Database Reference
In-Depth Information
(a) (b)
Figure 10.2 Presence distribution between 11 A.M. and noon, (a) survey, (b) GPS data;
frequent locations plotted with lighter shades. (See color plate.)
interval. Such values were averaged over all (regular) working days available.
Figure
10.2
b shows the results.
The two distributions match well in most locations, including some particular
areas along main streets and suburban residential areas, confirming again the
coherence of results obtained with survey and mobility data. The main deviation
occurs in the inner city center, where a high density spot found by surveys
is significantly lower in Milano2007: this is explained by the strong access
restrictions on private cars in the city center, as well as by the limited capacity
of roads and traffic, which causes an underrepresentation in the GPS data of the
people who reach their workplaces in the center with public transportation.
These two first analyses are useful to have a first insight of the data. The
next aspect to analyze is the exploration of the movement dynamics, that is,
identifying the movement quantities represented in the trajectory data sets: the
length of a trip, and the duration of a trip, the correlation of length and speed of
trips.
Trip Length and Duration
Figure
10.3
a shows the distribution of trip length (in km), as estimated from
GPS trajectories. The heavy-tailed distribution of trip length highlights how
there are many short trips of a few kilometers, and few, but nonnegligible, very
long trips of tens or even hundreds of kilometers; a similar consideration applies
to the distribution of trip duration, shown in Figure
10.3
b. The lesson learned
here confirms how mobility is a complex phenomenon, where a simple notion
of
average behavior
may be misleading. In fact the variance of the distribution
is so large that the representativeness of the average value is limited.
