Multilevel biomarker studies are now a classical approach for the monitoring of contaminated environments. Nevertheless, even if the population constitutes a level of interest for environmental management, the characterization of in situ effects of contaminants at this level reminds difficult notably because it also integrates plastic, adaptive and neutral variability related to physical-chemical habitat or phylogeograhical constraints. For predictive approaches in chemical risk assessment, the experimental difficulty to assess impacts of toxic compounds at the population scale is currently addressed by modelling techniques. With the environmental relevant species G. fossarum, we develop this methodology in a diagnostic framework by coupling population dynamics modelling and in situ caging. Hence, we expect to enhance the ecological relevance of in situ toxicity assessment and to develop biologically integrated ecotoxicological indicators. Our approach proceeds in two steps. (i) Population models are developed to mechanistically link life history traits (survival, growth, reproductive activity, fertility) and the dynamics of reference native populations all along a year. For this, we use a dual approach combining laboratory and field experiments. The latter consisting in a demographic follow-up based on monthly population census (densities, size distributions) and an individual approach with field organisms caged in situ (reproduction, growth, survival). (ii) Effects of contamination, observed at the organism level through the exposure of transplanted individuals caged in contaminated stations, are projected into the dynamics of these reference populations in order to propose a diagnostic assessment of water quality at the population level.