​In the context of accidental or chronic releases from nuclear installations to the environment, it is important to predict the future of radionuclides, particularly at the plant scale, the first link in the human food chain. IRSN and EDF are developing a simulation platform for the transfer of radionuclides in the environment called SYMBIOSE, in order to estimate the dose received by humans. Tritium is an example of a radionuclide released by humans into the environment, whose behavior is particularly critical to predict because of its high mobility and biological affinity depending on the chemical form considered. The TOCATTA-khi model, implemented in the SYMBIOSE platform, focuses on tritium transfers within a grassland ecosystem. The main objective of this study is to improve and evaluate the existing TOCATTA-khi model by studying the persistent uncertainties of tritium transfers within a prairial environment from in situ and laboratory experiments.