A Simulation Model for Integrating Multidisciplinary Knowledge in

Natural Sciences

Heuristic and Application to Wild Rodent Studies

Jean Le Fur

, Pape A. Mboup

and Moussa Sall

Institut de Recherche pour le Développement (IRD),

Centre de Biologie pour la Gestion des Populations (CBGP), Campus Baillarguet, CS 30016,

F-34988 Montferrier-sur-Lez, France

Dépt. de mathématiques et d’informatique, Univ. C. A. Diop/18522 Dakar, Senegal

Dépt. Informatique, Univ. G. Berger/Saint-Louis, Sénégal and lab. IRD-BIOPASS, Campus Bel-Air, Dakar, Senegal

Keywords: Multidisciplinary Knowledge, Heuristic, Object-oriented Modelling, Case Study, Agent-based Model,

Simulation Model, Robustness, Rodents.

Abstract: Knowledge about rodents has been obtained at multiple observation scales and covers many functional

levels, from DNA to ecosystems. We have developed an object/agent-oriented simulation model to

represent these elements and phenomena in an integrated manner. Given the diversity of domains, items,

processes and scales to be considered, we used an incremental approach to model development, with

contrasting case studies successively represented and connected within the same model. Each study enriches

the model and benefits from previous developments. The results emerging from this compilation are

reflected into a shared class tree composed of three broad domains of variability: (i) concrete agents, (ii)

specific genomes that instantiate the characteristics of each type of agent and (iii) agent containers that can

be described on several scales. The classification that appears is characterized by the triviality of the

categories obtained. It resembles natural partitioning, which lends it certain robustness, facilitating its

extension. The essentially transitory nature of the construction is discussed, together with its dependence on

the formalisms used. The current model, built on a combination of eight case studies, appears to be

sufficiently robust to address new aspects and to serve as a basis for the further construction of an integrated

view of the complex dynamics associated here with rodents.

1 INTRODUCTION

Disciplinary specialization is, logically, seen as a

necessity for deciphering the increasingly specific

facets of natural systems (Newell, 2001). In the

context of research on small rodents, for example,

knowledge is obtained at all levels of life, from

genes, through cells, organisms, populations and

communities, and up to ecosystem level. Each level

can itself be broken down into numerous

departments corresponding to individual disciplines.

This discipline-based diversification is observed in

many domains of research on rodents, from

biogeography to taxonomy, morphometry or

epidemiology.

The increasingly acute compartmentalisation of

approaches can sometimes become problematic

(Rouxel, 2002), particularly if it leads to a

disconnection from concrete situations (Burger and

Kamber, 2003) where the complex dynamics

displayed integrate several levels and scales with a

diversity of elements and relationships (Bar Yam,

1997). The development of an integrative model or

at least of a suitable approach for achieving such a

model, would facilitate characterization and

comprehension of the complex systems studied

(Newell, 2001). We focused on this issue of

integrated representation of disciplinary knowledge

in the ‘restricted’ domain of the biology of small

rodents and their parasites with a view to improve

our understanding of the phenomena observed in

situ.

Object-oriented paradigms have long been

considered a powerful approach for the modelling of

complex systems (Clapham and Crosby, 1992).

When associated with simulation models and ‘agent’

340

Fur, J., Mboup, P. and Sall, M.

A Simulation Model for Integrating Multidisciplinary Knowledge in Natural Sciences - Heuristic and Application to Wild Rodent Studies.

DOI: 10.5220/0006441803400347

In Proceedings of the 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2017), pages 340-347

ISBN: 978-989-758-265-3

formalism (Casti, 1997), this approach constitutes a

flexible and relevant tool (Goldspink, 2002) that has

already been used for the formal integration of

disciplinary knowledge.

The approach adopted is based on the empirical

construction of a model by the successive

aggregation of sub-models of concrete case studies.

