The Astroworld music festival in Houston, Texas, kicked off last Friday, but tragedy struck when Grammy-nominated rapper Travis Scott—who launched the festival in 2018—took to the stage around 9 pm. The enthusiastic crowd surged toward the stage and packed the mosh pit so tightly that people couldn’t breathe and began to pass out. There was no space to move, and in the end, at least eight people were killed, and another 25 were hospitalized.
Concert promoter Live Nation issued a statement saying it was “heartbroken for those lost and impacted at Astroworld,” and the company pledged its full cooperation with local authorities who are investigating. As for Houston native Scott, he pronounced himself “just devastated” in a video posted to his Instagram account last Saturday night and said he had not realized how severe the situation had become from his vantage point onstage. The rapper seems equally reluctant to take the stage in the immediate aftermath of the tragedy: Scott has reportedly canceled a planned set at the hiphop “Day N Vegas” festival, with sources telling Vulture the rapper is “too distraught to play.”
There is still a great deal we don’t know about the conditions at Astroworld and what actually happened that night, pending the results of an official investigation. But deadly crowd surges are a far too common occurrence all over the world. For instance, back in 1979, 11 people were trampled to death during a Who concert in Cincinnati. In 2000, nine people were trampled to death at a Pearl Jam concert during Denmark’s Roskilde Festival. And in April of this year in Meron, Israel, 45 people died in a crush at the Lag B’Omer religious festival, with 150 more injured.
Scientists have been studying crowd dynamics for decades in hopes of developing better strategies to avoid these kinds of tragedies. Typically, they have used computer simulations. Access to archival videos of these kinds of incidents can help, like the footage from the January 2006 Hajj to Mecca. Over 2 million Sunni Muslims were making their way along an established route to the Saudi city. As the route narrowed at the Jamaraat Bridge, the density of the crowd increased drastically, as people rushed to complete the final symbolic stoning ritual at Mina before sunset. A stampede ensued, killing 363. (That death toll, while high, pales in comparison to the roughly 2,400 pilgrims killed in another stampede near Mecca in 2015.)
Dirk Helbing and Anders Johansson of the Dresden University of Technology were able to analyze the video footage and developed a computer algorithm to track the position and velocity of every person in the crowd over a 45-minute period. They identified three distinct phases of the crowd’s motion. The crowd initially moved toward the bridge at a steady rate, but as the density increased, there was an abrupt phase transition to a kind of “stop-and-go” motion. This spread like a wave in the same direction the pilgrims were moving. The crowd density continued to increase until another sudden phase transition occurred, whereby pilgrims started moving randomly in all possible directions.
Helbing and Johansson dubbed this phenomenon “crowd turbulence,” or a “crowd quake,” and they found the critical threshold seemed to be about six people per square meter (10 square feet). “The researchers believe that turbulence may have been brought on by individuals panicking and pushing in all directions to increase their personal space,” Hamish Johnston wrote in Physics World in 2007. “This caused violent pressure waves to surge through the crowd, tossing individuals several metres, tearing off clothing and ultimately leading to the trampling of hundreds of pilgrims.”
The Jamaraat Bridge scenario is an example of a bottleneck. A similar bottleneck occurred in Eastern Germany during the 2010 Love Parade, a popular music festival. The bottleneck in this case was a 200-meter-long tunnel, through which attendees had to pass in order to get to one of the festival events. But the passageway was too small to handle such an immense crowd, and the density soon increased to dangerous levels. Police tried to block more people from entering the jammed parade grounds, triggering a stampede. People began to suffocate around 5 pm as thousands of other revelers danced to techno music, unaware of the tragedy that was unfolding nearby. In the end, 21 people died, and 651 were injured.
The Astroworld tragedy seems to have been centered on the crowd packed into the mosh pit rather than a more typical bottleneck scenario. There was a 2013 study on mosh-pit dynamics by a group of physics students at Cornell University, inspired when co-author Jesse Silverberg attended a heavy metal concert with his girlfriend. He wisely avoided the mosh pit and, like a true physicist, found himself fascinated by the motion of the crowd, which struck him as resembling the disordered collisions of molecules in a gas.
Silverberg and his co-authors decided to simulate mosh-pit dynamics. They drew on footage from rock concerts posted on YouTube and used a particle tracking program to convert everyone in those crowds into individual particles, dubbed MASHERS (Mobile Active Simulated Humanoids). There were two kinds of MASHERS: passive ones who stayed stationary after an accidental collision, and active ones who bounced after a collision. The researchers found that, when there were more active MASHERS than passive ones, the crowd did indeed behave like molecules in a gas, with random collisions. But sometimes, spontaneous “flocking” would occur, in which MASHERS began following their neighbors’ motions. In that scenario, vortices would form—basically human whirlpools.
Of course, people aren’t particles, and Silverberg et al. freely admitted they were using very simple mathematical models. Human beings are complicated and unpredictable, which is why there has been a great deal of recent work attempting to incorporate the human factor into the modeling of crowds.
For instance, a 2015 study by scientists at the University of Technology in Iran created a simulation that included so-called “emotional contagion.” In it, the simulated people became increasingly fearful and panicked—expressed as increasingly random movement—as they failed to find an exit from the crowded virtual environment. Similarly, a 2018 study by University of Plymouth researchers figured out how to measure the kinetic energy of crowds in real-time videos, using that as a gauge to identify areas where the crowd was transitioning into a dangerous emotional state.
Dinesh Manocha, a computer scientist at the University of Maryland, has conducted several studies on crowd behavior. He has sought to incorporate not just physics and physiology but also psychology into his models. That said, “In many ways, we don’t have access to the exact data, situation and crowd movement that occurs at such tragedies,” Manocha told Ars. “You typically hear the experiences of a few attendees or some isolated pictures and videos that do not provide all the details.” Nonetheless, there are two factors he has observed in his research over the years that seem common to all such tragedies.
The first, as we’ve seen, is density—specifically, situations where crowd density reaches more than four people per square meter. “In many ways, each human or pedestrian loses his/her capability to move independently in such density, but rather they become part of a macroscopic flow,” said Manocha. “So crowd tragedies are more likely to happen in such scenarios, as humans lack the ability to escape from the crowd flow.”