Image Processing Reference
In-Depth Information
Bristol region; this distinction was based on percentage of commuters traveling
daily towards the core urban area. The core urban area is identified differently for
each of the case cities, on the basis of national classification methods while the first
and second rings (suburb and hinterland) consist of zones where more or less than
40% of commuters' trips are directed towards the core area.
The second zoning systems, illustrated for all six cities in Fig.
2.3
, consist of
sub-zones representing the smallest statistical unit for which consistent and compa-
rable data are available. In the UK context, these sub-zones are based on wards and
parishes and aggregations thereof.
The generalized shift-share method computes for each small sub-zone the growth
rate of each variable (population, employment and commuting distances). In a second
step the deviation of each small sub-zone's growth rate from the regional growth rate
is also computed. In the SCATTER project the shift-share method is used to identify
the role played by the two growth components, the overall growth rate, l
()
a
and a
t
time depending factor g ()
i
representing zonal deviations from the average growth
path, in the actual growth of each small zone.
The analysis is carried out in three steps:
a
1. Estimation of the average growth rate as
a
1
Xt t
(
+
D
)
l
a
( )
t
=
ln
(2.1)
D
t Xt
a
()
where
XtXt D
represent the total volume of the variable over the
entire urban region at times
t
and
+
a
( ),
a
(
+
)
t
D
t
respectively.
2. Estimation of the zonal deviations of the average growth path as:
a
1
Xt t
(
+
D
)
(2.2)
a
i
a
g
()
t
=
ln
−
l
()
t
i
D
t
Xt
a
i
()
3. The estimated parameters
i
g
may exhibit some noisy structure,
due to possible data uncertainties. Therefore appropriate data ilters are applied
to the mean growth rates and the deviations of the growth rates in order to smooth
out such disturbances.
a
and
a
l
()
t
()
Tables
2.3
and
2.4
show the values of the parameters for population and employ-
ment growth rate in the six case cities. The values are smoothed using a Gaussian
moving average procedure.
The quantitative analysis has also applied more traditional spatial statistical
measures, such as the indicators of local and global spatial autocorrelation. For a
value of a particular variable (e.g., population density), indicators of spatial autocor-
relation make it possible to estimate whether a zone
i
is surrounded by zones exhibit-
ing very similar or very dissimilar values, or is surrounded by a heterogeneous,
patchy pattern of similar and dissimilar values. To identify local spatiotemporal pat-
tern of variables the correlations between nearby values of the statistics are derived
and verified by simulations. There are many possibilities to test for the existence of
such pattern. One of the most popular is Moran's I statistic, which is used to test the