This heuristic aims to develop a joint formalization

of the diversity of fields, questions and knowledge

worked separately by each discipline, while

remaining limited to the components strictly

necessary for the modelling of the case studies. This

approach aims to use object-oriented programming

generalisation to reveal (“let emerge”) a generic

architecture accounting for the composite knowledge

relating to the rodents populations dynamics. Given

the hypothetical outcome, this approach is

implemented as a heuristic, to facilitate progress

towards a better simulation of events or processes

really occurring in the field.

This article presents the heuristic we have used

and the case studies underlying the development of

the model. We then present two examples and the

overall class hierarchy obtained before discussing

the results in the light of our objectives.

2 DESIGN OF THE MODELLING

ARCHITECTURE

2.1 Presentation of the Heuristic

The approach adopted is based on the incremental

accretion of case studies specific to each research

discipline (ecology, genetics, geography, etc.). Each

time a new study considered, it is selected on the

basis of displaying as large a contrast as possible

with the previous studies, so as to capture a diversity

of components and processes, thereby testing the

robustness of the approach. For each particular case

study, modelling represents only the relevant

elements. The general model ‘grows’ through

progressive aggregation of the concepts resulting

from the modelling of each case study and by

seeking the lowest common denominator, by

identifying similarities and then factorising them

(see below). The resulting model must be

systematically back-compatible with all the previous

case studies represented, and capable of integrating

others.

For each case study, modelling is carried out in

collaboration with the researchers in biology,

ecology, geography, and other relevant disciplines

responsible for carrying out the study concerned.

Each case study is approached as an agent-based

model, in the ‘usual way’ (e.g., Macal and North,

2006), by reifying data and knowledge with objects,

environments and processes required by the

particular context of each of the studies considered.

At each addition of a formalised study, new types of

objects may need to be implemented, but some of

the existing elements can be reused.

During the transition between studies, there is a

return to the 'generic' model, during which the code

is essentially revised through three types of

operation: (i) if common structures emerge, classes

are generalised by abstraction and interfacing.

Shared methods are factorised, through

polymorphism and inheritance. Packages may also

be defined or rearranged; (ii) the method names are

then refactored, to clarify the logic constructed, and

time and space unit conversions are checked; (iii)

tuning or encapsulation relating to new added

features is performed, to ensure the back-

compatibility of the code, which underlies the

validity and robustness of the approach.

2.2 Simulation Environment

The architecture presented was developed in Java,

with the Repast Simphony platform (North et al.,

2005) and some of its primitives (schedulers,

contexts, continuous space, and graphical interface).

The representation of studies and the continual

overhaul of the model also apply to its three

dependencies, which benefit from this approach of

continuous improvement and are also the subject of

methodological studies relating to the taking into

account of interdisciplinary knowledge.

A ‘data’ package is responsible for collecting

externalities to the case studies. A class system is

built up progressively to process data (maps, tables

etc.), constants, parameters and seeds of random

number generators.

An 'observation' package aims to ensure

multimodal rendering (display, data storage,

connection to other software) of the functioning of

the system. This module makes use of the principle

of epiphytic or recommender systems (Richard and

Tchounikine, 2004) based on a set of specialized

observers/inspectors that simulate the data collection

(surveys, trapping, counting etc.) actually carried out

in the field by scientists and compute output

indicators.

Finally, each case study aggregated in the model

is characterised, distinguished and implemented

autonomously by a 'protocol' that realises the

A Simulation Model for Integrating Multidisciplinary Knowledge in Natural Sciences - Heuristic and Application to Wild Rodent Studies

341

‘world’ corresponding to the particular question

asked by the researcher (choice of scales, objects

and agents, output indicators selected, input data

sources, calendar and space manager).

2.3 Presentation of the Formalised

Case Studies

Rodents are the subject of many scientific studies,

because they are reservoirs of infectious diseases

(Taylor et al., 2008) or involved in the degradation

of goods and crops (Skonhoft et al., 2006). Rodents

have many characteristics and lifestyles. They may

be solitary, live in colonies, dig burrows or make

nests, or associate with humans. Rodent research is

extremely broad, ranging from laboratory

experiments to in situ studies, at both small and

large scales.

