Skip to content

Things I Have Learned

July 31, 2014

Well, it has been about four years since I have posted on this site. Since then, I have worked at Google, where I helped enhance the quality of Google Maps and Places, then The Climate Corporation, where I wrote the company’s first geospatial data standards (among other things) and now AcreValue, where one of the many hats I wear is that of a geospatial data scientist. In that time, I have learned a large number of lessons relevant to this site, some of which I will share now.

1) It is hard to keep up a blog (see the whole this is my first post since 2010 thing). It turns out that even when you hold a lot of enmity towards those who distort data, coming up with cases and writing posts adds significant work to the general work of being/becoming an adult with a balanced personal and professional life. Lesson Learned.

2) As much as I rant about the Mercator projection, there are very solid reasons (both technical and aesthetic) for it to be the default on the web. Basically, it does a very good job when you are zoomed in, as is the typical direciton-finding use case of Google Maps and it is not too hard for devotees to implement custom map projections for zoomed out levels. There are just strong network effects and default change aversion against doing so.

3) Spatial data is not special. Well, not as special as I thought it was. Maps are still brilliant and beautiful ways of presenting data, but the church and state separation I had in my head five years ago between spatial data and mere text etc. is largely artificial and has not served me well. That is to say that prematurely specializing in spatial data rather than learning general computer science and software engineering first is *not* advised.

4) It is much easier to hate than create. This should be obvious, but to a young soul who had not created altogether that much in this world, this blog was and may continue to be an easy vector by which to take down the work of others. However, in my creation of a large number of maps and models, and especially in the use of the Google Maps API for AcreValue, I have learned the wisdom that many of the things I have criticized here do exist for a reason. I still hope to post though in future updates when I see egregious misuses of my favorite data visualization.

And with that, I am off to create some more maps! I will try to make it <4 years before my next update.


The Oil Spill in Maps – Best and Worst

May 20, 2010

One month ago today an explosion rocked the Deepwater Horizion, beginning one of the greatest environmental disasters in U.S. history.  Since then, the spill has galvanized the environmental movement, brought oil company CEOs before congressional hearings and of course, has been mapped over and over again, at times beautifully and elegantly, and at times in ways very ugly and fashioned simply to score political points.

The worst:

Fun task: Try telling where the oil is predicted to be on May 3rd on this map. Especially if you are red-green colorblind. This shows why stacked, partly transparent polygons are a poor choice to show a multi-day event. Bad stuff.

Next up we have a map that shows us the complex pipeline and drilling infrastructure set up in the Gulf. The message seems to be “wow, look at this completely out of control system we have down there” and while I agree with the author’s point, I disagree with some of the ways he tries to convince us of it. First, all pipelines (other than those owned by BP) are shown the same here, there is no distinction between a 6-inch and a 4-foot diameter pipe. Imagine if  someone made a map of the Midwest, put every single gravel road and interstate on there and marked them all using the same symbol – it would look like absolute anarchy, just like this. Secondly, the author includes a lot of irrelevant on land information. Maybe it is just because he is from ESRI and wants to show off their nifty basemaps, but there is no reason to have interstates on a map of a water feature no matter how grayed out they are. Lastly, he has the same, multiple overlapping polygon problem as the NOAA map, although his way of dealing with it is significantly better.

The last bad map of the bunch has some of the unnecessary complexity problems of the second bad map, however what is uniquely bad about it is its contempt for the map user. All I have to say to the map author is please, trust us to use your scale bar. If you need to give specific distance figures, tuck them neatly into a text box somewhere but why, oh, why do you think we cannot tell how big things are, or how far they are from each other, when you have given us the key to doing so, right there in the bottom left corner?!

The best:

The New York Times map is simply amazing. In order to give users a feeling for how the situation changes from day-to-day, it runs as an animation. It contains relevant data  – habitats affected, states affected, fishery closures, and benchmarked estimates of the spill size, but still does not feel overwhelming. Quality stuff.

A Google employee also created a good animated offering in the mix – – with a nice comparison of the spill to many metro areas worldwide. Simplistic, sure, but also quite effective.

Lastly, to prove that NOAA can make good maps we have :

A daily forecast map of the spill’s trajectory and status. Great, meaningful legend and coding for the legend at the bottom. Extremely clear, and without unneeded on-land complexity. Nice.

The takeaway message from all this for map authors is yes, I know we want info on the disaster, but if you shove too much at us at once it ends up like that all-you can eat buffet in Vegas, horribly unappetizing.

Meaningless Indices

May 16, 2010

So you have made a beautiful map. Correct projection. Great colors. Simple symbology. Extremely readable. How do your ruin it now? Make it rely on a meaningless index.

But wait! You might say, “the Human Development Index is extremely meaningful, the UN uses it all the time.” Sure, but as discussed here, this map wasn’t actually made using comparable data. Instead, the data used here came from a wikipedia dataset that closely mirrors work done by the American Human Development Project, but cannot be sourced. Therefore, while perhaps plausible,  the map is meaningless.

Next up:

As I stated in the last post, this map is the worst map I have ever seen. Red-green color scheme, mercator (or maybe even worse) projection, and worst of all, a meaningless  “Any Apple Index”  that is not explained by the map’s source – . I guess we have to pay to see what any of this actually means.

In sum, if you are going to make a map, and your map is based on an index, please explain how you got there. Otherwise you are useless.

