For the last 40 years, continuous research has been dedicated to understanding the fate of radionuclides and metals through subsurface aquifers [1], [2]. Chernobyl and Fukushima disasters have resulted in a high release of fission products (especially 137Cs, 90Sr) in groundwater below the damaged nuclear power plants. Risk associated with the release of radionuclides from nuclear facilities to groundwater tends therefore to motivate additional intensive research efforts. Because of their geochemical properties, fission products like 137Cs and 90Sr migrate usually slower than water in contaminated porous media. Since initial research dedicated to remediate groundwater contamination, one major issue that arose quite early still remains unsolved: why do reaction rates observed at laboratory scale in porous media not usually match those observed in the field for such contaminants? Omnipresent chemical and physical heterogeneities, as well as discrepancies in the scale of model discretization relative to the scale of the chemical/physical process, are usually mentioned to explain this issue. Research carried out to account for heterogeneities in porous media (e.g. [3]–[5] ) tends moreover to indicate that the hydrogeological modelling approach developed for risk analysis needs to be tailored to the type of risks that is to be evaluated. Chronic or acute exposure scenarios should require different models relying on different assumptions and parameterizations. As demonstrated by [4], assessing a mean concentration in space or time in porous media does not require the same level and type of characterization than assessing extreme values (maximum concentrations, time of breakthrough, etc.). Within this framework, IRSN has been exploring the following research directions for better assessing the risk of subsurface radionuclide transfers under accidental situations: 1/ development of an adaptive mesh strategy in transport simulations for allowing heterogeneous parameterizations in porous media ensuring reasonable calculation times and accuracy of model predictions, 2/ field and modelling works over a Pilot Site in the Chernobyl Exclusion Zone for gaining insight into field scale processes controlling Sr and Cs migration, and 3/ laboratory experiments on sorption/desorption processes for better assessing, at this scale and in well-controlled conditions, the fate of Cs and Sr and for improving the prediction of their behaviour by developing alternative approaches to the Kd model.