Since visual detection is applied, it is noted that each type can be further
sub-divided. For example, since the urban village type also includes within it denser
settlements (known as slums), the slums could also have been extracted as a
sub-type. The rural type includes areas comprising mainly farmlands and fruit
trees, as well as areas containing mainly paddy fields. Most of the regular form
type areas are those developed by real estate companies; however, some areas
include the traditional Chinese town in the city center, which has a narrow frontage
and a long land strip at the back. The high-rise type includes mid-rise residential
buildings of 4–10 stories, and most of them have been developed by the National
Housing Construction Corporation (PERUMNAS) since the 1980s. They might be
extracted as the sub-types; however, as explained in the next section, we were able to
demonstrate the validity of classifying Jakarta’s residential areas into four types
using a multinomial logit model. Therefore, we adopted only four types in this study
to show the distribution of the residential landscape in Jakarta (Fig. 4.5).
4.3.2 Verification of Classification Validity
As discussed in Sect. 4.2.2, the data that directly represent the four indicators are not
prepared in many megacities, including the Jakarta Metropolitan area. Therefore, we
should prepare data for each indicator alternatively to predict the visual detection
results by the multinominal logit model and confirm the validity of visual detection
results in terms of consistency and reproducibility. In this study, the data used for
this validation are as follows (Table 4.2):
1. Classification of regular and irregular housing within the 2003 land-use data
2. Population census data for 2000
3. Category of complex housing within the land cover data
4. Normalized soil index based on LANDSAT ETM images
5. Newly generated 5-m-mesh building height data from ALOS PRISM satellite
images.
We have limited our study area to DKI Jakarta as this was the only municipality
that published land-use data, the latter of which was required for our visual detection
model. These data are used as explanatory variables to estimate a multinomial logit
model, that is, a type of discrete choice model (Ben-Akiva and Lerman 1985), that
explains the residential type of each mesh. The usefulness of applying a logit model
to estimate land-use patterns has been discussed by many researchers, for example,
Xie et al. (2005). The equation used to represent the observed component of the
utility for each residential type in each mesh is as follows:
V urban villege,n ¼ β k1 þ β k2 Population density n þ β k3 Irregular dummy n
4 Diversity and Historical Continuity of the Residential Landscape of a. . .
53
sub-divided. For example, since the urban village type also includes within it denser
settlements (known as slums), the slums could also have been extracted as a
sub-type. The rural type includes areas comprising mainly farmlands and fruit
trees, as well as areas containing mainly paddy fields. Most of the regular form
type areas are those developed by real estate companies; however, some areas
include the traditional Chinese town in the city center, which has a narrow frontage
and a long land strip at the back. The high-rise type includes mid-rise residential
buildings of 4–10 stories, and most of them have been developed by the National
Housing Construction Corporation (PERUMNAS) since the 1980s. They might be
extracted as the sub-types; however, as explained in the next section, we were able to
demonstrate the validity of classifying Jakarta’s residential areas into four types
using a multinomial logit model. Therefore, we adopted only four types in this study
to show the distribution of the residential landscape in Jakarta (Fig. 4.5).
4.3.2 Verification of Classification Validity
As discussed in Sect. 4.2.2, the data that directly represent the four indicators are not
prepared in many megacities, including the Jakarta Metropolitan area. Therefore, we
should prepare data for each indicator alternatively to predict the visual detection
results by the multinominal logit model and confirm the validity of visual detection
results in terms of consistency and reproducibility. In this study, the data used for
this validation are as follows (Table 4.2):
1. Classification of regular and irregular housing within the 2003 land-use data
2. Population census data for 2000
3. Category of complex housing within the land cover data
4. Normalized soil index based on LANDSAT ETM images
5. Newly generated 5-m-mesh building height data from ALOS PRISM satellite
images.
We have limited our study area to DKI Jakarta as this was the only municipality
that published land-use data, the latter of which was required for our visual detection
model. These data are used as explanatory variables to estimate a multinomial logit
model, that is, a type of discrete choice model (Ben-Akiva and Lerman 1985), that
explains the residential type of each mesh. The usefulness of applying a logit model
to estimate land-use patterns has been discussed by many researchers, for example,
Xie et al. (2005). The equation used to represent the observed component of the
utility for each residential type in each mesh is as follows:
V urban villege,n ¼ β k1 þ β k2 Population density n þ β k3 Irregular dummy n
4 Diversity and Historical Continuity of the Residential Landscape of a. . .
53