Discussion
The results of the model runs above indicate that the
model is capable of representing land-use change in
accordance with the characteristics of complex systems
described earlier in this paper.
The results are not simple linear extrapolations of
trends: connectivity between locations and competition
between land-use types cause the outcomes of the
model to be complex patterns, typical of non-linear
systems. Setting the decision rules of the elasticity towards
conversion creates stability in the land-use pattern.
The results show that the model is sensitive to the
settings of these decision rules. Care should therefore
be taken while specifying these rules. For land-use types
that are known to have different spatial behavior different
land-use classes should be created. This can be the
case in forested areas were a lot of regrowth of secondary
vegetation is found. Primary forest can only decrease
on its present area whereas secondary vegetation
will show a more dynamic spatial behavior: some secondary
vegetation might be cleared for agriculture
while at the same time new secondary regrowth occurs
on abandoned lands or logged-over primary forest.
Similar differences can be found in rice growing systems.
A lowland rice field with permanent irrigation
facilities will show totally different dynamics and relations
with driving forces than upland rice cultivation.
Subdivision of this type of land-use classes should be
considered for appropriate modeling of the dynamics
of these land-use types.
The setting of the elasticities for conversion are now
based on expert knowledge and can be modified by
calibration of the model, if a second data set for land
use is available. A sensitivity analysis has shown that
these settings have an important influence on the resulting
land-use patterns as they are directly related to
the trajectories of change and land-use histories. This
specification needs, therefore, considerable attention.
Future research should find methods to help the definition
of these conversion elasticities, based on the
analysis of historic land-use data and/or better insights
into the decision-making process of the actors of landuse
change.
The model structure clearly represents the hierarchical
organization of land-use systems, allowing for a
continuous iteration between regional level demands
and local-level land suitabilities. In addition, driving
factors operating at spatially aggregated analysis levels
can be taken into account. In this sense, the model has
an appropriate structure to study the scalar dynamics of
land-use systems. The exact interactions and feedbacks
between scales and the causal processes underlying
these interactions are, however, still largely unknown
and are an important topic of research (Gibson and
others 2000, Root and Schneider 1995, Wilbanks and
Kates 1999). It is especially difficult to comprehend the
link between the decision-making process by the individual
actors of land-use change and the emerging
patterns of land use (Geoghegan and others 1998,
Mertens and others 2000). When the system-based approach
described in this paper is combined with actororiented
studies (e.g., Bilsborrow and Okoth Ogondo
1992) and agent-based modeling (Bousquet and others
1998, Manson 2000) it is possible to gain further understandings
in the multi-scale dynamics of the landuse
system.
The CLUE-S model is clearly different from models
solely based on an empirical analysis of land-use change
(e.g., Mertens and Lambin 1997, Pijanowski and others
2000). The advantage of this model is the explicit attention
for the functioning of the land-use system as a
whole, the capability to simulate different land-use
types at the same time and the possibility to simulate
different scenarios. Models that rely heavily upon statistical
relations between land use and driving factors
are frequently criticized for their lack of causality (Irwin
and Geoghegan 2000, Kaimowitz and Angelsen 1998,
Lambin and others 2000b). The selection of driving
factors for the CLUE-S model should, therefore, be
based on the theoretical relationships between driving
factors and land use. Only driving factors are taken into
account for which a theoretical relationship with land
use is known, in order to avoid spurious correlations.
We have chosen not to base the selection of variables
on one single theoretical framework because of the
differences in dominant processes between case-studies.
In some case-studies is will be possible to base the
selection of driving factors solely on economic theory,
but in other cases other processes are important as well.
In such situations we need to also incorporate factors
based on other theories. The use of expert knowledge
is essential, both for the determination of the dominant
processes and selection of the potential driving variables
as well as for the evaluation of the outcomes of the
regression analysis.
Conclusion
The model can easily be applied to a wide range of
study areas and land-use change situations. The main
limitation of applying the model is its incapability to
simulate land-use dynamics in areas without a land-use
change history, e.g. deforestation in a pristine forest
area. This is because the model uses empirically-derived
relations based on existing land-use patterns for the
Discussion
The results of the model runs above indicate that the
model is capable of representing land-use change in
accordance with the characteristics of complex systems
described earlier in this paper.
