Database Reference
In-Depth Information
vary but the general consensus is that approximately 1.5 million (nonunique)
visitors attended the festival in 2010. Other than this rough figure and the use
of video technology by the police department to give a qualitative indication of
crowdedness or other safety issues, little is known about the general movement
patterns of these visitors within and around the festival site: how long they stay
at the festival, the number of days they visit the festival, how they reach the
event, and so on.
Given the limited range of Bluetooth scanners and the size of the event, a
full coverage of the entire study area was impossible from a practical point
of view. Instead, a careful selection of strategic coverage sites was made after
consultation with local policy makers and urban experts, with the purpose of
collecting as many significant individual movements as possible. In 2010, 22
locations were covered, including the large public squares in the city center, a
selection of points of access into the event zone, two train stations, and a tram
station located next to a park and ride facility. In 2011, we were able to capture
visitor movements in the center in a more finely grained way by employing 43
scanners exclusively in and around the center of the city.
As overcrowding is usually regarded as the main danger at mass events, we
started by using Bluetooth tracking as a counting methodology instead of a
tracking methodology as such. In order to extrapolate from counts of detected
devices to real numbers of peoplewithin the detection range of a scanner, we need
to know the fraction of visitors that are detected by our system (corresponding
to individuals carrying devices that have a discoverable Bluetooth interface). To
this end, we compared visual head counts with the number of unique Bluetooth
devices in a number of narrow passageways during a certain amount of time
(usually 15 minutes), and divided the latter by the former. This penetration rate -
also referred to as
detection ratio
- usually varies slightly from event to event,
but in 2010 it amounted to 11.0
±
1.8%. Using this figure, we could extrapolate
and roughly estimate crowdedness levels. As an example of this use as a counting
methodology, the daily and hourly variations in crowdedness of the event zone
are illustrated in Figure
14.4
.
The hourly variation is characterized by a very smooth curve with sharp
troughs in the morning (usually around 7 A.M.). The peaks are also usually sharp
and situated around 11 P.M. except for on days 2, 5, and 9, where a broader
peak in the late afternoon is observed. These correspond to two Sundays and
the national day of Belgium (July 21st), and these days are known to attract
more daytime visitors (such as working couples with children). As a result, the
sharp peaks around midnight do not appear because of the relatively greater
crowdedness earlier in the afternoon. The three busiest days are immediately
visible, with the fourth day being the most crowded with almost 10,000 detected
phones or around 90,000 unique visitors in the festivities zone between 11 P.M.
and 12 A.M. To aggregate over daily periods, we had to carefully consider how
