Authors:
Jean Le Fur
1
and
Moussa Sall
2
Affiliations:
1
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 and France
;
2
Dépt. Informatique, Univ.G.Berger/Saint-Louis Sénégal and lab. IRD-BIOPASS, Campus Bel-Air, Dakar and Senegal
Keyword(s):
Agent-based Model, Time Scale, Rodent, Discrete Time Simulation, Sensitivity Analysis.
Related
Ontology
Subjects/Areas/Topics:
Agent Based Modeling and Simulation
;
Complex Systems Modeling and Simulation
;
Environmental Modeling
;
Multiscale Simulation
;
Sensor Networks
;
Simulation and Modeling
;
Simulation Tools and Platforms
;
Software and Architectures
Abstract:
Identifying parameters value is a major issue in model engineering. In discrete time agent-based models, time step is an important one as it determines the frequency at which agents realize their activity step. This parameter is commonly defined as a fixed constant during the model design stage. In particular cases, this may lead to biases as it may be sometimes difficult to determine if agents efficiently realize their activity step once each 1, 2 seconds, hour or the like. A simulation model of a rodent population has been used to study the effect of using a flexible time scale on its outcomes. Three types of processes have been considered as time dependent in the model, environment sensing, movement and life cycle (maturity, gestation...). A time step sensitivity analysis constitutes the principal result of this study. For the widest range of time step values, model’s behaviour is unrealistic and bound to algorithms artefacts. A very small range of time steps leads to simulation o
f a perennial rodents’ population. Biases bound to variable time step implementation are discussed. Using flexible time scale approach proved efficient to get insight into the model’s behaviour and fruitful clues to assess agents’ processes frequency in the actual ecosystem.
(More)