Correct the Map

In July 1995 more than 700 people died in a Chicago heat wave, and the neighbourhoods where they died are not the hottest places in the city. Not by day, anyway. This episode joins satellite thermal imagery with 1995 business records and today's map of places to gather, and finds that neither map on its own predicts where the deaths clustered. The overlap does. Four lessons in reading maps: building a proxy out of independent sources, why the available measurement is not the right one, how two weak signals can beat one strong one, and how to test whether a blank space on your map is real.

Think of somewhere you know that is not on the map. A shop that opened last year, a path everyone in your neighbourhood uses. You know it is there and the map says nothing is. Every blank space on every map is one of two things: nothing is there, or nobody has told the map yet, and a map draws those two exactly the same. Four lessons in reading maps, tested against 14,510 humanitarian mapping projects and every UN humanitarian appeal since 2014. Volunteered geographic information and why it exists where volunteers are rather than where it is needed. Why zero and blank are different facts. How to test an idea you want to be true. And what happens when you divide by something that was measuring a different thing.

In ten days the world's map got redrawn twice, by two completely different mechanisms, and only one of them needed a vote. The UN General Assembly adopted the Equal Earth projection 164 to 1. Eight days earlier, one signature renamed Lake Ontario and was on a billion phones inside a week. Why every projection is wrong, what Mercator was actually built for, and why OpenStreetMap was the only map that did not move.