Sep 14, 2020
Challenging India’s news narrative of a Chinese plane crash using OSINT and geolocation
All information within was obtained using open-source information. I have provided evidence and want you to look at this as a starting point for your own investigation, rather than as proof to take without question.
On Friday September 4th, 2020, a Chinese plane crashed into a building, the pilot ejected and is rumoured to have been injured. Twitter suddenly lit up about the crash, and in its usual fashion, details diverged with every account. Footage of burning debris and a semi-destroyed building was shared, alongside several other videos of the smoke from further away. No actual video of the tumbling plane, just the aftermath.
Very quickly, the information was muddled. There was a lot of speculation of Taiwanese involvement, suggesting Taiwan shot down the plane.
It was clear that if some of these rumours were true, it would be huge news. China and Taiwan aren’t in a forgiving relationship and such an event would spark massive consequences. It was also clear that the majority of the news and reports were coming from Indian news sites and accounts. Both these facts rang alarm bells for me, particularly due to the recent activity of Indian forces along the Chinese borders due to disputes.
I set out to geolocate the crash to investigate whether any of the rumours were true. This became all the more important when floods of Indian accounts started posting about it before any official confirmation had happened. Shortly later, Taiwan’s Ministry of Defense completely denied any responsibility, citing that it was fake news.
Looking at the images and the rumours, I had several starting points. Firstly, I had Taiwan, one of the alleged locations of the crash. Secondly, I had China — at the time limited to the coastal regions near Taiwan.
The videos being shared on twitter were initially from three positions. The first location (LEFT) was of the building that was significantly damaged. It was a narrow view and had few significant landmarks to set it aside.
What it did show was a distinct pylon, a medium-sized, black-roofed building, and crucially, two plumes of smoke. One smoke plume was black, the other was a lighter grey.
This was an important key to linking the videos.
Since they would all be taken within the same rough time frame so I would expect to see two plumes of smoke and the other landmarks.
These cues held the initial search together.
The second video (RIGHT) was filmed from a height, probably a residential block. It showed the clue I needed to lock in on a location.
The running track had several distinct features, and more importantly, is a significant thing to see on satellite imagery. It is hard to miss a running track unless you are really zoomed out on Google Earth, and at the resolution needed to see it, you can have a much larger area in view than if you were trying to locate a single building.
Confirming that they were of the same incident, I referenced the pylon and the smoke plumes.
I also started to take note of the defining features of the video location. Small things that may differentiate it from other race tracks.
The third video I saw was filmed from a similar tall building, but with another tall building in front of it too, so you could just see the plumes of smoke. This video never really saw a moment of glory because it added no additional information.
Finding the location from those two pictures
This was the tricky part. I foolishly decided to start my search in Taiwan, the area that I was most skeptical about. Not only did I already have my doubts about the plane crashing on the island, but I didn’t know about race tracks in Taiwan.
Nearly every school has a racetrack. There were some screens on Google Earth where I could see ten or more tracks, all in a resolution high enough to try and match them to the photo above — so really zoomed in.
I narrowed down the search to areas near green hills and ended up searching most of the length of western Taiwan before I gave up.
I decided that after several hundred race tracks, I hadn’t even covered a fraction of what I needed and actually, a World War would be a nice break from staring at ovals so I did some deeper Twitter and news research.
My saving grace was from the Jewish Press which had got hold of the story. The original article posted that “Taiwanese Twitter users on Friday morning reported a ‘Chinese Communist Party’s People’s Liberation Army Sukhoi Su-35 fighter plane crashed in Guangxi.’”
This led me on a new search for sources on Twitter and I was able to get hold of some useful information about rough locations. As it turns out a lot of the videos and early reports had been lifted by Indian profiles and actually had crucial information removed.
By searching only Chinese and Taiwanese sources, I was able to (with the help of Google Translate), narrow down the search to a town called Guilin.
Let’s take a brief intermission to acknowledge that it wasn’t just small Indian civilian accounts posting the fake information, even senior figures were sharing the footage with the crucial information removed. Perhaps the worst culprit (outside India’s media channels) was Major Dr. Surendra Poonia, a politician from Bhartiya Janata Party in India and an actual doctor.
Once the search was narrowed down to Guilin, the real work began. I had reasonable information to work with and it was just in time. Twitter was becoming more and more flooded with Indian accounts sharing very politically motivated posts which were saturating out the useful information.
The Daily Express also released a very widely shared article about Taiwan denying responsibility, but the headline was popular and offered little clarity beyond sensationalism that was easy to share.
After a search of the much fewer running tracks in the area, I had a good guess of the one which I believed was featured in the video. It was time to match it up with other footage and verify the location.
