A third train could use the Borealis improvements

A stringline plot of currently-scheduled train trips on the St. Paul–Milwaukee–Chicago corridor shown in red, with three options for extending Hiawatha trains that would benefit from the track improvements in Winona through La Crosse

Yesterday, Star Tribune reporter Greta Kaul posted:

That's great news, especially on the ridership front! I am, however, very accustomed to politicians and reporters misstating and misinterpreting anything and everything about train services, so I'm hoping there will be a follow-up article clarifying a few things. For today, I'll take it as a win and hope that we'll soon see another trip pair following the current route.

While writing my previous article on where track improvements are being made with the initial Borealis-related funds, the thought did occur to me that future trips on the corridor could make use of those new sidings and rearranged rails between Winona and La Crosse, even if the current Borealis trips are largely moving through there without trouble.

Since the Borealis is an extension of Hiawatha trains 1333 (westbound/northbound) and 1340 (eastbound/southbound), I took a look at other Hiawatha trips to see which ones would be scheduled to pass by other Amtrak trains in that general area.

Train 329 which departs Chicago at 6:10 am would meet eastbound Empire Builder no. 8 around there, and train 331 departing Chicago at 8:25 am would meet Borealis train 1340 in that zone. Assuming the schedules are duplicated exactly from existing Borealis train 1333, they would be scheduled to arrive at St. Paul Union Depot at 1:39 pm and 3:54 pm.

In the other direction, extending Hiawatha train 342 back to St. Paul would have it departing from here at 1:40 pm, and it would meet westbound/northbound Borealis train 1340 right around La Crosse.

These are not the only times that an added train could run, but are ones that would best leverage the existing investment.

If Amtrak is only able to provide one set of train equipment, the best option here would probably be to use the 6:10 am train 329 from Chicago, then have it only stay in St. Paul long enough to head back at around 3 or 4 pm, and arrive back in Chicago in the 10:30 pm to 11:30 pm time frame. That would add a later trip between Milwaukee and Chicago on the Hiawatha corridor, though who knows if that trip would require additional track improvements there. That's somewhat appealing since the return trip roughly mirrors the Empire Builder schedule from Chicago to St. Paul.

I'd like to see some morning arrival options for St. Paul, though I'd be hesitant to try and run a night train with Horizon coaches, for the sake of passenger comfort, but I'm sure there would be people who'd take advantage of it.

There are many different possibilities for when a third daily train pair could run, and that's not even counting the potential variations through Eau Claire and Madison that are in the planning pipeline. Both of those will require significant investment in track rehabilitation (and the most straightforward Eau Claire routing actually has a gap of about 10 miles where track was removed and needs to be fully replaced)

I'm anxious about funding and, well, ~everything~, but let's hope things really can happen quickly this time around. Next year, when Winona–La Crosse track improvements are done? Sure would be nice.

Where Borealis improvements are being made

The train station building in Winona, Minnesota. The nearby rail siding will be rebuilt to mainline standards as part of upgrades to accommodate the new Borealis service.

As part of the startup of the new Borealis service on the Twin Cities–Milwaukee–Chicago, track improvements are planned near the stations in Winona, Minnesota and La Crosse, Wisconsin, as well as across the Mississippi River from La Crosse in La Crescent, MN. The primary host railroad, Canadian Pacific Kansas City (CPKC), is requiring upgrades to be made to preserve their capacity to run freight trains.

In isolation, these seem like good places to upgrade tracks—especially La Crosse and La Crescent, since they bookend a long, single-tracked river crossing that's limited to 25 mph. Frustratingly, these areas don't seem to be major contributors to congestion or delays for the Borealis itself, which makes me question whether these are really the right places to start. But, before I complain more about that, let's get into where things are happening.

The map above shows the general areas expected to receive upgrades, according to what was presented by WisDOT to area residents in November 2022. (WisDOT has been tasked with public engagement for all three cities about the upgrades, even though much of the work is happening in Minnesota.)

The areas to be changed look pretty small for a 410-mile route, even with a view focused just on the Winona–La Crosse area, though they do add up to around 8 miles of added or changed track.

Winona

Moving from west to east, the first area expected to be altered is around a spot known as Tower CK, where a short bit of Union Pacific Railroad track (an isolated part of the Winona Subdivision) branches off from the CPKC main line (their River Subdivision). The Union Pacific branch as well as the CPKC's short Third Street Spur nearby will have a short bypass added by extending a nearby siding.

