[en] Environmental management decisions require the use of models with varying degree of sohpistication to represent the evolution of natural and engineering systems. These models are of a predictive kind and they are based on our interpretative knowledge of of time between 1.000 to 10.000 years. In order to build up our confidence in the possibilities of using the models to predict the geochemical behaviour of the future repository systems, we study natural systems than reflect processes and/or materials used in the repositories. These studies are known as Natural Analogues. We have undertaken, within these investigations the checking of geochemical models to describe the aqueous speciation and the solubility of trace elemento, i.e. U, Th, Sr... The mobility of these elements is critical to assess the behaviour of nuclear waste repositories. These tests are known as Blind Predictive Modelling exercises and attempt to simulate the modelling of radionuclide solubilities and speciation in a repository environment. It can be seen in the paper that the concentrations of uranium in solution are more accurately predicted when using the formation of a mixed solid phase of U(VI) and Fe(III) than assuming a pure U(VI) solid phase. On the other hand, the concentrations of Ba predicted are slightly higher than the actual ones if assuming a congruent codissolution of a mixed solid phase of Ba with calcite. This is very useful result when trying to asses the environmental risk. The same results are obtained when predicting the concentration of other trace metals, such as Mn and Sr