14% for the high-rise type. The rural type has a slightly lower hit rate compared to
other types because the population density data, which is a proxy variable for
building density, has a lower resolution than a 250 m  250 m mesh. Further, the
high-rise type has a considerably low hit rate due to the insufficient accuracy of the
building heights created from the ALOS images. Since it is relatively easy to visually
distinguish between high- and low-rise structures from satellite images, there is not
much error in visual detection for the high-rise type. On the other hand, Fig. 4.6
compares the spatial distributions of the model prediction results and visual detection results. Except for the number of high-rise types and the areas indicated by the
circle, there was no significant difference in the overall distribution trend. Hence, it is
Table 4.2 Data used to verify the validity of residential classification
Variable used
Corresponding
indicator
Original data
Scope of
data
Resolution
Data
type
Classification of
regular housing and
irregular housing
Block shape
Land-use data
in 2003
DKI Jakarta –
GIS
(Vector)
Population density
Building
density
Population
census in 2000
Jabodetabek Each
Census
Block
(2000)
GIS
Classification of
complex housing
Land use
Land cover
data by
geospatial
authority
Jabodetabek –
GIS
(Vector)
Normalized soil
index
Land use
LANDSAT
ETM
Jabodetabek 30-m
mesh
Image
(TIFF)
Average building
height
Building height ALOS PRISM DKI Jakarta 5-m mesh Image
(TIFF)
a Geographic information system
b
Tagged image file format
Table 4.3 Model prediction and visual detection results for the four types of residential areas in
Jakarta
Model prediction (number of meshes)
Visual
detection
Hit rate
(%)
Urban village
type
Rural
type
Regular form
type
High-rise
type
Total
Urban village
type
76
1755
247
294
12
2308
Rural type
65
393
879
83
2
1357
Regular form
type
78
302
312
2260
20
2894
High-rise type
14
57
55
55
28
195
Total
73
2507
1493
2692
62
6754
4 Diversity and Historical Continuity of the Residential Landscape of a. . .
55
other types because the population density data, which is a proxy variable for
building density, has a lower resolution than a 250 m  250 m mesh. Further, the
high-rise type has a considerably low hit rate due to the insufficient accuracy of the
building heights created from the ALOS images. Since it is relatively easy to visually
distinguish between high- and low-rise structures from satellite images, there is not
much error in visual detection for the high-rise type. On the other hand, Fig. 4.6
compares the spatial distributions of the model prediction results and visual detection results. Except for the number of high-rise types and the areas indicated by the
circle, there was no significant difference in the overall distribution trend. Hence, it is
Table 4.2 Data used to verify the validity of residential classification
Variable used
Corresponding
indicator
Original data
Scope of
data
Resolution
Data
type
Classification of
regular housing and
irregular housing
Block shape
Land-use data
in 2003
DKI Jakarta –
GIS
(Vector)
Population density
Building
density
Population
census in 2000
Jabodetabek Each
Census
Block
(2000)
GIS
Classification of
complex housing
Land use
Land cover
data by
geospatial
authority
Jabodetabek –
GIS
(Vector)
Normalized soil
index
Land use
LANDSAT
ETM
Jabodetabek 30-m
mesh
Image
(TIFF)
Average building
height
Building height ALOS PRISM DKI Jakarta 5-m mesh Image
(TIFF)
a Geographic information system
b
Tagged image file format
Table 4.3 Model prediction and visual detection results for the four types of residential areas in
Jakarta
Model prediction (number of meshes)
Visual
detection
Hit rate
(%)
Urban village
type
Rural
type
Regular form
type
High-rise
type
Total
Urban village
type
76
1755
247
294
12
2308
Rural type
65
393
879
83
2
1357
Regular form
type
78
302
312
2260
20
2894
High-rise type
14
57
55
55
28
195
Total
73
2507
1493
2692
62
6754
4 Diversity and Historical Continuity of the Residential Landscape of a. . .
55