Computer simulations, or models, are commonplace in the scientific world. They are used to predict hurricanes, forecast the spread of epidemics and calculate how long it would take people to leave a crowded building. Computer models are considered scientifically reliable, yet they are a rarity in the world of computer-assisted reporting.
In the course of covering the Rhode lsland fire that killed 100 people in February 2003, The Providence Journal developed a database of people who were inside the building when it caught fire. This database was used for stories that documented how many people were inside the building, and for a narrative of the six minutes from when the fire started until the last survivor escaped. We also used it for a computer model that examined how the crowd fled the and how things could have been different with a few minor changes.
Crowd evacuation simulation software is well accepted in the building design and fire safety industries. In most jurisdictions, such software can be used to demonstrate that a building complies with codes governing emergency evacuations. An owner whose building might otherwise fail to meet fire codes that prescribe, among other things, the number and size of exits, can still pass inspection if an evacuation simulation shows building occupants would reach safety in time.
The Station fire was the ideal situation for a newspaperto use a computer model. The newspaper, through interviews with more than 200 survivors, had excellent data regarding where people were when the fire broke out and how they got out of the building. Because of those interviews and because the first minute of the fire was videotaped from inside the building by a television camera operator, the newspaper had a good idea of what actually happened, providing an excellent test of the computer model against the real world.
A model would produce results, both in terms of visualizing what happened and in exploring “what ifs”, such as what if the exits had been configured differently.
The Journal researched several evacuation simulation software packages online before selecting two Simulex and STEPS for evaluation.
Though the software is normally expensive, the developers of each package, after learning what the newspaper wanted to do, offered free licenses and technical support. Both packages were similar in their features and calculations, though they had substantial differences.
STEPS produced excellent three-dimensional color animations of the evacuation, allowed the user to have somewhat greater control over the movements of building occupants and easily rendered the animation in .AVI video format.
Simulex was easier to set up, the movements of individual people in the model appeared more lifelike, and most important for the Journal the overhead, two-dimensional animation of the evacuation was better suited for conversion to a newspaper graphic, though it had to be done by hand by a graphic artist.
Screen-grab shareware Screen Movie Studio was used to generate .AVI animations that were posted on the newspaper’s Web site because Simulex could only create video in a proprietary format.
Simulex, as does STEPS, starts with a building floor plan, which is created in a Computer-Aided Design (CAD) program and imported into Simulex. The Journal had used public records requests to obtain excellent floor plans of the on paper that were recreated using CAD software by the Journal employee who oversees remodeling of the newspaper’s building.
Next we defined the exits by entering their widths and graphically placing them on the floor plan. Then we entered information about the people inside the to account for gender, body sizes, walking speeds and alcohol consumption.
We placed the people on the floor plan, along with the exit they would use and a defined delay factor. The delay factor was critical in The Station fire. Because rock band Great White started its show with fireworks and many in the crowd had been drinking, few were alarmed to see flames shooting up the stage walls. The television video showed it took more than 30 seconds for the crowd as a whole to start moving toward the exits.
After we placed the people, we ran the simulation, which records an animation of the evacuation, and provides statistics that are updated 10 times a second.
The Journal employed a few tricks in getting Simulex to mimic what happened the night of the fire. Because the program is designed to calculate how long it takes to evacuate a building, it does not account for the “crowd collapse” that led to people piling up in the front door.
Crowd collapse can occur for many reasons. In The Station fire, people at the back of the crowd were the first to feel the effects of heat and smoke, and rushed forward to get nearer the exit. Meanwhile, people at the front of the crowd slowed down as they reached the outer doors, a normal reaction once safety is reached in an evacuation.
This difference in speed led to people in front being knocked over, and, as the crowd continued to surge forward, more people tripped and fell on top of them. This pileup effectively closed the front door of the club.
But the Simulex software was not designed to model doors that close during an evacuation. The Journal tricked Simulex into doing that by programming several large extra people who tried to push their way into the door at the moment the crowd collapsed. Having people trying to go both ways through the same door effectively jammed it.
In another instance, we had to compensate for a window broken after the start of the fire. We did that by adding some other people into the model and setting their delayed reaction to the time that the window broke. These people were then placed right in the window, making that area impassable until the delay had elapsed.
In the case of both tricks, the extra people were electronically removed from the graphics published in the paper and the video published on the Web.
The model showed that, in addition to the pile of people at the front door, fire victims were trapped deeper inside the club, at the beginning of a hallway leading to the front door.
The model also showed that if the club had a rear exit if, the hallway at the front exit had not been built or if fewer people were inside the club, which was over capacity, virtually everyone would have had a chance to escape.
The stories deveIOped from the computer models were part of The Journal’s package that was recognized as a finalist for the Pulitzer Prize for public service and in the lRE Awards.
Crowd evacuation models such as Simulex and STEPS could be useful for a variety of computer-assisted reporting projects, such as evaluating existing or proposed buildings or even, with a few tweaks, examining outdoor venues.
Contact Paul Edward Parker by email at [email protected]