Numerical models allowing to simulate the transfer of radionuclides over long distance in the environment must take into account the relevant interfaces where bio-physico-chemical processes may modify the fluxes exchanged between compartments. Estuaries are one of such hot spot area, since they may trap contaminated particles or release radionuclides from these particles through desorption. A box-model has been developed and applied to the Rhone River estuary (France) with field data in order to couple a river (Casteaur) and a sea dispersion model (Sterne) for simulating the transfer of 137Cs along this continuum. This model, also suitable for other estuaries, allows to estimate the hydrodynamic conditions acting at the mouth (saltwedge into the estuary vs external plume on sea) and the freshwater/seawater mixing. It focuses on dissolved 137Cs and take into account the possible desorption from particles, which is described in the estuary box-model with kinetic rates calibrated from adsorption/desorption experiment in laboratory. This experiment was designed to specifically study the influence of salinity and of the ageing effect. They particularly showed that desorption occurred above a salinity of 3 to 4, and that its importance increased inversely with the duration of the adsorption phase (ageing effect). Such river-sea continuum modelling can be used to anticipate the dispersion of a 137Cs release in case of accident from one of the nuclear installations along the river. For this a statistical approach through fuzzy c-mean clustering of a 10 years series of river discharge and wind speed collected at the estuary was set up to explain the variability of the surface plume. The method allows to define 6 scenarios corresponding to the most usual occurrence of these two hydrodynamic forcing, which were used to simulate the extension and dilution of a release fixed at 1Tbq over a temporal window of 48 h. Finally, databases reporting biophysical and socio-economic environments for this area were collected and used to define sensitivity indicators. They were associated to 5 classes of vulnerability relatively to the impact of a radioactive contamination, and aggregated to constitute a final sensitivity basemap. This map was confronted to the simulations of 137Cs release under the 6 hydroclimatic scenarios, for which the seawater activities were converted into fish activities and separated into 4 classes partly based on regulatory limits for food contamination. This work is in relation with the challenge 1 of the ALLIANCE SRA: predict human and wildlife exposure in a robust way by quantifying key processes that influence radionuclide transfers and exposure.