Image Processing Reference
In-Depth Information
necessary to conduct the analysis in parts since the
FRAGSTATS software has a limitation of approximately
4,000-5,000 input classes per calculation run. Applying a
uniform grid to the land cover data guarantees a uniform
external limitation on different test areas throughout the
dataset (Netzband and Kirstein
2001
). It also facilitates
raster-based intra-regional characterizations of features.
Gradients between the city center and the outskirts, or
through other divisions of urban agglomerations, can be ana-
lyzed and quantified in a differentiated manner. Furthermore,
different patterns of residential or green areas can be recon-
structed, if they are larger than the pixel size of the remotely
sensed data (minimum 15 m/pixel for ASTER).
gridding of the
land cover and
landscape metric
data provides a
uniform
framework for
regression
analysis, and
allows for
rigorous
comparison with
results from other
urban centers
12.3
Results and Discussion
The general urban structure of Phoenix is the product of only 150 years of develop-
ment. The majority of the current built-up area was constructed after 1940, and cur-
rently defines a northwest-southeast trending agglomeration set along a regular
north-south and east-west transportation grid (Fig.
12.2
). This development pattern
is the result of physiographic constraints (mountains), availability of water delivery
infrastructure, the dominance of individual automobile transport rather than mass
transit, and political boundaries related to federal and Native American lands
(GP2100
2003
). The Phoenix metropolitan area has a fairly well defined urban core
defined by the skyscrapers of downtown Phoenix proper. This is surrounded by a
heterogeneous mixture of residential, commercial/industrial, and parkland areas of
varying age and spatial extent comprising the numerous surrounding municipalities.
Older residential developments within the urban agglomeration tend to be more
open and occupy larger plots of land. More expensive and exclusive developments
located in desert areas and next to scenic landforms also tend to be less dense. A burst
of construction of dense multi-unit residential developments that began in the 1990s
defines the location of the current urban fringe together with a fragmenting halo of
agricultural fields and farms undergoing land use conversion to residential/com-
mercial uses. Most of the socially and economically weak segments of society in
Phoenix are concentrated well within the urbanized region rather than along the
outskirts (Gammage
1999
).
The majority of variance in the metric results is expressed in the urban grid cells
(Fig.
12.4
), as the nonurban grid cells include relatively little variation in land cover
types. Representative results for the CA, ED, and IJI metrics for the Built aggregate
land cover class are presented as greyscale raster images in Fig.
12.5
. An example
of the Mean Patch Size results is not included as there is little to no variation in this
metric at the 1 km analysis scale (leading to monotonic raster results). Raster
images were generated for each aggregate land cover class (data not shown) with
the exception of the Water class.
