Hajj crush: how crowd disasters happen, and how they can be avoided
Mass panic? Stampedes? Nonsense, say the experts trying to stop another disaster like last week’s in Mecca: they’re failures of management, and they aren’t inevitable. So why aren’t they a thing of the past?
Saturday 3 October 2015
My story comes from the Notting Hill carnival in 1999, though you may have your own. It was sunny and there wasn’t really a plan, other than to wander the streets and enjoy ourselves. Towards the end of the day, we came to a crossroads flooded with thousands of people. Like everybody else, we headed in.
Gradually the crowd tightened as we neared the junction. My friends and I had to go single-file, which made it difficult to see each other. If we wanted to talk we had to shout, because everybody else was shouting. Soon the crowd was so tight that it was impossible to turn back, or even choose direction. My arms were pinned by my sides. My head was wedged like an egg into a box of others. Mostly we all just stood and waited, but every so often the pack would slip forwards and we’d all have to shuffle with it. Sometimes we’d be carried.
I’m 5ft10in, so I could just about peer around. Smaller people were caged between chests and backs, their only view upwards at a ragged patch of blue. A few police were stationed behind crash barriers at the side shouting helpful things like, “Keep moving, please!” At one point I remember asking one of them how much longer this would last, only to be yelled at angrily. I don’t remember what he said, just that he was not in charge of his emotions.
Now and then there’d be a cry of panic, somewhere, followed by angry calls to stop pushing. I imagined some people were getting claustrophobic, understandably. I was pretty keen to get out myself, but I don’t think I ever doubted that I would. The idea that I was in danger seemed silly, and indeed some people were laughing. We were outside. There was no urgency. How could anybody die from lack of space beneath this empty sky?
Nobody was crushed to death at the 1999 Notting Hill carnival, but I was wrong about crowds. Since 1999, crush accidents killing 10 or more people have happened on at least 44 recorded occasions, or about one every four months. They’ve happened in South Africa, Ghana, Japan, the US, China, India, the Philippines, Tanzania, the Netherlands, Germany, Cambodia, Brazil, Egypt … everywhere. In 2005, 953 people died in the Al-Aaimmah bridge crush in Baghdad – a loss of life far greater than in any of Iraq’s terrorist attacks.The list of crowd disasters associated with the Hajj in Mecca, Saudi Arabia, is especially appalling. In 1990, 1,426 people were crushed to death, in 1994 it was 270, in 1998 it was 118 (at least), in 2001 it was 35, in 2003 it was 14, in 2004 it was 251, in 2005 it was three, in 2006 it was 360 (at least). More surprising than the 769 deaths so far confirmed from last month’s crush is the fact that the Hajj got through eight consecutive quiet years.
Why do crowd crushes happen? Why do they keep happening? That is easy. These are not natural disasters. They happen because no one stops them. Indeed, Professor Edwin Galea thinks they are happening more frequently. “Not all of them get reported,” he says. “But I think they are increasing. We’re trying to service more people in less time, which means higher crowd densities, which is leading to increased frequency. More urbanisation, I think that’s the key thing.”
Galea and his Fire Safety Engineering Group at the University of Greenwich use behavioural experiments and mathematical modelling to understand how crowds move in different scenarios. The aim is to prevent dangerous densities from building up. At the table in his narrow office, lined with box files, I tell him about my time at the junction in 1999. “That’s extremely dangerous,” he says without hesitation. “You should never be in that situation. That should never happen.”
Galea is Australian by birth and an astrophysicist by training, having originally specialised in the fluid dynamics inside stars. He became interested in the spread of fires after the disaster on the tarmac at Manchester airport in 1985, in which 54 people died when a Boeing 737 caught fire; then he became interested in the movement of crowds, then in crowd behaviour. Crowd crushes used to be thought of as a fluid-dynamics problem, but that isn’t quite right. When people are packed very tightly, they do move somewhat like a fluid, but understanding the dynamics of that kind of crowd isn’t going to keep it safe. By the time a crowd is moving like a fluid, it is already in danger. Knowing what it is going to do is like knowing how a plane will move without wings – interesting, maybe, but too late.
