Logo Cover Cover Cover

Chapter 6: Building expertise through UGC verification

Eliot Higgins is an investigative journalist and researcher, specialising in open source investigations. He has achieved worldwide recognition for his work, that has included investigating the use of cluster munitions in Syria, the smuggling of weapons to the Syrian opposition, the August 21st Sarin attacks in Damascus, and the downing of MH17 in Ukraine. His recently launched website Bellingcat aims to spread the use of open source investigation techniques to NGOs, media organisations, and other groups. He is @EliotHiggins on on Twitter.

During the later stages of the Libyan civil war in 2011, rebel groups pushed out from the Nafusa Mountain region and began to capture towns. There were many contradictory reports of the capture of towns along the base of the mountain range. One such claim was made about the small town of Tiji, just north of the mountains. A video was posted online that showed a tank driving through what was claimed to be the center of the town.

At the time, I was examining user-generated content coming from the Libyan conflict zone. My interest was in understanding the situation on the ground, beyond what was being reported in the press. There were constant claims and counterclaims about what was happening on the ground. There was really only one question I was interested in answering: How do we know if a report is accurate?

This is why and how I first learned to use geolocation to verify the location where videos were filmed. This work helped me sharpen the open source investigation techniques that are now used by myself and others to investigate everything from international corruption to war zones and plane crashes.

The video in Tiji showed a tank driving down a wide road, right next to a mosque. Tiji was a small town; I thought it might be easy to find that road and the mosque.



Until that point, I hadn’t even considered that you could use satellite maps to look for landmarks visible in videos to confirm where they had been filmed. The satellite map imagery below clearly showed only one major road running through the town, and on that road there was one mosque. I compared the position of the minaret, the dome and a nearby wall on the satellite map imagery to that in the video, and it was clear it was a perfect match.



Now that the likely position of the camera in the town was established, I could watch the whole video, comparing other details to what was visible on satellite map imagery. This further confirmed the positions matched.

Building expertise in satellite map based geolocation was something I did over time, using new tricks and techniques as I moved onto new videos.


Matching roads

After the Tiji video, I examined a video purportedly filmed in another Libyan town, Brega, which featured rebel fighters taking a tour of the streets. At first, it appeared there were no large features, such as mosques, on a satellite map imagery. But I realized there was one very large feature visible in the video. As they walked through the streets, it was possible to map out the roads along the route they took, and then match that pattern to what was visible in satellite map imagery. Below is a hand-drawn map of the roads, as I saw them represented on the video.



I scanned the satellite imagery of the town, looking for a similar road pattern. I soon found a match:



Hunting shadows

As you become more familiar with geolocating based on satellite map imagery, you’ll learn how to spot smaller objects as well. For example, while things like billboards and streetlights are small objects, the shadows they cast can actually indicate their presence. Shadows can also be used to reveal information about the comparative height of buildings, and the shape of those buildings:



Shadows can also be used to tell the time of day an image was recorded. After the downing of Flight MH17 in Ukraine, the following image was shared showing a Buk missile launcher in the town of Torez:



It was possible to establish the exact position of the camera, and from that, it was possible to establish the direction of the shadows. I used the website Sun Calc, which allows users to calculate the position of the sun throughout the day using a Google Maps based interface. It was then possible to establish the time of day as approximately 12:30 p.m. local time, which was later supported by interviews with civilians on the ground, and with social media sightings of the missile launcher traveling through the town.

In the case of July 17, 2014, and the downing of MH17, it was possible to do this by analyzing several videos and photographs of the Buk missile launcher. I and others were able to create a map of the missile launcher’s movements on the day, as well as a timeline of sightings.



By bringing together different sources, tools and techniques, it was possible to connect these individual pieces of information and establish critical facts about this incident.

A key element of working with user-generated content in investigations is understanding how that content is shared. With Syria, a handful of opposition social media pages are the main sources of information from certain areas. This obviously limits the perspective on the conflict from different regions, but also means it’s possible to collect, organize and systematically review those accounts for the latest information.

In the case of Ukraine, there’s few limits on Internet access, so information is shared everywhere. This creates new challenges for collecting information, but it also means there’s more unfiltered content that may contain hidden gems.

During Bellingcat’s research on the Buk missile launcher linked to the downing of MH17, it was possible to find multiple videos of a convoy traveling through Russia to the Ukrainian border that had the same missile launcher filmed and photographed on July 17 inside Ukraine.

These videos were on social media accounts and several different websites, all of which belonged to different individuals. They were uncovered by first geolocating the initial videos we found, then using that to predict the likely route those vehicles would have taken to get from each geolocated site. Then we could keyword search on various social media sites for the names of locations that were along the route the vehicle would have to had to travel. We also searched for keywords such as “convoy,” “missile,” etc. that could be associated with sightings.

Although this was very time consuming, it allowed us to build a collection of sightings from multiple sources that would have otherwise been overlooked, and certainly not pieced together.

If there’s one final piece of advice, it would be to give this work and approach a try in any investigation. It’s remarkable what you can turn up when you approach UGC and open source information in a systematic way. You tend to learn quickly by just doing it. Even something as simple as double-checking the geolocation someone else has done can teach you a lot about comparing videos and photographs to satellite map imagery.



Published on: 15 April 2015
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.