Planned changes at Tower CK in Winona. The upper image is of an area slightly northwest, showing the junction with the Union Pacific Winona Subdivision, and the lower one shows the Third Street Spur connection. (PDF)

The purple triangle is a foundry property that has a small part of its building and some land behind it that might be taken as part of the upgrade.

Extending the siding so it goes past these two spur connections is a good idea, though I'm kind of baffled by the inclusion of a pair of crossover tracks in between the two spurs, only half a mile from the new end of the siding. It seems like it would have made more sense to have the Third Street Spur branch off from the UP's connection

It seems to me like this would have been a better idea, which would have removed a switch from the mainline rather than adding more of them:

An idea for removing the direct connection of the Third Street Spur from the mainline (blue) and adding a short stretch of track off of the Union Pacific line instead (yellow)

But maybe the Union Pacific rejected that, or there were other challenges. Or it's still possible that the design has been changed to something other than what was provided in 2022.

The next section is an upgrade to an existing 2.5-mile siding beginning near Winona station and extending southeast to the edge of the city. The switches for the current siding are manually controlled, so this would upgrade that to proper mainline centralized traffic control (CTC), along with appropriate signaling. The track itself should be upgraded to continuous welded rail instead of jointed rail. Apparently the grade crossings are also planned to receive improvements, though I peeked at them in Google Maps, and they looked like they already had the configurations shown here, so there must be some subtlety with them that I'm missing.

Three panes showing the upgraded siding at Winona station and points southeast. Top pane is farthest northwest, and bottom pane is the southeast end. Blue spots at the ends of the siding indicate where low-speed manual switches will be replaced with higher-speed automated ones. (PDF)

It's a bit strange that there's another siding here, less than two miles away from the siding that's just west of Tower CK. They're usually spread 6 to 10 miles apart on the rest of the line, but the Winona area already has three signaled sidings in the immediate vicinity: One north of Minnesota City, the one in Goodview that ends at Tower CK, and another one just southeast of Winona around Homer—I guess the next step will be to someday fill in the short 1-2 mile gaps between them and make the line properly double-tracked again through the area.

Preparation for eventual full double-tracking of the line through Winona might explain the double-crossover at Tower CK, since that would probably be a good place for one to be added once that happens. It's still a head-scratcher to do it now, though.

But back to this siding. I think the greatest benefit of upgrading the siding by the Winona station will be for passengers, since Amtrak riders currently have to cross over the siding to reach trains that stop here. There are nearly a dozen crossings like this at the station, with panels like what you find at a grade crossing, and unsightly jersey barriers blocking the rest of the tracks.

An Amtrak conductor assisting passengers as they board the Borealis on May 26, 2024. Note the panels over the tracks, allowing passengers to board across the siding track nearest the station. The siding can't normally be used due to manually-thrown switches at the ends of the siding.

Allowing Amtrak to actually use the track next to the station will make it much easier to build a more ADA-compliant low-height platform, something that the company is doing all across the country due to a settlement reached in late 2020. Winona station is due to receive upgrades for that as well, though it's not clear what the exact design will be (work on that project appears to have been dependent on when these track upgrades would happen). It would be nice if the station gets upgraded to have platforms on both sides of the tracks, though that probably won't be necessary until additional daily trips are added along the corridor.

I haven't found any information yet about planned design features for the Winona station, other than it's expected to get a better platform, better lighting, better connecting paths, and things like that.

La Crescent

On to La Crescent, where there's what I like to call a "super-wye" and a "mini-wye," although the latter is just a regular-sized one. At the north end of a large triangular arrangement of track is a point known as River Junction, roughly marking the end of the River Subdivision (although the actual end appears to be a few miles north). Continuing straight south, trains enter the Marquette Subdivision, which continues along the river into Iowa. The Tomah Subdivision begins slightly north and veers east through the junction, then meeting a short east-west connection from the Marquette Subdivision at Bridge Junction, where it of course leads onto the Mississippi River Bridges that carry the line through La Crosse and let it continue east to Portage. 

The Marquette Subdivision leg of the wye includes the River Junction Yard. A number of changes are planned here, including converting one track to more easily allow through-running.

