Image Processing Reference
In-Depth Information
The creation of an urban-rural dichotomy requires that
the researcher decide upon the criteria that will go into an
algorithm for assigning each place to either the urban or
rural category. The creation of an urban-rural gradient
requires that we adapt such an algorithm to tell us how
urban or how rural a place is (a “soft” classification),
rather than simply assigning it to one category or the other
(a “hard” classification). There are several issues that must
be dealt with in the creation of an index, including: (1) the spatial unit of analysis
to be used; (2) the variables to be combined in the index; and (3) how the variables
will be combined to create an index.
the creation of an
urban
-
rural
gradient requires
a knowledge of
how urban or how
rural a place is
3.3.1
What Spatial Unit of Analysis Should Be Used?
If we are able only to circumscribe some large geographic zone (e.g., the contigu-
ously built-up area in a region) then the ends of the rural-urban spectrum will be rela-
tively close to one another. On the other hand, if we are able to define the attributes
for relatively small and regular zones, such as a half-kilometer grid of land, then we
could better understand variability both between and within human settlements.
Furthermore, if we had a clearly defined spatial grid, then we could more accurately
measure change over time - to understand the process of urban change and evolution
that almost certainly has an important impact on human attitudes and behavior.
However, the preliminary set of calculations that helps to establish the utility of this
approach must of necessity be based on geographically irregular administrative
boundaries because the census data that we are using in the creation of the index are
readily available only at the level of those administrative boundaries.
3.3.2
What Variables Should Be Used to Define Urbanness?
I have suggested elsewhere (Weeks et al.
2005
) that the urban index should combine
census and survey data (to capture aspects of the social environment) with data from
remotely-sensed imagery (to capture aspects of the built environment). Let me focus
here on the latter part of the equation. The classification of an image is done at the level
of the individual picture element (pixel), but in the creation of an index of urbanness we
are less interested in each pixel than we are in the
composition
and
configuration
of all
of the pixels within a defined geographic region (read further discussions in Chapters 5
and 12). This is the realm of landscape metrics, which are quantitative indices that
describe the structure of a landscape by measuring the way in which pixels of a particu-
lar land cover type are spatially related to one another (Herold et al.
2002
; Lam and
DeCola
1993
; McGarigal et al.
2002
). The structure of a scene is inferred by calculating
indices that measure composition and configuration of the pixels within an area.
