Ferralsols are highly weathered soils characteristic of the moist tropics. Many Ferralsols have kaolinitic horizons overlying more gibbsitic ones for which there has been no satisfactory pedological explanation. This has often been considered as resulting from sedimentary succession. However, studies of Ferralsols in the Manaus region of Amazônia showed that the kaolinite-over-gibbsite profile was developed in situ. To understand the processes Cornu et al. (1998) determined an annual balance of the transfer of dissolved Si, Fe, and Al by rain, through fall, stemflow, soil water and groundwater run off in an Amazonian ferralsol. They concluded that the inputs of dissolved Fe, Al, and Si to the topsoil were small but not negligible and that the balances of the elements in the upper soil horizons have to take into account translocation of dissolved elements in the slow percolating water. In order to achieve this we build up a soil column experiment considering suction at the column base in order to collect the slow percolating waters. This experiment was coupled with geochemistry modeling. With this approach we attempt to model the stabilities of the mineral phases kaolinite and gibbsite. The thermodynamic modeling used in this study provided a realistic approach on the stability of kaolinite and gibbsite in this ferralsol. We underline the importance of choosing the model for DOM complexation. It takes a role in the amount of Al-DOM complexes content, which in turns may or may not allow the formation of the different solid phases. These preliminary results are encouraging to simulate the geochemistry and water transfer interaction along the one dimension column by using inverse modeling and a coupled transport model.