GEOINT : Documenting destruction in Donbas

Predicta Lab
6 min readJan 25, 2023

--

Recently, I came across this tweet sharing a clip of a city devastated by the war :

Goal

This type of content gives us an opportunity to report on the state of devastation in Donbass after 11 months of war. It also allows us to participate in the general efforts to document the Russian invasion of Ukraine.

Hence we will find the locations shown in the video and reconstitute the route taken by the drone filming it.

What ?

As you can see this video is reportedly showing destruction in Marinka, Donbas: we will start by checking that information and to do so we will go where the video was first posted.

Who ?

Fortunately, the watermark embedded in the leads us straight to the Telegram channel that first shared the video. We can find all the videos posted by this channel in the profile.

Here we see only a few videos were posted in January and only two of them seem to show the same aerial view as the video we’re looking for.

However on Twitter the video lasts 2:20 and none of these have the same duration so the Twitter video is most likely a clip from the longest of these two (06:13).

After watching both videos we can confirm that the video posted on Twitter is a clip of the 06:13 telegram video: it shows the time range 02:18–04:38.

When ?

The video was posted on January 15th 2023 but that is not necessarily when it was shot. We can check the metadata (below) to find this information.

Unfortunately, the creation date isn’t indicated (or it was removed) so we cannot say exactly when the video was taken, only that it was between the 16/10/2021 (the date of the image on Google Earth) and the 15/01/2023 (when it was posted on Telegram).

To know more about metadata and how it can be used for OSINT purposes, you can have a look at our article on the topic.

Now let’s do some GEOINT.

Where ?

To start from the beginning, I will try to locate the building shown in the first frame. The view from the side that is first visible isn’t ideal for geolocation but fortunately the drone quickly comes around and shows a full-front view of the building. We get the image below at 00:14.

Luckily for us this building is rather recognizable : it is high, and in an orange-ish color with stripes. We can deduce the shape of the building from the video, and we can also distinguish a road that passes in front of it (indicated by the red arrows). Hence, when looking from above we have to search for the following shape :

So let’s get searching !

I start by going to Marinka on Google Earth. A simple search of “Marinka” gives us the right location in the Donetsk region as our first choice. In the satellite image 16/10/2021, we see that the town is organized around a main street (Druzhby Ave. - просп. Дружби) and most of the large buildings seem to be close around it. Hopefully that’s the road we have already spotted, anyhow it seems like a good place to start looking for our big orange building.

Indeed we can find the building along that main road, at the coordinates 47.940432, 37.501471

To verify this, my first instinct is to go to Yandex Maps because it tends to have a better street-view coverage than Google Earth for ex-USSR countries. But this is not the case here : no streets of Marinka are street-view on Yandex Maps whereas the main road is available in Google Earth. Here’s what we see there :

Although this is not the same view we get in the video (it’s the other side), the shape of the building as well as the matching color and stripes confirm this is a match.

This building is the clinic Tsentr Pervynnoyi Medyko-Sanitarnoyi Dopomohy (Центр первинної медико-санітарної допомоги).

We see a second recognizable place in the video at 01:39.

We can find the corresponding building on Google earth by following the trajectory of the drone. This leads us to the following coordinates : 47.944381, 37.504932.

This place is a school, the Mariinsky institution of general secondary education (Мар’їнський заклад загальної середньої освіти).

Source of the first picture

Finally, the last place that we see in the video at 02:12 is located at the following coordinates : 47.945603, 37.503805.

As mentionned, on Google Earth the default image of Marinka dates from 2021 and it’s rather clear but by going to the 2020 image we can also distinguish the distinctive blue staircase that remains in the video (indicated by the red circle above).

There is no street-view image available for this building, but Yandex maps indicates this building is a library.

The flight path

Overall, if we follow the flight path of the drone in the video we get (approximately) the following trajectory on Google Earth.

Before/after in a street block

To further estimate the extent of the devastation in Marinka, we can focus on one of the blocks that make up the town.

If we look at the block of buildings just south of the school, on Google Earth we can count 10 buildings.

Now if we go back to the video, we see what was left of this block between the 00:45 and 01:08 time stamps.

In this block, all 10 buildings have at least partially destroyed roofs and destroyed window frames.

  • 40% have holes in their roofs caused by rockets (3,4,6,9)
  • 20% only have some frames left of their roofs (5, 10)
  • 20% have half of their roofs collapsed (1,8)
  • 20% only have some walls that remain of the entire building (5, 10)

Thus, looking at this block and the rest of the town visible in the video, it seems that the majority of the buildings in Marinka have been hit and are badly damaged.

Conclusion

The large amount of media content we receive from Ukraine can be used to document the devastation endured by the country. Such efforts will surely be needed later and used in efforts of accountability.

For more content of the sort check out our article retracing the path of a car in Mariupol and our other OSINT articles on our Medium page.

If you want to be notified when we upload new content be sure to follow us on Twitter, Youtube, and LinkedIn.

--

--

Predicta Lab
Predicta Lab

Written by Predicta Lab

Official account of Predicta Lab

No responses yet