Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 164
8
Extraction and Modeling of Urban Attributes
Using Remote Sensing Technology
David ]. Cowen and John R. Jensen
Since the earliest developments in urban sociology and geography (Harris
and Ullman, 1945), researchers have recognized the essential spatial element of
urban development. Remote sensing provides an opportunity to measure at-
tributes of urban and suburban environments and record the data in accurate
digital maps and files suitable for analysis with geographic information systems
(GIS). These data, together with data available from ground-based observations,
can be used to monitor changes in space and time, to develop and validate dy-
namic models of urban development, and to forecast future land-use patterns and
changes in other urban attributes. Remotely sensed data are thus potentially
valuable both to social scientists and to urban planners and other public officials.
This chapter first identifies key attributes of urban and suburban environ-
ments and evaluates the capability of remote sensing technology to measure these
attributes accurately at the requisite levels of temporal, spatial, and spectral reso-
lution. It then presents a detailed case example that illustrates how measurements
of several of these attributes can be combined to address a social science prob-
lem: the development of an empirically based theory of urban residential expan-
s~on.
REMOTE SENSING OF URBAN/SUBURBAN ATTRIBUTES
Humans create complex urban landscapes that are composed of various ma-
terials (concrete, asphalt, plastic, shingles, water, grass, soil, and shrubbery) ar-
ranged in specific ways to build transportation systems, utility lines, housing,
commercial buildings, and public space in order to improve the quality of life.
164
OCR for page 165
DAVID J. COWEN AND JOHN R. JENSEN
165
Characteristics of many of these phenomena can be remotely sensed from subor-
bital aircraft or from satellites. The information thus derived may be both quali-
tative and quantitative.
Ten of the major urban/suburban attributes of significant value for under-
standing the urban environment are summarized in Table 8- 1. To remotely sense
these urban phenomena, it is necessary to understand the temporal and spatial
resolution required for each. Temporal resolution refers to how often managers
need the information; for example, local planning agencies may need precise
population estimates every 5 to 7 years to supplement estimates provided by the
decennial census. As an example of required spatial resolution, local population
estimates based on building unit counts usually must have a minimum mapping
unit of 0.3-5 m (0.98-16.4 ft). The information presented in Table 8-1 was
synthesized both from theliterature (e.g., Branch, 1971;Ford, 1979; Jensen,
1983a, b; Haack et al., 1997; Philipson, 1997) and from practical experience.
Ideally, there would always be a remote sensing system that could obtain images
of the terrain that would satisfy the temporal and spatial resolution requirements
specified in Table 8-1. Unfortunately, this is not always the case, as will be
demonstrated later in this chapter.
Information about urban attributes is also best collected using the specific
portions of the electromagnetic spectrum shown in Table 8-1. For example, land
cover (U.S. Geological Survey [USGS] Level III) is best acquired using the
visible (V: 0.4-0.7 micrometers Semi), near-infrared (NIR: 0.7-1.1 lam), and
mid-infrared (MIR: 1.5-2.5 Am) portions of the spectrum. Building perimeter,
area, volume, and height information is best acquired using black-and-white pan-
chromatic (0.5-0.7 Am) or color imagery. The thermal infrared portion of the
spectrum (TIR: 3-12,um) may be used to obtain urban temperature measurements.
The relationship between temporal and spatial data requirements for urban/
suburban attributes and the temporal and spatial characteristics of available and
proposed remote sensing systems is shown in Figure 8-1. Note that the codes
shown on this figure are defined in Table 8-1, while abbreviations used for the
various remote sensing systems are defined in the glossary in Appendix B. and in
Morain and Budge (1996~.
Land Use/Land Cover
Urban land-use/land-cover information is required for residential-industrial-
commercial site selection, population estimation, and development of zoning
regulations (Green et al., 1994~. For this reason, the USGS developed a land-use
and land-cover classification system for use with remotely sensed data (Anderson
et al., 1976~. Broad Level I classes may be inventoried using the Landsat Multi-
spectral Scanner (MSS), with a spatial resolution of 79 x 79 m; the Thematic