Two images showing the River Junction Yard area. The top image is the south end of the yard, and the bottom image is the north end (yard in the lower-right). North is to the left in both images. A thin yellow line shows a track that will be added/changed to a through track. Other tracks in the yard will be trimmed back a bit on the south end. It appears the westernmost track will also be set up similar to a through track. (PDF)

The changes there will go hand-in-hand with changes on the south leg of the super-wye, where the mini-wye also joins the Marquette Subdivision to the Tomah Subdivision.

Bridge Junction (inset, north is up) and the La Crescent wye and south end of River Junction Yard (rotated >90° counter-clockwise, north is to the left). (PDF)

In part, these changes are intended to allow trains to move faster through these junctions. Apparently freight trains are currently limited to just 10 mph when turning through Bridge Junction, and the limit may be similar for the La Crescent wye. I was a bit surprised looking at these plans that the proposed Bridge Junction curve actually seems sharper in spots, which you'd think would limit speeds more, but the real limiting factor has been the turnout curve of the switch.

Making a straighter segment for the switch itself and slightly sharpening the curves around it apparently will allow freights to move through it more quickly. It'll only increase their speeds to 15 mph, but that's a 50% boost for them. Replacing the switches on the La Crescent wye should allow freights to move at 25 mph that way.

This feels like the most significant improvement I've seen so far for actually speeding up trains, though I have reservations about whether now was the right time to implement it. I certainly would have preferred more smoothing of that Bridge Junction curve, such as if it could allow freights to go through there at 25 mph as well, but that would have required some real earth-moving, and would likely need extensive environmental review due to the nearby wetlands.

At some point, a much faster river crossing should be built, which will make these tweaks look pretty small.

La Crosse

Finally, we get to areas around La Crosse station.

Changes planned for the area around La Crosse station in the upper pane, and additional track changes leading into the nearby rail yard in the lower pane. The second pane shows an area northeast of the first. Some of the coloring for track changes seems incorrect for this one. (PDF)

While it looks like this area currently has two main tracks, I believe the second one is really just a lead track for trains getting assembled at the southwest end of the yard. This area will be upgraded to have two through tracks, and the lead track may be usable as a main track as well, as this example shows double-crossovers just west of the yard.

Before 2000, the station appeared to have three or four platforms, though I have a vague recollection of visiting for a school field trip and seeing everything in a very rickety condition. The station had once been planned to be a new Union Station for La Crosse, which ended up having at least five stations. They never consolidated operations, so those platforms probably went unused by passengers, but might have helped freight train crews.

After decades of decline, track was consolidated to the current configuration (or something close to it). A new platform was built roughly where the second platform had been before, and the gap between the station and the platform was filled in.

It's surprising that this layout didn't plan to undo that infill, changing the existing platform into an island between two tracks, although it would still need to be rebuilt higher to meet newer ADA requirements, and I'm not sure if a higher ADA platform would fit with the historic awning in front of La Crosse station (though if that was the issue, it likely says more about historic preservation rules than ADA requirements).

So, I'm a bit concerned about how this new layout will work if service reaches a point where two trains could be boarding here at about the same time. Perhaps with that crossover, they intend the empty area across the tracks from the station to be used as another platform. It would be a bit strange to have 3 tracks sandwiched between two platforms, but it could work.

Most of the public engagement around the La Crosse station has to do with the changes to sidewalks and roadway crossings, which create significant challenges. There are quite a lot of properties that may be impacted, either just during construction or permanently, although it's generally just a matter of a few feet of space being needed.

Impact on Borealis train performance

After going through this exercise, I'm feeling a bit better about these changes, and seeing some of the positives, but I'm still skeptical that they're very impactful in terms of traffic flow as things stand today. Is the real-world performance matching expectations from the models that drove the decisions to update these particular spots?

The Winona station siding is bubbling to the top of my list as a useful improvement, since it facilitates building a better platform. But it also would have been fine to build a platform on the opposite side of the tracks from the station building. As long as there's clear communication with passengers about where to be and when, along with a short path to get across, that could work well.

I don't have good access to details on exactly when freight trains move through their network or on the other lines that share or cross their tracks, so it's difficult to say as an outsider how this one new train affects their operations.

Delays by station for the eastbound Borealis train 1340 from start of service until June 22nd showing delays mostly happening between Columbus/Milwaukee and Chicago. (interactive link)

Eastbound, there's not much delay happening getting to La Crosse. Going by median values, trains are delayed by about 8 minutes shortly after getting out of Union Depot. By La Crosse, that delay has only increased to 9 minutes, and the run to Tomah regains 3, so the median delay there is only 6 minutes.

