UFO Sightings of 2016: A Recreation of a John Nelson Infographic

For some time now, I’ve been wanting to create a visualization showing UFO sightings. I found that the National UFO Reporting Center (NUFORC) maintains a database of sightings dating back many years, so I extracted the data and experimented with a few different ways to visualize it. I’d play with it a little, put it on the shelf, then return to it, but I never quite found a compelling way to tell the story, so it never really went anywhere. Eventually, however, I came across the following infographic, created by John Nelson and posted on his UXBlog in June, 2015. The infographic shows self-reported UFO sightings dating back to 1925, using the NUFORC data set.

Game Over! Case Closed! After seeing this, I knew there was simply no way I could visualize this data in a way that comes anywhere close to Mr. Nelson’s work. This is such a fantastic visualization (or is it art?). It’s both incredibly visually appealing and incredibly insightful. I mean, he even analyzes seasonality!!

Well, if I couldn’t top this, I figured I could at least try to recreate it. As the cliché says, “imitation is the sincerest form of flattery.” So, I set out to replicate the visualization to the best of my ability using Tableau (with John's permission). Since he had already shown data up through 2014, I figured I’d simplify my visualization and show only 2016 sightings.

Warning: The rest of this post will contain a very wonky explanation of some of the technical challenges of creating this visualization. If you’re not interested in all the details or simply don’t want to hear me complain about how difficult this was to create, I don’t blame you! Feel free to go straight to the visualization on Tableau Public. 

A Few Challenges

This visualization was incredibly challenging for many reasons. First is the use of Albers projection maps. Tableau’s default map projection is the Mercator projection, which does not have quite the same visual appeal when looking at the US only. As you can see below, the curvature of the Albers projection just adds a bit of flair that is absent in the Mercator projection.

Mercator Projection

Albers Projection

For a good primer to different map projections and how to use them in Tableau, check out Fun with Alternate Map Projections, by Ben Jones.

To create the Albers projection maps, I referred back to some work done by Josh Tapley and Jake Riley, which Josh discussed in a blog post called At it again…Getting Alaska and Hawaii on the Map. This post details the process of creating Albers projection maps in Tableau using polygons. I applied this method to my maps and it worked fantastically for the filled maps (choropleths), but I really struggled to figure out how to plot the location of each individual sighting as shown on the following map.

For this, I once again reached out to Jake Riley. Jake was incredibly helpful and willingly took on the challenge. By the next day, he had performed his magic and provided me with a data set with the x and y coordinates for each sighting and a method of using dual axis charts to plot them (essentially, one axis shows the outline of the states, while the second shows each individual sighting).

The next big challenge was the mapping of “sightings dimmed by population density.”

Now I have to admit that, at first glance, I didn’t have the slightest clue what this meant. Fortunately, Mr. Nelson discussed it briefly on his blog:

In order to visualize the actual sighting phenomenon, I needed to normalize by the underlying population.  The first, more prominent map shows a simple ratio of the sightings by population. A per-capita approach. The second, smaller and slightly more complex map, shows a bi-variate mapping of sightings in the color dimension (dark slate for low-sightings and bright green for high-sightings) and population density in the opacity dimension (denser populations are more transparent). The result is a map that is more nuanced regarding the problem of variable populations and area sizes.

Maybe I’m in the minority here, but I had never heard of bivariate mapping before. But, in the case of the UFO data set, it definitely has an interesting application. The bivariate approach puts a little more emphasis on those areas with low populations but comparatively high numbers of sightings. In a sense, this more heavily weighs low population areas, but it definitely gives a slightly different perspective and provides some additional insight.

But, how to create it? Fortunately, Mr. Nelson referred to a great tutorial by Joshua Stevens in a blog post called Bivariate Choropleth Maps: A How-to Guide. Mr. Stevens’s details the creation of 9 separate colors from the combination of two variables—in my case, population density and number of sightings—where each variable starts with 3 colors and are combined with each other to create 9 (I can tell I’ll be terrible at explaining this, so it may be best to just read his blog). The result is a color key that looks something like this:

When I followed this approach (then rotated it to match Mr. Nelson’s key), it looked something like this:

Of course, you’ll notice that this doesn’t look quite like the key on the original infographic. My best guess is that he then applied some sort of blurring effect to make the color transitions look more smooth. So I did the same, applying a Gaussian blur (I used Paint.NET for this), which resulted in something a bit closer to the original:

A final challenge was the sheer density of the infographic—it has a ton of different elements visualized using a number of different techniques. Granted, most of the remaining work was relatively straightforward—line charts, bar charts, unit charts, text tables, plus some custom color pallets and custom shapes—but there was just so much of it. Fortunately, after quite a bit of time and effort, I was finally able to piece together a completed dashboard, which I think looks quite a bit like the original. Below is a static screenshot, but since this is Tableau, the full version is much more interactive—you can hover over any county to see the number of sightings and the population; you can hover over individual sightings and get more details about each; you can get more details on the month-to-month trends for each shape, etc. You can find the fully-interactive version here.

I’d love to get your feedback, so please let me know what you think in the comments section.

Ken Flerlage, February 20, 2017

Twitter | LinkedIn | Tableau Public