Mapper (TM), with a resolution of 30 x 30 m; the Systeme pour ['Observation de
la Terre (SPOT) High Resolution Visible (HRV) (XS), with a resolution of 20 x
OCR for page 166
166
MODELING OFA'ITRIBUTES USING REMOTE SENSING TECHNOLOGY
TABLE 8-1 Relationship Between Selected Urban/Suburban Attnbutes and
the Remote Sensing Resolutions Required to Provide Such Information
Minimum Resolution Requirements
Spatial Spectral
Temporal
Land Use/Land Cover
L1 - USGS Level I 5 -10 years20 - 100 m V-NIR-MIR-Radar
L2 - USGS Level II 5 -10 years5 - 20 m V-NIR-MIR-Radar
L3 - USGS Level III 3 - 5 years1 - 5 m V-NIR-MIR-Pan
L4 - USGS Level IV 1 - 3 years0.3 - 1 m Pan
Building and Property Line Infrastructure
B 1 - building perimeter, area, volume, height 1 - 2 years0.3 - 0.5 m Pan
B2 - cadastral mapping (property lines) 1 - 6 months0.3 - 0.5 m Pan
Transportation Infrastructure
T1 - general road centerline 1 - 5 years1 - 30 m Pan
T2 - precise road width 1 - 2 years0.3 - 0.5 m Pan
T3 - traffic count studies (cars, airplanes etc.) 5 - 10 min0.3 - 0.5 m Pan
T4 - parking studies 10 - 60 min0.3 - 0.5 m Pan
Utility Infrastructure
U1 - general utility line mapping and routing 1 - 5 years 1 - 30 m Pan
U2 - precise utility line width, right-of-way 1 - 2 years 0.3 - 0.6 m Pan
US - location of poles, manholes, substations 1 - 2 years 0.3 - 0.6 m Pan
Digital Elevation Model (DEM) Creation
D1 - large scale DEM 5 - 10 years 0.3 - 0.5 m Pan
D2 - large scale slope map 5 - 10 years 0.3 - 0.5 m Pan
Socioeconomic Characteristics
Sl-localpopulation estimation 5- 7years 0.3- Sm Pan
S2- regional/nationalpopulationestimation 5 - 15 years 5 - 20m V-NIR
S3 - qualityoflifeindicators 5 - lOyears 0.3 - 30m Pan-NIP
Energy Demand and Conservation
E1 - energy demand and production potential 1 - 5 years 0.3 - 1 m Pan-NIP
E2 - building insulation surveys 1 - 5 years 1 - 5 m TIR
Meterological Data
Ml-daily weather prediction 30min-12hr 1-8 km V-NIR-TIR
M2 - currenttemperature 30 min - 1 hr 1 - 8 km TIR
M3 - current precipitation 10 - 30 min 4 km Doppler Radar
M4 - immediate severe storm warning 5 - lOmin 4km Doppler Radar
MS - monitoring urban heat island effect 12 - 24 hr 5 - 10 m TIR
Critical Environmental Area Assessment
C1 - stable sensitive environments
C2 - dynamic sensitive environments
1 - 2 years1 - lOm V-NIR-MIR
1 - 6 months0.3 - 2 m V-NIR-MIR-TIR
Disaster Emergency Response
DE1 - pre-emergency imagery 1 - 5 years1 - 5 m V-NIR
DE2 - post-emergency imagery 12 hr - 2 days0.3 - 2 m Pan-NIR-Radar
DE3 - damaged housing stock 1 - 2 days0.3 - 1 m Pan-NIP
DE4 - damaged transportation 1 - 2 days0.3 - 1 m Pan-NIP
DES - damaged utilities 1 - 2 days0.3 - 1 m Pan-NIP
OCR for page 167
DAVID J. COWEN AND JOHN R. JENSEN
107_
8
5~
3
2
L
6
8
5_ .
3_
2
105_
8- .
5- .
3
2
.= 104
8
o
ca c
~J
3
ct
5- 2
O
~ 103
E ~8
s
3
2
8
5
2
10
8
5
2- .
167
0.3 0.5 1 m 5 10 20 30 100
.. .. .. . .. . .. .. . .
..................................... , , , ~
i,.,.,.,.,.,~,.,.,.,.,., .,.,.,.y~ 52 ~y,~
~',
U1 1 DE1, E2l. T1 }.
J C1 ~/
.. .. . .. .. .. . . ..
DE3 ~~`
~DE4
_ lday
. ,t,t,~ ~:~ ~,~,^
J 11~A ~
.-12hr . .
100 min
. _ 3mn. . ..
.. .. .. .. .. .. . .. .. .. . . .
1 m 2 3 5 10 15 20 30 lOOm
~_~
.-.-.-......
..................
. . ..
Pan, 1 x 1 .
hASS4x4. . ..
I RS-Pi (1 999) " "
PĒn 2.,'5 x 2.5 ,. .. ..
ORBIMAGE
Orbilieu'3 (1999)
.Pan,.1 x 1 . .. .. ..
..... ,,,,,,:~S,.,S.,.40.,x,,,4,, , ,,,,,,,,,, ,,,,,,,,,, ,,,,:,:,
,. . .. .. .
...... ............. ...... ...... ..... .............. ....... ....... ....... ...... ..
|(erial Photog.rapl~y .
~,, 0.3,x 0.3 rn (0.98,,x 0.98 ft')
1 g 1 m (3.281 x 3.28'i ft.
........... , ........... .......... ........... ...... ...... ..... , .... ..... ..... ..
,. .. . .. ..
,. .. . .. ..
.,: ........... ........... ........... ........... ~, ........... , ........... ........... , , , : ........... ...........
. ,.
.
SPOT HRV"1,2,3 and 4 (1998)
lPain 10 x10
__MOS RGx(220Oo2) IRS-1 AB
,,,., ,,, , ,, --~- - --' MSS 10 x=' MIR 20 x 20 LISS-2 36.25 x 36.25
-- - -----: ----- ----- ------ ~-.-- ~ ' IRS-1 CD
.-.-.-.,.-.-.-.-.-.,.-.-.-.-.- -.-., -.-.-.-., _ Pan5.8 x5.8
LISS-3 23 x 23 MIR 70 x 70
LANDSAT 7 (1998)
............. Pan ~5 x 1,h,
, ............ ........... ........... ........... ........... ....
......................................................................
MODIS*
Land 0.25 x 0.25 km
'''''Lancl'0.'5O'x O:50'krn''
Ocean 1 x 1 km
Atmo 1 x 1 km
TIR1 x1 km
,.
,.
,.
GOES
. VIS0.9X 0.9 km
., TlR8.0X8.0km
. ,.
..... ........... ........... ........... ........... ............ ........... ........... ...........
. ,.
, .......... .......... ........... ........... ........... ........... ........... ........... .
,.
,.
..: .-.-.-.-.. ..-.-.-.-.
. ,.
NOAA AVHRR
LAC1.1 x1.1 km
,. GAC4x4kn~.
