Assessing and modelling the coastal plume dispersion of nuclearized rivers is strategic in case of accidental releases, but taking into account the variation of main hydrodynamic forcings is challenging. This study uses fuzzy c-mean clustering of a 10 years series of discharge and wind speed at the Rhone River estuary (France) in order to explain the variability of its plume. The method allows to classify the data into 6 scenarios of hydrodynamic forcings that were related to different spatial extensions of the plume, as well as to surface currents measured in-situ. These scenarios were used to simulate the extension and dilution of a radioactive release issued from the river. Based on threshold values of the forcings, a decisional tree is proposed to provide a quick decision tool identifying, in real time, which climatological scenario occurs at the river mouth and the potential plume pattern.