Centers for Disease Control and Prevention (CDC)
This CDC-funded project modeled the effect of an earthquake in downtown Los Angeles on both the buildings and the people working there, during and immediately after the event. The damage to a typical office building was calculated and that information was used to simulate the behavior of the occupants as they evacuated the building.
A natural disaster, such as an earthquake in Southern California, can cause an immense amount of damage and overwhelm an inadequate disaster response system. By building accurate computer models of this type of event we can study its potential impact and plan ahead for when a real disaster strikes.
This CDC funded project modeled the effect of an earthquake in downtown Los Angeles on both the buildings and the people working there. We were able to calculate the damage to a typical office building and use that information to simulate the behavior of the occupants as they evacuated the building and looked for medical help. We were able to do this by thinking of the building as a complete system, with walls, staircases, furnishings, and people, and creating an interconnected set of computer models describing it all.
The first model is an accurate reconstruction of a typical steel-framed office building with concrete floors, which is shaken by the type of ground movements that can be expected during an earthquake. The shaking causes a part of the building to collapse and we are able to calculate the size and strength of the movement over every part of its frame.
These movements are fed into the second model, which represents the furniture, windows, staircases, ductwork, ceiling panels, etc. that are typically found in an office building. This model allows us to calculate the interior damage – collapsed ceilings and interior walls, blocked stairwells, broken furniture – and use this information to calculate injuries and fatalities among the occupants.
The third model uses all of the information we have gathered on the structural damage and possible injuries to simulate the behavior of the people as they make their way out of the building and into the street. To do this we use a technique called Agent Based Modeling (ABM). In this method, we can build entire virtual societies of software individuals, each with the power to make choices based on the information they get in their virtual world. These “cyber populations” can be highly diverse, varying by age, gender, medical condition, and attitudes, for example. They interact with one another in realistic spaces—cities, transportation systems, neighborhoods, what have you. Agent interactions can be viewed as animations, in living color, as though you were looking down at the action from above, and their behavior can be studied over time.
For this project the agents were given the characteristics of a typical population of office workers – working age, a variety of health conditions, etc. – some of whom have been injured or killed by the falling walls and ceilings calculated in model two. The simulation represents their most likely behavior as they rush to escape the building, taking into account the damage they encounter on the way. Varying the initial conditions of the models can allow us to study different scenarios, such as crowd pressure at doors and stairwells or altruistic behavior toward the injured.
Once outside the building, we used a less detailed model of several city blocks, complete with traffic and a range of different buildings, to follow the behavior of the evacuees as the injured made their way to medical facilities and the crowds spilled out into the roadways.
This suite of models represents every aspect of an earthquake evacuation scenario; buildings, contents, traffic, and people. We can vary any part of the system and observe its effect on everything else as the action unfolds. As such, we have a very powerful tool for disaster planning, enabling local authorities and medical services to make informed decisions in advance of the next big quake