AN INTEGRATED ECOLOGICAL MODELLING AND DECISION
SUPPORT METHODOLOGY

António Pereira1,2, Pedro Duarte2, Luís Paulo Reis1

1NIAD&R - LIACC
Faculty of Engineering - University of Porto
Rua Dr. Roberto Frias, s/n 4200-465 Porto, Portugal
E-mails: {amcp, lpreis}@fe.up.pt


2CEMAS – UFP, University Fernando Pessoa
Praça 9 de Abril, 349, 4249-004 Porto, Portugal
E-mail: pduarte@ufp.pt




KEYWORDS
Ecological Modelling, Decision Support System,
Intelligent Agents.
ABSTRACT
This work presents a methodology for modelling and
simulating aquatic ecological systems and an associated
methodology and software for creating Decision Support
Systems (DSS) for this kind of environment. Both are
based on object oriented programming (OOP) and, in
the case of the DSS, on Autonomous Intelligent Agents.
The modelling software (EcoDynamo) includes several
object dynamic link libraries to simulate different
physical and biogeochemical ecosystem processes.
These libraries were designed to be possible their
linkage with different model shells, for the sake of
portability and usability. A high level communication
language (ECOLANG) was developed to allow the
communication between EcoDynamo and the agents.
The DSS uses one methodology of discrete multi-
criteria, where given a users preference structure
(elicited with the Analytic Hierarchy Process
methodology - AHP), a priority ranking of the pre-
processed scenarios is achieved. To validate the
approach, EcoDynamo was used with the DSS to
simulate different management scenarios in a coastal
lagoon at the South of Portugal – Ria Formosa. The
experiments performed showed that these tools may be
widely used by people involved in the management of
coastal areas to integrate environmental, economic and
social issues in the decision process, without an in-depth
knowledge of modelling methodologies.
INTRODUCTION
Models of aquatic ecosystems include hydrodynamic
and biogeochemical processes, such as photosynthesis,
nutrient cycling and grazing. The former are responsible
for the transport of pelagic variables across model
domain and across model boundaries. The latter are
responsible for local changes of state variables, such as
concentration of chemical constituents and biomass of
different species or groups of species.
Different modelling teams tend to adopt different
modelling tools, due to the knowledge of their
researchers and the inherited software they use. The
most common approaches are the structured
programming - e.g. the EMS Dollard (Baretta and
Ruardij 1988), COHERENS model (Luyten et al. 1999)
- or object oriented programming - e.g. EcoWin,
(Ferreira 1995), individual-based modelling (IBM)
(Breckling et al. 2006).
Models based on structured programming consist of a
main program, where some general state variables are
defined and calls are made to several sub-routines, at
each model time step. These sub-routines calculate all
processes represented in the model and the fluxes that
affect each state variable. At the end of each simulation
cycle, state variables are updated as a function of the
mentioned fluxes. In some cases, sub-routine
calculations depend on general scope variables. In these
situations, it is difficult to reuse these sub-routines from
other source codes. Moreover, depending on the
compilers used, it may be difficult to combine sub-
routines written in different source codes to build new
models (Pereira et al. 2006).
In the case of object-oriented software, it is possible to
define objects representing several groups of variables
and processes that may correspond roughly to the sub-
routines of the previous approach. The elimination of
general scope variables helps avoiding programming
errors. Furthermore, object properties, such as
reusability, inheritance and polymorphism (Meyer 1997;
Weiss 2000) make them very useful in ecological
modelling (Pereira et al. 2006). For example, one may
define a zooplankton object that has a set of parameters
and procedures to change its state variables (e.g.
biomass) by simulating relevant physiologic processes.
However, for the implementation of a particular model it
may be necessary to include several zooplankton
species, differing in the values of some physiologic
parameters, everything else being equal. In this
situation, it is not necessary to create code or variables
for each new zooplankton species. Using polymorphism,
several dynamic instances of the same object may be
created, each with their own parameter values and
corresponding behaviour. There is no need to create
more state variables, as it would be the case in a
Proceedings 21st European Conference on Modelling and Simulation
Ivan Zelinka, Zuzana Oplatková, Alessandra Orsoni ©ECMS 2007
ISBN 978-0-9553018-2-7 / ISBN 978-0-9553018-3-4 (CD)

