Monday, August 31, 2015

16th Century Maps for 21st Century Data Science

Maps bother me. I love them, and I'm not a geospatial GIS coding specialist, but I do visualizations, and we keep using the wrong maps. Greenland is a lot smaller than all of Africa, ok?

This is the map in my office kitchen:

It's the typical Mercator projection (projection, since you have to "project" a sphere onto a flat surface, which doesn't work well). Mercator came up with this map view in 1569, according to Wikipedia. Yet we still use it for 21st century data science! Granted just because something is old doesn't mean it's not useful, but in this case the Mercator projection was created primarily for navigation, that is, sailing the seven seas. When you present geospatial data the only thing your viewers are navigating is your data. As such this is totally the wrong mapping projection to use. Totally. Don't do it. Data visualizations are about accuracy, and using the Mercator projection starts you off with a completely inaccurate mapping. Greenland and Africa? "Africa's area is 14 times greater" than Greenland according to that Wikipedia article! Fourteen!

So what to do instead?

Wikipedia has a page of many different projections, I'd vote for one of the equal-area ones, and am a fan of the Gall-Peters projection (which was the centerpiece of a great segment on The West Wing), but you'll need to decide what's best for your use.

So, I'm a little upset about the giant Mercator map in my office, but with good reason.