Geoscience Reference
In-Depth Information
9.5 RECENT TRENDS IN AB GEOGRAPHIC MODELLING
During the last decade, the popularity of AB modelling has grown steadily, and more and more
geographic phenomena are being investigated in this way. This includes recent topics (Marceau
and Benenson, 2011; Helbing, 2012; Heppenstall et al., 2012; Zhang, 2012) and numerous papers.
References to these papers, organised by major modelling topics, are provided in Appendix 9A.
9.6 GUIDANCE FOR STUDENTS OF AB MODELLING
The objective of geographic modelling is to understand, explain and forecast geographic phenom-
ena. Usually, there are no reasons to prefer a certain model to another one since most models
deal with different aspects of the system. This section considers operational problems that I have
encountered and resolved in my AB modelling exercises as well as operational problems raised by
my students. I focus on problems encountered at the second stage of development, when you have
already built some model prototype, formulated an initial view of the system and now aim to inves-
tigate the model in depth.
9.6.1 B
efore
S
tarting
d
eVeloPMent
of
the
l
ong
-l
iVing
V
erSion
of
y
our
M
odel
Think in terms of object-oriented programming
: An object-oriented view of the system is nec-
essary for proper software architecture and guarantees ease of modification to the never-
ending changes in the behavioural rules of the agents.
Think in terms of an entity-relationship database model and GIS
: A database/GIS approach
to information management enables loose coupling between the model and the data and
makes it easy to substitute one dataset with the other. In geographic modelling, an entity-
relationship model works hand in hand with object-oriented programming and facilitates
the location of agents and objects by relating them to other objects.
Specify the goal of your modelling exercise
: The world around us is infinitely complex but you
are interested in very specific aspects of this complexity. Formulate these aspects before
you build the model. Plan your model experiments before you start development. During
the development, you will see numerous ways of defining every piece of the model. Your
Occam's razor should be ready.
9.6.2 w
hen
d
eVeloPing
and
c
oding
the
S
iMulation
Start with the minimal model
: Start with the minimal version of the model that you are able
to specify and code quickly, and study it in depth. Only then decide on the level of realism
and the amount of the details that you would incorporate and investigate in the future.
Define agents, objects and relationships before you define behaviours
: A database view
of the AB model enables you to define the model agents, objects and the relationships
between them before deciding on their attributes and behaviours. Do not forget that you
can locate geographic objects by relationship. Therefore, decide which of the agents and
objects should be located explicitly and become geographic features and which should be
located implicitly, using relationships. The fewer geographic features the better; GIS layers
demand essentially more work than just tables.
Programming style 1
: Development and debugging of the time-driven model is always sim-
pler than that of the event-driven so start with the time-driven version. Event-driven model-
ling demands essential programming experience.
Programming style 2
: Start with sequential (asynchronous) updating and do not forget to
randomise the order of considering agents at every time step. If you deal with the agents
of two or more types, be very careful regarding the order of updating of the agents of dif-
ferent types. Consider several options, starting from the random mixing of the agent lists and
