The ability to predict the consequences of an accidental release of radionuclides relies mainly on the level of understanding of the mechanisms involved in radionuclides interactions with different components of agricultural and natural ecosystems and their formalisation into predictive models. Numerous studies and databases about contaminated agricultural and natural areas have been obtained but their use to enhance our prediction ability has been largely limited by their unresolved variability. Such variability seems to stem from incomplete knowledge about radionuclide interactions with the soil matrix, soil moisture, biological elements in the soil and additional pollutants, which may be found in such soils. In this project, we investigated mainly the role of the biological elements (plants, mycorrhiza, microbes) in: radionuclide sorption/desorption in soils and radionuclide uptake/release by plants. Because of their chemical nature importance, the bioavailability of three radionuclides: caesium, strontium, and technetium have been followed. The role of one additional non-radioactive pollutant (copper) has been scrutinised. Role of microorganisms (Kd for caesium and strontium in organic soils is much greater in the presence of microorganisms than in their absence), plant physiology (changes in plant physiology affect radionuclide uptake by plants), the presence of mycorrhizal fungi (interferes with the uptake of radionuclides by plants), have been demonstrated. Knowledge acquired from these experiments has been incorporated into two mechanistic models CHEMFAST and BIORUR specifically modelling radionuclide sorption/desorption from soil matrices and radionuclide uptake by/release from plants. These mechanistic models have been incorporated into an assessment model to enhance its prediction ability.