On the Ongoing Ubiquity of the Mercator Projection
As anyone who has ever looked at a globe can tell you, Greenland is not as big as Africa, even though the Mercator projection would lead you to believe it is.
The amount to which the Mercator projection distorts relative size, combined with how often we see the maps using Mercator projection being used to the point where it has come to represent the generic image that many people carry in their mind when thinking about the world map, has real implications about our collective understanding of the world.
As an overview on the various issues surrounding the Mercator projection I can recommend this article by Saumya Kharbanda:
Considering that the biases that are caused by this projection have been well known for a long time already, one might wonder why the Mercator projection is still so ubiquitous nowadays. This article is an attempt at answering this question, at least partially. It is not my intention to dismiss anyone’s legitimate objections to this projection, nor do I want to support the Eurocentric view perpetuated by it. Instead the text is intended for readers who are aware of the sociopolitical factors and wonder what reasons there could possibly be for its continued use.
Many of the most popular map services on the Internet, sites like Google Maps, use a slightly modified version of the projection, Web Mercator, to display their maps (I won’t go in to the distinction between Web Mercator and Mercator, for our purposes they can be considered to be identical, hence I’ll just be using the term “Mercator” from now on).
Now obviously, Google doesn’t just use the Mercator projection because they are lazy or because they “don’t really care about maps”. After all, they run a massive product that is all about maps. There must be a reason why Google and other companies like Bing or Yandex choose to use the Mercator projection despite all its flaws.
To understand that, we must first consider one of the fundamental problems of cartography: All map projections are a trade-off between different inaccuracies. The Mercator projection doesn’t have all of its flaws just because the cartographers at the time didn’t know any better, or because they purposefully set out to diminish the visual importance of Africa on the map. The Mercator projection is defined the way it is because while it carries these issues, it also holds significant advantages over other projections.
At the time, around the sixteenth, seventeenth century, one of the main reasons why people needed maps was for navigation on the seas. The Mercator projection is great for this purpose, because rhumbs (lines of constant bearing) are straight lines in this projection, or in other words: Say you wanted to sail from Cape Town to New York. You take out your Mercator map, draw a straight line between the two, measure the angle, and then just sail away and make sure your compass stays at that angle. This won’t necessarily give you the shortest path, but it’s a very simple path as you don’t have to do any trigonometry calculations.
“Sure, that’s useful for navigation,” you say, “but how is this relevant to online Maps?”
Well, let’s consider the challenges a digital mapping service faces, that do not arise for paper maps. The main difference between a digital map and a paper map is, that you can’t zoom in to the latter one. Cartographers typically use different projections for world maps and maps of just one country. Many smaller countries, for example the Netherlands, use a stereographic projection, which is nice for a small area, but not so much when looking at the entire planet. Unfortunately, Google can’t just switch the projection when the user zooms beyond a certain level or moves to a different place, which brings us to one of the main requirements for a projection for digital maps:
Online map services needs a projection that works reasonably well at all scales, all across the globe (excluding the north and south pole).
While defining the subjective criterion of “works well” is a bit more difficult, we have some intuition about how a map is supposed to work. For navigating locally, in an area over which the curvature of the earth is negligible, it’s useful if the projection doesn’t distort angles, stretch the map in one direction, or bend straight lines. Essentially: if I go forwards 60 metres, turn 90 degrees to the right, and walk another 60 metres, the trace I leave in the snow corresponds to the shape I would draw on the map. Nothing fancy here, this is kind of like saying that maps for small areas should look like aerial photos without any distortion. Furthermore, it would be nice if north was always up.
This is basically what we would get if we were to use a stereographic projection centered around the place we’re looking at. Now the problem with this projection is, that it gets distorted more and more as you move outside the zone of negligible curvature. Which means that in order to preserve the properties listed above, the stereographic projection would have to be constantly recentred as you move across the map, and that would be inconvenient for several reasons. It would significantly impact performance, since the service couldn’t just store fixed tiles, but would effectively have to recalculate the projection every time, and more importantly, I’d imagine this to be a very nauseating experience if you’re moving around on a scale where an entire continent is in view, because while the centre that you’re currently looking at would look alright, everything around it would warp as you’re moving around.
We are thus looking for a map projection where local angles remain accurate, without having to change the projection. Well this sounds pretty much like what we’ve heard about the Mercator projection. A small square area should remain square in the projection, since the bearing from the bottom left to the top right corner (45°) has to correspond to the angle on the map, which is how we know that the square won’t turn out as a rectangle with unequal side lengths in the projection.
This now unfortunately leads to Africa and Greenland to appear to be of roughly equal size when you zoom all the way out on Google Maps, bringing us back to the problems mentioned at the beginning of the article. Of course, Google kept looking for a better alternative, and so they did what every reasonable person would do when they realize that you can’t just take a 3d sphere and turn it in to a 2d map without any artifacts:
Google Maps giving users the option to show not a flat map, but instead a globe, solves the problem of map projections. This solution acknowledges that there’s no way to see all sides of the planet at once without introducing some trade-offs. However, not all views use a globe, some still show the Mercator projection. Especially on weaker mobile devices, performing the calculations to show an actual globe might not always be an option.
An alternative solution, which the South African map service AfriGIS opted for on their website, is to use a different map projection and accept the distortions that are caused by this. Their map seems to use the plate carrée projection, which means that streets that should be at a right angle to each other are sometimes a bit skewed, but apparently that was deemed an acceptable trade-off to make in order to get closer to the goal of an equal-area projection. This is however only possible because their map is only intended to be used for South Africa, since the distortion would be stronger in other locations.
Hopefully this article sheds some light on the mathematical background of why the Mercator projection is still being used sometimes. I personally believe that the Mercator projection should not be used in classrooms or other physical places, and instead an equal-area projection should be chosen. For digital maps however I can see why Mercator might be a necessary evil. Around the internet I have found various articles explaining the maritime history of the projection, but few of them mention the connection to zoomable maps, which is why I felt the need to write my own article on the topic.
Note: an earlier version of this article was worded in a way that could be interpreted as a dismissal of any criticism of the Mercator and other Eurocentric projections. This was not my intention. I have reworded the article and apologize for any offence that the previous text caused.