The boundaries of the study are defined by the

work carried out by a multidisciplinary team of

researchers working in the domain (the ‘Rodent’s

group’ of the French Biology Center for Population

Management, UMR 022 INRA-IRD-Cirad-SupAgro,

France). We used eight selected case studies for

model development and tests of the robustness of the

approach. These case studies were selected on the

basis of their being as different as possible, in terms

of the questions tackled and their temporal and

spatial extents, ranging from laboratory cages to the

eco-climatic zone (Table 1).

Table 1: Case studies from particular disciplines

successfully modelled (chronological classification) and

diversity of the corresponding scales.

Main Features

Space extent

(m)

Time extent

(year)

1. Common voles in

agricultural landscapes

566

2. Cage hybridisation of

African rodents

3. Catch-mark-recatch

experiment in an African

reserve

441

4&5. Epidemiology and

transportation of black rats

817,810

471,432

100

6. House mouse invasion in

Senegal

681,120

7. household habitats

exploration by mice

138

8. Sahel invasion by a gerbil

species

1,148,831+

zooms

In the first study, the notions of landscape,

agricultural operations, crop rotation, rodent

burrowing, reproductive and social behaviour were

included. In the second study, the chromosomal and

gene levels were represented, as well as the cellular

processes of fertilisation (meiosis, fusion, etc.). In

the third case study, the trapping devices and their

manipulation have been formalised. The next three

studies were devoted to the transportation of

commensal (associated with humans) rodents at

several spatial and temporal scales. Road networks,

transport vehicles, human carriers, cities, markets,

and economic zones were therefore added to the

model. The seventh study led to the introduction of

daily activity rhythms into the model.

The final study led to formalization within the model

of vegetation, its growth and the impact of rodents

on its dynamics, as well as the integration of data

derived from remote sensing or taking the effects of

predatory owls into account.

The model obtained and described hereafter is

constituted of 81 classes, 132 attributes, 143

relations, and 562 operations. As we focus here on

abstracting and interfacing, we have chosen to prefix

interfaces with ‘I_’, abstract classes with ‘A_’ and

standard classes with ‘C_’ in the source code and the

article. Moreover, to ensure the integrity of the

multiple scales and units (more than 30 are used in

the current version) dealt with in all case studies, we

have suffixed most methods or properties with

‘_Uxxx’ where xxx is the unit (e.g., meter, day, cell,

tick, gramPerDay) of the method or property.

3 RESULTS

The model makes it possible to represent contrasting

simulations, addressing diverse aspects of dynamics,

for several rodent species, over various spatial and

temporal scales, within different simulation contexts.

For instance, in the example shown in Figure 1a,

only aspects relating to the cross-breeding of rodents

were considered, whereas, in the example shown in

Figure 1b, studies of the diffusion of black rats over

a century required the simulation of a rich historical

and geographical environment, including all forms

of commercial transport in the country concerned.

In this study, the main result lies in the model’s

structure that emerges from the compilation of case

studies. This model has three necessary and

sufficient domains of diversity: simulated concrete

entities, genomes associated with living organisms

and different types of substrate in which objects and

agents can be located.

SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

342

Figure 1: Displays of simulations for two extreme case

studies from the eight considered in this work (table 1).

a) case n°2: simulation of an animal room protocol for the

crossing of rodents from sibling species in different

configurations. The challenge here is to study the barriers

to fertilisation linked to chromosomal differences between

the species. b) case n°4: simulation of the colonisation of

Senegal (West Africa) by the black rat, over a century.

The biological, historical and geographical aspects relate

to the likelihood of invasion through the transport of

rodents in commercial vehicles (trucks, trains, boats).

For each of these domains, various classes have

appeared successively, corresponding to separate

functions compatible with several case studies. Class

methods were aggregated, abstracted, repositioned

or refined when integrating successive case studies.

The first domain constitutes the principal tree

(Figure 2). It describes the agents that can intervene

in the model.

At the root of the tree, any element within any

system represented in the model is considered to be

a nearly decomposable system (NDS) (Simon,

1962).

