Streaming Crayoning

I’m going to stream on Twitch at 18:00 tonight my time (UTC+1). This is usually intended as a platform for streaming playing video games, but can also be used for other things, such as discussions that would benefit from video and screen sharing. This may turn into a regular feature, covering a different topic every time – probably monthly, I don’t see myself having enough time to do this weekly unless it’s extremely impromptu.

Today’s topic is crayons:

  • How I make maps such as these ones: how I grab the base map, how I draw lines, etc.
  • Good tips for how to figure out which lines are useful, including very rudimentary cost-per-rider analyses.
  • An example of making a map, probably a US-wide high-speed rail map based on my Metcalfe’s law posts but maybe something else if there’s popular demand.


  1. Tonami Playman

    This was a great session Alon. Really enjoyed it. I did miss the first hr of streaming though so I did not catch the process of grabbing the base map. Is there a way to view the recorded video on Twitch?

    • Eric2

      I don’t know how Alon does it, but the slightly stupider way I would do it is as follows:
      1) Go to at the desired resolution
      2) Screen shot, as large as possible
      3) Move to a different place, take a screen shot there too, with a little overlap to the previous screenshot
      4) Repeat as necessarily for the entire map area
      5) Open all screenshots in a graphics program (e.g. Photoshop or Gimp), paste them into one image as different layers, align them to form a big map, flatten image
      6) Open a new SVG in Inkscape (or Adobe Illustrator), make the map a background image in the SVG
      7) Draw rail lines as Bezier curve paths, adjust for the proper curvature or straightness, change color and width
      8) Add stations as shaded-in circles

      • Alon Levy

        I have code for grabbing tiles from OSM and pasting them together. Pasting things as you say takes an incredibly long amount of time if you want a high-resolution base layer.

        • Tronçon

          Why don’t you just use ArcGIS Online?

          Pretty sure it gives you the necessary layers

        • Eric2

          Makes sense. Personally I have never crayoned anything detailed enough that copying and pasting was a major imposition.

      • Ben She

        Alternatively, I would strongly recommend using Google My Maps if you have a Google account, which is basically a very easy-straightforward GIS shapefile editor. Any mapping exercise really ought to be GIS-compatible at minimum, and it’s really easy to export high-resolution rasters when you’re done.

  2. adirondacker12800

    probably a US-wide high-speed rail map based on my Metcalfe’s law post
    Do it, it would come with “there is a whole lot of nothing between I-35 and I-5”
    Very very very roughly half the people in the U.S. live in the Eastern Time Zone. Also very very roughly two thirds of them live in the Eastern Time Zone and the eastern edge of the Central Time Zone.There is a whole lot of nothing west of I-35.
    It shouldn’t take much time. There’s Denver, Salt Lake City and Phoenix. Omaha? Metro Omaha is less than a million people. Albuquerque! Albuquerque’s MSA is less than a million and the CSA is less than 1.2. …..

    • barbarian2000

      the US is extremely flat between the mississippi and the rockies, which makes it much easier to build HSR

      • adirondacker12800

        Passenger railroads need passengers. There aren’t a lot of them out there.
        A very quick glance at a list of the Metropolitan Statistical Areas of the U.S., I found twelve MSAs in the Mountain Time Zone that managed to get into the top 200. Tuscon appears to be the fourth largest and it’s just over a million. There isn’t anyone out there.

    • Alon Levy

      I did it, but it’s not yet done, I’ll have to finish it next Saturday, same time (which is 13:00 your time because you’re in daylight savings and we aren’t yet for another 2 weeks). And yeah, it’s not a connected map for this reason.

      • adirondacker12800

        I-35 is being generous. An I-35 corridor usually comes up. Tulsa-Oklahoma City-Dallas-San Antonio might makes sense but OKC-Kansas City-Minneappolis is 800 miles/1300 km. Wichita, metro area of just under 650,000 is along the way and Des Moines at 700.000.
        ,,,,, and metro Kansas City is only 2.5 million. Maybe if Warren Buffet decided he wanted a full scale model railroad to play with. But then it would be Omaha not Kansas City. According to Wikipedia “Omaha is the anchor of the eight-county, bi-state Omaha-Council Bluffs metropolitan area. The Omaha Metropolitan Area is the 59th largest in the United States, with an estimated population of 944,316 (2018).”
        At least it gets less complicated the farther west you go.

        • Matthew Hutton

          Between the I5 and I35 there are a whole series of massive destinations for leisure travel, and those destinations aren’t exactly easy to get to so there would be lots of demand. Plus if you built the line following an existing straight road and perhaps had at grade crossings then the construction cost would be extremely low.