When crowds are not too tightly packed – less than four people per square metre – the individuals in them have room to make decisions, and they do not move like a fluid. When you think about this, it’s obvious: for any number of reasons, people aren’t particles. You don’t usually go to a music festival alone, you go in a group, which means you are not just trying to get somewhere, you are also trying to stay together. Sometimes you turn round after forgetting something, or you veer off when you recognise a friend. Galea’s research suggests that people are generally decent and calm, and behave rationally, even in a crisis, but their decisions can only be based on what they know – which is rarely what simple models expect.
Instead of treating them as “ball bearings” therefore, Galea and a few others around the world have developed software that models crowds as what they are – a mass of complex agents, simulating as closely as possible the choices and limitations of real people. “Each of the agents has a respiration rate,” Galea explains, “and [in a fire] they’re inhaling the toxic products and being exposed to the heat. We’re calculating and predicting the effect of the exposure to the products and the exposure to the heat had on the person. They’ll try and avoid hot areas, for example. If the smoke descends, they’ll fall to the ground and try to crawl … ” It sounds intensely complex. “It is.”
He shows me the result on his computer – blocky animations, many of them free to view on YouTube, which show hundreds or thousands of stick figures trying to deal with crushes and fires – sometimes crushes and fires that actually happened. They make unpleasant viewing. In one, Galea’s people try to escape the Station nightclub fire on Rhode Island in 2003. You can see how they mostly ignore the emergency exits and head for the front, where a severe crush develops, killing some and blocking the path of the others, who try to break windows, then try to crawl, then die. After three minutes there are 100 corpses on the floor, just as there were after three minutes of the real thing.
If crowd densities rise above four people per square metre, and especially if they get to six, they become prone to two broad types of accident. Both – be warned – are horrifying. A “progressive crowd collapse” often begins when one person falls over, perhaps because of a shock wave travelling through the crowd, or just because they slip. Suddenly denied the fallen person’s body to lean against, but still under pressure on the other side, the adjacent people fall on top of them, or are knocked over trying to help. This in turn creates a larger hole, which ever more people are forced into, and more, and more, piled high on top of each other until the pressure eases. It is being buried alive in bodies, perhaps bodies that you know. A progressive crowd collapse often happens when a large crowd is moving steadily through a confined route, driving ever forwards. Early indications are that the most recent disaster in Mecca was a progressive crowd collapse, seemingly caused by the meeting of two crowds that were moving against each other – something that should simply never be allowed to happen.
The other type of crowd disaster occurs when people are simply squeezed to such an extent that they can no longer inflate their lungs, and die gradually of compressive asphyxiation. This is a “crowd crush”, and it typically happens when ever more people push into a confined area – either on the way in, or trying to get out. Often those who die in crowd crushes are the ones against the wall. Hillsborough was one example, but many terrible fires are crowd crushes, too. No matter how calmly a crowd behaves, it can only fit through a narrow exit at a certain rate. Those caught at the back have a choice between pushing harder and waiting to die. In 1997, 243 people were killed on the Hajj when a fire among the tents caused a crowd crush.
For all their complexity, however, crowd disasters are as much a political problem as a technical one. A common reaction – indeed the usual reaction – is to evoke the idea of an indiscriminate mob, of mass panic. To blame, in short, the crowd. In the case of Hillsborough, this was done deliberately by the police and the Sun newspaper. In other cases, it may just be assumed and implied. People who have never seen mass panic find it easy to imagine, but in fact that’s almost everybody, because mass panic virtually does not exist. Indeed believing in mass panics is dangerous, because it means the authorities sometimes conceal alarming but important information for fear of starting one. “Utter, complete rubbish,” is what Galea thinks of that strategy. “All the evidence shows that people will be able to react and take sensible decisions based on the information you provide. You don’t want to provide them with too much information so they can’t process it all … You just want to provide them with accurate, simple information they can act on.”
Saudi Arabia under pressure to improve safety at Mecca after fatal hajj crush
One word bears a lot of blame here, at least in English. Mention a “stampede” in front of Galea and he starts to look pretty wild-eyed. “This is just absolute nonsense,” he says. “It’s pure ignorance, and laziness … It gives the impression that it was a mindless crowd only caring about themselves, and they were prepared to crush people.” The truth is that people are only directly crushed by others who have no choice in the matter, and the people who can choose don’t know what is going on because they’re too far away from the epicentre – often reassuringly surrounded by marshals and smiling faces. Like I was, in fact, when I entered – and helped create – that dangerous carnival crowd.
