Due to the constant fluxes of water, sediment, and chemical compounds, rivers are one of the most open ecosystems of the earth. They often react on a short time scale to any change of human activities within the whole catchment (land use) or within the floodplain and/or the river bed itself. Nevertheless, the ways by which such a co-evolving system of humans and nature can be restored from a highly altered state are complex and surprising responses are the rule. Interactions between biotic and abiotic factors that have developed during the period the system was modified by humans can make the system insensitive to any restoration of the historical environmental features and remaining in an alternative stable state. Lessons from historical changes in rivers ecosystems can help nowadays to define realistic restoration plan for rivers and to better understand to which point re-establishing previous abiotic conditions (i.e. suppression of specific human pressure on the system) can really allow the expected recovery of biotic community. The aim of the presentation is to highlight the way by which historical studies on rivers can be fruitful for the establishment of future Water Basin Management Plan with the two examples of: (1) the evolution of the fish community of the river Seine, facing several anthropogenic alterations of their biota, and (2) the development of statistical models to predict the past occurrence of fish species. The evolution of fish communities since the middle of the 19th century to the end of the 20th century is analysed for several river stretches of the Seine Basin exhibiting different evolution trends of pressures (channelization for navigation purpose - modification of hydrologic and thermal condition from construction of reservoirs, transformation of land use without profound modifications of the river itself). Different evolution patterns, sometime opposite ones, are observed for fish communities, in relation to the dominant pressures involved. For fish species from the Rhône catchment, our objective was to evaluate the efficiency of statistical models calibrated on the present day species occurrence to correctly predict the past distribution of the same species during the first half of the twentieth century. Historical maps were compiled and implemented in GIS and in association with the main environmental variables describing the habitat at the reach scale. Our ability to reconstruct the past distribution of 10 species is discussed, in relation with the alteration of the river system (past and present), the quality of historical data and the technical limit of such approach.