Figure 2: UML-based class diagram of the agents of the

model, as determined from the case studies and engineered

according to natural classifications. For the sake of clarity,

only the relevant methods, properties or relationships are

presented. Legend: A_: abstract class, I_: interface, C_:

Class, _Uxxx: unit of the method or property, NDS: nearly

decomposable system (Simon, 1962), see text.

This notion is used as a proxy for complex systems

organised into a hierarchy (Le Fur, 2013). It is

operative in the sense that a system, if it can be

identified as an emerging entity resulting from the

interaction of its components, automatically acquires

individuality. Once its existence is revealed, the

entity can be identified immediately (i.e., named),

and the beginning of its existence can be noted.

Finally, its existence necessarily has a duration that

is formalised by ageing, a method that can then be

recursively overloaded by the daughter classes, i.e.,

agents belonging to the leaf classes accumulate

‘skills’ for ‘growing older’ all over the steps of the

specialisation chain. They thus acquire sophisticated

competencies for action or capacity to respond to

their environment. These basic and minimalist

notions outline the definition of the model over time:

they make it possible to encapsulate here the

dynamic aspect of the model.

cage

male species 1

fem. species 1

male species 2

fem. species 2

hybrid

truck

trainroad Transported rodent

original

simulation

A_VisibleAgent

bornCoord_Umeter

sensing_Umeter

A_NDS

thisId

thisName

birthDate_Utick

age_Utick

age_Uday

A_Organism

getGenome()

I_ExistingThing

getThisName()

getBirthDate_Utick()

getAge_Uday()

actionGrowOlder_Utick()

C_Vegetation

vegetationType

biomass_Ugram

I_ReproducingThing

actionMate()

actionGiveBirth()

I_SituatedThing

getCoord_Umeters()

setCurrentSo ilCell()

getCurrentSoilCell()

A_Amniote

updatePhysiologicStatus()

actionMate()

actionSpawn()

isSexualMature()

isPregnant()

getReadyToMate()

getAgeOfMatur it y_Uday()

I_Container

see fig.4

I_DiploidGenome

see fig.3

perception()

C_Rodent

C_RodentDomestic

C_RodentCommensal

C_RodentCagedC_RodentFossorial

C_RodentFossorial

C_Human Carrier C_BarnOwlC_TaxiMan

Colonial

A_Animal

speed_UmeterByTick

nextMove_ Umeter

currentTarget

hasToDisperse

deliberation()

A Simulation Model for Integrating Multidisciplinary Knowledge in Natural Sciences - Heuristic and Application to Wild Rodent Studies

343

The first specialisation (A_VisibleAgent) relates

to space and makes it possible to distinguish

between objects located in their environment and to

identify the fundamental properties relating to this

characteristic. Visible objects and agents implement

the I_SituatedThing interface with abstracted

procedures for localisation. In this model, as soon as

an object is located, the notion of environment takes

on a meaning. Perception then establishes a faculty

attributed to all agents. Beyond this class, NDS are

either the agents (organisms) or their containers

(Figure 4).

The second specialisation (A_Organism) is

linked to an essential characteristic of living

organisms: the determination of the properties of the

organism by the genes it carries, to which values can

be assigned. Each agent from this class or its

daughters bears a genome (see Figure 3) with

‘genes’ that can be transcribed into properties or

parameters (‘phenotypes’ or ‘life traits’ in Biology)

defining how agents carry out their action in their

environment. An agent belonging to the organism

hierarchy also implements I_ReproducingThing,

enabling it to reproduce with other organisms of the

same species and to transmit part of its DNA to its

descendants. Various agent properties from these

three superclasses were gradually introduced into the

model. Any function identified within a class was

matched systematically to a biological justification

within the hierarchy of life sciences for its

qualification: animals (A_Animal) are able to move,

burrowing rodents (C_RodentFossorial) dig

burrows, colonial rodents (C_RodentFosso-

rialColonial) have social interaction, and so on. The

agents operate according to a PDE (perception-

deliberation-execution) behavioural scheme (e.g.,

Macia Perez et al., 2014).