Delays by station for the westbound Borealis train 1333 from start of service until June 22nd, showing delays mostly happening between Milwaukee and Columbus and between Red Wing and St. Paul instead of around Winona–La Crosse. (interactive link)

Westbound is a similar story, with trains having a median delay of 27 minutes in Tomah, 26 minutes in La Crosse, 24 minutes in Winona, and recovering one more minute arriving in Red Wing.

When studies for a new TCMC train came back around 2015, they arrived with eye-watering cost estimates of over $200 million in infrastructure costs just to add a single trip pair, with another $46 million tacked on for the trains themselves. Through years of work, the overall cost was trimmed down to $53.3 million.

It may be that the train volumes in this area are unusually low lately. While recent filings with the Federal Railroad Administration's grade crossing safety system indicate there are 21 trains per day on the northern part of the River Subdivision, that apparently drop to about 11 trains per day at the south end. Submitted data suggests only about 4 or 5 trains per day on the west end of the Tomah Subdivision, and 13 trains per day on the north end of the Marquette Subdivision.

The merger of Canadian Pacific with Kansas City Southern is expected to add six or seven trains per day running north-south on the River and Marquette Subdivisions versus the pre-merger conditions. The River Sub is expected to reach about 18 trains per day and Marquette is supposed to see about 14.

One of the big purported benefits of the merger is that freight customers' cargo can bypass Chicago by running along the Mississippi instead, so we'll see if that ends up fully coming to fruition, and whether that really makes these investments worthwhile.

But, it certainly gives me an uneasy feeling that we've spent more than a decade of time from advocates and well-intentioned government officials only to find that this project may have done more to facilitate a $31 billion railroad mega-merger than to actually get passengers where they need to go on time.

The curves of the River Route

Sharp curves along the Mississippi River rail route used by Amtrak's Empire Builder and Borealis. Red, orange, and yellow denote zones where curves dictate speeds of about 75 mph or less. Most of the route is currently constrained to 65 mph for passenger trains instead of the usual 79 mph limit.

I've recently been trying to wrap my head around the speed zones for passenger trains between St. Paul and La Crosse along the Mississippi River. A post from Korh on the UrbanMSP forums got me interested in taking a closer look at track curvature and the concepts of superelevation (banking or cant) and cant deficiency (underbalance, the amount a vehicle wants to swing outward).

I dug up Alon Levy's post about calculating speeds, and ended up using Ari Ofsevit's method linked at the end, which involves tracing lines from the outer rail to the inner rail and back out again to create chords to estimate the degree of curvature.

Example of drawing a chord between outer and inner tracks as a way of deriving approximate curvature, from Ari Ofsevit's spreadsheet

While observing Borealis trips on the asm.transitdocs.com tracker, such as this run on June 12th, I'd noticed that much of the route appeared to be limited to 65 mph, with only limited sections allowing 75 or 79 mph.

I was reminded today that the Midwest Regional Rail Initiative (MWRRI) had produced a report for 110-mph train service in the corridor, which included travel time modeling graphs. The graphs also show the attainable speeds for standard Amtrak trains in light green, confirming the mostly-65 mph speed zones:

Passenger train speed plot from Tomah, Wisconsin to north of Red Wing, MN showing a hypothetical 110-mph train in dark green (limited to 90 mph along most of the river corridor) and what a current Amtrak train speed would be in light green. (One of three graphs, travel time is likely for Milwaukee through St. Paul to Minneapolis)

 

At the speeds we're concerned about here, increasing superelevation by one inch or allowing one extra inch of cant deficiency in the rail cars translates to a speed limit increase of about 5 mph. For example, going from a combination of 3 inches of superelevation and 3 inches of cant deficiency to 4 inches of superelevation and 5 inches of cant deficiency would likely allow a 65-mph zone to become a 79-mph zone.

Selected probable speed limits given a combination of superelevation and cant deficiency in inches, color-coded to match the map at the top of this article. Increasing by the equivalent of 3 inches increases possible speeds by 10 to 20 mph in most cases. See the full calculator.

These appear to be the main factors that the MWRRI study used to justify higher speeds along the Mississippi River route—they planned to increase the height of the outside rail to 4" of banking whenever possible (the starting point is unclear, though they do mention 2.5" as an "optimal" amount on page 3-5 of the main study document), and hoped to gain another 2" from enhanced cant deficiency of tilting trains, going from a standard of 4" to 6".

