networks—have impacted our understanding of the scale and resolution issues.
Because little has been done in the past, more researchs is also needed to better
understand temporal resolution, change and evolution of urban features over time, and
temporal requirements for urban mapping (Weng, 2012). There is not a simple answer
to any of the questions discussed above. In this review, I start with the requirements
for mapping three interrelated entities or substances in the urban space (i.e., material,
land cover, and land use) and their relationships. Spectral resolution is a common
consideration in imaging and mapping and is closely associated with the categorical
scale. Then, the relationship between spatial resolution—which is termed the
observational scale of remote sensing in this chapter—and the fabric of urban
landscape is examined. Central to this relationship is the problem of mixed pixels
in the urban areas. The pixel and subpixel approaches to urban analyses are thus
discussed. Next, the author’s two previous studies are discussed, both assessing the
patterns of land surface temperature at different aggregation levels in order to find out
the operational scale/optimal scale for the studies. Section 4.5 is developed to review
the issue of scale dependency of urban phenomena and to discuss two case studies,
one on LST variability across multiple census levels (block, block group, and tract)
and the other on multiscale residential population estimation modeling. Section 4.6
provides a summary of the discussions and reflects on future developments.
4.2 URBAN LAND MAPPING AND CATEGORICAL SCALE
Urban remote sensing should consider the requirements for mapping three interrelated entities or substances on Earth’s surface (i.e., material, land cover, and land
use) and their relationships (Weng and Lu, 2009; Weng, 2012). Urban areas are
composed of a variety of materials, including different types of artificial materials
(i.e., impervious surfaces), soils, rocks and minerals, and green and nonphotosynthetic vegetation. These materials comprise land cover and are used in different
manners for various purposes by human beings. Land cover can be defined as the
biophysical state of Earth’s surface and immediate subsurface, including biota, soil,
topography, surface water and groundwater, and human structures (Turner et al.,
1995). Land use can be defined as the human use of the land and involves both the
manner in which the biophysical attributes of the land are manipulated and the
purpose for which the land is used (Turner et al., 1995). Remote sensing technology
has been applied to map urban land use, land cover, and materials. Their relationships are illustrated in Figure 4.1. Each type of land cover may possess unique
surface properties (material). However, mapping land covers and materials have
different requirements. Land cover mapping needs to consider characteristics in
addition to those coming from the material (Herold et al., 2006). The surface
structure (roughness) may influence the spectral response as much as the intraclass
variability (Gong and Howarth, 1990; Myint, 2001; Shaban and Dikshit, 2001;
Herold et al., 2006). Two different land covers, for example, asphalt roads and
composite shingle/tar roofs, may have very similar materials (hydrocarbons) and
thus are difficult to discern, although from a material perspective these surfaces can
62
ON THE ISSUE OF SCALE IN URBAN REMOTE SENSING
Because little has been done in the past, more researchs is also needed to better
understand temporal resolution, change and evolution of urban features over time, and
temporal requirements for urban mapping (Weng, 2012). There is not a simple answer
to any of the questions discussed above. In this review, I start with the requirements
for mapping three interrelated entities or substances in the urban space (i.e., material,
land cover, and land use) and their relationships. Spectral resolution is a common
consideration in imaging and mapping and is closely associated with the categorical
scale. Then, the relationship between spatial resolution—which is termed the
observational scale of remote sensing in this chapter—and the fabric of urban
landscape is examined. Central to this relationship is the problem of mixed pixels
in the urban areas. The pixel and subpixel approaches to urban analyses are thus
discussed. Next, the author’s two previous studies are discussed, both assessing the
patterns of land surface temperature at different aggregation levels in order to find out
the operational scale/optimal scale for the studies. Section 4.5 is developed to review
the issue of scale dependency of urban phenomena and to discuss two case studies,
one on LST variability across multiple census levels (block, block group, and tract)
and the other on multiscale residential population estimation modeling. Section 4.6
provides a summary of the discussions and reflects on future developments.
4.2 URBAN LAND MAPPING AND CATEGORICAL SCALE
Urban remote sensing should consider the requirements for mapping three interrelated entities or substances on Earth’s surface (i.e., material, land cover, and land
use) and their relationships (Weng and Lu, 2009; Weng, 2012). Urban areas are
composed of a variety of materials, including different types of artificial materials
(i.e., impervious surfaces), soils, rocks and minerals, and green and nonphotosynthetic vegetation. These materials comprise land cover and are used in different
manners for various purposes by human beings. Land cover can be defined as the
biophysical state of Earth’s surface and immediate subsurface, including biota, soil,
topography, surface water and groundwater, and human structures (Turner et al.,
1995). Land use can be defined as the human use of the land and involves both the
manner in which the biophysical attributes of the land are manipulated and the
purpose for which the land is used (Turner et al., 1995). Remote sensing technology
has been applied to map urban land use, land cover, and materials. Their relationships are illustrated in Figure 4.1. Each type of land cover may possess unique
surface properties (material). However, mapping land covers and materials have
different requirements. Land cover mapping needs to consider characteristics in
addition to those coming from the material (Herold et al., 2006). The surface
structure (roughness) may influence the spectral response as much as the intraclass
variability (Gong and Howarth, 1990; Myint, 2001; Shaban and Dikshit, 2001;
Herold et al., 2006). Two different land covers, for example, asphalt roads and
composite shingle/tar roofs, may have very similar materials (hydrocarbons) and
thus are difficult to discern, although from a material perspective these surfaces can
62
ON THE ISSUE OF SCALE IN URBAN REMOTE SENSING