On the extremely rare occasions that a real stampede happens – that is, people running over you – it is unlikely to be fatal. “If you look at the analysis, I’ve not seen any instances of the cause of mass fatalities being a stampede,” says Keith Still, professor of crowd science at Manchester Metropolitan University. “People don’t die because they panic. They panic because they are dying.” In Still and Galea’s small but growing field, this is now the consensus view. “Crowd quakes [or collapses] are a typical reason for crowd disasters, to be distinguished from those resulting from ‘mass panic’ or ‘crowd crushes’,” says Dirk Helbing, a computing professor at ETH Zurich. “The idea of the hysterical mass is a myth,” says Paul Torrens, a professor at the Center for Geospatial Information Science at the University of Maryland.
Still advises architects, police and event planners – including the planners of the royal wedding – on handling large events, and he insists that crushes are wholly “preventable, predictable and avoidable”. In 2004, he was hired by the Saudi authorities to consult on designs for the new Jamarat bridge in Mecca, in order to ease what had become an obvious bottleneck in front of the pillars. As he explains, however, that may not solve the problem. “They are improving one part of a complex system, making it much more efficient,” he says. “It used to only handle 200,000 people per hour. Now it can handle around 500,000 or 600,000. So that load puts pressure on the rest of the system” – in other words, a greater number of people arriving at potential bottleneck points further along the route. Since the new bridge was built, the layout of the Mina valley encampment, where the recent disaster happened, has not changed.
A redesign of the whole complex could, Still believes, make the Hajj safe, but Galea has his doubts. “If you’re designing an event to handle that crowd density, it’s inherently dangerous,” he says. The challenges, as always, are the pinch points. Everybody must get within throwing distance of the pillars, for instance (where the deaths occurred in 1994, 2003, 2005 and 2006), and everybody must circle the Kaaba within the Masjid-al-Haram, so there will always be a limit to how many people can safely be present in these spaces in a day – and numbers at the Hajj are quite extraordinary. For comparison, 500,000 people an hour over Jamarat bridge is the largest-ever football crowd once every 24 minutes, or the whole of Germany once a week. One possible solution, of course, would be to spread the Hajj over more days.
The problem for crowd management generally, however, is that no two places are alike. In 2010, 21 young people were crushed to death in Duisberg, Germany, during a music festival called the Love Parade, which was taking place on a new site. A plan had been made, but for some reason it expected the crowds to flow in both directions through a single tunnel, where the disaster happened. In fact, by dividing the flow into three streams, Galea’s model suggests that twice as many people could actually have moved safely through the area.
Still speaks highly of local knowledge and experience in crowd management at large events, but also warns of the false confidence they can create. “For instance,” he says, “I saw a gold [police] commander who took over Wembley stadium with a tactic that worked well at Arsenal. He wanted to try it that way at Wembley, and 30,000 lives were at risk … Nobody was even in distress, but it was one of those moments where you thought, ‘They’re all going through a 6ft-wide gate, and there’s nothing to stop them.’ I was in the control room at the time, and everybody was sitting there watching it unfold. The strategy worked in geometry A, but it did not work at all in geometry B. I witnessed not just that, but a number of close calls.” What infuriates Still especially is that we fail to learn from them. “There is nothing in the industry that records that information and seeks to improve or better it in following years,” he says. If Notting Hill 1999 is recorded anywhere, it’s here, only, and maybe in a police officer’s nightmares.
If such a system were implemented, if the law more generally required organisers to “ensure crowd safety”, if we could stop calling them “stampedes”, and if new research and technology could help – perhaps by putting real-time simulation software on to marshals’ phones – then crowd crushes might stop happening. In the meantime, the near-misses – and the catastrophes – will continue. Just a few months ago Galea himself experienced a dangerous crowd at London Bridge station. “I feared for my safety and other people’s safety,” he says. “I thought this could easily end up being a tragedy. Very easily.” He made a formal complaint, hopefully with some effect.
Next time you approach a large crowd, Still has some basic rules to follow. “Be aware of your surroundings,” he says. “Look ahead. Listen to the crowd noise. If you start finding yourself in a crowd surge, wait for the surge to come, go with it, and move sideways. Keep moving with it and sideways, with it and sideways.” Usually, of course, people survive. Usually not realising. They check their pockets and go home, telling their friends how – for a minute there – they were nearly scared.