................................. . ~
M ETEOSAT
VISIR 2.5 x 225 km
TIR 5 x 5 ~m
.,., ...................... ~, ............................................ , : ........... : ...........
Ground
.. Doppler
Radar
.,.,.,., ~, , : ,
4x4km l
..... ..... ...... ...... ...... ...... ...... ...... ...... ..... ..... .
1 km . ..
lOOOm Skm lOkm
. M3
.M4
0.2 0.3 015 .8 1.0 2 3 5 1 110 2 3 5 1112 2 3 5 1113 2 3 1 5 8 1o4
Spatial Resolution in meters
FIGURE 8-1 Spatial and temporal resolution requirements for urban/suburban attributes
overlaid on the spatial and temporal capabilities of current and proposed remote sensing
systems.
OCR for page 168
168
MODELING OFA'ITRIBUTES USING REMOTE SENSING TECHNOLOGY
20 m; and the Indian LISS 1-3 (72 x 72, 36.25 x 36.25, and 23.5 x 23.5 m,
respectively). For example, Plate 8-1 (after page 182) depicts USGS Level I
urban vs. nonurban information for Charleston, South Carolina, extracted from
Landsat data for 1973, 1981, 1982, and 1994. Sensors with a minimum spatial
resolution of 5-15 m (e.g., SPOT panchromatic [pan] at 10 x 10 m; SPIN-2 TK-
350 at 10 x 10 m; proposed Landsat 7 pan at 15 x 15 m) are required to obtain
USGS Level II information, which includes specific types of man-made struc-
tures. USGS Level III classes may be inventoried using sensors with a spatial
resolution of 1-5 m, such as Indian Remote Sensing (IRS) pan (approximately 5
x 5 m) and the SPIN-2 KVR-1000 (2 x 2 m). Future sensors may include
commercial ventures such as EOSAT/Space Imaging IKONOS (1 x 1 m pan),
OrbView 3 (1 x 1 m pan), and Indian IRS P5 (2.5 x 2.5 m) (Montesano, 1997~.
USGS Level IV classes may best be monitored using high-spatial-resolution
sensors, including aerial photography (0.3-1 m), and proposed EarthWatch
Quickbird pan (0.8 x 0.8 m) and IKONOS (1 x 1 m). A sensor that collects
panchromatic data of 0.3-0.5 m resolution is required to provide detailed Level
IV information. RADARSAT provides data with 11- 100 m spatial resolution for
Level I and II land-cover inventories, even in cloud-shrouded tropical landscapes
where conventional sensors would not be able to penetrate (Leberl, 1990~.
Urban land-use/land-cover classes in Levels I through IV have temporal
resolution requirements of 1-10 years (see Table 8-1 and Figure 8-1~. All of the
sensors mentioned have temporal resolutions of less than 22 days, and thus sat-
isfy these requirements.
Building and Cadastral (Property Line) Infrastructure
Data on building perimeter, area, volume, and height are best obtained using
stereoscopic (overlapping) panchromatic aerial photography or other remote sens-
ing data with a spatial resolution of 0.3-0.5 m (Jensen, 1995; Warner et al., 1996~.
The stereo images are required to visualize features in three dimensions. For
example, panchromatic stereoscopic aerial photography with a spatial resolution
of 0.3 x 0.3 m (1 ft) was used to extract building perimeter and area information
for a residential area in Covina, California (Figure 8-2~. Each building, tree,
driveway, fence, and contour can be extracted from this type of data. In many
instances, the fence lines are the cadastral property lines. Accurate photogram-
metric surveys can meet the new draft Geospatial Positioning Accuracy Stan-
dards (Federal Geographic Data Committee, 1997~. If necessary, the property
lines can be surveyed by a licensed surveyor and the information overlaid onto
the photographic or planimetric map database to represent the legal cadastral
(property) map. Many municipalities in the United States are moving toward
using such high-spatial-resolution imagery as the source for some cadastral infor-
mation and as an image backdrop upon which to depict all surveyed cadastral
information.
OCR for page 169
DAVID J. COWEN AND JOHN R. JENSEN
169
Detailed data on building height and volume can be extracted from high-
spatial-resolution (0.3-0.5 m) stereoscopic imagery (Jensen et al., 1996~. Such
information can then be used to create three-dimensional displays of the terrain
that one can walk through in a virtual-reality environment if desired (Wolff and
Yaeger, 1993) (see Figure 8-3~. Such information provides an extremely useful
way to visualize the density and arrangement of structures in a neighborhood, and
architects, planners, engineers, and realtors are beginning to use this information
for a variety of purposes. It is expected that in the next few years, Space Imaging
(1997) and EarthWatch (Quickbird, 1998/1999) will provide such stereoscopic
images from satellite-based platforms with approximately 0.8-1 m spatial resolu-
tion. Unfortunately, such imagery will still not provide the detailed planimetric
(perimeter, area) and topographic (terrain contours, building height and volume)
details that can be extracted from high-spatial-resolution large-scale aerial pho-
tography. Therefore, a satellite sensor system with 0.3-0.5 m spatial resolution
may be required, but it will not be available in the immediate future.
Transportation Infrastructure
Transportation studies have long relied on remote sensor data to (1) examine
the origin and destination of trips; (2) study traffic patterns at choke points such
as tunnels, bridges, shopping malls, and airports; (3) analyze metropolitan traffic
patterns; (4) conduct parking studies; and (5) evaluate the condition of roads
(Mintzer, 1983; Haack et al., 1997~. The general updating of a road network
centerline map is a fundamental task that is often done once every 1 to 5 years. In
areas with minimum tree density, this task can be accomplished using imagery
with a spatial resolution of 1 to 30 m (Lacy, 1992~. If more precise road dimen-
sions, such as the exact width of the road and sidewalks, are needed, a spatial
resolution of 0.3-0.5 m is required (Jensen et al., 1994~. Currently, only aerial
photography can provide such planimetric information (see Figure 8-2~.
