SMOKESCREEN: Behind the scenes of our visual investigation into Iraq’s lethal use of tear gas grenades

In October 2019 mass protests erupted in Baghdad and across southern Iraq. Amnesty’s Crisis Evidence Lab was asked to use open source techniques to support the organization’s Iraq research team to investigate the crackdown against protesters. This soon included videos of the most gruesome deaths: protesters being killed with 40mm military-style grenades embedded in their skulls. 

After identifying the make and provenance of the tear gas and smoke grenades, Amnesty wrote up its findings and campaigned about the issue extensively, using cutting-edge methodologies to show how security forces in Iraq were using these grenades deliberately to kill and maim. 

As part of this investigation, Amnesty approached SITU Research to collaborate on building a platform to show just how deliberate these attacks were. The result is SMOKESCREEN— an immersive 3D environment that synthesizes geolocated eyewitness footage and narrative into an intuitive user experience. This post by SITU Research is a behind-the-scenes look at how the platform was built, and an explanation of the methodology used.


The goal of the three-month collaboration was to contextualize and analyze an archive of social media videos showing use of the tear gas grenades. There were two main challenges. First, how to communicate the lethality of these grenades without including the gruesome videos and photographs. Second, how to convey the performance characteristics of the munitions in relation to the public spaces and urban environment in which the protests were happening. The solution was, in equal parts, a mix of content curation, user experience design, event reconstruction, and technical analysis.  


Using commercially-sourced 30cm-resolution satellite imagery and reference photos of the area, we constructed a digital model of the urban environment, extending 3.5 kilometres around Tahrir Square. We used this model and known landmarks to estimate the videographers’ locations and cones of view in the open source software Blender. This software’s motion tracking functionality also allowed us to compute camera motion and thus capture the micro-movements of the videographer. 

The result is a seamless visual and spatial integration of primary source material with our diagrammatic and representational content. The model and 3D camera motion extends the field of view beyond what is captured in the primary videos, which adds yet another layer of spatial information to the platform. It is the multimodal and continuous spatial awareness that allows the user to draw crucial conclusions about the relationship between the urban environment and the misuse of the weapon. For example, we know that these canisters are particularly lethal when fired directly at protesters. By combining the digital model with the video footage we can reveal the Iraqi security forces’ low angle of fire in relation to the location of crowds gathered on the Jimhouriya Bridge in central Baghdad.


In designing the user experience, we knew we wanted to interweave the open source video  with explanatory text, while losing neither the spatial context nor the narrative continuity. By placing the footage in a 3D model of its original urban environment — the area around Tahrir Square and the Jimhouriya Bridge — we were able to create added value: the built environment became the glue to connect disparate videos shot at different times by different people. Next, we reduced the video archive from over fifty videos to 10 clips of under 5 seconds each, thus isolating key moments from the demonstrations. In some of the clips the 40mm grenades are clearly visible in flight. Others give context to the locations of protesters and security forces, or highlight the volume of grenades deployed. This video analysis was only made possible because of the extensive video documentation done by activists in Baghdad. 


The platform features a scroll-based navigation that allows visitors to move forwards or backwards through the textual, visual and spatial content at their own pace and with a consistent motion.

SMOKESCREEN maximizes user control, allowing viewers to manipulate the platform’s video content as they scroll through. Playback cues are linked with page navigation, such that the user scrolling seamlessly translates to video seeking. The mechanism is akin to placing a hand on the platter of a vinyl record player, thereby pausing playback and potentially pushing the record in either direction and, in this case, allowing viewers to examine video content frame by frame. SITU Research worked with consultant Studio Scasascia to refine this technique.


In order to deliver a smooth scroll-based video seeking, we had to adopt an unorthodox video playback implementation. Due to optimizations of HTML5 video in the browser, video seeking is time-based, not frame-based, as might be the case in standard video editing software.

This means that quickly changing the playback position will generally cause the player to drop frames, and the viewer will likely miss content. Since each and every frame of the videos embedded within the platform shows critical visual evidence, it was essential not to lose any of this content. As a solution, we rendered each clip as a sequence of discrete images: a digital flipbook. This approach guarantees a smooth viewing experience with a consistent frame rate, and reveals the granular, frame-by-frame analysis that goes into digital verification and event reconstruction.


An important element of SMOKESCREEN is to show the lethal impact of the 40mm grenades on the human body. To represent this without showing graphic footage of actual injuries, we used an analytical tool from the engineering field. This analysis also allowed us to understand and corroborate first-hand reports of just how deadly these nominally ‘less-lethal’ weapons can be. We produced a numerical model using a finite element analysis tool that simulates the effects of physical forces on interacting objects. Our methodology was to simulate the impact of projectiles against a proxy of the human body. We used ballistics gelatin as a medium — a material often used to quantify and measure ballistic impact. While the human body is much more complex, ballistic gelatin is used because its elastic and tensile properties are broadly comparable to human flesh and muscle. We created two digital models: one from the manufacturers’ specifications of the M713 40mm tear gas grenades found in use in Baghdad, and a second from a standard 12-gauge shotgun slug used for hunting, which is expressly designed to kill.

While there is a lot of research documenting the performance of less lethal weapons when discharged in manner consistent with the stated intent of deploying tear gas (i.e., at a high angle) there is a surprising dearth of analysis of the same weapons when misused. All one has to do is lower the angle of fire to turn this 40mm canister into a deadly munition. It is essential this fact is understood and the nominal classification of “less lethal” does not become a cover for the lethal and excessive use of force.

SMOKESCREEN shows the importance of bringing different elements of analysis from a variety of fields together in the human rights space today. By doing so, Amnesty International and SITU Research, along with other partners, were able to show how the Iraqi security forces deliberately targeted protesters by firing the canisters around head height, with devastating results.

This guest post was written by Jeevan Farias, Researcher at SITU Research