Partial description of the guidelines used when modeling train speed and trip time, from the beginning on page 4-1 of the main MWRRI study document

Even with that, they only planned to have 90-mph speeds along the river, with a lot of curves requiring drops down to about 85 mph (although they neglected a few spots where 110-mph running appears reasonable: around Kellogg, around Frontenac, and between Red Wing and Hastings).

Alon Levy argues against tilting trains these days, saying they don't really provide enough speed benefit to justify their extra maintenance cost. They also note that other countries are running non-tilting trains with 180 mm (about 7") of cant deficiency, which is the same as what Amtrak has been allowed on their tilting Talgos and Acelas. (One of the limiting factor there may have been the heavy non-tilting locomotives rather than the passenger compartments, which are capable of much greater tilt and underbalance.)

As Alon notes, cant deficiency is primarily an issue of how the suspension system of a train behaves, so equipment with higher centers of gravity and softer springs/dampers will tend to swing further outward on curves than other train cars. Alon referred to a "magic HSR waiver" proposal, which might be in this federal register notice, where Amtrak proposed allowing higher cant deficiency on their trains following testing. In that notice, it suggested operators could apply to allow 4" on Superliner equipment like on the Empire Builder and 6" on single-level equipment like what the Borealis runs.

This Federal Railroad Administration freight+passenger superelevation study from 2019 didn't include rail height standards from Canadian Pacific, but did note that they limit cant deficiency to 3". Considering current speed limits on the river route, they presumably limit most curves to 2 to 3" of superelevation.

I believe the only Amtrak trains that have been using significant amounts of Canadian Pacific track until this year have been the Adirondack in New York, the Hiawatha between Milwaukee and Chicago, and the Empire Builder on our familiar route. It's quite likely that they've set standards based on the Builder's Superliners, so it would be worth having MnDOT, WisDOT, and even NYSDOT (since the Adirondack uses single-level trains) get together with CPKC and the FRA to determine if 1" to 2" of additional cant deficiency can be allowed with existing equipment, boosting speeds by 5-10 mph on curvy sections.

I haven't been able to find any allowable cant deficiency information for Siemens Venture / Amtrak Airo passenger equipment that is being planned/procured for many lines across the country.

There are some true limits when it comes to cant deficiency—you don't want a rail car to swing outward from an inner track to hit a slow or stopped freight train leaning inward, for instance. However, that's not an issue on single-tracked parts of the route, so those sections would be the best ones to increase speeds on to begin with. Most of the single-track sections along the river are 6 to 10 miles in length, allowing several miles of faster running at a time before needing to worry about adjacent freight cars.

If half of the speed gains on this section were just supposed to come from better equipment rather than major investment in the track itself, we should make sure we're using our existing and planned train cars as effectively as possible. It doesn't remove the need to maintain track well enough to remove bumps and wobbles on the smaller scale, but it would substantially lessen the need to increase banking on curves.

The flatness of California HSR, vs. someday Driftless rail

I recently did a scan through California's Central Valley on Google Maps to trace out the California High-Speed Rail construction zones currently visible there. The imagery there will be lagging the real world significantly, but shows a nearly continuous stretch being built from Shafter, CA outside of Bakersfield up to Fresno. It's mostly just right-of-way and parts of bridges so far, with a few that are structurally complete.

What struck me is how unbelievably flat the area is. The default view of this embedded map uses a "terrain" background layer, and it's hard to find any topographical contour lines (Depending on zoom level, Google appears to mark differences of 40 or 80 feet). You can change the background to an aerial ("satellite") view by clicking the upper-right icon to pop out the sidebar, and then an icon should appear at the bottom of that to use an alternate backdrop.

Someday we'll need true high-speed rail on the Twin Cities–Milwaukee–Chicago corridor, which will involve getting through the Driftless Area of southeastern Minnesota and western Wisconsin. Even getting the Empire Builder / Borealis route upgraded to higher speeds will require dealing with it (those trains are mostly limited to 65 mph between Red Wing and La Crosse / La Crescent today, and historic trains only got to between 70 and 80 mph for most of that segment).

The area is pretty flat compared to much of California, but will be a much bigger topographical challenge than the Central Valley, considering the bluffs were created by the Mississippi River and its various feeders digging a channel that drops more than 600 feet below the surrounding landscape.