The root classes were rapidly identified and then

progressively refined by displacement or refactoring

of the methods or properties. Between case studies,

several rodent agent classes were successively

added, on the basis of functionality criteria. The

classification obtained (bottom of Figure 2) was

ultimately shown to correspond to the various

known social statuses in rodents (commensal,

fossorial, etc.). Once the nature of the classification

was identified, it was strengthened.

The second hierarchy used (Figure 3) represents

all aspects relating to the genetics of living

organisms. It includes the mechanical elements

(genes, alleles, etc.) required for genomic operations

such as meiosis, segregation, fertilisation, mutation,

and recombination.

Figure 3: UML-based class diagram of the genetic part of

the model, as obtained from the case studies. Left: Genetic

structure allowing the transmission of the genetic heritage

of a simulated agent to its 'progeny' following mating. The

structure of the tree on the right-hand site resembles a

knowledge base for instantiating, in the general scheme,

any type of species (rodent, predator, man, etc.) and its

characteristics (life expectancy, age at maturity, etc.).

Legend: identical to Figure 2, LUCA: last universal

common ancestor, the root of the species phylogeny

(Forterre et al., 2005).

The natural ‘gene’ - ‘chromosome’ - ‘pair of

chromosomes’ - ‘genome’ composition is adapted

from the work of Shaw and Wagner (2008) on locust

genomes. In this sequence, any genome aggregates

various pairs of ‘chromosomes’ that can be

recombined and inherited, in part, during

‘reproduction’. Each agent is associated with a

genome (I_DiploidGenome) that corresponds to its

species. This branch makes use of the inheritance

principle of the object paradigm to instantiate the

characteristics (called ‘traits’) of living agents in a

heritable and cumulative manner up to species-

specific values. By exploiting the analogy between

the biological approach (phylogeny) and object-

oriented programming (inheritance), this part of the

tree can also be considered as an object-oriented

knowledge base (right-hand part of Figure 3),

making it possible to value the movement speed,

litter size, age at sexual maturity, etc., as genetically

encoded characteristics for differing species of

agents. Here, too, the mother classes are established

by generalisation: when a biological trait

corresponds to a natural classification, the

corresponding class is created. For example, the

duration of gestation is coded as a property of

I_Recombinator

I_GeneMutator

C_GenomeAmniota

C_GenomePoacea

C_GenomeRattusRattus

C_GenomeMicrotusArvalis

C_GenomeMastomys

C_ChromosomePair

A_GenomeLUCA

C_GenomeSpermatophyta

C_GenomeHomoSapiens

C_GenomeEucaryote

C_Chromosome

I_DiploidGenome

mate()

makeGametes()

getDiploidNumber( )

isHybrid()

C_ChromosomePair-

etc.

C_GenomeBalanites

C_GenomeMastoEryth roleucus

C_GenomeMastoNatalensis

(see fig.2)

C_Gene

allele

mapLoc

mutate()

copy()

compareTo()

Sexual

etc.

SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

344

amniotes, which constitute the branch of animals in

which a foetus is formed, or eukaryotes are

characterised by the possession of a genome

consisting of pairs of chromosomes.

The third variable domain concerns space. Most

agents are located at a particular point in space, the

status of which may be associated with diverse

contexts, depending on the case studies considered.

Successive rearrangements of the objects

characterising the substrates and elements of the

substrate in which agents evolve led to the

development of the arbitrary concept of a 'container'

(Figure 4). I_Container is implemented by all

structures that can contain agents. Containers are

defined as a recursive system, in that they may

themselves contain other containers.

Figure 4 UML-based class diagram of the types of ground

organisation, as determined from the aggregated case

studies. Every instance in this hierarchy shares the

I_Container features. Three successive levels of

composition have been distinguished: landscape, land plot,

soil cell. Each level of the composition is itself the root of

a hierarchy that can be further extended (on the right) to

describe various features of the worlds simulated.

A_Container implements I_Container and allows

containers to contain I_SituatedThing objects.

Containers are visible agents and, as NDS (Figure

2), they also have an age and an ability 'to age', or, in

other words, to change over time according to

specifications. Within this model, environment is

discretised into elementary cells of the same size

that describe the type of terrain (road, river, crop,

etc.). Cell size depends on the level of detail chosen.

