​Following the explosion at the Chernobyl nuclear power plant, about 800 shallow trenches were dug to bury radioactive waste in the exclusion zone. However, these trenches were built in permeable aeolian sand and do not prevent the migration of radionuclides in the superficial aquifer. Trench T22 allows us to explore general research problems such as the uncertainty linked to the volume and the activity of radioactive waste, and to the water quality in groundwater downstream the trench. Previous estimations of the trench inventory assume that the correlation between specific activity measured in soil samples and in situ count rate is linear. Geostatistical simulations are used to analyze this correlation and its sensitivity to the activity spatial variability and to its surrounding environment. If the correlation can be considered as linear, the study of field measurements undertaken in 2015 demonstrates that it is better to apply cokriging to estimate the activity by the count rate rather than transforming the count rate data. The inventory of 137Cs calculated for 1999 is then compared to a previous estimation. The trench boundaries are interpolated using the results of ground penetrating radar profiles and gamma logging carried on boreholes drilled into the trench. The new estimation is completed by geostatistical simulations and enables us to quantify the uncertainty of 137Cs trench inventory. Finally, the effect of the source term spatial variability is explored with the 90Sr migration modeling. The previous 90Sr transport model did not take into account the water table fluctuations which may cause some discrepancies between model predictions and field observations. They are thus reproduced in a 2D non saturated model. A sensitivity analysis on the flow and transport parameters as well as the source term variability is undertaken.