Next to meteorological investigations, traffic-count studies of automobiles,
airplanes, boats, pedestrians, and people in groups require data of the highest
temporal resolution, ranging from 5 to 10 minutes. It is difficult to resolve the
type of car or boat using even 1 x 1 m data; for this purpose, high-spatial-
resolution imagery (0.3-0.5 m) is required. Such information can be acquired
only via aerial photography or video sensors that are (1) located on the top edges
of buildings looking obliquely at the terrain, or (2) placed in aircraft or helicop-
ters and flown repeatedly over the study areas. Parking studies require the same
high spatial resolution (0.3-0.5 m) but slightly lower temporal resolution (10-60
minutes). Road and bridge conditions (e.g., cracks, potholes) can be documented
using high-spatial-resolution aerial photography (<0.3 x 0.3 m) (Stoeckeler, 1979~.
OCR for page 170
170
OCR for page 171
171
Cal
o
o
A
Ct
Cal
Cal
Ct
.=
o ~
o
·
· - d
O
O
O ~
.0
Ct
·=
.0
Cal ~
so -4
;^ I.,
d ~ · ~
Ct
lo, ~
~O
·0 ~ lo,,
Cal · _~
O ·= .0
sly
~ ·0 ·0
,=0 ~ ~
lo, .= ·>
-~ so
Cal O O
·= O
Cal ~
. so O
- , so,
00
Cd
O
o
OCR for page 172
172
OCR for page 173
173
Ct ~
= :>
o ~ ~
= L~ ~
~ ~ o
.
.=
o ~
Cd
sit o
·_.
. o ~
~ i= ~
o
sit
.0 o
.= o -~
° =.=
ho O ~ .
~ ~ o o
cd^ t33 ~ 3
o ~ ~ ~
sit ~ . ~
° ~ .= ~
~ o sit ~
.= at, at, cd
~ ~ ~ -
~ o ~ ~
.
·
l
Do son
~ ~ ~ 3
OCR for page 174
74
MODELING OFA'ITRIBUTES USING REMOTE SENSING TECHNOLOGY
Utility Infrastructure
Urban/suburban environments are enormous consumers of electrical power,
natural gas, telephone service, and potable water (Haack,1997~. In addition, they
create great quantities of garbage, waste water, and sewage. The removal of
storm water from urban impervious surfaces is also a serious problem (Schultz,
1988~. Automated mapping/facilities management (AM/FM) and GIS have been
developed to manage extensive right-of-way corridors for various utilities, espe-
cially pipelines (Jadkowski et al., 1994~. The most fundamental task is updating
maps to show a general centerline of the utility of interest, such as a powerline
right-of-way. This task is relatively straightforward if the utility is not buried.
Major utility rights-of-way can be observed well in imagery with a spatial resolu-
tion of 1-2 m and obtained once every 1-5 years. However, when it is necessary
to inventory the exact locations of footpads or transmission towers, utility poles,
manhole covers, the true centerline of the utility, the width of the utility right-of-
way, and the dimensions of buildings, pumphouses, and substations, a spatial
resolution of 0.3-0.6 m is required (Jadkowski et al., 1994~.
Creation of a Digital Elevation Model
Almost all GISs used for socioeconomic or environmental planning include
a digital elevation model (DEM) (Cowen et al., 1995~. Analysts often forget that
DEMs are derived from analysis of stereoscopic remote sensor data (Jensen,
1995~. It is possible to extract z-elevation data using SPOT 10 x 10 m data and
even Landsat TM 30 x 30 m data for terrain that has not been mapped previously
(Gugan and Dowman, 1988~. However, any DEM to be used for an urban/
suburban application should ideally have a z-elevation and x, y coordinates that
meet Draft Geospatial Positioning Accuracy Standards (Federal Geographic Data
Committee, 1997~. At a minimum, the data should meet the old USGS national
map accuracy standards. The only sensor that can provide such information at
the present time is stereoscopic large-scale metric aerial photography with a
spatial resolution of 0.3-0.5 m. The terrain elevation does not change very
rapidly. Therefore, a DEM of an urbanized area need be acquired only once
every 5 to 10 years unless there is significant development, and the analyst
wishes to compare two DEMs for different dates to determine changes in terrain
elevation and identify unpermitted additions onto buildings or changes in build-
ing heights. DEM data can be modeled to compute slope and aspect statistical
surfaces for a variety of applications (Jensen, 1996~. They can also be used to
predict the optimum sites for locating various utilities, as shown earlier in Figure
8-3(d). Digital desktop soft-copy photogrammetry is revolutionizing the creation
and availability of special-purpose DEMs by minimizing the need for expensive
specialized steroplotting equipment (Petrie and Kennie, 1990; Jensen, 1995~.
OCR for page 178
178
MODELING OFA'ITRIBUTES USING REMOTE SENSING TECHNOLOGY
every 5 to 10 minutes in severe storm warning mode. Early warnings provided by
these meteorological radars have also saved many lives.
Finally, daytime and nighttime thermal infrared remote sensor data with high
spatial resolution (5-10 m) represent one of the primary methods for obtaining
quantitative spatial information on the urban heat island effect (Lo et al., 1997~.