This meant going back to the original footage and marking all the features I could use to ID the matches. The following is what I got:
Annotating seemingly minor details becomes key later when the burden of proof lies on you to match every detail.
I started with a Google Earth project and started adding pins to my project to get a rough idea, expanding on it when I found key points to match up and identify. This was my first iteration:
As you can see, the track which I had matched up with reasonable confidence became the focal point of identifying the other locations. From it, I had several features to gain directions and distances.
For example, the straight part of the track gave me an orientation and a rough angle to direct my search, the distance between the smoke plumes gave me a rough distance to look (very rough guesswork here).
I was then able to use the close-up of the burning wreckage to find the exact building and camera location from the first video.
Details such as the pylons and the building walls, sheds, roof colour, and open land around it all played a factor. Finding the pylons was a tricky affair since they were frames and hard to see with the imagery resolution. Luckily the approximate locations and odd shapes on the map gave me a clue where to look, and I just needed to match the shadows to verify that there was a tall structure there.
There were a series of questions and concerns that I needed to address with the image. Firstly, the building that I suspected to be the one damaged needed more evidence. The satellite angle wasn’t great, but there was a nearby building which was of a similar height and I could easily count the number of levels. It had four stories and so I compared the length of the shadow to that of the suspected building and found that they were very similar compared to shadows of taller/shorter buildings. Mapping of the shape and comparing played a large role too.
The second issue I had was that there were clearly buildings in between the camera looking at the track and the crash site. These were not visible in the race track footage. Luckily the satellite data showed a lot of the rooftops under tree cover and with shadows cast on them by the trees. This meant that I could reasonably assume they were shorter than the tree line which was fairly uniform.
Verification
The next stage was where things really kicked off. A number of other Twitter users had started posting images and grid references matching my location and were posting the same analysis, which was good to see as it was independent corroboration of my results.
These steps became a case of combining knowledge to maximise analysis and minimise reasonable doubt on our accuracy.
One important feature that I hadn’t considered was the hill in the background. As a fairly boring and low-profile hill I was worried that it would be difficult to match, but based on the knowledge that I had pretty accurate locations for the cameras I was able to match the hill too.
A key feature of this hill, as pointed out to me by Twitter user Evergreen Intel, was the definition of the trees. We had to be looking at a hill relatively nearby to get that level of definition in the image.
Luckily, with the camera position, building position, and the pylon position known, I was able to find a good match which fell within the parameters of “reasonably close”.
The hill in Google Earth’s 3D mode matched the scale expected and a calculation showed that the distance (yellow line) was around 900–1000m.
Additional analysis
The geolocation itself was just the first step to achieve a good overall picture on what was being reported. It was crucial to determining the likelihood of Taiwan shooting down the plane.
The confidence in having a crash site identified means we can look at the context of the claim and challenge it.
Because the claims were that Taiwan shot down a fighter jet over the island or in the Strait, we could look at the capabilities and practicality of the attack.
Firstly, calculating the distance between the crash site and Taiwan led to a rough distance of 1000km. This could be matched up with intuition and data on Taiwan’s surface-to-air missiles and anti-air defence systems.
Public data shows that even their best missile systems have a range of around 200km giving the plane an unlikely 800km crash trajectory before hitting the ground.
To put that into context, an SU-35 (one of the alleged models of the crashed plane) would be able to travel that far if it maintained top speed in around 17 minutes.
Due to critical damage, it probably wouldn’t be going top speed with afterburners on full blast. It also would pass over a number of airports and bases during that time to be able to make an emergency landing.
Near to the landing site is an airbase. That airbase has been home or a stop-off for what have been identified as J-10s, the second likely option for the plane’s model.
Historical analysis shows a J-10 on the base in the past 12 months. This doesn’t prove the identity of the plane by any means, but it may be important for future developments.
It also opens up the possibility of an early-flight issue, which could have caused the plane to crash shortly after take-off, but without further data, this is only speculation.
Conclusion
Taiwan’s Ministry of National Defence wrote: “In response to rumors online that claim a Chinese Su-35 fighter jet had been shot down by Taiwan air defense systems, #ROCAirForce would like to categorically state this is fake news.”
While there are still unknowns, such as why the plane crashed, what it’s final moments were like, and where the pilot landed. It is very unlikely that the Indian media coverage of the story matches with the findings published here.
The fact that sources of the video clearly stated the general area of the crash, prompts suspicion of a malicious effort to distort the nature of the crash when Indian news sites republished the footage without that information.
The massive outpouring of support for Taiwan shows that the narrative suited the wants of the Indian public, rather than accuracy.
