Information Technology Reference
In-Depth Information
4 Discussion and Conclusions
Although the accuracy of two CA models is only 55%, simulation model accuracy, to
some extent, depends on the complexity and stochasticity of real city and also the
availability of more detailed information. From the previous part of this research, the
accuracy of global pattern model that is based on logistic regression analysis with 10
explanatory variables is only around 70%. If more detailed data like control plan
scheme is available, more rigorous model calibration will become possible. From the
angle of spatial modelling, as criticized by other researchers, CA is not an appropriate
tool on micro scale, we need to integrate agent-based techniques e.g.
[19]
. Models of
complex systems with geographic properties, such as city and ecology systems, usu-
ally involve spatial and temporal processes, which are difficult to embed within pro-
prietary GIS. Most CA software available such as AUGH, DUEM either lack GIS
functions or do not fit specific complex city. A loose coupling strategy is still pre-
ferred, which is also adopted in this research.
We can not ignore the fact that any advanced modelling techniques including CA
must be based on the proper understanding and abstract of the system studied. The
more proper, the more accurate it is. The ability of science to understand the real
world is to a large extent dependent on knowledge constrained by the limits of our
understanding of complexity.
CA is only a simulation tool for testing user's understanding. Limited by existing
GIS theory and technique, the identification of spatial and temporal heterogeneity can
not be completed without the assistance of local knowledge as rich historical data
layers do not guarantee the improvement of model calibration. It implies that local
knowledge is an important ancillary data sources for CA modelling. During modeling,
temporal control, dynamic weighting, and manual test need more local knowledge. For
the division of temporal process, due to limited temporal resolution, local knowledge
is a key source of qualitative information.
The major purpose of CA simulation is to generate alternative scenarios for deci-
sion support in a smart growth management. Apparently, the methodology developed
here can be extended in this direction. As it is based on the soft systems thinking,
which stresses the role of users' subjectivity; Local planners' intention can be trans-
formed into spatially and temporally explicit weight values and certain parameters.
References
1.
Xie, Y. and Batty, M., Automata-based exploration of emergent urban form. Geographical
System, (1997) 83-102.
2.
Allen, P.M., Cities and regions as evolutionary, complex systems. Geographical systems,
(1997) 103-130.
