Faced with ever increasing environmental perturbation, quantitative risk assessment comes up today against theoretical and methodological barriers linked to changes in temporal, spatial and demographic observation scales. One key issue is the link between the scales at which effects are experimentally observed and the relevant biological levels to protect (individual, population, community or ecosystem). Within this framework, team MEPS promotes an interdisciplinary and integrative approach for cognitive and predictive purposes, based on dynamical and mechanistic models, as well as on the development of statistical inference methods. In particular, we assess environmental risks by integrating all sources of variability (inter-cellular, inter-individual, or inter-species heterogeneity) and uncertainty (gaps in modeller knowledge) within appropriate hierarchical models. We develop Bayesian approaches to estimate parameters of complex models (e.g., dynamical systems involving latent variables) from data of different nature, sometimes uncertain (expert knowledge, metadata…). We promote the use of two dimensional Monte Carlo simulations, enabling the separation of variability and uncertainty recommended for risk assessment. In order to provide an easy access to our methods we develop several R-packages in combination with a dedicated web-site.

Key words: Quantitative risk assessment of environmental hazards, Dynamical systems, Matrix population models, Sensitivity analysis, Multilevel/hierarchical mechanistic modelling, Bayesian inference, Ecotoxicology