International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B2, 2012 
XXII ISPRS Congress, 25 August - 01 September 2012, Melbourne, Australia 
101 
DIGITAL ARCHIVING OF PEOPLE FLOW BY RECYCLING LARGE-SCALE SOCIAL 
SURVEY DATA OF DEVELOPING CITIES 
Yoshihide Sekimoto 3 ’ *, Atsuto Watanabe 3 , Toshikazu Nakamura 3 , Teerayut Horanont b 
a Center for Spatial Information Science, the University of Tokyo, Japan 
- (sekimoto, atsuto, ki_ki_gu)@csis.u-tokyo.ac.jp 
Institute of Industrial Science, the University of Tokyo, Japan - teerayut@iis.u-tokyo.ac.jp 
KEY WORDS: Reconstruction ot people flow, Social survey data, Large-scale spatio-temporal analysis 
ABSTRACT: 
Data on people flow has become increasingly important in the field of business, including the areas of marketing and public services. 
Although mobile phones enable a person's position to be located to a certain degree, it is a challenge to acquire sufficient data from 
people with mobile phones. In order to grasp people flow in its entirety, it is important to establish a practical method of 
reconstructing people flow from various kinds of existing fragmentary spatio-temporal data such as social survey data. For example, 
despite typical Person Trip Survey Data collected by the public sector showing the fragmentary spatio-temporal positions accessed, 
the data are attractive given the sufficiently large sample size to estimate the entire flow of people. In this study, we apply our 
proposed basic method to Japan International Cooperation Agency (JICA) PT data pertaining to developing cities around the world, 
and we propose some correction methods to resolve the difficulties in applying it to many cities and stably to infrastructure data. 
1. INTRODUCTION 
Recently, the monitoring of dynamic changes in people flow 
has become necessary in order to mitigate secondary disasters 
following earthquakes, fires, or other major events, as well as to 
relieve congestion at nodes in terminal stations. For example, 
more than 12,000 people were killed while trying to escape 
from the tsunami of the Great East Japan Disaster on March 11, 
2011. Moreover, about 4 million people daily ride trains from 
the Shinjuku Station in Tokyo, the most crowded station in the 
world. For public facility managers to design safe and 
comfortable spaces as well as appropriate urban transport 
policies, it is necessary that they grasp comprehensively the 
people flow. In the commercial fields of outdoor advertising, 
price systems, which support effective advertising, depend on 
each location's traffic volume. 
Previous work has been done on population distribution data. 
For example, the National Center for Geographic Information 
Analysis (NCGIA) produced the Gridded Population of the 
World (GPW) in 1995, as the first raster global dataset ot 
population [1]. A second version (GPW2) was produced in 
2000 by the Center for International Earth Science Information 
Network (CIESIN) at Columbia University and used a higher 
resolution 1. Data are gridded on the basis of original census 
units, at the highest spatial resolution for which they are 
available by country; two methods are used, the latter of which 
involves a smoothing method that assumes grids closer to units 
with high density are higher. Similar information is provided by 
LandScan, developed by the Oak Ridge National Laboratory 
(ORNL). LandScan models are tailored to match the data 
conditions and geographical nature of each individual country 
and region, and at approximately 1km resolution (30" x 30") they 
provide the finest resolution of global population distribution 
currently available [2]. Whereas GPW provides nighttime 
population, LandScan represents ambient population, meaning 
the average over 24 hours. 
However, according to recent developments in sensing 
technology, the following examples are ways to measure people 
flow more dynamically from various dimensions. Especially, 
mobile phone sensing can be widely applied for taking the 
above measurement, and some research exists [3-5]. However, 
such research cannot be regarded as providing infrastructure 
data that can give an overview of the mass flow by integrating 
various acquired data mentioned in the previous paragraph. This 
is true in terms of comprehensive qualities including 
spatial/tcmporal accuracy, acquisition/process costs, and value 
to the user as a service. For example, our group had pointed out 
the necessity of three aspects of people flow data sets including 
"sufficiently large scale," "temporal completeness," and 
"realistic spatial accuracy" [6]. Moreover, mobile phone data 
for everyone is not usually available, compared with social 
survey data acquired for public benefits such as transportation 
or disaster prevention planning. 
Hence, our group had proposed a data process for the 
reconstruction of the spatio-temporal positions of large numbers 
of people using existing person trip survey (hereinafter referred 
to as PTS) data, conducted by various transportation planning 
agencies or commissions consisting of several Japanese cities 
[6-7]. Some results are summarized on the "People Flow 
Project" web site. However, some interpolations and corrections 
will be necessary when applying our proposed core method [6] 
to various social survey data globally, because each city's zones 
and network data vary in size. 
In this study, we apply the proposed process to Japan 
International Cooperation Agency (JICA) PTS data pertaining 
to developing cities around the world, and we resolve the 
difficulties in preparing infrastructure data such as road network 
data. Section 2 gives an outline of our proposed method and 
JICA-PTS data, and a problem statement of our existing method. 
* Corresponding author.