The results are not simple linear extrapolations of
trends: connectivity between locations and competition
between land-use types cause the outcomes of the
model to be complex patterns, typical of non-linear
systems. Setting the decision rules of the elasticity towards
conversion creates stability in the land-use pattern.
The results show that the model is sensitive to the
settings of these decision rules. Care should therefore
be taken while specifying these rules. For land-use types
that are known to have different spatial behavior different
land-use classes should be created. This can be the
case in forested areas were a lot of regrowth of secondary
vegetation is found. Primary forest can only decrease
on its present area whereas secondary vegetation
will show a more dynamic spatial behavior: some secondary
vegetation might be cleared for agriculture
while at the same time new secondary regrowth occurs
on abandoned lands or logged-over primary forest.
Similar differences can be found in rice growing systems.
A lowland rice field with permanent irrigation
facilities will show totally different dynamics and relations
with driving forces than upland rice cultivation.
Subdivision of this type of land-use classes should be
considered for appropriate modeling of the dynamics
of these land-use types.
The setting of the elasticities for conversion are now
based on expert knowledge and can be modified by
calibration of the model, if a second data set for land
use is available. A sensitivity analysis has shown that
these settings have an important influence on the resulting
land-use patterns as they are directly related to
the trajectories of change and land-use histories. This
specification needs, therefore, considerable attention.
Future research should find methods to help the definition
of these conversion elasticities, based on the
analysis of historic land-use data and/or better insights
into the decision-making process of the actors of landuse
change.
The model structure clearly represents the hierarchical
organization of land-use systems, allowing for a
continuous iteration between regional level demands
and local-level land suitabilities. In addition, driving
factors operating at spatially aggregated analysis levels
can be taken into account. In this sense, the model has
an appropriate structure to study the scalar dynamics of
land-use systems. The exact interactions and feedbacks
between scales and the causal processes underlying
these interactions are, however, still largely unknown
and are an important topic of research (Gibson and
others 2000, Root and Schneider 1995, Wilbanks and
Kates 1999). It is especially difficult to comprehend the
link between the decision-making process by the individual
actors of land-use change and the emerging
patterns of land use (Geoghegan and others 1998,
Mertens and others 2000). When the system-based approach
described in this paper is combined with actororiented
studies (e.g., Bilsborrow and Okoth Ogondo
1992) and agent-based modeling (Bousquet and others
1998, Manson 2000) it is possible to gain further understandings
in the multi-scale dynamics of the landuse
system.
The CLUE-S model is clearly different from models
solely based on an empirical analysis of land-use change
(e.g., Mertens and Lambin 1997, Pijanowski and others
2000). The advantage of this model is the explicit attention
for the functioning of the land-use system as a
whole, the capability to simulate different land-use
types at the same time and the possibility to simulate
different scenarios. Models that rely heavily upon statistical
relations between land use and driving factors
are frequently criticized for their lack of causality (Irwin
and Geoghegan 2000, Kaimowitz and Angelsen 1998,
Lambin and others 2000b). The selection of driving
factors for the CLUE-S model should, therefore, be
based on the theoretical relationships between driving
factors and land use. Only driving factors are taken into
account for which a theoretical relationship with land
use is known, in order to avoid spurious correlations.
We have chosen not to base the selection of variables
on one single theoretical framework because of the
differences in dominant processes between case-studies.
In some case-studies is will be possible to base the
selection of driving factors solely on economic theory,
but in other cases other processes are important as well.
In such situations we need to also incorporate factors
based on other theories. The use of expert knowledge
is essential, both for the determination of the dominant
processes and selection of the potential driving variables
as well as for the evaluation of the outcomes of the
regression analysis.
Conclusion
The model can easily be applied to a wide range of
study areas and land-use change situations. The main
limitation of applying the model is its incapability to
simulate land-use dynamics in areas without a land-use
change history, e.g. deforestation in a pristine forest
area. This is because the model uses empirically-derived
relations based on existing land-use patterns for the
การแปล กรุณารอสักครู่..