It's nothing compared to the roughly 3,400 feet that California HSR will need to climb from Bakersfield to the Tehachapi Pass, or the apparent 2,500-foot drop from Palmdale to Burbank (mostly happening in tunnels under the San Gabriel Mountains). But, the soft karst (dolomite, limestone, and sandstone) geology has made many small valleys that could create the need for lots of smaller bridges and tunnels to run through and between them.

On one hand, we've had some very good experiences tunneling in similar conditions at MSP airport, where the twin 1-mile tunnels apparently only cost about $117 million (probably 2004 dollars, which is still a bit shy of $200 million today if going by consumer inflation), but we've also seen costs skyrocket for Southwest LRT, which has had to deal with much swampier conditions along much of its route.

Eroded bluff faces will probably be more challenging than the flat land the airport tunnels were built through, but hopefully the rock will still be favorable for tunneling through. I'm sure we'll have to create some significant ones if we ever hope to get 90-mph, 110-mph, or faster trains running along the river corridor.

Some potential for "hot rocks" in railroading

A set of switches and crossover tracks along BNSF's double-track mainline west of Minot, near Lonetree, North Dakota (Google Maps)

While looking around on Google Maps to better understand some aspects of railroad operation, these little white blobs began catching my attention in aerial shots. After a while, I saw through Street View that they are propane tanks. Okay, maybe fuel for backup generators? Seems strange, but okay.

Then, the other day, I went back to look at some of the infrastructure BNSF Railway built in North Dakota about a decade ago. The whole 100-mile stretch between Minot and Williston received major upgrades and became fully double-tracked. At the site above, they installed fairly high-speed switches featuring movable-point frogs, which allow faster train operation with smoother ride quality while reducing the chance of derailment.

There are four switches here (or turnouts, as they're sometimes called), and 8 spots where there are "switch machines," the motorized units that push or pull on a piece of rail, and also include interlocking components and sensors to ensure the switch actually moved into the correct position. Switches usually have one of these motorized units, just for the main switch points, but the movable frog also has a motor attached in this situation.

Moving a set of switch points back and forth does take some effort, but why in the world would you need four 1,000-gallon propane tanks for that? Looking them up, I saw they mostly used standard 110-volt or even 24-volt power, for which that amount of stored energy is extreme overkill. After a while, I finally remembered it's all about snow and ice in the winter, and the heat needed to fend that off.

Metra in Chicago famously uses open-flame heaters on certain switches in their network, particularly those at the busy A-2 interlocking near Metra's Western Avenue station. There, a set of three tracks crosses another set of four tracks at a shallow angle, each intersecting with double-slip switches.

The rear end of a commuter train that has just passed through the orange flames shooting up around switches at the A-2 interlocking in Chicago, with the control tower visible to the right (Metra)

There are multiple different ways of heating up rails and other components that make up switches, including direct flame like this, resistive nichrome wire attached to the side of the rail, and forced air systems that use blowers to push air across gas, electric, or infrared heaters and duct it throughout different parts of the switch.

It seems that forced-air systems are the most common, and that's the type of heater at the installation in North Dakota. Many forced-air heaters have thin gray ducts that run lengthwise through the switch area, and the blower/heater assembly and its main output duct sit off to the side of the switch.

Ducts running lengthwise between the points of a railroad switch (Rails Company via railsystem.net)

Diagram of a switch heater system. 1) Heater and blower. 2) Main duct 3) Distribution duct sitting under the tracks like a railroad tie with small angled ducts heat the tips of the switch points and 4) parallel ducts running lengthwise to heat the track bed and rails. 5) Switch motor being heated by a duct outside the rails (Rails Company via railsystem.net)

Metra says that the A-2 interlocking area is too tightly packed for forced-air heating systems to work. I'm skeptical, since the track is on an elevated embankment, and something could probably be piped underground, but it's certainly true that the standard setup with heater/blower units sitting off to the side of the tracks would be a no-go at that site.

Someday, those double-slip switches and their fiery heaters will disappear when the A-2 interlocking finally gets replaced with a flyover (a project that should also add another West Loop Metra station). Switches in other locations will still need to be heated, however.

I count 22 of the 1,000-gallon tanks at the BNSF crossover site, which represent a pretty massive amount of energy. Other similar sites seem to sometimes have 24 or even 32 tanks—three or four for every switch and frog. These types of tanks get filled to about 800 gallons, leaving room for fuel to expand and for vaporization to occur. Propane tanks are recommended to be refilled once they get down to 20% full, otherwise they're susceptible to freezing up, so that gives about 600 gallons of usable energy per tank.

