A pluralistic approach to defining and measuring urban sprawl - Urban Remote Sensing: Monitoring, Synthesis and Modeling in the Urban Environment - page 166

Sprawl Indices Type A Type B Type C Type D

Gross density 22,356 Compact 9,213 Compact 1,304 Sprawl 2,470 Sprawl

Density growth rate
(1985-2004)

37% Compact 52% Compact 1% Sprawl − 8% Sprawl

Shape index 1.68 Compact 3.15 Sprawl 1.95 Compact 4.63 Sprawl

9 %

Leap frog index 0% Compact 2% Sprawl 0% Compact Sprawl

Population size 2004 138,900 211,600 2,500 13,300

FIGURE 12.2 A comparison of four urban areas and their ranking according to four selected sprawl measures. Each urban
area is illustrated with an estimate of built area derived from maps and verified with aerial photographs (left; for which the
sprawl measures were calculated) and with a visually abstracted illustration (right) to simplify its geometry (right).

new typology for urban land use patterns is needed in place of a
misleading 'more or less sprawled' dichotomy.
Most of the sprawl measures score high in objectivity ;that
is, they can be quantified and the methods by which they are
obtained can be replicated. While the value obtained for any
measure is subject to user interpretation (e.g., what density
value constitutes high or low density), some measures have an
added layer of subjectivity in that they require user decisions
prior to calculating the value of the measure. For instance, some
measures require deciding apriori what constitutes a low-density
neighborhood, a suburb, or a central business district so that
their area or distances between them can be quantified. This is
straightforward in theory, but can be challenging in practice and
subject to much deliberation. Other examples are those measures
that depend on user-input for determining thresholds by which
the measure will be calculated. Continuity index received a lower
ranking for this reason because its calculation depends on the
designation of threshold distances by which to calculate whether
a patch is continuous (adjoining) with a similar, nearby patch.
This decision is not trivial, as the choice of threshold may have
significant effect on the outcome of the calculation of the measure.
The other variables that depend on measuring patch types are
objective as long as there is general apriori agreement regarding
what characterizes a patch and what differentiates it from other
patches.
With regard to the criteria of applicability to a large number
of diverse research sites , we found that some of the measure-
ments were formulated to fit uniquely to urban development
patterns in the specific country being researched (primarily for
the United States or Europe). Suburbs and low-density residen-
tial areas, for example, are fairly distinct to the United States
context, and may have a different meaning or even irrelevance in
other country case studies. Quantifying developable land is also
very specific to particular countries - some countries may have
statutory plans that define what is developable, while in some
countries, topography and ecological conditions or indigenous
land tenure rules may dictate what is developable. Thus, this
variable would be difficult to use in international comparative
work. Similarly, using measures that depend on a central busi-
ness district (CBD) is becoming increasingly difficult. In recent
years there has been great change in the evolution of big cities
and metropolitan regions from the classic spatial monocentric

pattern into polycentric pattern (Gar-on Yeh and Wu, 1997; Parr,
2004).
Spatial - geometric measures, on the other hand, are consid-
ered to be suitable in most places, especially when investigating
landscapes partitioned into dichotomous built and open space.
Their use appears most frequently in the interdisciplinary lit-
erature focusing on ecology and urban/regional planning (e.g.,
Leitao and Ahern, 2002; Taylor, Brown and Larsen, 2007). A
caveat to this is that the relevant type of patches to measure may
differ greatly from site to site. Because of this, comparative stud-
ies between sites using these measures maybe more difficult and
therefore we ranked them in the middle range for this criterion.
Regarding our multiple-scale applicability criterion, some
measures are excellent for a particular spatial scale, but diffi-
cult to apply or not applicable at another scale. Our ranking
system for this criterion is based on whether the given measure
is appropriate for many spatial scales (A) or only a very specific
scale (C). We suggest that the researcher must carefully select
an appropriate measure for the spatial scale under investigation
and make no assumptions with regard to the application to other
spatial scales. For example, at the scale of a single neighborhood
within a city, measuring growth in residential area is a relatively
straightforward task (as distinct from industrial or commercial
areas).
However, scaling up to the level of an entire metropolitan
area, residential area cannot always be reliably differentiated from
other forms of development using standard data sources (satellite
imagery, aerial photographs) unless researchers have access to
detailed ancillary data sets (Vogelmann et al ., 1998; Yang and Lo,
2002). This is even more relevant at broader spatial scales like
regions and countries. From our research in Israel, we find that
using suburban development as an indicator does not work at
the scale of an individual city because suburbs, defined as low
density, residential neighborhoods, generally occur outside of the
urban locality jurisdiction in satellite ''bedroom'' communities.
So in our case, using suburban versus urban population growth
rates is relevant mostly at the metropolitan or regional scale.
Using patch measures can be useful at multiple spatial scales,
though the patch types may vary depending on the scale of analy-
sis. Some patch types become difficult to measure or irrelevant to
sprawl characteristics at certain spatial scales. Contiguity between
patch types, for example, is important for regional-scale analyses