Critical Environmental Area Assessment
Urban/suburban environments often include sensitive areas such as wet-
lands, endangered species habitats, parks, land surrounding treatment plants, and
land in urbanized watersheds that provides the runoff for potable drinking water.
Relatively stable sensitive environments need be monitored only every 1 to 2
years using a multispectral remote sensor collecting data with a resolution of 1 - 10
m. For extremely critical areas that could change rapidly, multispectral remote
sensors (including a thermal infrared band) should obtain data with a resolution
of 0.3-2 m every 1 to 6 months.
Disaster Emergency Response
Recent floods (Mississippi River in 1993; Albany, Georgia, in 1994), hurri-
canes (Hugo in 1989, Andrew in 1991, Fran in 1996), tornadoes (every year),
fires, tanker spills, and earthquakes (Northridge in 1994) have demonstrated that
a rectified predisaster remote sensing image database is indispensable. The
predisaster data need be updated only every 1 to 5 years; however, multispectral
data with high spatial resolution (1-5 m) should be obtained if possible.
When disaster strikes, high-resolution (0.5-2 m) panchromatic and/or near-
infrared data should be acquired within 12 hours to 2 days. If the terrain is
shrouded in clouds, imaging radar may provide the most useful information.
Postdisaster images are registered to the predisaster images, and manual or digital
change detection is performed (Jensen, 1996~. If precise, quantitative informa-
tion about damaged housing stock, disrupted transportation arteries, the flow of
spilled materials, and damage to above-ground utilities is required, it is advisable
to acquire postdisaster panchromatic and near-infrared data with a resolution of
0.3-1 m within 1 to 2 days. Such information was indispensable in assessing
damages and allocating scarce cleanup resources during Hurricanes Hugo, An-
drew, and Fran (Wagman, 1997) and the recent Northridge earthquake.
USE OF REMOTE SENSING FOR FORECASTING
URBAN RESIDENTIAL EXPANSION
The study of residential expansion has a long history that is closely linked to
early models of the internal structure of cities in which an urban area is viewed as
a series of concentric rings, sectors, or multiple nuclei (Harris and Ullman,1945~.
OCR for page 179
DAVID J. COWEN AND JOHN R. JENSEN
179
In those early models, the rate of expansion of the city was treated as a struggle
between a series of centrifugal and centripetal forces. In recent decades, re-
searchers have attempted to model this process using empirical data.
Models ranging from those that emerged from urban ecology literature in the
mid-1920s to those based on urban economics of the 1960s, which were founded
in rent theory. The general assumption of these models was that land prices
would be highest in the center of the city where accessibility was greatest.
Wealthier and more mobile residents would trade off accessibility for more space
at the periphery, while poorer residents would live near the center of the urban
area at higher densities. This model was articulated by Alonso (1964), but was
directly related to much earlier work on agricultural land-use theory.
An important model of residential growth was developed by Chapin and
Weiss (1968~. This model was based on the concept of priming actions that
trigger secondary actions and together produce land development. This residen-
tial location model was designed to allocate residential units to areas experienc-
ing growth.
Recent research by Batty and Longley (1994) has taken a fresh look at these
urban models and attempted to rework them in light of emerging research in
fractal geometry. Most models of urban growth have modeled this diffusion
process as a manifestation of random events; however, it is likely that dynamic
urban systems are not random, but deterministic in nature. Therefore, a good
time series of events that can capture the underlying patterns needs to be estab-
lished. Remote sensing that can monitor changes approximately every 3 weeks
can provide data not available from traditional sources, such as the decennial
census of population and housing.
In light of previous attempts to model residential expansion, a major research
effort funded by the National Aeronautics and Space Administration was under-
taken. This effort focused on the development of an integrated remote sensing
and GIS model that could be used to predict urban expansion between census
periods (Jensen et al., 1994~. This model was based on a systematic method of
capturing and analyzing a wide range of data sources that are indicators of urban
development. Unlike previous residential models, this model incorporated cen-
sus data, land use/land cover, raster-based satellite imagery, building permit data,
and postal code geography.
An important goal of the model was to forecast not only future growth
patterns, but also the specific number of new single-family homes that might be
constructed. To accomplish this goal, it was necessary to measure available land,
density of housing units, and residential growth rates. The first component of the
model was estimation of the change in the amount of available land for develop-
ment. This was done using USGS land-use/land-cover polygons from 1976 and
SPOTJ classified multispectral imagery from 1989. A land-use change detection
resulted in a data set showing land that was urban in 1976 and land that had been
converted to urban by 1989 (Figure 8-4~. The land-use data provide a basis for
OCR for page 180
180
MODELING OFATTRIBUTES USING REMOTE SENSING TECHNOLOGY
~<,-
\
_ ~ . ~
~ ~ ~ ~ ~ Hi,
/~
a
IKE
~e ~'0
\ ~^'5^
..t
_
~ ~:
i ~ ~
~ a; ~ ~ ~
Ja ckso:
~ ~",/1 'at ~
2 0 2 4 hales ~ .
fag
. ~
4. ~
01
A
A
~\, C ~ ~ ~ ~ .~
~ COMA
b
OCR for page 181
DAVID J. COWEN AND JOHN R. JENSEN
. ~ .
