Geoscience Reference
In-Depth Information
values for two macrolevel indicators, the Moran's index of spatial autocorrelation on
the one hand, and the share of Arab population on the other hand. The objective is to
evaluate the mechanisms introduced in the model about the residential preferences
of individuals. Thus, the calibration of the model is based at the emergent level
rather than at the level at which the interactions induce change.
Figure 4.5.
Emergence of a hierarchical system of cities simulated
with the SimPop model [SAN 97b]
While pointing out the differences between the statistical and agent approaches,
we are regularly reminded of their complementary nature. Indeed, faced with a given
spatio-temporal question, whatever it relates to (an environmental change, the
dynamics of past settlements, the social or economic evolution of regions or cities),
the statistical models allow us to show trends in the relationships between different
phenomena through the analysis of empirical data. Similarly, the computing models
allow us to explore the mechanisms that have led to these observed trends. Thus,
formalisms such as the CA and the MAS are adapted to explore the hypotheses
concerning the effects of interactions between objects, whether they are simple or
composite, and the interactions between the considered micro- and macrolevels. The
next section is dedicated to microsimulation models. They have an intermediate
position relative to the fields of statistics and agent-based modeling.
4.2. Microsimulation models
Microsimulation models are based on the characterization and follow-up in time
of the elementary entities of a society, most often individuals, but also households or
firms, with the objective of understanding global phenomena. The first models of
microsimulation came from economists and sociologists [ORC 57] who advocated
to formalize hypotheses about change at the household level in order to better
understand the changes in the American economy by major sectors of economic
