CA-BASED SIMULATION OF ASIAN URBAN DYNAMICS: A CASE STUDY OF
TAIPEI METROPOLITAN AREA, TAIWAN
Bing Sheng Wu, Daniel Z. Sui
Department of Geography, Texas A&M University
1. INTRODUCTION
Urban regions are considered as non-linear, complex, self-organizing, and dynamic systems. Cellular automata
(CA), serving as a powerful metaphor for the complex interaction of urban dynamics [1], is viewed as useful as a
tool for modeling urban spatial dynamics [2]. Derived from conventional western urban theories, most CA-based
urban models pay much attention to the relationship between physical driving factors, for instance, accessibility to
city center or road network, and urban expansion. Nonetheless, under the wave of globalization taking place in
late 20th Century, rapid urbanization in Asia has exhibited a process distinctively different from that of the West
[3]. Factors contributing to time-space compression of Asian urbanization include not only traditional physical
drivers but also global economic features (e.g. capital flows from developed countries to developing countries).
However, current CA-based urban simulation does not systematically discuss the importance of globalization in
Asian urban dynamics. In this paper we focus on foreign direct investment and shift of economic activities as
essential indicators to represent influences of globalization on Asian urbanization, and combine these features
with GIS and remote sensing data to develop a CA-based Asian urban model.
2. STUDY AREA AND DATA ACQUISITION
Taipei metropolitan area, a rapid urbanizing region in Asia, is selected to examine this model in terms of export-
oriented policies and deep interaction with global economy. We use two SPOT satellite images taken in 1993 and
2000 to extract constructed/non-constructed pixels. The two classes are used to examine expansion of urbanized
areas. Other GIS data including boundary of Taipei metropolitan area, road network, and Taiwan DEM data are
also acquired for estimate traditional physical constraints on Asian urbanization. Socioeconomic data applied in
this model contain: foreign direct investment data, population data, and industrial activity data from
district/county/national level. The temporal scale of the socio-economic data ranges from 1985 to 2004.
3. METHODOLOGY
In our CA model, we develop transition rules in terms of GIS data, classification results from remote sensing
images, and socioeconomic data. First we discuss how population growth is affected under globalization in county
and district levels. We apply stepwise linear regression model to develop relationships between population,
industrial activity data and foreign direct investment. Second, we estimate the probability of conversion from a
non-constructed pixel to constructed pixel. We apply the following physical constraints into a multinomial logistic
model and develop a probability rule for every cell: slope, accessibility to roads, and accessibility to urban centers.
After extracting non-constructed and constructed pixels from the SPOT images, we define the relationship
between population growth in district level and increase of constructed pixels. Then we simulate the change of
constructed pixels according to population growth, and located the constructed pixels based up Monte Carlo
mechanism.