Commissioned By
Additional Funding
Collaborators
Methodologies
Forums
On 7 April 2018, the city of Douma, Syria, was targeted by two chemical weapons attacks. At that time, the city and its surrounding areas had been under siege by the Syrian military since 2013. At least 70 people died in the attacks, according to reports.
Two distinctive yellow canisters, similar to those previously linked to chlorine gas attacks across Syria, were found following the strikes: one on a rooftop balcony near to Al Shuhada square, and the other in a bedroom at an unknown location in the city.
Days later, following the negotiated surrender of the remaining rebel forces in Douma, Russian media were the first to be allowed access to the attack sites.
Reporters from RT and TV Zvezda quickly claimed that the attacks had been staged, and that the canisters had been carried into place by the rebels, rather than dropped from regime-controlled airspace above the city.
Commissioned by and working closely with the New York Times, we searched for physical evidence concerning how the canisters had arrived at each scene.
We reconstructed the two sites as 3D models using available images and video material. We also reconstructed the canisters as digital objects and analysed the physical traces inscribed on them.
Our analysis supports the assessment that the canisters were dropped from the air. At both sites, we identified a harness consistent with the aerial delivery of munitions documented elsewhere in Syria. Image and sound analysis revealed further evidence at each site.
In June 2018, the New York Times published our analysis as part of their own video investigation and an interactive ‘augmented reality’ feature.
Update
24.09.2019
An ‘immersive investigation’ by the New York Times’ visual investigations team, which features our analysis, is awarded a News & Documentary Emmy. You can view the piece, One Building, One Bomb: How Assad Gassed his Own People, here.
Update
01.03.2019
The OPCW’s ‘fact-finding mission’ assigned to investigate the Douma incident (OPCW-FFM) publishes its final report into the Douma attacks, including precise measurements of the size of the cylinders at Locations 2 and 4 (our sites 1 and 2, respectively), the size of the crater in the roof at Location 4 (our site 2), and dimensions of the spaces in which they were found.
Using these precise dimensions, we scaled and refined our models of both sites.
We used a technique called ‘image-matching’ to verify the dimensions provided in the OPCW report. Image-matching is a way of juxtaposing photographic evidence with spatial measurements.
In the process of image-matching, given measurements are supported when a photograph is matched to the model from the same camera position, lense angle and direction, as could be seen in the video above. Our image matching supports the measurements provided by the OPCW.
We also refined our model of the cylinder at Location 2, the balcony location that our original analysis called ‘site 1’.
You can find our updated digital models with camera positions of site and cylinders relating to Locations 2 and 4 on our Github.
Update
13.05.2019
A report leaks from within the OPCW, authored by Ian Henderson, a long-time employee of the agency as a member of an ‘engineering sub-team’ within the OPCW-FFM.
The report argues that it was more likely that the cylinders were placed at Locations 2 and 4, than that they were delivered there by air. In part, this conclusion partially relies on Henderson’s assessment of the relative sizes of the cylinder found at Location 4, and the crater in the ceiling.
Forensic Architecture and Bellingcat worked together to examine and compare the two assessments: the final report by the OPCW-FFM, and the leaked ‘Henderson report’. Forensic Architecture examined the claims made by both reports about the (relative) dimensions of the cylinder found at Location 4, and the crater in the roof through which the cylinder apparently entered the room.
According to the OPCW-FFM report, the cylinder found at Location 4 was 1.4m in length. The report described the dimensions of the crater in the roof as 1.66 x 1.05m. The Henderson report does not independently define the dimensions of the cylinder, but provides the following diagram (from the Henderson report, Appendix 4, Figure 9), based upon a drawing from the OPCW-FFM report.
The diagram, which Henderson’s report says was produced for ‘illustrative purposes’, notes a scale of 1:10 at the top, but appears to have been photocopied and reduced in size, and is as such not provided to any clearly defined scale.
The diagram places the front of the cylinder over the edge of the crater, appears to support Henderson’s claims that this cylinder could not have passed through this hole. ‘it was not possible to establish a set of circumstances where the post-deformation cylinder could fit through the crater with the valve [the pressurised sealed opening at the end of the cylinder] still intact… and the fins deformed in the manner observed”’.
The Henderson report claims that its analysis was ‘conducted using sources of information available to the FFM’. However, if we follow the OPCW-FFM’s dimensions, there appears to be an error in the diagram included in Henderson report.
According to our analysis, if the crater in the diagram is 1.66×1.05m (as stated by the OPCW-FFM report), the cylinder drawn by Henderson would be approximately 1.48m in length; that is, 8cm too long. This difference, of almost 6%, is significant, since it challenges Henderson’s assessment that the cylinder could not have passed through a crater of the dimensions specified by the OPCW-FFM report.
Alternatively, when we scale the Henderson report’s diagram to the size of the cylinder as determined by the OPCW-FFM report — that is to 1.4m length — the dimensions of the crater are around 6 percent smaller than those provided by the OPCW.
Given that Henderson’s drawing has no scale or measurement we found that the proportion between the cylinder and the crater were drawn incorrectly: either the cylinder is drawn too large, or the crater too small.
We used our model of Location 4 to ‘image-match’ cylinders of 1.48m and 1.4m in length. The study below confirms that the photographic evidence from Location 4 is consistent with a cylinder of 1.4m in length, and is clearly inconsistent with a cylinder of 1.48m. This image-matching process suggests that the Henderson report diagram has most likely exaggerated the length of the cylinder, and not the crater.
The Henderson report’s diagram superimposes over the crater an outline of the cylinder’s ‘harness’ in its damaged, deformed state. In this state, not only is the cylinder’s dimensions exaggerated but the cylinder and harness together have a larger ‘footprint’ than the state in which it would have impacted the roof.
This way of drawing is misleading: if the diagram is drawn to illustrate or test the hypothesis that the cylinder and harness could not have fitted through the crater, it should have taken into account the fact that the deformation would have been caused by the impact of the harness and the ceiling, and by the subsequent impact on the floor of the room below.
This study does not change our previous determination that the hole in the ceiling and the location of the cylinder on the bed do not easily align. It is possible that this cylinder was moved after arriving at the site from the air, as, our previous investigation showed, appears to have been the case at Location 2.
Update
06.02.2020
Following a number of leaks from within the organisation concerning its investigation of the Douma attacks, the OPCW publishes the report of an ‘Independent Investigation into Possible Breaches of Confidentiality’.
Methodology
Methodology
We used open-source software called Blender to build 3D models of the attack sites and the canisters themselves. The dimensions of each 3D model were established from satellite imagery and corroborated by architectural details in images captured at ground level.
To reconstruct the canisters’ exterior surfaces, we projected multiple images from different angles onto a cylindrical surface. In doing so, we took into account variations in the position and focal length of different cameras, and corroborated details on the canisters’ surfaces across multiple images.
Download Blender here, and our models here: Site 1 / Site 2.