          • Herbert

            At grade crossings on a high speed rail line are a spectacularly dumb idea

          • Henry Miller

            @Matthew Hutton
            Because HSR is moving fast enough to hit/kill someone who stops, looks both ways, sees nothing, and then starts crossing! since people cannot correctly judge when it is safe to cross you need to do better than “stop look and listen”. Technically the train often is in sight, however it is so far away that it is hard to see at all and by the time you realize it is close and coming so fast that it is too late.

            You could possibly make it work with gates, but someone will see the gate close with no train in sight and incorrectly assume it is a gate malfunction and try to go around. No school bus will ever be allowed to cross at such a gate, nor will any truck with hazardous materials.

            There is also the cost of disaster. First, at the speed the train is going the odds that anyone survives is zero: I don’t care to put a value on life. Second, the disaster will do damage to the train which means expensive repairs (and you need to take it out of service to do them). Third, investigation and clean up means the line will be closed for hours to weeks while customers who have places to go cannot get there.

          • Matthew Hutton

            Yeah I didn’t realise they were that fast on the ground, but of course it makes sense!

        • Henry Miller

          Omaha means you slip des Moines. Either way you end up in Kansas city. If you think Omaha, you probably want a timed transfer somewhere, going north to Souix falls, Fargo, Winnipeg Canada. None of those cities are very big, but they arare more or less a straight line. The timed transfer is either in Kansas city with the line going to Minneapolis, of Omaha with the line going to Chicago: both is your only chance to make this stretch work, because the network effects of all three lines isn’t captured in metcalfe’s law.

          I live in the area where the lines in question would be of potential use to me, but I still can’t cheerlead it much, since there isn’t strong enough local transport in any of the named cities, if I’m driving to the train, and then renting a car when I get there why don’t I just drive?

          • adirondacker12800

            There’s only 250,000 people in metropolitan Fargo, 270,000-ish if you add in micropolitan Wahpeton. There aren’t enough people there to build a station. Or hundreds of miles of railroad.There aren’t enough people in Manitoba and the distances are too long to be building HSR to anywhere.

          • Henry Miller

            @adirondacker12800 your conclusion is correct, but your analysis is wrong.

            I said Winnipeg, not Manitoba, that city has a large part of the population of Manitoba and it isn’t that far from the US, and it sits on more or less a straight line to Kansas City through all of the cities mentioned. A 260km/h train makes the whole trip in 5 hours, meaning it is close enough to take a train, far enough that driving is annoying. These speeds are fast for HSR, not unreasonable. Of course it is an international line (which under performs in general).

            Be on the way is very important. Fargo doesn’t get a stop on its own, for sure but if you are building the line anyway it does, it is on the way and the largest thing around. So long as we do fast turns (one minute total stopped) the total cost in time isn’t that much and a few people will get on.

            They key to making this line work is low cost of construction. I didn’t apply Metcalf’s law to all the city pairs, but the trend seemed to be working towards everything working out with break even profits if we can build the entire 1300km line for $10 billion. This works out to 7-8 million/km assuming 3% ROI, about half what Spain needs to build their system. While I don’t think cost control can get down that far, we can see hope. If we can accept a 2% ROI the whole thing works out assuming a reasonable improvement on Spanish costs (the terrain is easier than Spain so we should be able to do better).

            Politically, between KS, NE, SD, and ND that is about 6 senate votes (never assume unanimous). The line doesn’t actually touch IA, MO, or MN, but it will be useful enough to those states to get a few votes that are otherwise slightly against. Therefore this is an important line as a compromise to get more useful lines in other states. Canada’s cost to this line is pretty small, so they will probably agree to pay for their part, and once the international agreements are signed it will be really hard to back down if costs don’t overrun by a lot.

            The whole point of this exercise is Minneapolis to Kansas City should work out, and Des Moines is on the way, while Omaha – even though it is a bigger city – is far enough out of the way that you wouldn’t go through it. So we assume the Kansas city to Minneapolis via Des Moines is built (which we should – the ROI is there if we use Spain construction costs), Omaha is a city that will feel excluded. Omaha alone doesn’t justify anything, but the entire line Kansas city to Winnipeg via Omaha and the other small cities on the way just adds up to possible. There is the beauty of HSR, it can stop at places in between at low cost and thus Metcalf’s law can include cities that alone are nothing. (I didn’t calculate the Des Moises to Souix Falls via Kansas city, or Wichita to Omaha, but those network effects are important and might even get us to 3% ROI)

            In conclusion, if the US can build for Spanish costs (or less where the terrain is better) there are a lot of lines that can justify HSR service. So long as the US is extremely expensive we will find it hard to justify even what should be obvious.

