Towards a Comparative Science of Cities: Using Mobile Traffic Records in New York, London, and Hong Kong - Computational Approaches for Urban Environments - page 360

Greater London − Calls

a

3

2

1

b

0

3

2

1

0

−1

Apr−01 Apr−08 Apr−15 Apr−22 Apr−29 May−06 May−13 May−20 May−27 Jun−03 Jun−10 Jun−17 Jun−24 Jul−01

Fig. 15.3 Decomposition of Greater London “calls” timeline. ( a ) The mean normalized
timeline can be decomposed into ( b ) a repeating weekly pattern (the city calls signature S Calls

London ,in

red ) and a residual part (in blue )

trends. For example, we observe a slight overall increase in calls in late spring /early
summer which may be due to the arrival of tourists in London at this time of the year.
Overall, the typical week signature captures the main temporal patterns within
the cities, by reducing noise and getting rid of long-term trends. Incidentally, it can
provide a good predictive baseline of expected mobile device usages. In the example
presented in Fig. 15.3 , the average absolute ratio between residuals and timelines is
approximatively equal to 8 %, but is typically included between 5 and 10 % for the
different activity type either in whole cities or at each pixel level. Based on this
timeline, an operator could detect irregular operations, as also anticipate upcoming
special events.

15.4.2 Comparing Cities' Signatures

Let us first explore the temporal patterns at the macro, citywide scale. The
city signatures of the different activity types are displayed on Fig. 15.4 which
emphasizes their similarities and differences.
All three studied cities display a broadly comparable rhythm, common to all
components of activity. Mobile activity rapidly ramps up in the morning between 6
and 10 AM, followed by rather steady activity levels within the day and a slower
decrease of activity at night between 9 PM and 2 AM. The same pattern appears
on workdays and with somewhat slower amplitude on weekends. On workdays, we
can also observe small peaks of activity at commuting hours and at lunch times.
It thus seems reasonable to associate this common rhythm to a simple daily cycle,
corresponding to people waking up, going to work, having lunch, and then heading
back home in the evening.
Let us now turn to the differences appearing on Fig. 15.4 when comparing the
cities.