To evaluate the feasibility of a partially connected ad hoc network established by mobile phones for disseminating
emergency information, we have conducted a data collection experiment involving one million mobile phone users of a
US telecom operator in the Boston metropolitan area, whose positions have been anonymously traced during the month of
July 2009. The large user set allows us to limit biases in the obtained results, which are found in previous datasets where
participation was restricted to students or conference attendees who volunteered to be tracked (see Section 2).
In this paper, we select as an event representative of a social gathering event the Boston Independence Day Celebration
on July 4th, 2009,2 when people usually congregate around the Charles river (the area depicted in Fig. 2) to attend the
concert and watch fireworks organized by the city administration. The choice of Independence Day is not arbitrary, since:
(i) it is a crowd gathering event, and thus it allows us to trace human mobility during a social event; (ii) although it is not an
emergency event, it is a major US holiday, and so the human mobility traces are likely not to be affected by routine behaviors.
In the following, first we present the procedure adopted for measuring real world human positions. Then, we describe
how the contact events have been inferred from the mobility patterns. Finally, we illustrate how the emergency alarm has
been spread among the people.