VIP

Virtual reconstitution of a Prairie

Cendrine Mony(*), Marc Garbey(**), Malek Smaoui(**) and Marie-Lise Benot(*)
(*) UMR CNRS 6553 ECOBIO, University of Rennes1, FRANCE
(**) Department of Computer Science, University of Houston, USA


Both phases are analyzed by an in silico multiscale model of the prairie validated by experiments; phase one is achieved through the construction of an agent-based model of individual plants and the extraction of a family of rules from a set of controlled experiments. The parameter space is very large and evolutionary programming is a natural match for this ecological study. In the second phase, the dynamic interaction with thousands of individual needs to be simulated to observe potential emerging properties from this complex ecological system. Both phases require very large scale simulations with embarrassing parallelism that can be friendly achieved with BOINC.
Link for the project
Journal For our Virtual Prairie Explorers

A coarse approximation of the parameter space requires of the order of 1 000 000 simulations which should take 10 years on a single PC. The study of one evolutionary path should take of the order of a year at least on a single PC (phase 1). There are indeed dozens of paths involved in evolutionary dynamics. The complexity of the simulation is about two orders of magnitude higher for a prairie (phase 2). A run that can take advantage of the idle time on 10 000 PCs during a month should bring new results in the ecology of a prairie that has never been accessible in the past.

<< Prev  1  2  3  4