A set of contiguous elementary cells of the same

type is a C_LandPlot. This class provides

information about the delimitation and identification

of different areas within the simulated environments

(e.g., crop, city, region, road, river). Finally, for each

case study, an object of type C_Landscape is defined

to contain the set of visible objects within the

simulation: this class contains one continuous space

representing the topology over which the agents

move and a grid (matrix) of I_Container, the

elements of which constitute a discretisation of the

environment and a topology used by the agents to

perceive their neighbourhood (Moore

neighbourhood).

As the model was gradually refined, it became

increasingly clear that almost all the relevant objects

of the represented systems (burrow, nest, vehicle,

trap) could be assimilated as containers for agents, a

single class (A_SupportedContainer) facilitating

their integration into this scheme.

4 DISCUSSION

The model developed here encompassed all the

components linked to the different domains

modelled between case studies. Every case studies

was simulated using the same model. Through time,

each of them both enriched, and benefited from, the

mechanisms and components included in the

simulator and hence, evolved following the model’s

development. This led, for example, to the

progressive inclusion of cell processes, social

behaviour, spatial structures, and several types of

movement.

According to the heuristic, only classes that

provide functions specific to their rank are

implemented. Several characteristics of the

architecture of the model emerge from this

approach. The first of these characteristics is the

partitioning of the model into three main areas of

diversity: the diversity of agents (concrete objects),

specific traits (genomes) and spaces (containers).

Other elements were gradually defined and

confirmed, such as the encapsulation of time-related

aspects in the root of the class tree (that may appear

convenient or appropriate for a simulation model),

the partitioning of rodent types according to a

behavioural classification (fossorial, commensal,

etc.) similar to that used in natural sciences, the

functional breakdown of genomes (eukaryotes,

amniotes), which also follows the natural tree, the

composition of space around the concept of a

container or the factorization of most of the objects

of the domains modelled via the notion of a

supported container (Figure 4).

I_Container

agentIncoming()

agentLeaving()

getOccupantList()

getFullOccupantList()

A_Container

C_SoilCell

C_Burrow

C_Vehicle

C_Nest

C_LandscapeNetwork

C_Crop

C_City

C_Market

C_Trap

C_Country

A_SupportedContainer

C_Landscape

etc.

(see fig.2)

C_LandPlot

A Simulation Model for Integrating Multidisciplinary Knowledge in Natural Sciences - Heuristic and Application to Wild Rodent Studies

345

Two salient features emerge from the results

presented. The first relates to the evidence-based or

trivial nature of the categories resulting from the

heuristic used. The second concerns the constantly

unfinished and perfectible aspect of the architecture,

which adapts and evolves over time.

4.1 Obviousness of Emerging

Categories and Associated

Robustness

A posteriori, the arborescence that has emerged

appears trivial. This is the case for the root

categories, which sequentially encapsulate time,

space, biology, motion, and reproduction, for any

domain agent. Similarly, the typology of the

diversified behaviour of rodents obtained (Figure 2

bottom) reconstitutes a widely accepted

classification of known types of behaviour that is

common to zoology (e.g., Nowak, 1999). The same

is also true for the representation of genomes, for

which, thanks to the similarity between natural

phylogeny and OOP inheritance, the characterization

of the agents follows a universal biological

classification.

The a posteriori evidence of the classifications

obtained can be seen as a sign of robustness, an issue

of prime importance for the proposed approach. This

is the case for the ‘time / space / biology / movement

/ reproduction / behaviour’ hierarchy, which may, a

priori, be applicable to a wide variety of domains.

This is also the case for classifications converging

on the hierarchies recognized in Nature, which are

inherently the most robust. The approach would be

thus analogous to those used in bio-inspired

modelling (Egan, 2015), but in this case dedicated to

the representation of living systems.

4.2 Transitory and Perfectible Nature

of the Construction

The incremental approach adopted assumes both

continuous re-engineering of the code and control

over back-compatibility. With this approach, all the

case studies modelled, even the oldest ones, remain

active and constantly updated as the model is built.

