Information Technology Reference
In-Depth Information
the driving forces of land distribution at the grid scale and allocation of changing land
use areas [4]. Analyzing both the driving forces behind land distribution and the spatial
allocation of land use change is the important component of the DLS model.
Mechanism analysis of the DLS model aims to estimate the statistical relationship
between the pattern changes of land uses and its driving factors. Theoretically,
mechanistic analysis provides a reaction function of each land use type. Corresponding
weights are given to all driving factors according to principles that can be assumed to
be fixed for a short period, but driving factors change over time. With the reaction
function determined, reasons for differences between simulated and observed distri-
bution of land use types can be summarized as follows: values of some driving factors
have changed, such as population growth or temperature; competition exists among
different land use types; and restrictions occur between local historic conditions and
current demand. Driving factors include natural environmental conditions, socioeco-
nomic factors and land use management policies, all of which are closely linked to
pattern changes of the land uses. Driving factors behind land use pattern can be ana-
lyzed with the explanatory linear model of land use pattern (ELMLUP) and explanatory
nonlinear model of land use pattern (ENMLUP) built at the pixel level.
3 Explanatory Linear Model
Linear regression is the model most commonly used in researching driving mechanisms
of land use patterns as it explores driving factors at wide ranges and with high spatial
resolution. The explanatory linear model of land use pattern at the pixel level,
or
ELMLUP, contains a demanding and a distribution module. The target variable of the
ELMLUP is the proportion of the area of land use type
k
(
k
=1, 2…
M
) in grid
i
(
i
=1, 2…
n
) at time
t
abbreviated as
Q
i
kt
. The explanatory variable of the model is a covariant
vector of driving factors composed of a series of natural environmental conditions
and
socioeconomic factors that are tightly related to the pattern changes of land uses (with a
significance level of 5%):
t
t
t
t
t
T
(1)
Xx
=
(
,
x
,...,
x
,...,
x
) .
i
i
12
i
il
iL
To measure the impact of spatially autocorrelated land use types, several variables,
including
ˆ
ˆ
kt
and
X
,
are defined in the ELMLUP. The quantitative relationship be-
Q
ˆ
i
ˆ
kt
tween
and
is
developed through the following multiple linear regression
X
Q
model:
ˆ
ˆ
kt
k
k
t
(2)
QaaX
=+
.
i
0
i
aaaa
k
=
(
k
,
k
,...,
k
L
)
is the coefficient matrix of
ˆ
i
a
is a constant
k
where
,
and
X
12
0
ˆ
()
term. Regarding
grid
i
at time
t
, the result
reg Q
estimated by the model is naturally
employed to reflect the average proportion of area of land use type
k
under natural and
socioeconomic conditions
ˆ
kt
i
X
.
Search WWH ::
Custom Search