~ : ~s ~
C
181
~ 1
.N
O ~ 4 Miles ~:
FIGURE 8-4 Land-use maps for Columbia, South Carolina, based on (a) 1976 USGS
land-use and land-cover data, and (b) 1990 SPOT 20 x 20 m multispectral data. Shaded
areas indicate urban land uses. (c) Land converted to urban uses 1976-1990 (shaded
areas).
determining where development can be expected to occur. In fact, the SPOT
multispectral data at a resolution of 20 x 20 m provided a basis for classifying all
the land in South Carolina as either developable or undevelopable (Plate 8-2,
after page 182~. Developable land consisted of agricultural land, scrub/shrub,
and forests. Undevelopable land consisted of bodies of water, wetlands, and
publicly owned lands such as parks and military installations. This approach
provided a useful static view of potential areas of development throughout the
state. Furthermore, government-owned lands and other protected areas that had
been identified by the Bureau of the Census were identified in the Topographi-
cally Integrated Geographic Encoding Reference (TIGER) line files, and those
land-use polygons were extracted from the developable land areas to present a
more realistic estimate of the amount of land available for development.
Statewide analysis of the 20 x 20 m multispectral data is not economically
feasible on a regular basis. However, our research indicates that the higher-
resolution 10 x 10 m panchromatic band of SPOT data can be used to detect
OCR for page 182
182 MODELING OFATTRIBUTES USING EMOTE SENSING TECHNOLOGY
changes on a local level. For example, land-cover changes in the Columbia,
South Carolina, metropolitan area have been monitored on a 2-year cycle for the
past 8 years. These remotely sensed data were integrated with data on 1990
developable land to provide timely updates that clearly identify where distur-
bances are occurring (see Figures 8-5 and 8-6~. This type of synoptic view is
more efficient than traditional windshield surveys and much less expensive than
aerial photography missions.
Once a measure of the amount of developable land had been determined, it
was necessary to estimate the average amount of land per housing unit. This
component of the residential forecasting model was calculated for each block
group on the basis of the 1990 Census of Housing figures at the block level. The
average lot size was adjusted on the basis of the actual urban land use, not the
1000 0 1000 2000 Valeted
I 1 1 1 1
FIGURE 8-5 SPOT 10 x 10 m data overlaid with developed land (diagonal lines). Note
the airport and the Interstate highway system.
OCR for page 183
183
_ - , ~ ~ ';- ),. - ~ -. BAJA-, ~ if ~ ;
FIGURE 8-6 SPOT 10 x 10 m derived urban land-use changes within census block
group polygons derived from SPOT 10 x 10 m data. 1990-1992 changes are in dark gray;
1992-1994 changes are in light gray.
total land area of the block group. Therefore, it could be assumed that the figure
was representative of the average lot size in that neighborhood at that time.
The final factor in the model required a spatially detailed empirical estimate
of the residential growth rate throughout the metropolitan area. The best indica-
tor of housing changes was a record of 15,303 new single-family building permits
for an 11-year period between 1980 and 1990. The permits were geocoded and
assigned to block groups. The result was a time-series database for estimating the
rate of land-use conversion for small areas. The building permits represent an
excellent resource for analyzing spatiotemporal change. The pattern can also be
treated as a demographic process a series of births, deaths, and migrations that
result in a changing spatial point pattern. From the viewpoint of real estate
developers, the housing market progresses through a life cycle that involves the
density of houses, or lot size, and the availability of land, or land absorption rate.
When all of these data sources are combined, it is possible to visualize the series
of events occurring within an urban area with a considerable amount of detail
(Plate 8-3, after page 182~.
Rather than trying to fit a generalized expansion model to the entire region,
OCR for page 184
184
MODELING OFA'ITRIBUTES USING REMOTE SENSING TECHNOLOGY
Pre d i`:;t
O- 119
q20- 587
588 - 81 ~
1 1 - 1 ~ SO
~-e
~ 41`i
FIGURE 8-7 Number of new houses forecasted in 2005 for census block groups.
it was deemed more useful to summarize the temporal aspects of the permit data
for each of the 393 block groups. A regression analysis was performed for each
block group to determine the relationship between the number of building per-
mits and time. The parameters of the regression models (slope and intercept)
actually became attributes of the polygons. The Y intercept represented an esti-
mate at the start of the study period. The regression coefficient provided a
summary of the rate and direction of change throughout the period. The correla-
tion coefficient measured the strength of the trend. With the parameters for the
regression model for each polygon, this model has the ability to predict the
number of housing units through time. By incorporating the availability of land
and housing unit size into the model, it is possible to estimate the period when
developable land within a block group will become saturated. These models
were used to forecast the number of housing units, the amount of available land,
and the year of saturation for the years 1992 through 2005 (see Figure 8-7~. This
effort lays the foundation for future models that will incorporate spatial informa-
tion extracted from remotely sensed data (Halls et al., 1994~.
CONCLUSIONS
The future interface between social science research and remote sensing
depends on what kind of features can be detected and how often the data can be
obtained. Remote sensing may be used not only for monitoring change, but also
OCR for page 185
DAVID J. COWEN AND JOHN R. JENSEN
185
for conducting surveillance. For example, it may become possible not just to
count houses, but to count the number of stories and detect changes in structures.
Thus remotely sensed data may provide the ability to check on building regula-
tions. It may also be possible to develop some new surrogates for socioeconomic
conditions. For example, factors such as lot size, the condition of lawns, numbers
of swimming pools, and numbers of vehicles may be used to provide insight into
residential quality. These capabilities will also provide extremely valuable inputs
for models of residential water usage and the demand for other public services.