          • Henry Miller

            Arrgh, i just realized I carried the wrong column in my spreadsheet over when calculating ROI. The line is looking at about 1/3 to 1/2 less profits overall. Divide my ROI in half…. It still can work out with network effects, but it is even more marginal.

            Like I said, I don’t think it works out to build, it just looks even worse.

    • jcranmer

      The big nothing area is between I-35/I-29 and I-25 (you can extend these to Winnipeg and Edmonton in Canada if you’re feeling inclusive). I-35 corridor, as you note elsewhere, contains some MSAs in the 500k-1m range. The I-25 trace of the eastern edge of the Rocky Mountains nets you even larger cities: El Paso, Albuquerque, Denver, all counting a millionish or more. The Canadian bracket is even stronger: 1.5 million each in Edmonton/Calgary, a bit more than half that in Winnipeg, all in top 10 Canadian cities. Between them? You’re rocking Lubbock and Saskatoon, about a third a million each, as the largest, with the most of the states north of Texas struggling to post half even that as their largest city in that region.

      In between I-5 and I-25, you have some clear identifiable loci of populations: Phoenix, Las Vegas, Salt Lake, Tucson, Boise, Provo, Spokane, and Reno (in descending order of size). In between those cities, there’s also a whole lot of nothing, but unlike the High Plains gap, there are actually identifiable large cities. Some of these are even reachable from the main I-5 corridor (Phoenix/Las Vegas/Tucson for sure, Reno is reachable but probably not large enough to justify the mountains unless leisure demand for Lake Tahoe is enough to push it over).

      • adirondacker12800

        The combined population of Colorado and New Mexico is less than 8 million. It’s 450 miles between Denver and Albuquerque on I-25. There aren’t enough people for that much railroad. They can get on airplanes. Small ones.

        • Henry Miller

          The combined population isn’t what counts, it is the concentrated population in the cities that get service. Too bad the El Paso Area population is 3/4ths in Mexico, with all the big political baggage that brings. If we could allow the Mexican side of the border to count with El Paso we could get to 1% ROI with Spanish costs. Of course these are bigger mountains than Spain has to deal with so you can’t do that. Also the Mexico side of the border is extremely poor so in practice Metcalf’s law probably doesn’t apply even if the border was easy to cross. Still there is a real potential here if someone can innovate on cost control.

          • adirondacker12800

            The actual population is important because passenger railroads need passengers . They run trains because people want to use them. The fewer people there are, the fewer who will want to get on the train.
            It’s 266 miles on I-35 between Albuquerque and El Paso. The maintenance would cost too much for so few people.
            ….just stop. there is no network effect. No one is going to take the train from Albuquerque to San Antonio because it’s 551 miles from El Paso to San Antonio and since it’s roughly twice as far the maintenance would be twice as much. They can get on airplanes. Small ones because there aren’t many people

          • Matthew Hutton

            But national parks are also destinations. If LA to Texas is worth 10 trains a day, LA to Chicago and the east coast is 5 days a day, then if the Grand Canyon gives you 3 trains a day, and Joshua tree national park gives you one and the Navajo nation gives you 1 then maybe the line starts to make sense.

            Don’t also forget that the national parks might be a 5-6 hours drive from the nearest place with intercontinental flights, and then you might have to change at Heathrow or Paris or wherever from wherever you live which adds several hours of time so for Stockholm-New York-Grand Canyon for example might well actually be substantially quicker than now.

          • Alon Levy

            Grand Canyon is not worth this many trains per day, though. National Parks are set up for car travel, with diffusion of spots to stop at. Rail-based vacation travel is different and doesn’t really thrive on the American National Park system but on things like Niagara Falls, trails that intersect commuter rail lines, and lots and lots of urban tourism.

          • adirondacker12800

            None of his fantasies are worth any trains a day. there are airplanes.

          • Matthew Hutton

            How exactly do you fly to US national parks? The general way to get to the Grand Canyon would be to fly to Las Vegas and then do the 5 hour drive to the national park. And generally when you get to the national park you then leave your car in a car park and use national park land train to get around – and if you don’t do that for some destinations the national park land train could easily be extended to the 1-2 extra places you might drive to – as well as also linking to the hotel accommodation that typically sits just outside the park.

          • Eric2

            Popular national parks are generally in mountainous areas, not suited for HSR.

          • Matthew Hutton

            That’s kinda half true. The Grand Canyon is only an hours drive from the I40 which is pretty flat. Yosemite village is ~2 hours from the California high speed rail proposal and the northern entrance to Yellowstone national park is an hour from the I90 where the northern trans-continental railroad goes.

            Plus as those areas are relatively rural a bus service should be able to achieve journey times that are pretty similar to car journey times.