structured dynamic model (at least one per species),
because each instance of the zooplankton object has
pointers to the corresponding state variables. Therefore,
using the same objects, it is possible to build models of
different complexity. Furthermore, if it becomes
necessary to change some calculations, create new
variables or parameters, for one of the modelled species,
a descendant object may be easily created by
overloading only the procedures that have to be
changed, everything else being inherited from the
ancestor objects.
Objects may interact between each other. For example, a
zooplankton object may graze a phytoplankton object
after inspecting its properties, such as biomass, and
“informing” phytoplankton how much will be grazed.
This is done by having methods in each object that are
responsible to show “public” properties (e.g. biomass)
and methods that are responsible to change them as
when phytoplankton is grazed by zooplankton (e.g.
Ferreira 1995).
Ecological models are frequently developed for
management purposes. The set up of a model with
existing modelling software may be a relatively complex
task, where the user must prepare several configuration
files that define several ecosystem characteristics,
parameters and initial values for the variables – this is
usually called a scenario. Model outputs consist of
results, regarding several variables that reflect, for
example, water and sediment chemical quality and
production of several species. Therefore, the usage of
such models by non-modellers, involved directly on
management issues, may be discouraged by the
complexity mentioned above. Moreover, results
obtained with different scenarios may be difficult to
order according to some quality criteria, unless some
decision support rules exist. Accordingly, the purpose of
this work is to describe a methodology allowing the
linkage of a modelling software – EcoDynamo (Pereira
et al. 2006) - through a specific communication
language – ECOLANG (Pereira et al. 2005) – with a
Decision Support System (DSS) software and to present
a practical application of this combined methodology. It
is noteworthy that this work presents a first description
of a new tool, which is the DSS software, and a first
synthesis of tools described elsewhere (e.g. for
ECOLANG see (Pereira et al. 2005); for EcoDynamo
see (Pereira et al. 2006)).
The rest of the paper is organized as follows. Section 2
describes the system conceptualization with emphasis on
our ecological modelling software - EcoDynamo.
Section 3 briefly describes ECOLANG – a high-level
communication language developed for this type of
ecological simulations. Section 4 analyses the Decision
Support System developed and is followed by a
description of the Analytic Hierarchy Process. The
paper ends with an analysis of our case study: Ria
Formosa lagoon and some conclusions and future work.
SYSTEM CONCEPTUALIZATION
EcoDynamo - Ecological Modelling Software
EcoDynamo is an example of object oriented modelling
software, built in C++ that was designed to simulate
thermodynamic, hydrodynamic and biogeochemical
processes of aquatic ecosystems. Details about
EcoDynamo may be found in (Pereira et al. 2006). One
important characteristic of this software is that its
objects are compiled as Dynamic Link Libraries (DLLs)
with an interface allowing their linkage with other
modelling software codes written in FORTRAN or C.
This allows different modelling teams to share software
for the calculation of specific processes, independently
of the programming language preferred, and enabling
bidirectional code reutilisation. The EcoDynamo shell
(Figure 1) manages the graphical user interface, the
communications between classes and the output devices,
where the simulation results are saved. Simulated
processes include:
(i) hydrodynamics of aquatic systems - water
elevations, current speeds and directions;
(ii) thermodynamics - energy balances between water
and atmosphere and water temperature;
(iii) biogeochemical - nutrient and biological species
dynamics;
(iv) anthropogenic: e.g. biomass harvesting.
Ecosystem characteristic properties are described in a
model database. EcoDynamo has an interface module
that enables communications with other programs for
external control. For example, the simulations runs can
be controlled by commands like start / stop / pause /
restart / step simulation. Simulation activity can be
monitored with the help of log files, activated before the
simulation run.


Figure 1: EcoDynamo Shell

Some of the output options provided by EcoDynamo
make use of MatLab® functions. The implemented
interface DLLs integrate graphics output features and
HDF (Hierarchical Data Format) file functionalities.