Table 8-1 and Figure 8-1 reveal that there are a number of remote sensing
systems that currently provide some of the desired urban/socioeconomic infor-
mation when the required spatial resolution is poorer than 5 x 5 m and the
temporal resolution is 1 to 55 days. However, data with very high spatial resolu-
tion (< 1 x 1 m) are needed to satisfy many of the requirements for socioeco-
nomic data. In fact, as shown in Figure 8-1, the only sensor that currently
provides such data on demand is aerial photography (0.3-0.5 m). Neither EOSAT/
Space Imaging (1997) with its 1 x 1 m panchromatic data nor EarthWatch
Earlybird (1997) with its proposed 3 x 3 m panchromatic data nor Quickbird with
its 0.8-0.8 m data will satisfy all of the data requirements. None of the sensors,
except repetitive aerial photography, can provide the 5-60 minute temporal reso-
lution needed for traffic and parking studies. It may be necessary to have satellite
remotely sensed data with higher spatial resolution (0.3-0.5 m) and temporal
resolution (1-3 days) to provide much of the desired detailed urban/suburban
socioeconomic information, or to utilize aerial photography. Fortunately, the
GOES constellation (East and West) and the European Meteosat provide suffi-
cient national and regional weather information at reasonable temporal (30 min-
utes) and spatial (1-8 km) resolution. Ground-based Doppler radar provides
sufficient spatial (4 x 4 km) and temporal (5-30 minutes) resolution for precipita-
tion and intense storm tracking.
Finally, while remote sensing provides a valuable way to monitor changes on
the earth's surface, it can only suggest details about human activity. To obtain
this type of information, it is necessary to have a source of data for monitoring the
movement of people; consumer behavior; and a wide range of events relating to
crime, health, and other matters. It is also important to note that while such
knowledge helps route emergency vehicles to our houses and can help utility
companies and urban planners prepare for future developments, there is no ques-
tion that our individual rights to privacy may be jeopardized. It is clear that
improvements in the spatial and spectral resolution of sensing systems have the
potential to impact our privacy by providing public and private organizations
with visual clues regarding the activities in our houses or on our property. From
a social science perspective, we will soon have the ability to monitor human activity
much more closely than has been possible to date. The key question is whether this
type of information can be used to create more efficient urban environments and
provide for a more equitable distribution of resources and services.
OCR for page 186
186
Alonso, W.
MODELING OFATTRIBUTES USING REMOTE SENSING TECHNOLOGY
REFERENCES
1964 Location and Land Use. Cambridge, Mass.: Harvard University Press.
Anderson, J.R., E.E. Hardy, J.T. Roach, and R.E. Witmer
1976 A Land Use and Land Cover Classification System for Use with Remote Sensor Data
U.S. Geological Survey Professional Paper 964. USGS, Washington D.C.
Angelici, G. L., N.A. Bryant, R.K. Fretz, and S.Z. Friedman
1980 Urban Solar Photovoltaics Potential: An Inventory and Modeling Study Applied to the
San Fernando Valley Region of Los Angeles. JPL Report. 80-43, Pasadena, Calif.: Jet
Propulsion Laboratory.
Batty, M., and P. Longley
1994 Fractal Cities: A Geometry of Form and Function. London, England: Academic Press.
Branch, M.C.
1971 City Planning and Aerial Information. Cambridge, Mass.: Harvard University Press.
Chapin F.C., and S.F. Weiss
1968 A probabilistic model for residential growth. Transportation Research 2:375-390.
Clayton, C., and J.E. Estes
1980 Distributed parameter modeling of urban residential energy demand. Photogrammetric
Engineering and Remote Sensing 45:106-115.
Colcord, J.E.
1981 Thermal imagery energy surveys. Photogrammetric Engineering and Remote Sensing
47:237-240.
Cowen, D., J. Jensen, J. Halls, M. King, and S. Narumalani
1993 Estimating Housing Density with CAMS Remotely Sensed Data, Proceedings, ACSM/
ASPRS :35-43.
Cowen, D., J.R. Jensen, G. Bresnahan, D. Ehler, D. Traves, X. Huang, C. Weisner, and H.E. Mackey
1995 The design and implementation of an integrated GIS for environmental applications. Pho-
togrammetric Engineering and Remote Sensing 61(11):1393-1404.
Eliasson, I.
1992 Infrared thermography and urban temperature patterns. International Journal of Remote
Sensing 13(5):869-879.
Federal Geographic Data Committee
1997 Draft Geospatial Positioning Accuracy Standards. Washington, D.C.: Federal Geographic
Data Committee.
Ford, K.
1979 Remote Sensing for Planners. Rutgers: State University. of New Jersey.
Green, K., D. Kempka, and L. Lackey
1994 Using remote sensing to detect and monitor land-cover and land-use change. Photogram-
metric Engineering and Remote Sensing 60(3):331-337.
Green, N.E.
1957 Aerial photographic interpretation and social structure of the city. Photogrammetric Engi-
neering 23:89-96.
Gugan, D.J., and I.J. Dowman
1988 Topographic mapping from SPOT imagery. Photogrammetric Engineering and Remote
Sensing 54(10):1409-1404.
Haack, B., S. Guptill, R. Holz, S. Jampoler, J. Jensen, and R. Welch
1997 Urban analysis and planning. Pp. 517-553 in Manual of Photographic Interpretation.
Bethesda, Md.: American Society for Photogrammetry and Remote Sensing.
Halls, J.N., D.J. Cowen, and J.R. Jensen
1994 Predictive spatio-temporal modeling in GIS. Pp. 431-448 in Advances in GIS Research,
Vol. 1, T.C. Waugh and R.G. Healy, eds., London, England: Taylor and Francis.
