Scientific Opinion on good modelling practice in the context of mechanistic effect models for risk assessment of plant protection products
The Panel has interpreted the Terms of Reference as a stepwise analysis of issues relevant to both the development and the evaluation of models to assess ecological effects of pesticides. The regulatory model should be selected or developed to address the relevant specific protection goal. The basis of good modelling practice must be the knowledge of relevant processes and the availability of data of sufficient quality. The opinion identifies several critical steps in order to set models within risk assessment, namely: problem formulation, considering the specific protection goals for the taxa or functional groups of concern; model domain of applicability, which drives the species and scenarios to model; species (and life stage) selection, considering relevant life history traits and toxicological/toxicokinetics characteristics of the pesticide; selection of the environmental scenario, which is defined by a combination of abiotic, biotic and agronomic parameters to provide a realistic worst-case situation. Model development should follow the modelling cycle, in which every step has to be fully documented: (i) problem definition; (ii) model formulation, i.e. design of a conceptual model; (iii) model formalisation, in which variables and parameters are linked together into mathematical equations or algorithms; (iv) model implementation, in which a computer code is produced and verified; (v) model setup, including sensitivity analysis, uncertainty analysis and comparison with observed data, that delivers the regulatory model; (vi) prior to actual use in risk assessment, the regulatory model should be evaluated for relevance to the specific protection goals; (vii) feedback from risk assessor with possible recommendations for model improvement. Model evaluation by regulatory authorities should consider each step of the modelling cycle: the opinion identifies points of particular attention for the use of mechanistic effect models in pesticide risk assessment. It is recommended that models be documented in a complete and transparent way, that a feedback platform be established involving risk assessors and model developers, and that a set of agreed models be made available.