Environmental Engineering Reference
In-Depth Information
clear. The spatial distribution pattern became more dispersed.
Low-density urban use continued to grow after 1987. Most
of the new development, however, took place in the exterior
metropolis. This widely spread-out pattern is a major indicator
of suburbanization. Quantitatively, the low-density urban use
occupied 373 685 hectares or 35.75% of the total area in 2007,
indicating a 386.19% increase between 1973 and 2007. The daily
increment was about 24 hectares or 59 acres for the same period.
Tables 2.3 and 2.4 also indicate that the continuing decline
in cropland/grassland and forest land. The shrinking pattern of
these two classes was proportional to the growth of the two
urban classes. In general, the decline of cropland/grassland and
forest land predominately took place in the interior metropolis
before 1987 but in the exterior after 1987. Quantitatively, crop-
land/grassland occupied 159 366 hectares or 15.25% of the total
study area in 1973. It declined to 69 608 hectares (or 6.66%) by
2007. This represents a decrease of 56.32%, or a daily rate of
approximately 7 hectares (17 acres). Similarly, forest declined
from 750 880 hectares (or 71.84%) in 1973 to 454 962 hectares
(or 43.53%) in 2007, thus representing a decrease of 39.41%, or
a daily rate of about 24 hectares (59 acres) in land area.
ThenatureofchangeisquiteclearfromTable2.5.From
1973 to 2007, the loss of forest land contributed to 68.90% of the
urban growth while the loss of cropland/grassland accounted for
28.35%. The high-density urban use had a net addition of 79 285
hectares, among which 60.53% came from the loss in forest land
(C5) and 34.59% resulted from the loss in cropland/grassland
(C3). The loss in cultivated/exposed land only contributed to
4.89% of the increase in high-density urban use (C1). For the
low-density urban use, 70.01% (C6) and 26.79% (C4) of the
increase came from the loss in forest land and cropland/grassland,
respectively. The loss of cultivated/exposed land only accounted
for 2.20% (C2) of the net addition in low-density urban use.
of dispersal to more dispersed pattern for low-density urban use
(mainly residential) in Atlanta.
2.4 Discussion
2.4.1 A generic urban growth
monitoring workflow
Urban spatial growth can be mapped and measured by using
time-sequential satellite imagery such as archival Landsat data,
as demonstrated in our case study and some other studies (e.g.,
Pathan et al ., 1993; Gomarasca et al ., 1993; Green, Kempka
and Lackey, 1994; Hill and Hostert, 1996; Masek, Lindsay and
Goward, 2000; Seto and Fragkias, 2005; Liu et al ., 2010). Based
on these studies, a generic workflow for urban growth moni-
toring by remote sensing can be summarized in Fig. 2.5, which
includes several major components: defining research questions,
data acquisition and collection, image preprocessing, change
detection, and interpretation and analysis. Appropriate handling
each of these components is not a trivial task by any means.
Defining research questions involvesomenon-technicalissues
such as the purpose and scope of research as well as the physical
and cultural characteristics of a study site, which will ultimately
determine specific technical procedures to be adopted. While
we will not delve into these non-technical issues, a moderate
understanding of the physical and cultural characteristics of the
area under investigation can improve the design and implemen-
tation of subsequent technical procedures (Yang and Lo, 2002).
Moreover, some essential knowledge on urban geography and
landscapeecologycanhelpidentifyappropriateremotesensor
data or information extraction techniques (Herold, Scepan and
Clarke, 2002; Longley, 2002; Yang and Lo, 2003; Lo, 2004). For
example, urban physical composition varies greatly across the
world, which is linked with underlying social, economic, and
cultural circumstances (Kaplan, Wheeler and Holloway, 2009).
Compared to American cities, most non-American counterparts
are compact and clustered by nature, which may need to use
data with much higher spatial resolution (Welch, 1982; Jensen
and Cowen, 1999). Our further discussion will focus on some
issues relating to the three technical components in the generic
workflow (Fig. 2.5).
2.3.6 Summary
By using Atlanta as a case, this study has demonstrated the use-
fulness of satellite remote sensing for urban change mapping and
measurement. Central to this study was a time series of satellite
images acquired by three Landsat sensors, namely, MSS, TM and
ETM
, which has been used to produce land use/cover maps
through image classification. The combined use of unsupervised
classification and GIS-based spatial reclassification procedures
has been quite effective to resolve the spectral confusion among
different land classes, a typical problem with broad-band sensors
and within an urban image scene. The time series of land classifi-
cation maps was further used to analyze urban spatial growth and
the nature of change. This was built upon the combined use of
post-classification comparison and GIS-based overlay techniques
that made possible the production of single-theme change maps,
which emphasize spatial dynamics. On the other hand, this study
has established a well-documented regional case study focusing
on Atlanta, a typical postmodern American metropolis having
undergone rapid demographic and economic growth during the
past several decades. This study reveals a rampant growth in urban
land during 1973 - 2007, which substantially outpaced the rate of
population growth in Atlanta. Urban spatial form experienced a
transition from linear concentration to multinucleated pattern
for high-density urban use (mainly commercial, industrial and
transportation) and from concentration mixed with some degree
+
2.4.2 Image resolution and land
use/cover classification
Among the technical components (see Fig. 2.5), acquiring an
appropriate remote sensor data set can be quite challenging.
Theoretically, data used for urban studies must meet certain
conditions in terms of spatial, spectral, radiometric, and temporal
characteristics (Lo, 1986; Jensen and Cowen, 1999; Jensen, 2007).
However, the choice of available data sets can be quite limited
for a study covering a decades-long period. For many such
studies, Landsat data become the only choice, particularly during
the period of early 1970s to early 1980s when there was no
any alternative available because other Earth resource satellite
programs had not been initiated. Although within the range of
Search WWH ::




Custom Search