1- Scientific context and project objectives

The increased importance of sustainable development as a challenge for society increases demand for environmental dynamics models to understand and anticipate change to be able to help define future policies. At the same time, the increase in memory and computing power of machines means that ever more detailed models can be developed, particularly individual-centred (IC) models representing all of the system’s entities and their interactions. But such models often behave in a complex way that is difficult to understand, which considerably hinders their practical use.
Over the last few decades, countless research studies show that the complexity of these models comes from the difficulty in establishing a link between the individual dynamics of entities and the overall behaviour of the aggregate variables. In other words, it is like comparing one person with a population. The general procedure involves studying the behaviour of these variables by referring to systematic simulation experiments. The challenge is to understand the influence of individual, local dynamics on global, statistically robust behaviour, likely to help in decision-making.
Although there are currently lots of tools to help modelling specialists design and set up their models, their exploratory and analytical stages are much less systematised. Moreover, few tools take account of the massive exploratory stages of simulation models, which requires a wide range of simulations to be launched. The rare few that do, do it simply. To meet these needs in the laboratory and in other teams, a new tool, SIMEXPLORER, has been developed to enable laboratory staff and partners to make better use of their algorithms, access, manage and process their data more effectively.

2- Project description

SimExplorer is a platform that explores ecosystem simulation models by producing experimental designs on the model’s parameters and on its initial conditions. It can be used for any model that can be used on command lines, for which the inputs and outputs are specified in the form of files in some type of format. Here, the simulation model is considered to be a black box receiving one or more files specifying the parameters and initial conditions and which produces results files. SimExplorer:

• generates experimental designs defined by the user for a set of input parameters,
• launches corresponding simulations on a cluster or a computing grid,
• retrieves and processes results, by computing statistical indicators, for example.

More specifically, SimExplorer is a programming environment that can:

• program experimental designs on the parameters and initial conditions of models,
• program the processing to carry out at output,
• produce these experimental designs.

The project has two objectives:

• to develop a robust, durable version of the programming environment from existing prototypes,
• to attach to SimExplorer an information system with which the various experiments carried out on different simulation models can be managed, and processed more effectively in particular.

LifeGrid, the regional information system