They, thus, continually generate new results,

enhanced by the new functions added.

The evidence of robustness for this heuristic

makes it possible to consider diverse extensions and

improvements to the model. For example, coupling

agents with species-specific genomes would make it

possible to simulate diverse species, with a view to

addressing community ecology issues (Chesson,

2000), a key element in the understanding rodent

spread. The model could also be adapted to consider

the representation of viruses or parasites in the

transmission or non-transmission of diseases to

humans, the perception of odours by rodents or the

taking into account of energy balances. This would

open up as many new possibilities for this approach

as there are functionalities to be added.

The class tree can, therefore, continually be

tuned step-by-step, by taking new studies into

account, provided that these new studies are

compatible with the previous ones, or that the

previous studies can be rendered compatible with the

constraints of the new ones. Consequently, the

construction is, by nature, unbounded or perpetually

transitory (as long as it remains robust), with

methods and properties that can be modified or

repositioned.

4.3 Influence of Formalisms on Results

Despite its generic nature, this model may not

provide a canonical representation of these diverse

worlds. Indeed, each of the choices made in

computer modelling, such as the choice of a

Perception-Deliberation-Execution paradigm for

agent activities or the choices made for the

discretisation of space, is only one of the many

possibilities that could have been proposed. For

example, it would have been possible to base the

model on other formalisms, such as ‘Agent / Group /

Role’ (Ferber et al., 2004) or ‘Belief / Desire /

Intention’ (Caillou et al., 2015). Logically, even

using the heuristic presented here, the choice of one

of these other possibilities could have resulted in a

different architecture.

5 CONCLUSION

The proposed approach is based on the incremental

articulation of contrasting case studies within a

single model. The continuous consolidation and

questioning of the model through new case studies,

including studies based on other disciplinary

approaches, appears to satisfy the robustness

requirements for long-term integration.

The results obtained from the accretion of case

studies show that each approach adds to the others

already included in the model, to yield an integrated

system. The challenge, however, is articulating the

model so that it can reveal new processes or

dynamics through the multidisciplinary integration

of items, concepts or processes (McMurtry, 2009).

SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

346

The chosen approach must, by definition, be

considered to be in a state of continual improvement.

However, it can be used to identify the components

that are shared, could be shared, or are

irreconcilable, and, possibly, the ways to achieve the

mutually beneficial integration of points of view

specific to different disciplines.

ACKNOWLEDGEMENTS

We would like to thank J.F. Cosson, J.P. Quéré, C.

Berthier, B. Gauffre, J.M. Duplantier, L. Granjon, G.

Ganem, J. Britton, O. Ninot, J. Lombard, P.

Handschumacher, S.Piry, the scientists who kindly

agreed to decipher their disciplinary expertise for the

formalization of thematic case studies. This study

owes much to the work done by Q. Baduel, A.

Realini, J.E. Longueville, A. Comte and M.

Diakhate, as part of their student internships. The

SimMasto project (http://simmasto.org) was

supported, through the various case studies of the

Chancira projects (grant IRD-ANR-11-CEPL-0010),

Cerise (grant IRD-FRB no.AAP-SCEN -20B III), and with

occasional support from the French National

Research Institute for Sustainable Development

(IRD) and the ‘Centre de Biologie pour la Gestion

des Populations’ (CBGP, UMR no.22 INRA/IRD/Cirad/Sup-

agro). We also wish to thank J. Sappa and L. Granjon

for their help during the revision process.

REFERENCES

Bar-Yam, Y., 1997. Dynamics of Complex Systems,

Addison-Wesley, Reading, Mass., vol. xvi, 848p.

Burger P., Kamber R., 2003. Cognitive integration in

transdisciplinary science: knowledge as a key notion.

Issues in integrative studies, 21:43-73.

Caillou, P., Gaudou, B., Grignard, A., Chi Quang Truong,

C.Q., Taillandier, P., 2015. A Simple-to-use BDI

architecture for Agent-based Modeling and

Simulation. The Eleventh Conference of the European

Social Simulation Association (ESSA 2015), Sep 2015,

Groningen, Netherlands, 14p.

