Vegetation and Disturbance - Environmental Modelling: Finding Simplicity in Complexity - page 232

enhanced capabilities that are well-suited for ecological
modelling work (Muetzelfeldt and Massheder, 2003).
A new software tool named SIMARC (www.ecoap.
unina.it) was recently developed to interface models
created by SIMILE with the ArcView (ESRI Inc.) GIS
environment. The tool enables users to link a model
to input data from the GIS database. This essentially
amounts to running a model in each polygon element of
the ArcView shapefile and providing the ability to create
new GIS layers from any selected model output variable.
Considering the intrinsic power of SIMILE for ecological
modelling, its combination with the spatial analysis power
of a GIS has great potential for many applications.
The system described in the following modelling
exercise illustrates a step in the direction outlined above.
We made use of two modelling environments: the SIM-
ILE system and a raster-based spatial modelling system
named 5D (Mazzoleni et al ., 2006). The former system
has been used to model plant community-level processes,
whereas the latter system handles spatial processes at
the landscape level. An additional software tool named
LSD (Landlord Simile Dynamic link) allows the dynamic
integration of models produced by the two systems.
The advantage of using a new dedicated spatial tool
instead of an established GIS package lies in its enhanced
modelling capabilities, i.e. the possibility of making use
of temporal simulations at both local and spatial scales
in a highly integrated way. The graphical interfaces of
the integrated system are quite simple and user friendly
and allow the flexible creation of different modelling
workspaces of increasing complexity in a transparent and
modular way.
This line of reasoning refers to a major school
of thoughts in ecology known, since the 1990s, as
'Individual-Based Ecology' (IBE). The related modelling
paradigm - Individual-Based Modelling (IBM) - was
formalized as 'a bottom-up approach, which starts with
the parts (i.e. individuals) of a system (i.e. population)
and then tries to understand how the system's properties
emerge from the interaction among these parts' (Grimm,
1999). The reasons to choose either the state-variable
approaches or individual-based modelling can be differ-
ent (Schieritz and Milling, 2003). In short, we can summa-
rize that IBM is useful when individual-scale processes are
important to explain the behaviour of a system accurately.
The structure of IBMs feels closer to what can be observed
in reality, and, in many cases, IBMs are capable of simulat-
ing population-scale phenomena in a more understand-
able manner through a limited set of rules dictating the

life of individuals. So building an IBM consists of isolating
the major algorithmic rules that all individuals follow,
and also from defining variable properties that will intro-
duce the heterogeneity between individuals that is often
necessary to reproduce the dynamics of the entire system.
A striking example can be observed in the ecological
applications of the very basic 'Boids' model (Reynolds,
1987) to simulate the spatial behaviour of fish schools
(Huth and Wissel, 1992). Tractability issues can rises
with IBM compared to mathematical models as the
number of individuals to simulate as well as the nature of
their interactions can rapidly increase the computational
complexity of models (DeAngelis et al ., 1990: 585). As
a young approach, IBM is often poorly used (Grimm,
1999) because of the absence of a strict methodology of
application (Grimm and Railsback, 2005: 17). Readers
interested in a framework to design, test and analyse IB
models should refer to the concept of 'Pattern-Oriented
Modeling' (Grimm et al ., 2005) (see also Chapter 13).
Even if conversions from existing equation-based
models to individual-based models have been described
(Borshchev and Filippov, 2004), they do not represent
general procedures that should be followed to build
individual-based models. Indeed, for this purpose, it is
necessary to rethink how the whole system works from the
perspective of an individual. The system's behaviour is to
be obtained as a product of individual interactions. Tech-
nically, in an IBM, individuals are autonomous entities
conforming to the same set of rules, but each potentially
in a different state defined by a collection of internal
variables. In our study case, we can imagine that the life
of a plant can be reduced to three basic processes growth,
dispersal and death. To handle these at the individual-
scale, it seems appropriate to introduce stochasticity. If
we pick randomly two plants inside of a population and
monitor their lives, we will probably notice that their
deaths cannot be calculated deterministically nor will
they happen simultaneously. This unpredictability is also
the case for the plants' dispersal of seeds. Another axis of
improvement will consist in allowing for a more accurate
representation of ecosystem components through the
combined use of complementary modelling approaches
(Vincenot et al ., 2011).
Future developments in ecological modelling will
include enhanced computing power through grid
computing and parallelization techniques. Moreover,
the use of web-based access to modelling servers will
make the development of much more complex ecological
models and simulation applications feasible.