For the last several weeks, I’ve been creating a visualization of past and predicted conflicts in Afghanistan. The data, which come from a statistical model developed by Jay Yonamine with help from John Beieler, show us what might happen based on a long history of what has happened.  The database behind this model is known as GDELT, or the Global Database of Events, Language, and Tone. GDELT has received a lot of attention lately and I’ve participated in two hackathons that have emerged around it. After Wired showed interest in publishing this story along with the visualization, I was thrilled.

What You Saw

When I heard that Wired had published the work this morning, I was excited to check it out and see how the visualization tied the story together.

And then I saw my map. Well, it was sort of my map. It had my name on it… but it was not at all how I had originally designed it. Here is the map as it is currently published:

Red-green colorblind? Too bad!

The Afghanistan map published in all its red-green glory.

Here’s my quibble: Not only was my work altered without acknowledging the changes from the original, the change resulted in a red-green color scheme – a classic map mishap that no professional cartographer would ever make.

Is that -4 or +100?

The Afghanistan map, as seen by someone with red-green colorblindness. Does bright yellow mean -4 or +100?

I’ve written about red/green colorblindness in the past. Others have too. In fact, advice on not using red/green color schemes on maps and other visualizations are practically all[1] over[2] the[3] web[4]. It’s not a new problem either: red/green color blindness has been written about by scientists for at least 120 years. We even have tools to help designers avoid common pitfalls with their color choices.

By being in red and green, this map is impossible to interpret by anyone affected by red/green colorblindness (almost 10% of  males, about 5% of the total population). Instead, readers with red/green colorblindness will see something like the image on the right (simulated protanopia).

What You Should Have Seen

The second problem is that I intended to show readers a lot more about the data – temporal trends, the propensity of conflicts near district and country borders, and the role of Operation Enduring Freedom in the high number of material conflicts over the past decade.

This is what I wanted you to see:

Visualization of Afghanistan: More info, colorblind-safe!

Click here to access a larger version.

This view shows much more about the material conflicts that have occurred in Afghanistan over the past 34 years. From more than a million geocoded conflicts (check out the activity at the border between Afghanistan and Pakistan) to temporal trends, readers instantly get a better sense of the situation and how the data have changed over time. As a professional cartographer, I also made sure it is red/green colorblind-safe (preview). Naturally, I would have liked to see this version published for the article’s readers.

In the end, I really can’t change the article. But I can share the original, and I hope it provides the utility and clarity that was intended – for all readers, colorblind or not.

Tagged with →  

9 Responses to The Map of GDELT Predictions in Afghanistan as You Should Have Seen It

  1. John Nelson says:

    Josh,
    You are taking the stoic road about it but you have much reason to be upset about this -this is more than a quibble. The comparison is a *magnitudinal* difference in utility and design and communication. The variability in what ran vs. what was submitted is a gaff of ineptitude and your exclusion from the process is a lapse in editorial integrity. You say that the article can’t be changed, but certainly the digital version can and should be updated to the submitted graphic (which is epic, by the way -a cartographic triumph). Ironically, to do so will make theirs an insanely popular post for Wired as it is sure to catch fire on the web. Jonathan Keats and his editors need to correct this.
    John Nelson

    • Josh says:

      John, you’re completely right. I’m trying to be positive and not rage (even though I think it’s justified).

      The crop doesn’t bother me; if the entire graphic didn’t make sense for their goal, no biggie. Wired and I shared a few emails about general layout.

      But changing the color scheme like that? Not acknowledging that heavy changes were made to the author’s original? I’m really surprised they went that route.

  2. Add to the misrepresentation of Josh’s work the fact that this is not my project at all: I’ve done no work specific to Afghanistan, and the forecasting work was done by Jay Yonamine without any involvement on my part except as a dissertation advisor. GDELT, meanwhile, is primarily the result of Kalev Leetaru’s efforts, and he isn’t even mentioned. Keating contacted us about a year ago and then did no followup, resulting in this complete hash of a report. This is in distinct contrast to some of the other journalists we’ve worked with, who have done extensive cross-checking with various people on the project before publishing.

    What’s with “Danger Room”? — their previous foray into event data involved extensively criticizing the DARPA ICEWS project for not forecasting the Arab Spring during a period when ICEWS was solely focused on Asia and wasn’t collecting data on (or at any point in the entire project tasked with analyzing) any countries in the Middle East.

  3. Andrew says:

    Very impressive! What platform did you use to build the visualization? (I may have missed it in the post.)

    And, to add my two cents, you have every right to be annoyed. At the very least, Wired should have run it by you before publication.

    • Josh says:

      The maps were made with plain old GIS tools, and then styled in TileMill. The calendar view was coded in D3, and then everything was brought together in Photoshop. I plan to write a tutorial on using these tools to create graphics like this soon!

  4. Varun Goel says:

    Hi Josh,
    I just graduated from the University of Illinois at Urbana Champaign in Geography and GIS, and have been trying to educate myself with open source tools for cartographic visualization. No wonder, your blog is one of my go to places for tips and examples. I am fascinated by this map and wanted to know how you generated the heat map for the time series visualization.
    Thanks
    Varun

    • Josh says:

      Hi Varun – congrats on the graduation!

      The calendar view was made with D3, following the code example here. I’ve created a quick demo of this using some GDELT data you can interact with and view the source of here.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code class="" title="" data-url=""> <del datetime=""> <em> <i> <q cite=""> <strike> <strong> <pre class="" title="" data-url=""> <span class="" title="" data-url="">