          • ARC

            If we’d like to talk about specific destinations, Grand Canyon NP received 2.9 million visitors in 2020, and Zion NP received 3.6 million visitors in 2020.

            Zion is a freebie if you build a line from Las Vegas to Salt Lake City, as it’s almost directly on the way. Zion is worth 10K visitors per day.

            GCNP requires a large detour on the Las Vegas to Phoenix line, but there are other synergies. Las Vegas to Phoenix direct is 300 miles. Las Vegas to Phoenix via Prescott is 350 miles, and Prescott MSA has 220k people. Adding Flagstaff and GCNP to the LV-Prescott-Phoenix route takes 100 miles of additional route, but gains you a 140k pop MSA in Flagstaff, and an 8000 person per day destination in GCNP.

            Local transportation can be handled by car rentals or municipal or NPS buses, in theory.

            1. Adding side destinations makes the Las Vegas-Phoenix terrain much rougher.

            2. My feeling, without doing the math, is that the Las Vegas to Phoenix Deluxe route, and the Las Vegas to SLX routes, borderline make sense only if you can build tracks at Spanish costs or similar. Las Vegas to Phoenix Basic is more resilient: the terrain is simpler and both cities have rail ROW into their cores.

            Do I think hitting National Parks is a good idea? Not having run the math, my gut says probably not. Do I think it’s a crazy suggestion? Not any more.

          • adirondacker12800

            Las Vegas to Phoenix via Prescott is 350 miles
            It’s 350 via Blythe California where the tracks could meet up with the Phoenix-Los Angeles line. It’s only 200 between Las Vegas and Blythe. Saves 150 miles of new construction. If it was bit farther east it could have a stop in Lake Havasu City for the tourists who want to see London Bridge.

            8000 person per day destination in GCNP.
            Grand Canyon National Park is reallly really big. There is more than one entrance to it. They aren’t all going to Grand Canyon Village on the south rim.
            You have noticed that 2020 was quite unusual in many ways? Perhaps you’d want to average out the visitor number from 2014 to 2019.
            It’s 8.000 a day if they all are evenly spaced all year and all going to Grand Canyon Village. And automobiles and airplanes have been banned. It gets quite wintry in northern Arizona. So wintry that they close the North Rim. Discourages tourism on the south rim too.

          • Henry Miller

            If you need to rent a car, why not fly into a more important airport and drive all the way? Either way you have the cost of the car, and so all you save is gas. Maybe a single person can work it out, but for a family renting is cheaper than train tickets, and you can see something else along the way.

            For a train to national parks to work there needs to be good non-car transit within the park. Hiking only fills a small part of that.

  3. terry

    Not sure if you knew but there’s an entire new game that allows you to do crayoning. It’s called NIMBYRails.

  4. RVA_Exile

    Thanks Alon, I concur that last Saturday was an interesting session.

    I was one of the ones that suggested the Southeast Corridor, since the Northeast has been crayoned to death.

    I would suggest a bit more granularity on the Southeast in the future, as and when you have time:
    – The Charlotte Gateway Station is already under construction, so no need to reinvent the wheel ( It will have streetcar, light rail and maybe commuter rail connections before anything like 300kmh HSR reaches Charlotte. (CLT is still a top 30 O&D airport so also worth a non-express stop.)
    – Saturday’s crayon went to the east of Petersburg and through Petersburg National Battlefield (site of the Siege of Petersburg during the Civil War). Historic preservationists already stopped the building of a circumferential beltway around Richmond and through this site – the route of I-295 purposefully avoids the Richmond and Petersburg battlefields. The existing corridor west of Petersburg could tie into an I-95 alignment up to Main St Station in Richmond and easily connects to the existing line to Norfolk (arrow straight for over 50 miles). The wye already exists, and all Norfolk trains should probably stop at Petersburg, so no need to worry about this curve’s speed. There is also the opportunity for direct Norfolk-North Carolina service as shown at the link below
    – There was a lot of handwaving around using the I-95 corridor in Northern Virginia. The I-95 corridor has a lot of curves and hills, requiring a lot of takings of expensive properties (that come right up to the interstate RoW) if curve straightening were attempted, aerials/earthworks to smooth out the grades and several high river bridges. A more granular examination of the I-95 corridor may show that this is unrealistic.
    An alternative could be to go via Southern Maryland, offering MD something in the process for all the disruption – regional/commuter rail on a corridor that has the worst radial transit to downtown Washington in the DMV region. This corridor offers mostly straight and uncomplicated ROWs until leaving the DC built-up area, and could be paired with a freight rail bypass of downtown Washington (which has been a long-time goal), allowing for more intense conventional-speed rail along the RF&P:

Leave a Reply

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

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

Facebook photo

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

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.