Casti J. L., 1997. Would-be Worlds: How Simulation is

changing the Frontiers of Science. John Wiley & Sons,

New York, 242p.

Chesson, P., 2000. Mechanism of maintenance of species

diversity. Annu. Rev. Ecol. Syst. 2000. 31:343–66.

Clapham, W. M., Crosby, C. J., 1992. Utility of Object-

oriented Programming in Complex System Modeling

Math. Comput. Modelling, 16(617): 45-50.

Egan, P., Sinko, R., LeDuc, P.R., Keten, S., 2015. The role

of mechanics in biological and bio-inspired systems.

Nature Communications, 6:7418, DOI: 10.1038 /

ncomms8418.

Ferber, J., Gutknecht, O., Michel, F., 2004. From Agents

to Organizations: An Organizational View of Multi-

agent Systems. In: Giorgini P., Müller J.P., Odell J.

(eds) Agent-Oriented Software Engineering IV. AOSE

2003. Lecture Notes in Computer Science, vol 2935.

Springer, Berlin, Heidelberg.

Forterre, P., Gribaldo, S., Brochier, C., 2005. LUCA : à la

recherche du plus proche ancêtre commun universel.

MIS: Médecine Science, 21(10):860-865.

Goldspink, C., 2002. Methodological Implications Of

Complex Systems Approaches to Sociality: Simulation

as a foundation for knowledge. J. Artificial Societies

and Social Simulation, 5(1).

Le Fur, J., 2013. Extending life concepts to complex

systems. Interdisciplinary Description of Complex

Systems, 11(1):37-50.

Macia-Perez, F., Lorenzo-Fonseca, I., Berna-Martinez,

J.V., 2014. A formal framework for modelling

complex network management systems. J. Network

and Computer Applications, 40:255-269.

McMurtry, A., 2009. Knowers and phenomena : two

different approaches to interdisciplinarity and

interprofessionalism. Issues in integrative studies,

27:1-16.

Newell, W. H., 2001. A theory of interdisciplinary studies.

Issues in integrative studies, 19:1-25.

North, M. J., Howe, T. R., Collier, N. T., Vos, J. R., 2005.

The Repast Simphony Development Environment. In,

Proc. Agent 2005 Conference on Generative Social

Processes, Models, and Mechanisms, Macal C. M.,

North M. J. and Sallach D. (eds.), ANL/DIS-06-1,

Oct.:13-15.

Nowak, R.M., 1999. Walker's Mammals of the World. 6th

Ed. The John Hopkins University Press, Baltimore

Page, S., 2011. Diversity and complexity. Princeton Publ.,

290p.

Richard, B., Tchounikine, P., 2004. Enhancing the

Adaptivity of an Existing Website with an

Epiphyte Recommender System. The New Review of

Hypermedia and Multimedia, 10(1):31- 52.

Rouxel, A., 2002. Identité disciplinaire: cohérence ou

éclatement ? Tréma, Littérature enseignée et

reconfigurations, Montpellier, n°19: 29-36.

Shaw, K. L., Wagner, K., 2008. Cricketsim: a Genetic and

Evolutionary Computer Simulation. J. Artificial

Societies and Social Simulation, 11(1).

Simon, H. A., 1962. The Architecture of Complexity.

Proc. American Philosophical Society, 106(6):467-

482.

Skonhoft, A., Leirs, H., Andreassen, H. P., Mulungu, L.

S., Stenseth N. C., 2006. The bioeconomics of

controlling an African rodent pest species. Environ.

Dev. Econ., 11 :453-475.

Taylor, P. J., Arntzen, L., Hayter, M., Iles, M., Frean, J.,

Belmain, S., 2008. Understanding and managing

sanitary risks due to rodent zoonoses in an African

city: beyond the Boston Model. Integr. Zool. 3:38-50.

A Simulation Model for Integrating Multidisciplinary Knowledge in Natural Sciences - Heuristic and Application to Wild Rodent Studies

347