TRAJECTOIRE aims to establish past and predictive trajectories of contaminants at the outlets of the major French watersheds (Rhône, Loire, Seine, Garonne, Rhine, Meuse, Moselle) for substances brought about by human activities in these environments during the technological, industrial and environmental development that punctuated the 20th century: radionuclides, microplastics and their additives, and critical metals. These non-legacy substances are currently at the heart of reflections on the energy transition. By considering these three families of contaminants we expect to draw lessons learned based on the differentiated and successive time scales of their inputs, which were governed by institutions and public policies according to distinct management modes. By considering them, we expect to draw general lessons on the environmental resiliency regarding contaminants, and then to positioning or repositioning current environmental concerns face to new and future technologies. In other words, it will be a question of evaluating how society can be an actor of the resilience of the environment following anthropic disturbances from economic choices, political decisions and collective actions. In river systems, sediments convey and store most of contaminants introduced in the catchment. Therefore, sedimentary archives in perennial storage areas, such as riverbanks or alluvial margins, give testimonials on previous contaminations and anthropic pressures. Feedbacks on the ability of large rivers to absorb or remove anthropogenic pressures will be established by reconstructing time-series of: 1) contamination levels based on sedimentary records and 2) pressures exerted on environments and responses provided by institution and society, based on analyses of documentary archives. The difficulties of such retrospective exercise, requiring to cross multiple and complex information, are all the greater as the statistical sources concerning contaminants are rare or confidential. The causal links between the observed contamination levels in sedimentary archives (quantitative data sets) and the anthropic pressures determined from documented archives (qualitative and semi-quantitative data sets) will be assessed using neural network analyses for time series prediction. Time series models are purely dependent on the idea that past behavior and patterns can be used to predict future behavior and trends. By using these models on data sets acquired at the outlets of major French rivers, various anthropic pressures will be considered and their consequences on the concentration of contaminants over time will be identified. Socio-historical events, acquired from documented archive analyses, will be characterized regarding their impact on concentrations, the time-lag between their occurrence and the environmental impact, and the duration of the environmental perturbation. The values of these three parameters associated to the best fittings between the data and times series models will define key pressures to be implemented in a predictive model based on scenario in order to forecast the levels of contaminants in river systems and estimate trajectories and resiliencies for the short, medium and long terms. Our results will put forth the environmental changes that succeeded over the last industrial era, and will help to predict those expected depending on our future conduct. We consider that society needs such feedbacks as well as predictive vision in order to reinforce environmental awareness and future decision making related to the sustainability of ecosystems. Our project aims to give quantitative feed backs and predictive models based on scenarios in order to inform stakeholders on environmental impacts of their past and future decisions, over short and longer term time periods. It aims to demonstrate that society can act on environmental resiliency.