A Generalization Ramble

Spatial accuracy is not always your friend.

A few weeks ago I wrote about the potential downsides of making an overly-detailed shaded relief image. It’s easy for readers to miss the major landforms, as they are hidden under the highly-detailed noise of the individual hills and bumps.

I truly understand the urge to show all that detail. It can feel wrong to hide it, to smooth it away, to present a gross reduction of nature’s amazing complexity. We are awash in so much great data nowadays, so let’s show off all of the exquisite detail that we finally have about land and water and populations and incomes — information that previous generations would have loved to access, but couldn’t.

But more importantly, it feels safer. 

Putting every possible detail on the map means we can avoid difficult choices about what to show and what not to show. We can tell ourselves that we’re just showing what is, without applying influence. We can say to the reader “I’m just putting everything out there and you can decide; I’m not trying to interpose myself.” It lets us imagine that our maps are objective, even if that’s not actually possible; nothing authored by humans (or authored by human-authored software) can be anything but subjective.

If we don’t make too many choices, hopefully it also means no one will disagree with us. Map generalization is a form of artistic vulnerability. If I reduce the detail in a coastline, I am keenly aware that another person might have done it a bit differently. Did I do it wrong? Should I have used different parameters? How much is too much? Does mine look stupid, or confusing? I grapple with this constantly.

We all commit sins against the landscape when we fix it into a cartographic form. I know that it can feel uncomfortable; it does to me. But it’s necessary if we want to serve our readers well. Think again to the example above, of the broad landform structures vs. the tiny details. The “right” answer will depend on the map that you are making, but in many, many cases, it’s better for the reader to focus not on the thousands of individual peaks and valleys, but the great ridges that they form — the large structures that affect the weather, determine the course of streams, and constrain the movement of animals and people.

Here’s an example I often share with students. Fictional cities, in a fictional country (but based on a real-life example I encountered as a student).

Notice how Border City appears to be just inside the country of Pineland. But, if we make the dots semitransparent, you can see that, due to a dip in the border, Border City is actually in Utopia.

One easy solution is to just move Border City. I mean, you could restyle the dots, too, but if you don’t feel like doing that, just move things.

When I tell students this, I wonder if some of them find it shocking to hear. I think a lot of us, early on in our careers (certainly my own), are very attached to the idea that we must make maps “accurate” and “objective,” and so this sort of generalization feels very wrong. But sometimes the lie is more accurate. The reader now will correctly understand what country Border City is in. Just like, with the terrain example above, they will now correctly understand the big shapes of the earth, rather than getting lost in details that are technically accurate but serve no greater purpose.

I tell my students to contrast “spatial accuracy” with “narrative accuracy.” Instead of obsessing about the exact proper position of everything, the latter focuses on asking: What will the reader actually learn from the map? What will they remember five minutes after they’ve put it down? We’re generally not landing spacecraft with these maps, so it’s OK if something is gently smoothed or nudged (or even made up!), so long as it’s not so extreme as to leave people with the wrong impression.

It’s been a long journey, but I mostly do not hesitate anymore to move or reshape cartographically inconvenient features. Many times, though certainly not all of them, it doesn’t matter if you get it “wrong.”

This post ended up being a bit of ramble, much less structured than I usually try to present my thoughts here. Thanks for coming along!