The soil adsorption behaviour of atrazine and its photodegradation intermediate 4,6-diamino-2-chloro-1,3,5-s-triazine (I), was evaluated by measuring partition coefficients on five representative European topsoils using classical batch experiments. From this study, (I) appears to be less immobilised than atrazine itself, especially for typical agricultural soils. Consequently, (I) has a higher tendency to reach the saturated zones, and hence, the groundwater. In order to remove atrazine and its degradation products from groundwater prior to making it drinkable, photocatalytic degradation onto membranes immobilizing 30±3 wt.% TiO 2 has been investigated as a possible technology to mineralize these waters satisfactorily. For this purpose, experiments have been carried out using a pilot-plant facility, with 0.02 -0.10 mmol L -1 solutions of atrazine, as well as of compound (I) and of 2,4,6-trihydroxy-1,3,5-s-triazine (II). When employing an excess of hydrogen peroxide as the oxidizing agent, the monitoring of chemical oxygen demand (COD), as well as control by HPLC, showed that (II) was a photostable degradation product, owing to its resistance to attack by hydroxyl radicals. On the other hand, when kinetic runs were carried out onto membranes in the presence of dissolved oxygen only or of sub-stoichiometric quantities of H 2 O 2 , (II) could be quantitatively mineralized. A mechanism has been proposed, based on the possible direct involvement of holes in the irradiated semiconductor surface, together with superoxide anions, these latter being produced by the reaction of conduction band electrons with dissolved oxygen. This mechanism, which is inhibited by an excess of hydrogen peroxide, owing to its reaction yielding hydroxyl radicals by any route, is also substantiated by previous experiments, carried out in the presence of ozone, with and without photocatalytic promoters.