A 1,000-gallon tank is big enough to supply a house with heat and hot water for a year or more, so it's stunning that a single heater may require four times as much fuel. While a large house might have a 100,000-BTU/hr furnace (about 29 kW), switch heaters may be rated at 400,000 to even 900,000 BTUs/hr (117 to 264 kW). At full tilt, one could be consuming up to about 10 gallons of propane per hour (propane contains roughly 91,500 BTUs of energy per gallon).

Steel rails are enormous heat sinks, of course, with these rails likely weighing 132 lbs per yard or more on their own, and being able to rapidly transmit any heat that gets dumped into them. While a lot of heat is needed to counteract that, these sites may also be sized according to some worst-case temperature conditions that affect the fuel itself.

Propane has a boiling point of −44°F (−42°C), so it's typically a gas at room temperature. Since propane tanks are pressure vessels that contain the boiled-off gas, the container is able to reach equilibrium at several to a dozen or so times atmospheric pressure, with most of the fuel staying in liquid form.

However, as the outdoor temperature drops, the liquid boils less readily and less vapor is able to be fed into the burner. As the temperature declines from 30°F (−1°C) zero to −20°F (-29°C), the vaporization rate drops by 90%, so only about 1 gallon worth of fuel becomes available as a gas per hour. Of course, vaporization devices exist to preheat the fuel, so it's more likely that these mini tank farms are sized for the seasonal heating needs.

Are there ways we could avoid burning stuff to keep these pieces of track free of ice? It's a lot of heat, although perhaps not quite so crazy considering the size of the area. The whole rail crossover site is more than a quarter-mile long, and covers around 4 acres, around the size of a small city block.

If only low-grade heat is necessary, heat pumps could probably work, though they would need to be tied to underground loops in order to be effective on really cold days. On warmer days, the heat pumps could pull heat from outside and charge up the underground reservoir somewhat, which would be an interesting idea to think through.

I originally started this article thinking of the growing number of companies doing above-ground thermal storage, also known as heat batteries. One that gets a lot of press is Rondo, who say their primary model of heat battery can store 100 MWh of energy, and is able to provide heat output at a rate of 7 MW, which would be ample for a site like this.

If it could be hooked up in the right way, that type of battery could provide heat for this site for several days even in very cold temperatures. Just using hot sand or rocks is one of the cheapest ways we've found to store energy. 

Propane is annoyingly cheap when comparing against green alternatives, especially in places like North Dakota, but it would be interesting to see if someone can make the financial cost worthwhile without even needing to think about the environmental benefits.

It's clear that there are technical options for making greener energy sources work for this purpose, though finding the right combination of energy transmission and storage solutions to make it cheap and robust will take some work to figure out.

Northstar v. COVID (and the people who've always wanted to kill it)

Passengers wait to board an approaching Northstar train at Ramsey station on June 4, 2024. About 30 people boarded from the platform for this trip.

When COVID lockdowns gripped the country and the world in March 2020, Metro Transit's Northstar commuter trains received perhaps harshest service cut of any commuter rail system in the US. On March 23rd, a week and a half after Minnesota stopped business as usual, Northstar service was cut from the already paltry 72 train trips per week down to just 20—a cut of 72% on its own, of course combined with the loss of special event service.

This chart shows relative service levels for Northstar and several other commuter services, as measured in vehicle revenue hours (number of rail passenger cars and locomotives multiplied by the number of hours in service), with the peak service level from the decade of 2013-2022 pegged at 100%: 

Data from National Transit Database agency profiles. Somehow, Northstar's service level varied by 15% in the years before the pandemic and peaked in 2014, though I don't know of schedule changes that explain that. Special event service only adds about 3%. The most likely explanation is that it's from changes in train length.

Those are considered as Metro Transit's peer agencies and routes in the Northstar Corridor Post-Pandemic Study from March of last year: Utah Transit Agency's FrontRunner in Salt Lake City, Sound Transit's Sounder in Seattle, Trinity Railway Express in Dallas–Fort Worth, the North County Transit District's Coaster in San Diego, and oddly the Amtrak Downeaster—somehow classified as a commuter train even though it has a route length and regional-rail service pattern similar to the planned Northern Lights Express to Duluth. 

