Stampedes occur often, and not just in wild animal herds. Previously, physicists developed numerous models of crowd evacuation dynamics based on disasters such as the yearly Muslim Hajj or of the Love Parade disaster in Germany in 2010.

If crowd dynamics models were used for evacuation routes in those casualties, they weren't very good. A new study in EPJ B outlines a procedure for quantitatively comparing different crowd models, which also helps to compare these models with real-world data.

The trouble with models is that they use inputs, real-world data, of crowds that is limited and incomplete. Often only top-down, macroscopic scale measurements of data such as flow, density, and average speed are used. It is not clear whether these metrics can validate the modeled dynamics at the individual level.

Instead of using data that is so difficult to source, Vaisagh Viswanathan, a PhD student from Nanyang Technological University in Singapore, and colleagues adopted a quantitative study comparing the simulated congestion flow rates, among other things, of three so-called bottom-up models. These focus on the individual behavior of school children evacuating their classroom during the May 2008 Sichuan Earthquake.

They found that a model referred to as the social force model — based on the idea that pedestrians move in response to fictitious attractive or repulsive social forces — best matches the real-world data showing how pupils exit their classrooms.

They also identified a new macroscopic metric, 'the zoned evacuation time', as the one observable parameter that can best discriminate between these models, and also between models and real-world data.