Since 1964, the Temple sur Lot dam, built in south west of France in 1948, has presented continuous AAR induced displacements despite low and relatively constant alkali content in the concrete and non-significant residual swelling test results. The authors have shown, based on MEB observations and SEM analysis, that substitution process between alkali and calcium into the AAR gel could explain this long term dam’s behaviour. As calcium substitution phenomenon is very slow, it cannot be detected using a classical residual swelling test, so an original method to assess the AAR kinetic and the residual swelling capability is proposed. This method consists first in a laboratory test dealing with the silica consumption kinetic and secondly in a numerical finite element inverse analysis of the dam which include the consumption kinetic measured in laboratory. The final swelling amplitude is thus fitted from one observed structural displacement rate. The model prediction capability is validated through the comparison between the evolutions of displacement of instrumented points predicted by the calculations (unused for the fitting) and the evolutions measured on the dam. At last, calculations have been performed in order to predict the displacements and the damage fields of the dam for the next decades.