A huge frustration with Northstar has always been that it began operation as a half-built service, and in more ways than one. We are all familiar with the idea that it was intended to go to St. Cloud or even slightly beyond, and that it got cut back to Big Lake—39 miles instead of an targeted 70 or 80. Less well-known is that it was also supposed to have a broader service schedule, with 18 one-way trips per weekday instead of the 12 that we got (plus the six trips each day on weekends), as noted on page 6 of the Post-Pandemic Study.

Text and graphic from page 6 of the Northstar Corridor Post-Pandemic Study from 2023

How substantial of a difference did it make reducing the planned service by ⅓? It's hard to say for certain, but ⅔ of the initial 4,030 weekday rider projection is 2,687, slightly above the 2019 observed weekday ridership of 2,660, and approximately the level that Northstar had at least touched by 2013. The observed ridership also grew by 46% from 2009 to 2019, above the 39% growth predicted for the period up through 2025.

A chart showing Northstar weekday ridership on an upward trend from 2009 until 2019, with peaks around halfway between 2,500 and 3,000 in the years 2013, 2017, 2018, and 2019.

The chart above shows a nice trend of growth over the years for weekday service, though a depressing decline on weekends where there was an oddball schedule of three round-trips, which was done by running a single trainset from Big Lake down to Minneapolis and back. I've never seen weekend ridership charted for Northstar before coming across this study, so it's interesting to observe how that would have dragged down the total number of riders each year, even if the weekdays with more service were improving.

From the beginning, Northstar had the lowest level of service of any route in the Post-Pandemic Study, and has some of the lowest service of any in the country. One of the only lines to have less service was the WeGo Star in Nashville (formerly the Music City Star). Unfortunately, that line doesn't report numbers to the National Transit Database, so I can't directly compare figures, but it seems to now be exceeding Northstar in terms of ridership, probably just because our service cuts were so severe.

Another thing stoking my indignation as I look through this information is that Northstar historically has historically had quite good utilization in terms of seats filled vs. the number of seats that are available.

Northstar had 54.76 riders per vehicle hour in 2019 (the chart is mislabeled with 2017), second only to Sounder at 60.98. In 2021, the worst full year for post-pandemic ridership among most agencies, Northstar was very close to FrontRunner, Sounder, and TRE passenger loads per car.

The National Transit Database now has another full year of data available vs. the 2021 comparisons in the study. In 2022, Northstar pulled slightly ahead of all of its peers, even though it was still running with just two round-trips per day.

Now, if Northstar was supposed to have been hitting a ridership target 50 or 100% higher than it initially did, it would be well off the chart, up around 82 to 110 passengers per hour. Caltrain had passenger loads in the 80s and 90s prior to the pandemic, but that's an unusual exception as far as I can tell. It seems to have been much more common for commuter lines to be in the 30–60 range.

Anecdotally, it had felt when Northstar opened that a lot of people got burned by early experiences with heavy crowds and decided not to return, at least not for a long time. I think there was a lot of demand for the line that went unfulfilled.

2,600 people per day might not sound like much, but that translates to 217 people per train (with a median a bit higher since the reverse trips were usually quiet). You'd basically need a bus every 5 minutes or better to match the capacity of the old Northstar schedule.

I've been surprised to see that the study only estimated the cost of extending Northstar to St. Cloud to be in the $36 to $67 million range, a small fraction of the $320 million it originally cost to build (or around $475 million adjusted for inflation). That's the equivalent of a few apartment buildings with several hundred units total, and yet the train has spurred the creation of more than 3,200 housing units already.

Northstar has always had a noose around its neck from the politicians who've hated it from the beginning. They like to rag on its high subsidy per passenger, but that's driven almost entirely by the high operating costs to begin with, and the length of the trips (per passenger-mile, Northstar has subsidies somewhat higher than light rail, but much lower than urban buses). The limited schedule meant that the trains were under-utilized, leading to high overheads that can't be spread out easily. While it's likely inflated by the needs of the congested BNSF corridor the service runs on, the overall subsidy to Northstar is pretty small. The way it gets divided up among so few trains and the suppressed passenger levels just makes it look a lot worse.

The train is fine—it's doing as well or better than anyone could have expected considering its constraints. It was opened with unrealistic expectations. The right thing to do is to invest in it and turn it into the service its true supporters always meant it to be.

Passengers unloading at Target Field on June 4, 2024. People streamed out of the station for several minutes, with part of the crowd filling a Green Line train that arrived shortly after, and a steady line of others walking to their destinations