The evolution of toxic effects at a relevant scale is an important challenge for the ecosystem protection. Indeed, pollutants may impact populations over long-term and represent a new evolutionary force which can be adding itself to the natural selection forces. Thereby, it is necessary to acquire knowledge on the phenotypics and genetics changes that may appear in populations submitted to stress over several generations. Usually statistical analyses are performed to analyse such multigenerational studies. The use of a mechanistic mathematical model may provide a way to fully understand the impact of pollutants on the populations’ dynamics. Such kind of model allows the integration of biological and toxic processes into the analysis of ecotoxicological data and the assessment of interactions between these processes. The aim of this Ph.D. project was to assess the contributions of the mechanistical modelling to the analysis of evolutionary experiment assessing long-term exposure. To do so, a three step strategy has been developed. Foremost, a multi-generational study was performed to assess the evolution of two populations of the ubiquitous nematode Caenorhabditis elegans in control conditions or exposed to 1.1 mM of uranium. Several generations were selected to assess growth, reproduction, and dose-responses relationships, through exposure to a range of concentrations (from 0 to 1.2 mM U) with all endpoints measured daily. A first statistical analysis was then performed. In a second step, a bioenergetic model adapted to the assessment of ecotoxicological data (DEBtox) was developed on C. elegans. Its numerical behaviour was analysed. Finally, this model was applied to all the selected generations in order to infer parameters values for the two populations and to assess their evolutions. Results highlighted an impact of the uranium starting from 0.4 mM U on both C. elegans’ growth and reproduction. Results from the mechanistical analysis indicate this effect is due to an impact on the assimilation of energy from food. Both the mechanistic and the classic approaches highlighted individuals’ adaptation to environmental conditions. Despite this, differential evolutions of the individuals from the uranium-selected population were also highlighted. All these results were more in-depth described by the mechanistical analysis. Overall, this work contributes to our knowledge on the effects of pollutants on population dynamics, and demonstrates the contributions of mechanistical modelling which can be applied in other contexts to achieve in fine a better assessment of environmental risks of pollutants.