Environmental Engineering Reference
In-Depth Information
policy agents and the local developer agents. The policy agents,
including town-village governments, municipality governments,
and external developers such as investors from abroad and other
China municipalities as well as national policy-making organiza-
tions, are acting at the macro level. The developer agents are the
collection of individual entrepreneurs, small corporations, town
and village owned enterprises, and privately operated businesses
at the micro level. Therefore, SUGAM is specified at two levels.
The township policy agents who are making decisions and poli-
cies at townships and over higher spatial units are functioning
at the global level to determine the rates of growth in each of
the townships measured by changes in households which can be
converted into developable units. The rate of change,
R
k
(
T
)is
defined as
development, subject to the suitability of the land in question as
reflected in the measure
C
ik
(
t
). In principle, what each agent is
doing is converting the land in question to an urban use, to
P
ik
(
t
)
by maximizing
ρ
ik
(
t
) subject to the constraint posed by the land
suitability
C
ik
(
t
).
An agent's behavior of maximizing development probability
is realized through intelligent search in two levels of spaces. The
first level of space is a township, which is ranked by the agent
according to the policy (township competition) index,
S
k
(
t
). A
higher rank of the township competition index allows the agent
to choose a ''maximized'' probability of development, which
means that the agent can break the limitation of growth rate (the
rate of change), converting more open or agricultural land to
urban uses in that township. This also indicates that townships
with lower policy indices will not be able to keep growth rates
that were determined by their socioeconomic factors. Thus, a
competition among townships is generated. The second search
space is an extended neighborhood, which is taken in this case
study as a neighborhood of 20
X
k
(
T
),
X
k
(
T
),
...
}
R
k
(
T
)
=
f
{
(24.1)
where
X
k
(
T
),
=
1, 2,
...
,
L
are socioeconomic drivers associ-
ated with economic development and regional policy appropriate
to the township level. Equation 24.1 is in fact the basis for the
estimation of the importance of exogenous variables to the rates
of change fitted using linear regression methods in a later section
with these rates determining the amount of growth over the
macro-time period
T
. To generate a total growth over a period,
they are applied directly to the total households (as developable
units) at the initial time
×
×
2 km) surrounding
each economic center. A developer agent will look first in this
neighborhood for conjunctions of high accessibilities to both
economic centers,
E
ik
(
t
) and transportation facilities,
T
ik
(
t
). The
developer agent will assign maximized probabilities of develop-
ment to these conjunctions as new sources of development. The
agent will then look at high accessibilities to either economic cen-
ters or transportation facilities, but assign these locations much
lower development probabilities in comparison with the former.
The implementation of maximizing development probability
in this case study is very close to the notion of solving agents'
multi-criteria evaluation problem as an optimization problem
(Bennett and Tang, 2006; Manson, 2006; Xiao, Bennett and
Armstrong, 2007). However, the focus of this case study is
on township competitiveness, location advantage, and easy-to-
implement in simulation. Moreover, it is worth pointing out that
the maximization (or optimization) decision processes taken by
the developer agents in this model distinguish themselves from
the traditional cellular automata approaches. These processes
also provide heuristic methods to assign values to the three
parameters
μ
,
λ
,and
ψ
in equation 4.
20 cells (2
P
k
(
T
+
1)
=
[1
+
R
k
(
T
)]
P
k
(
T
)
.
(24.2)
We can then factor this rate into a rate per unit time period
t
=
[
t
+
1]
−
[
t
] by discounting the cumulative rate as
P
k
(
T
1
τ
+
1)
r
k
(
t
)
=
=
1
+
r
k
(
t
)
.
(24.3)
P
k
(
T
)
When applied cumulatively to the population
P
k
(
t
)
r
k
(
t
)
P
k
(
T
),
Equation 24.3 updates the totals at each time period
t
to meet
the constraint that
P
k
(
t
=
+
τ
)
=
+
P
k
(
T
1). In each macro-time
period
T
, the total change
P
k
=
P
k
(
T
)isbroken
into its finer temporal units using Equation 24.3 and each
subtotal
P
k
(
t
),
P
k
(
t
+
1),
...
,
P
k
(
t
+
τ
) forms the control
for the detailed urban development process at the cellular level.
At the micro level, the local developer agents take into consid-
eration accessibilities (to economic center and to transportation),
land cost reflected through suitability, and growth management
policies to allocate future growth. Land suitability in the fine cell
i
in township
k
is defined as
C
ik
(
t
), accessibility to economic
centers as
E
ik
(
t
), and accessibility to transportation facilities as
T
ik
(
t
). The interaction or feedback between the local developer
agents and the township policy agents is realized through a policy
index
S
k
(
t
), which is in effect a ''Township Competition Index''
derived from the rate of change in
k
,
R
k
(
T
),
P
k
(
T
+
1)
−
24.4.2
Themodel construction
The SUGAMmodel is implemented in the open source modeling
language
RePast 3
(Collier, Howe, and North 2003 [Online]
Available at: http://repast.sourceforge.net [accessed19November
2010]). One thing that needs to be pointed out is that
Repast
is using a different time concept in its simulation, which is
referenced as ticks. These ticks do not match the real times
t
and
T
for they are essentially used to track the movement of agents
across the space as they search for suitable cells to transform and
as such, reflect the various iterations that are used to achieve the
control totals from the global level.
∀
i
,
t
.Thislinksthe
effects of accessibility and suitability with respect to the growth
management and economic policies set at the township level
with the index being set in proportion to the rate of growth of
each township (Xie, Batty and Zhao, 2007; Xie and Batty, 2005).
These factors are used by the developer agents to determine a
probability for development
ρ
ik
(
t
) which is a form of utility
given as
24.4.3
Themodel calibration
ρ
ik
(
t
)
=
μT
ik
(
t
)
+
λE
ik
(
t
)
+
ψS
ik
(
t
)
.
(24.4)
The model calibration is done in two steps: first calibrating
urban growth rate in township
k
; and second calibrating the
probability of development in terms of the local parameters
μ
,
λ
,
In general, land is converted to urban uses by the developer agent
j
who for each cell
i
in township
k
evaluates the probability of
Search WWH ::
Custom Search