As shown by many laboratory studies since 1990s, a large range of chemical stressors can inhibit gammarids feeding rate. Feeding inhibition also constitutes in most case one of the first traits observed in responses to environmental pollutions. Studying the effects of pollutants on the feeding behaviour is of ecological concern; as it can be related to life history traits like growth, survival or fertility thus allowing assessing the effects of toxicants on higher biological organization levels, such as population. A mechanistic modelling approach is proposed as a perspective to perform this extrapolation between biological scales. Our work was focused on the crustacean Gammarus fossarum widely present in European rivers. In a first part, we illustrate how taking into account of the influence of biotic and abiotic factors(body size, temperature, water hardness) on feeding activity allows one to improve the interpretation of in situ feeding rate assays for the evaluation of water quality. For this, we performed a three-steps approach: (i) we characterized the influence of these important confounding factors in laboratory conditions, (ii) we validated the robustness of feeding activity reference values; these latter were established through in situ caging experiments with transplanted standard organisms in reference streams with contrasted abiotic profiles at different seasons (iii) finally, by considering in situ caging in contaminated streams, we underlined the importance of taking into account the influence of such factors for a better toxicological bio-monitoring of freshwater ecosystems. In a second part, we show that feeding activity can be linked to life history traits such as fecundity and we propose a modelling methodology to link impact on feeding activity to potential effects at the population level, taking into account the influence of environmental conditions. For this purpose, we developed an environmentally realistic Leslie population model, which allowed us to link individual-level demographic parameters to population dynamics. Ultimately, we illustrate how couple fitness related endpoints measured in bioassays procedures to population modelling.