OCR for page 187
DAVID J. COWEN AND JOHN R. JENSEN
187
Harris, C.D., and E.L. Ullman
1945 The nature of cities. Annals of the American Academy of Political and Social Science
244:7-17.
Henderson, F.M., and J.J. Utano
1975 Assessing general urban socioeconomic conditions with conventional air photography.
Photogrammetria 31 :81-89.
Holz, R.K.
1988 Population Estimation of Colonias in the Lower Rio Grande Valley Using Remote Sens-
ing Techniques. Paper presented at the annual meeting of the Association of American
Geographers, Phoenix, AZ.
Jadkowski, M.A., P. Convery, R.J. Birk, and S. Kuo
1994 Aerial image databases for pipeline rights-of-way management. Photogrammetric Engi-
neering and Remote Sensing 60(3):347-353.
Jensen, J.R., ed.
1983a Urban/suburban land use analysis. Pp. 1571-1666 in Manual of Remote Sensing, 2nd ea.,
R.N. Colwell, ed. Falls Church, Va.: American Society of Photogrammetry.
Jensen, J.R.
1983b Detecting residential land-use development at the urban fringe. Photogrammetric Engi-
neering and Remote Sensing 48:629-643.
1995 Issues involving the creation of digital elevation models and terrain corrected
orthoimagery using soft-copy photogrammetry. Geocarto International: A
Multidisciplinary Journal of Remote Sensing 10(1):1-17.
1996 Introductory Digital Image Processing: A Remote Sensing Perspective. Saddle River,
N.J.: Prentice-Hall.
Jensen, J.R., D.C. Cowen, J. Halls, S. Narumalani, N. Schmidt, B.A. Davis, and B. Burgess
1994 Improved urban infrastructure mapping and forecasting for BellSouth using remote sens-
ing and GIS technology. Photogrammetric Engineering and Remote Sensing 60(3):339-
346.
Jensen, J.R., X. Huang, D. Graves, and R. Hanning
1996 Cellular phone transceiver site selection. Pp. 117-125 in Raster Imagery in Geographic
Information Systems, S. Morain and S. Baros, eds. Santa Fe, NM: OnWard Press.
Lacy, R.
1992 South Carolina finds economical way to update digital road data. GIS World 5(10):58-60.
Leberl, F.W.
1990 Radargrammetric Image Processing. Norwood, Mass.: Artech House.
Lindgren, D.T.
1985 Land Use Planning and Remote Sensing. Boston, Mass.: Martinus Nijhhoff Inc.
Lo, C.P.
1986 The human population. Pp. 40-70 in Applied Remote Sensing, New York: Longman.
1995 Automated population and dwelling unit estimation from high-resolution satellite images:
A GIS approach. International Journal of Remote Sensing 16(1):17-34.
Lo, C.P., D.A. Quattrochi, and J.C. Luvall
1997 Application of high-resolution thermal infrared remote sensing and GIS to assess the
urban heat island effect. International Journal of Remote Sensing 18(2):287-304.
McCoy, R.M., and E.D. Metivier
1973 House density vs. socioeconomic conditions. Photogrammetric Engineering 39:43-49.
Mintzer, O.W., ed.
1983 Engineering applications. Pp. 1955-2109 in Manual of Remote Sensing, 2nd ea., R.N.
Colwell, ed. Falls Church, Va.: American Society for Photogrammetry.
OCR for page 188
188
MODELING OFA'ITRIBUTES USING REMOTE SENSING TECHNOLOGY
Monier, R.B., and N.E. Green
1953 Preliminary findings on the development of criteria for the identification of urban struc-
tures from aerial photographs. Annals of the Association of American Geographers,
special issue.
Montesano, A.P.
1997 Roadmap to the future. Earth Observation 6(2):16-19.
Morain, S.A., and A.M. Budge, eds.
1996 Earth observing platforms and sensors. Vol. 2 in Manual of Remote Sensing, 3rd Edition,
A. Ryerson, editor-in-chief. CD-ROM. Bethesda, Md.: American Society for Photo-
grammetry and Remote Sensing.
Petrie, G., and T.J.M. Kennie
1990 Terrain Modeling in Surveying and Civil Engineering. London, England: Whittles Pub-
lishing.
Philipson, W.
1997 Manual of Photographic Interpretation. Bethesda, Md.: American Society for Photo-
grammetry and Remote Sensing.
Raymondo, James C.
1992 Population Estimation and Projection: Methods for Marketing, Demographic, and Plan-
ning Personnel. New York: Quorum Books.
Schultz, G.A.
1988 Remote sensing in hydrology. Journal of Hydrology 100:239-265.
Stoeckeler, E.G.
1979 Use of aerial color photography for pavement evaluation studies. Highway Research
Record 319:40-57.
Tobler, W.
1969 Satellite confirmation of settlement size coefficients. Area 1:30-34.
Tuyahov, A.J., C.S. Davies, and R.K. Holz
1973 Detection of urban blight using remote sensing techniques. Remote Sensing of Earth
Resources 2:213-226.
Wagman, D.
1997 Fires, hurricanes prove no match for GIS. Earth Observation 6(2):27-29.
Warner, W.S., R.W. Graham, and R.E. Read
1996 Urban survey. Pp. 253-256 in Small Format Aerial Photography. Scotland: Wittles Pub-
lishing.
Wolff, R.S., and L. Yaeger
1993 Visualization of Natural Phenomena. Santa Clara, Calif.: Telos Springer-Verlag.
Representative terms from entire chapter:
aerial photography
