Environmental Engineering Reference
In-Depth Information
where ecological open space issues are important. At the local
scale, this measure has less utility.
Ranking of whether the measurement is
meaningful, useful
and
simple to understand
was conducted based on our assessment of
how much professional knowledge was required for a stakeholder
to understand the concept behind the measure. Many of the
measures are very straightforward (how much land is built, how
many people live in a certain block of land), but others require
more nuanced understanding, such as density gradients, and
certain geometric measurements. Fractal dimensions, as useful
as they may be, are difficult to explain to a diverse audience.
The sprawl quotient is a very popular measure, but we find
several instances where its application is problematic and its
meaning misconstrued. For example, in compact, high den-
sity cities with aging demographic profiles, a small amount of
spatial growth can result in an illogically high sprawl quotient
relative to sprawling, low density (but demographically young)
cities (Frenkel and Ashkenazi, 2008b). Similarly, the comparison
of percentage of population living in low density versus high
density urban areas can be heavily influenced by the particular
demographics of each area (e.g. young families in low density ver-
sus aging individuals in high density tracts). Values may change,
perhaps suggesting sprawl even in the absence of urban expan-
sion. Negative population growth has been shown to introduce
complications for the use of other per capita indices to measure
sprawl, as well (Hasse and Lathrop, 2003b). On the other hand,
sprawl has been shown to occur where the amount of developed
land grows, even while population falls (Kasanko
et al
., 2006),
thereby producing negative values of the sprawl quotient.
Finally, we consider
ease of application
. This consideration
depends on how much data is required, the need for software
and associated technical ability, and/or whether quantifying the
measures depends on complex calculations. At the extreme, data
for some measures can be extracted and estimated with a paper
map and a marking pen. Others demand access to digitized
maps, remotely sensed data, GIS software and survey/census
data. For instance using patch type measures may demand a
high degree of
apriori
knowledge about the area and ancillary
data sets to complement remotely sensed data, such that the user
can define each urban patch type (e.g., residential, industrial,
commercial, etc.). Still others demand proficiency at applying
complex computational or mathematical calculations using pro-
fessional software and programming. We give high ranking to
those measurements that could be used outside of a university
or well-funded government research institution, and low rank-
ing to those measures that would be difficult to collect without
large budgets and high levels of technical proficiency. Some
spatial metrics received lower rankings due to the challenge of
clearly defining and measuring patch types. Once patch types are
defined, however, the landscape metrics can be calculated using
the popular and free Fragstats software (McGarigal
et al
., 2002),
assuming GIS software proficiency.
Each of the studies employed a different sprawl measure or set of
measures and in some cases, unique datasets (see Table 12.2).
The results of the comparative analysis of the studies' findings
show many similarities, but also several differences (Table 12.3).
In many cases the same metropolitan region appeared at the
extremes (i.e., the most sprawled or the most compact) in all
of the studies. These regions have characteristics of sprawl or
compactness that were robust enough to manifest themselves
across many measures. On the other hand, there were multiple
inconsistencies in the rankings, where regions would rank highly
as either sprawled or compact in one or more studies, but fall
into the mid-range in other studies, being neither sprawled nor
compact.
Third, there were several cases where a metropolitan region
would be characterized as compact by one or more studies,
but sprawled in another. In these cases, it was generally the
study by Jordan and colleagues (1998) that provided a contrary
result for a given region, as it did with Los Angeles, Miami,
and Chicago metropolitan regions. In each of these cases, the
regions were considered compact according to the measures in
two or three of the other studies, while they rated sprawled
in the Jordan
et al
.study.Thismaybeduetoatleastthree
reasons. First, three studies were measuring
state
of sprawl at a
given time. The fourth study measured both state and
process
between 1970 and 1990. In the case of Miami and Chicago,
the areas were becoming more sprawled over time relative to
their status in 1970 and 1980. Likewise, the Oklahoma City
metropolitan region, rated as sprawling in one study and in
the middle range in two others, was becoming more compact
over time according to Jordan
et al
.
7
Second, spatial extent of
metropolitan areas may have been defined differently by each
author. Even though most of the researchers were working
with US Census Bureau definitions, there is room for selectivity
regarding which metropolitan boundaries to employ. Third, as
several authors have suggested, different sprawl measures can
yield disparate results regarding a single location (Frenkel and
Ashkenazi, 2008b; Torrens, 2008), as also shown in Fig. 12.1. It
appearsthatdensitygradients(usedbyJordan,RossandUsowski,
1998) capture elements of sprawl differently than the measures
used in other studies.
This emphasizes three parallel considerations for sprawl
research. First, the element of time deserves a more central
role in the study of sprawl. Some scholars discuss relative values
of sprawl measures, either changing in time or between places,
where the difference between sprawl and compact is not an
absolute, but rather, a relative change along a continuum (Pen-
dall, 1999; Johnson, 2001). They investigate temporal changes
in urban spatial development, such as increases or decreases
in residential density (Hasse and Lathrop, 2003a; Frenkel and
Ashkenazi, 2008b), or changes in density gradients from CBDs
(Jordan, Ross and Usowski, 1998) to determine dynamic pat-
terns of sprawl. Relative sprawl and the direction of sprawl
indicators over time are thus key considerations in sprawl studies
(Torrens and Alberti, 2000; Galster
et al
., 2001; Malpezzi and
12.4.3.1
Does choice of measures
matter?
In order to assess how important the choice of sprawl measures
is to the characterization of sprawl, we compared four studies
that estimated sprawl across metropolitan areas in the United
States (Jordan, Ross and Usowski, 1998; Razin and Rosentraub,
2000; Ewing, Pendall and Chen, 2002; Lopez and Hynes, 2003).
7
A fifth study which classified level of sprawl across US metropolitan areas, and
which makes use of remotely sensed data (nighttime satellite imagery) is Sutton
(2003). Sutton's results support, for the most part, the classifications in Table 12.3.
Little Rock, Knoxville, Greenville and Atlanta ranked as sprawled (support for the
consensus), as did Oklahoma City. Lincoln, Chicago and Miami were ranked as
relatively compact, although New York and Phoenix rank neither compact nor
sprawled, but near the national average. Syracuse ranked as relatively compact, as
did Los Angeles, adding to the ambiguity about that city.
Search WWH ::
Custom Search