urban planning. Unlike in Japan, adequate data are often not publicly available in
developing countries. Over the years, the development and public availability of
geospatial information have progressed in Japan, the United States, and other
developed countries, due to which it is possible to obtain various physical and social
environmental information and is easy to quantitatively categorize residential areas
in these countries. However, in developing countries, relevant data are not yet
available or accessible to the public. Further, the data that are globally available
remain limited. These conditions make the statistical classification difficult in developing countries. Therefore, it is important to develop a universal and practical
classification method which enables us to extract various residential types within
the current data availability constraints.
Accordingly, in this chapter, we propose a method to classify the city’s residential
areas through visual detection based on four visually readable indicators from
satellite images. We are able to obtain reliable classification results by applying
the multinominal logit model, that is, a type of discrete choice model (Ben-Akiva
and Lerman 1985), to determine the validity of visual detection results. We apply the
method to the Jakarta metropolitan area and classify its residential areas into four
types. Then, we construct the model to predict the visual detection results by using
available data as explanatory variables and verify the consistency in the classification criteria of visual detection based on the hit rate of model prediction results to
visual detection results. Through this process, we argue that this method effectively
provides reliable classification results and enables us to understand the diversity of
residential landscape in megacities.
Based on the classification result of Jakarta, we clarify the characteristics of
Jakarta’s residential landscape, especially focusing on the distributions of kampung,
an informal settlement in Jakarta (Jellinek 1991; Funo 1991; Sawa 1999; Colombijn
2011). Furthermore, by superimposing the classification results with an old map that
was published in the 1930s, we show that most of the present kampung have
historical continuity with indigenous residential areas in the early twentieth century
(Dutch colonial period). We discuss that the historical kampung can be considered to
have stronger social solidarity and maintain a lifestyle with a smaller environmental
impact than planned residential areas and hence is essential for the urban sustainability in Jakarta. Finally, we suggest applying the classification method of the
residential landscape to other megacities to reveal the essential residential areas
that could help achieve global sustainability.
4.2 Classification of the Residential Landscape
4.2.1 Method
A city’s residential landscape is a collection of various residential areas consisting of
individual dwellings. It can be divided into three scales: building, neighborhood, and
city scales. In this chapter, the neighborhood scale, which comprises multiple
4 Diversity and Historical Continuity of the Residential Landscape of a. . .
47
developing countries. Over the years, the development and public availability of
geospatial information have progressed in Japan, the United States, and other
developed countries, due to which it is possible to obtain various physical and social
environmental information and is easy to quantitatively categorize residential areas
in these countries. However, in developing countries, relevant data are not yet
available or accessible to the public. Further, the data that are globally available
remain limited. These conditions make the statistical classification difficult in developing countries. Therefore, it is important to develop a universal and practical
classification method which enables us to extract various residential types within
the current data availability constraints.
Accordingly, in this chapter, we propose a method to classify the city’s residential
areas through visual detection based on four visually readable indicators from
satellite images. We are able to obtain reliable classification results by applying
the multinominal logit model, that is, a type of discrete choice model (Ben-Akiva
and Lerman 1985), to determine the validity of visual detection results. We apply the
method to the Jakarta metropolitan area and classify its residential areas into four
types. Then, we construct the model to predict the visual detection results by using
available data as explanatory variables and verify the consistency in the classification criteria of visual detection based on the hit rate of model prediction results to
visual detection results. Through this process, we argue that this method effectively
provides reliable classification results and enables us to understand the diversity of
residential landscape in megacities.
Based on the classification result of Jakarta, we clarify the characteristics of
Jakarta’s residential landscape, especially focusing on the distributions of kampung,
an informal settlement in Jakarta (Jellinek 1991; Funo 1991; Sawa 1999; Colombijn
2011). Furthermore, by superimposing the classification results with an old map that
was published in the 1930s, we show that most of the present kampung have
historical continuity with indigenous residential areas in the early twentieth century
(Dutch colonial period). We discuss that the historical kampung can be considered to
have stronger social solidarity and maintain a lifestyle with a smaller environmental
impact than planned residential areas and hence is essential for the urban sustainability in Jakarta. Finally, we suggest applying the classification method of the
residential landscape to other megacities to reveal the essential residential areas
that could help achieve global sustainability.
4.2 Classification of the Residential Landscape
4.2.1 Method
A city’s residential landscape is a collection of various residential areas consisting of
individual dwellings. It can be divided into three scales: building, neighborhood, and
city scales. In this chapter, the neighborhood scale, which comprises multiple
4 Diversity and Historical Continuity of the Residential Landscape of a. . .
47