Vegetation and Disturbance - Environmental Modelling: Finding Simplicity in Complexity - page 231

(Turner and Bratton, 1987). Models of fire effects and of
resource selection by herbivores can be applied to pre-
dict the combined effects of the two factors. One of the
greatest difficulties associated with developing spatially
explicit models is the volume of data and computational
effort required to scale up from individual-level models
to landscapes. It is clear that different ecological processes
operate at different temporal and spatial scales (Levin,
1992; Legg et al ., 1998) and these can only be incorpo-
rated into ecological models through a hierarchical system
with disturbances and processes represented at the correct
scale (Forman and Godron, 1986; Chertov et al ., 1999).
One approach to this problem is described in Legg
et al . (1998) where models of vegetation response to
herbivory are constructed at three different spatial and
temporal scales: the individual level, community level
and landscape level. The dispersal of animals across the
landscape is determined by vegetation and topographic
characteristics at the landscape level. The selection by
herbivores of particular plant species, the effects of
dung and trampling on soils and seed dispersal by the
animals are modelled as community-level processes.
The response of plants to damage through allocation
of resources to roots, shoots and reproduction are
individual-level processes. These three spatial scales can
be modelled separately in a hierarchical system with each
level modelling only a sample of the area represented
at the level above, but a sufficient sample to capture the
behaviour of that level of the whole system.

scales. Engelen et al . (1995) recognized how GIS, in addi-
tion to a capacity for handling spatial information, has
intrinsic limits in representing dynamic processes. Thus,
these authors proposed a modelling system based on two
interacting components representing micro- and macro-
level processes. They reported, by way of example, an
integrated simulation of socio-economic and environ-
mental models applied at different scales (Engelen et al .,
1995). This work was a very good proof of concept but the
system was not fully accessible to users since the system
developed did not permit them to change input maps nor
the model's functions.
More recently, Boumans et al . (2001) pointed out
the need for the 'formulation of general unit models
for simulation of temporal processes' at the landscape
level, in contrast to specific models for particular habi-
tat types. They provided an example of the application
of a non-spatial ecological model within elements of a
river landscape. An interesting example of the coupling
of a biological model based on differential equations with
a spatial cellular-automaton model has been presented
recently by Aassine and El Jai (2002), who underlined
how this integrated approach is consistent with a natural
way of describing different processes. We fully agree with
their conclusion and believe that their approach needs to
be implemented to include integration of generic models
as pointed out by Boumans et al . (2001).
Another aspect to be considered is that traditional
model implementation has been done by conventional
programming languages, such as Fortran, Basic, and
C, whereas recent trends in the scientific literature
show an increasing application of generic modelling
environments. These software tools (see Costanza, 1998,
2001 for review) allow the construction of models by
graphical interfaces that do not require knowledge
of conventional programming languages because the
executable program is automatically compiled by the
system. Among these, STELLA (www.iseesystems.com/
softwares/Education/StellaSoftware.aspx) has a wide
application in the field of ecological and agricultural mod-
elling (e.g Pan and Raynal, 1995; Chivaura-Mususa et al .,
2000; Liddel, 2001). SIMULINK (www.mathworks.com)
is a supporting tool for the MATLAB package and is char-
acterized by strong computational capacity and access to
mathematical libraries. Other products to be mentioned
are POWERSIM (www.powersim.com) and ModelMaker
(www.cherwell.com). SIMILE (www.simulistics.com)
is a recent addition to the family of system dynamic
modelling environments and presents very interesting

14.3 New developments in ecological
modelling

In general, simple population models (see Renshaw,
1991 for a review) have described the competition
between species using few state variables, either without
or with poor explicit representation of spatial interac-
tions, whereas individual-based models of ecological
populations (e.g. De Angelis and Gross, 1992) have been
facing the complexity of competitive spatial interactions
among neighbours. More recently, an increasing interest
in spatial modelling has been associated with ecological
applications of geographic information systems (GIS)
and with the development of specific object oriented
methodologies

for modelling (Sequeira et al ., 1991;

Muetzelfeldt and Massheder, 2003).
The understanding of systems as complex as vegetation
or ecosystems requires an integrated analysis at different