3 Worlds: A multiscale simulation plateform for ecology and environmental science

The 3Worlds project is supported by the Agence National pour la Recherche (ANR-CIS) (French) as a 3 years project from 2008-2011. It is coordinated by Jacques Gignoux (ENS, Paris)

OBJECTIVES

Our objective is to promote numerical simulation in ecology on sound theoretical bases by

producing a user-friendly software based on a unified conceptual framework, giving access to

parallelism and computation intensive simulation. The software will target researchers, students and

teachers, and later decision makers in ecology and environmental sciences. Ecological systems will be

modelled as multi-scale hierarchically organized systems, and the software will organize the interaction

between physics and biology in space1 in such systems, allowing for adaptation of any particular model

to the available level of knowledge. Given the state of advancement of this project, this will involve all

the usual tasks of computer program development, but with a special emphasis on coding and

developing explanatory guide manuals with conceptual and application developments. To guarantee a

future wide use of this software, we intend to distribute it as a free / open source software.

Finding the right scale for a model is a particularly important problem in ecology. Providing a

common multiscale framework for the modelling of ecological systems is the first step towards the

definition of standardized scaling strategies: once models at different scales can be constructed within a

common formalism, comparison between coarse resolution and fine resolution models is easier. This

should help track scale invariance, emergence, and other properties of complex systems. Although

providing 'ready to use' scaling strategies is far beyond the scope of the current project, we will initiate

thinking on these questions during the course of the project.

For long, the dominant paradigm for ecosystem modelling has been that of physics, i.e. systems of

differential equations, applied to matter and energy flows or to population dynamics. Without an explicit

modelling of emergence, predicting changes of the whole system's behaviour from changes of its

components is impossible. 'Multi-agent simulators' (MAS) and individual-based models (IBM) have used

emergence as their central paradigm. The key difficulty of ecological modelling is to reconcile the

classical state-variable, physics based, approach, and the IBM or MAS, biology based, approach

because ecosystems are both physical and 'biological-social' systems. Emergence implies hierarchy: a

system must be described as a set of subsystems for properties to emerge. The originality of 3Worlds,

compared to other ecology or environment oriented simulation platforms (e.g. 'physics based': Earth

simulator, ORCHIDEE, SEVE; or IBM/MAS-oriented: ATLSS, LAMOS, MUSE, CAPSIS, AMAPSim) and

to computer science platforms (e.g. SWARM, DEVS, or the multiparadigm approach) is to propose a

common framework for representing hierarchical ecological systems, whatever the scale of observation.

3Worlds is in continuity with the MUSE and LAMOS softwares. Based on previous work, the major

features of the platform will be:

    1. a simulator able to encompass many different organizational scales through a unique set of classes

        describing a hierarchy of nested ecological systems (multi-scale model);

    2. the possibility to easily change the detailed representation of an ecological process through a

        programming interface (multi-model)

    3. the deployment of series of simulations in parallel over a network to reach high computation power;

    4. a library of ecological models applied to systems at different scales;

    5. a toolkit for analyzing simulation results according to widely accepted ecological theory, or more

        general tools not included in statistical software (like tracking emergence).

The project has been organized into 4 work packages (software deployment, implementation and

integration, simulator, model library, simulation analysis toolkit). A scientific committee will lead the

project. After an initial kick-off meeting helping to make most design and organization decisions, the

work packages will enter cycles of code development. During the 2nd or 3rd year of the project, a first

public version of the software will be released. A final meeting at an international conference and a

follow-up group will help perennize the achievements of this project.