1 Spatial Modelling of the Terrestrial Environment - Georeferencial

210 Spatial Modelling of the Terrestrial Environment
level, the null hypothesis cannot be rejected and the spatial distribution of x in terms of
n is considered to be random. If I < 0.0 and z I is significant, on the other hand, it may
be concluded that there is no significant spatial autocorrelation of n in terms of x (i.e., x
exhibits significant spatial dispersion). Finally, if I > 0.0 and z I is significant, n exhibits
significant spatial autocorrelation in terms of x (i.e., x exhibits significant spatial clustering,
such that similar values of x are likely to be found close to one another). In the analysis
that follows, a standard 1 weighting and a significance level of 0.001 are used.
d
10.3 Study Area and Data Pre-Processing
The town of Orpington in the London Borough of Bromley is used to evaluate the builtform
connectivity approach outlined above. Orpington contains numerous different types
of urban land use within a relatively small area. These include several different residential
districts, ranging in age from turn-of-the-century terraced houses (1900s), through
semi-detached developments built during the 1920s/1930s and 3–4-storey blocks of flats
constructed in the 1960s/1970s, to modern townhouses (1980s) and detached 3–4 bedroom
residences (1990s). There are also several major schools in the area, as well as a large hospital
complex. The existence of a 1:25 000 scale land-use map of the town and its environs,
generated for a previous study (Barnsley and Barr, 1996), allows an objective evaluation of
the analyses performed here.
The primary dataset employed in this study isa1mspatial resolution land-cover map that
has been generated from Ordnance Survey (OS) 1:1250 scale Land-Line.93+ digital map
data (Figure 10.5). This covers an area of 2 km × 2 km. The original Land-Line.93+ data
were provided by the OS as sixteen 500 m × 500 m tiles in standard National Transfer
Format (NTF). Each tile contains vector data for up to 32 feature codes. These are processed
using ARC/INFO to select the feature codes of interest from each tile, to mosaic the resultant
data together into a single 2 km × 2 km coverage, and generate a topologically structured
vector dataset. Each of the polygons in the resultant coverage is assigned a land-cover label
drawn from the set ROAD, BUILT, TREE (representing areas of continuous, extensive tree
cover), OPEN SPACE (representing all other extensive areas of non-wooded open space) and
WATER (representing all open water bodies, such as lakes and rivers). This coverage is then
transferred to the Grid module of ARC/INFO and resampled to a spatial resolution of 1 m
to generate a raster land-cover map, before being exported from ARC/INFO and converted
to the image format used by SAMS.
Based on the data shown in Figure 10.5, an XRAG is compiled for the spatial, topological
relation containment and the morphological properties area, compactness ( perimeter2 )
4πarea
and geographical centroid. Containment is selected as it is the key spatial relation used by
Kruger (1979a) to recognize built-form constellations. The three morphological properties,
outlined above, provide a means by which to evaluate region (building) size and
shape.
10.4 Recognition of Built-Form Constellations
To identify the built-form constellations present in Figure 10.5, we must first analyse the
containment relation for the corresponding node set in the XRAG. Figure 10.6 shows a