best user will be even more profitable.

With low land taxes, the cost of ‘holding out’ and waiting until a later time to develop is actually very low. 
Because the majority of the value of a land is actually  in its future value, the present value of land is largely irrelevant (especially in the context of high inflation), so interim uses are irrelevant—squatters and slums do not matter, insofar as they cannot claim legal title to the land, and can  be removed/cleared when the time is appropriate to develop the parcel.

Stair-Step Walking Distances

A study of Seattle-area park-and-ride lots found that for suburban lots, 50% of
the park-and-ride facility’s demand is typically generated within a 2.5-mile (4-km)
radius of the facility, and that an additional 35% comes from an area defined by a
parabola extending 10 miles (16 km) upstream of the lot and having a long chord of
10 to 12 miles (16 to 19 km).(R28) This market area is illustrated in Exhibit 3-8(a).

I need to make a diagram that is an analogue of this for walking. The 'stair-steps' of distance decay, and their trip-generation potential. The half mile radius around transit stations is simply not the whole picture. It's an average, rather than a distribution. And if there is a big, nice pocket of density, just beyond the half-mile circle, it's not going to capture the full effects of transit station accessibility.

Bus vs. Rail

I am going to punch the next feckless moron to conflate the costs of mixed-traffic bus system with a dedicated running-way rail system. It's simply not an apples to apples comparison, in either cost or quality of service. One is a Buick, and the other is a Cadillac. While both enjoy the same potential ridership (in terms of the built environment), there is a vast difference in performance. To call out one aspect in particular: system delay. As the Transit Capacity and Quality of Service Manual is good enough to point out:

This point of view also includes measures of
facility capacity in terms of the numbers of transit vehicles or total vehicles that can
be accommodated. Because transit vehicles carry passengers, these measures also
reflect the passenger point of view: passengers on board a transit vehicle traveling at
an average speed of 12 mph (20 km/h) individually experience this same average
travel speed. However, because these vehicle-oriented measures do not take
passenger loading into account, the passenger point of view is hidden, as all vehicles
are treated equally, regardless of the number of passengers in each vehicle. For
example, while a single-occupant vehicle and a 40-passenger bus traveling on the
same street may experience the same amount of delay due to on-street congestion
and traffic signal delays, the person-delay experienced by the bus is 40 times as great
as the single-occupant vehicle.

It's not a question of bus versus rail, (as many BRT projects are empirically proving) but a matter of right of way. Dedicated running-way provides value, but at a price. Whether the value is worth the price depends on the project, not the technology. 

Schedule Span

Catching up on Wikipedia transit updates, I happened across this chart of annual ridership, and ridership on a per-mile basis.  While I was pleased as punch about TRAX's success (#9 in the nation), I was less psyched about the ridership per mile. Reading up on other systems, something about the Seattle's 'Central Link' struck me--"Service operates seven days a week, from 5 am to 1 am Monday through Saturday and from 6 am to midnight on Sundays".

I spent most of last night reading chapter 3 of the 'Transit Capacity and Quality of Service Manual, 2nd Edition'. One of the ideas it brought to my attention was the idea of 'schedule span', which is the transit equivalent of hours of operation. The Central Link is operating 20 hours a day. TRAX calls it quit at about 10:30-11, and it starts later! I think that's reflected in the per-mile ridership numbers--almost 2000 vs. just over 1500 daily riders/mile. Central Link is 33% higher than TRAX, for ~33% higher hours.

I've previously commented on the lack of late-night service for TRAX, which (as one commenter noted) has actually been declining over time. Anyone who has read the history of transit is familiar with the ridership sabotaging effects of reducing service. Less service means fewer riders, which means less money, which means less service---it's a downward spiral.

Miles of track are expensive, running between $60m and $100m per mile. We should make as much use of it as possible.

Great Cities and Their Traffic

Thomson* notes an inverse relation between the number of center city jobs and the number commuting by car. The implications of Thomson's data are significant because of it's age. Most of it is drawn from 1963 and 1973. Detroit had only 80 thousand jobs (Thomson 35). That makes some of the  'center cities' of Thomson's age are equivalent in size to Garreau's 'Edge Cities' of today. Automotive sprawl had become the dominant land use pattern by 1970, and the public tram-ways were long since replaced by buses. Yet the relationship remained...the question remains--is it a matter of share of total regional jobs, or merely the geographic concentration of jobs?

*Thomson, Michael, J. Great Cities and Their Traffic. London: Victor Gollancz ltd. 1977

Growing a Transit Network

As far as I can tell, building a good rapid transit system is pretty simple: Good planning and incremental investment. That seems to be why San Diego has done well, Salt Lake has done well, and Denver's Fast-Track is a bit of a boondoggle. Doing it incrementally makes it possible to apply 'lessons learned' to subsequent expansions. It could be argued that SLC's last expansion (4 lines at once) 'overdid it', but I don't think so. While it was funded as a package, UTA has been pretty clever in phasing the construction time of the different lines so that everything hasn't been happening all at once. The South Jordan and West Valley lines started service on the same day, but neither the Airport nor Draper lines have completed construction. I suspect that made it possible to re-use construction equipment/staff time on the different lines.