Environmental Engineering Reference
In-Depth Information
driving the simulation from its start point.
TABLE 24.3
Demographic and socioeconomic variables
considered in the stepwise regressions.
R
k
(1995
→
2000)
⎬
⎭
0
.
99
+
0
.
24
TAX
k
(1995)
−
0
.
28
INA
k
(1995)
Description
Unit
(
−
0.17) (8.58)
(
−
4.65)
=
Agricultural population
IND*
+
−
+
0
.
17
GDP
k
(1995)
0
.
14
FA
k
(1995)
0
.
25
RE
k
(4.35)
(
−
4.04)
(2.37)
Non-Agricultural population
IND
(24.6)
Total population
IND
Land area in cultivation
MU*
TAX
k
(1995)is the total tax levied,
INA
k
(1995)income in the
non-agricultural sectors,
GDP
k
(1995) gross domestic product,
FA
k
(1995) the net value of fixed total assets, and
RE
k
(1995)
the expenditure in the rural economy, all at 1995, and defined
in each township
k
.The
t
statistics (
bold
) under each variable
imply that the
β
k
parameters are all significantly different from
zero at the 5% level. The variance explained by this equation is
88 percent which is particularly high, given the aggregation and
uncertainties posed by the quality of the data. Moreover, different
explanatory variables were chosen in Equation 24.6 from those
in Equation 24.5. The main cause was the distinct policy shift
between 1990-1995 and 1995-2000 (Xie and Batty, 2005).
Equations 24.5 and 24.6 are those used in the global model
which in terms of the simulation provides the parameters deter-
mining the overall rates of change from 1990 to 1995 and 1995 to
2000 in the 122 townships. These are used to compute the rates
input to Equation 24.2 that in turn is used to factor the total
urban change into its constituent components, which is then
allocated to the cells by the lower-level agent model.
Thevaluesofthethreeparameters
μ
,
λ
,and
ψ
in Equation
24.4 are determined by using our judgment and controlled by
the land suitability
C
ik
(
t
) as the scores used in obtaining a
probability of development (conversion) from rural to urban
land use. Thus it is their relative values that are important. In
fact, during this process because land suitability is taken into
account, developers will not develop a cell if the land suitability
is less than a certain threshold
k
(
T
), that is, if
C
ik
(
t
)
<
k
(
T
).
The reasons for this initial allocation step which is different
from the subsequent steps within the macro-time period, rests
on the fact that there was a strong shift in policy between 1995
and 2000 in this region and this needs to be reflected in the
initial placement. We call this first process
random allocation
but in subsequent time periods, the master developer agents are
used to ''spawn'' additional agents which add up to the total
required in subsequent micro-time periods. These agents begin
by considering development in the cellular neighborhood of each
master agent activating a process we call
neighborhood allocation
.
It is at this point that the probabilities defined in Equation 24.4 are
considered in neighborhood order: that is, the developer agent
begins by considering cells in the immediate band of eight cells
around the master agent - in the Moore neighborhood - and if
no suitable cell is found, then the agent considers the next band
of cells, and so on until a suitable cell is located. The reason for
this somewhat convoluted process is to ensure that development
remains ''close'' to existing development which reflects the need
for connectivity in the urbanizing system.
Total output value of agriculture, forestry, animal
husbandry and fishery
MY*
Gross domestic product value
MY
Gross product value of primary and secondary
industries
MY
Gross product value of tertiary industries
MY
Total value of fixed assets investment
MY
Total income of rural economy
MY
Income in agriculture, forestry, animal husbandry
and fishery
MY
Income in non-agriculture, non-forestry, non-animal
husbandry and non-fishery
MY
Total expense in rural economy
MY
Total income of the farmers
MY
Total value of industrial assets
MY
Net value of the fixed assets
MY
The number of factories
CNT*
The number of employed people at the year end
IND
Total tax value
MY
Sold ratio of the product value
Percent
IND*: individual count of all people; MU*: 1 mu = 1/15 hectare = 1/6 acre.
MY*: million Chinese yuan CNT*: count of all factories.”
and
ψ
associated with Equation 24.4. For calibrating township
growth rate, twenty socioeconomic variables were collected in
1990, 1995, and 2000 (Table 24.3). Stepwise regression was
employed to find out the best variables that could explain urban
growth at township level. The results are given below on the basis
of best fit:
R
k
(1990
→
1995)
⎬
⎭
15
.
68
+
0
.
627
RP
k
(1990)
+
7
.
69
P
k
(1990)
(0.89)
(2.48)
(4.54)
=
(24.5)
−
5
.
02
E
k
(1990)
(
−
3.49)
RP
k
(1990) is the rural (non-urban population),
P
k
(1990) the
urban population, and
E
k
(1990) the employment (labor force)
total at 1990. The
t
statistics (
bold
) under each weight and
variable make clear that the parameters
β
k
are all significantly
different from zero at the 5 percent level. The amount of variance
explained by this equation is 72 percent which is acceptable for
24.4.4
Themodel validation
The validation procedure used in this paper is a variation of the
map-comparison method. The steps of validation are following
Search WWH ::
Custom Search