Maps For the (Color)Blind

May 3, 2010

7-10% of males have trouble telling the difference between this and this. Why should a mapmaker care? Well, if you want your map to be intelligible to as many people as possible, you should probably not use a red-green color scheme. Unfortunately, many, many mapmakers have forgotten this fact.

2008 Virginia Democratic Primary Results - Obama in Green, Clinton in Red

Where Apple products are the most (and least) popular

The last of these maps being just about the most misguided map I have ever seen (but more on that later).

All of these were found in just a few minutes on Google, meaning that unfortunately, red-green maps are everywhere. Still, this is no excuse for adding to the pile. Even though it might be common to forget about them, the colorblind need maps too, and it is pretty damn easy to select any other color scheme than red-green. So please do.

Solar Distortion

April 30, 2010

So after my last post about the evil Mercator, I got an e-mail asking how this distortion has played a role in real life maps. Therefore, I give you:

So what is so wrong with this map?

It purports to show the area it would take to power the world on solar panels alone, but does so using a projection that is not equal area. This means that all those solar panels, most of which are gathered near or around the tropics, appear to be much smaller than they actually are.

This map got a lot of coverage in the blogosphere in late 2009, and I will not pretend that I am the first one to notice that it uses a bad projection to distort its message. In fact, the author of the map go so much flack over this distortion that he ended up having to redo the map using an equal area projection later.

Unfortunately, much of the power of his message – that we can power the world with a seemingly small area devoted to solar power (yes, these panels are actually quite large, but do seem small) got lost in the controversy over the projection he used.

Mercator Projection – Evil, or SuperEvil™?

April 22, 2010

Mapmakers take projections personally (see the Propaganda in Action section of this article for an example). Yes, they are simply methods used to represent our wonderful, slightly elliptical, Earth as a plane, but we mapmakers have to deal with them all the time. We wade through them, for hours on end. We bemoan the fact that hundreds of them exist and that no project ever has data based on only one of them (or even one datum for that matter, which is a slightly different issue).

So by now you may be wondering, “Why is the Mercator projection Evil?”

It comes down to distortion. The Mercator projection was created in the 1500s for navigation,

If only the distortion were at the equator...

and was great for sailors because any straight line on the map is a rhumb line, or a steady bearing from where you are to where you want to be. Sounds nice, right? But this projection distorts the world -a lot – especially around the poles. On the right is an approximation of what my picture would look like in the Mercator projection if I stood on Antarctica and my head went to Greenland. Note: I am not happy in said picture.

Why is the Mercator projection SuperEvil™?

For some reason (actually a number of reasons), the Mercator projection became the default projection used in classrooms around the world. So, for centuries we taught kids that Greenland was the size of Africa when, instead, it is the size of Algeria – not even the biggest country in Africa.

If that wasn’t bad enough, the Mercator became the projection for pretty much every web mapping service known to man. Thus, on Google Maps’ main page there is Greenland, staring right at you, looking all huge like it wants to fight. Russia looks pretty huge too (well, yes it is huge, but not thathuge) and people tend to think that the world actually looks this way, they do not notice that the scale bar changes in size in the same zoom view going from Alaska to Angola, and this distortion inevitably shapes how they view the importance of different countries are in the world. Also, it makes small scale web maps look just plain ugly.

What is better than the Mercator?

In order to change all this horrible, horrible distortion, some folks in the 60s decided that the Peters projection should be used for all world maps. Turns out the Peters’ projection was actually a ripoff of the Gall projection from over 100 years earlier (the Propaganda in Action section of the article referenced above has more info on this as well) and that a lot of these folks were tools. Personally I prefer the Winkel Tripel projection (seen at right) since it is much closer to what the world actually looks like.

ah, much better

The Winkel Tripel is a compromise projection, so distorts the world in a number of ways, but only slightly, as it has the smallest overall distortion of any projection. Also, National Geographic uses it and they have been my heroes since I was could walk.

Why change?

Some of the coming articles will show times when using the Mercator projection heavily distorts the message of a map. There is no doubt that in a number of places, the projection is used specifically for this purpose. But generally, misuse of the Mercator is a sign and symbol everywhere of mapping ignorance.  Its use is a sign that the mapmaker had little knowledge of the underlying geographic principles he or she was working with and that quite possibly, the remainder of the mapmaker’s analysis ignores similarly important principles.

So please, join me in abandoning this map projection for small scale maps. Leave it behind in the 16th century graveyard where it belongs.

Hating on Haiti Mappers

April 19, 2010

Why make a map when a graph is better?

In the wake of a disaster or a major world event, one of the first instincts of the media (and rightly so)

Little known fact - the U.S. actually invaded Egypt in 2003

is to make a map educating the public where the disaster is. In this initial rush, some mistakes are forgivable (such as the poor initial georeferencing on the this very cool NYtimes interactive feature that has now been corrected). Though some are pretty horrible (see right).

In any case, there is no excuse, a month after the disaster, for making the map below. (Original link so that you can see it in its full size/glory).

Why is this so bad you ask? Oh, let me list the ways:

1) It it meant to feature a tropical country, yet uses the Mercator projection to do so – this projection diminishes the size of equatorial countries and exaggerates those near the poles (more on why this projection is evil, and what it is, in a later post, I promise)

2) Why is Russia green? Why is Canada green? and why are some of the islands they own white?

3)It is utterly, utterly confusing – so yes, all lines go from donors nations to Haiti, but it is quite the tangled mess. Can you tell where all those lines are going? If so, I have a project for you.

So how could one remake this map for the better? How about a graph next time. You can still use the same data. I promise.