Contamination by pesticides may often be a cause of non achievement of the good status of water required by the Water Framework Directive for 2015. In France, the Ecophyto 2018 plan insists on the decreasing by one half of the quantity of pesticides used by agriculture before 2018. Although such a measure goes in the right direction, it will not prevent all the substances applied in the fields to reach the water resource. Taking advantage of the capacity of pesticides to be held back and degraded during their transfer from the field where they are applied to the receipting water table or river, buffer zones can represent useful tools limiting pesticides fluxes transferred towards surface waters (and groundwater in the case of karstic areas). Three main kind of buffer zones can be cited: buffer zones where infiltration processes are dominant (strips, meadows, woods ...) ; humid woods where runoff circulates in the humic surface layer and artificial (or restored) wetlands. In each case, the principles are similar: slowing down the pesticide in order to let time for it to be degraded, in an environment rich in organic matter and biological activity. Nevertheless, each kind of buffer requires specific conditions to be efficient (Gregoire et al., 2009). For example, vegetative strip are efficient when infiltration of water is not limited by a saturated soil or by concentration of flow (Carluer et al., Soumis; Lacas et al., 2005). As a consequence, the ensemble of mitigation solutions to be set up on a catchment must be designed conjointly, taking into account advantages and limits of buffers' ecological services, considering the local context to build a global solution where the different kind of buffers are complementary (CORPEN, 2007). Results coming from experiences realised by Cemagref teams are presented to illustrate these considerations and to give orders of magnitude of the efficiency of these different kinds of buffer zones. For example, an experiment was leaded in the Beaujolais vineyard (France), on a sandy-loam soil, with high slopes (15-25%). A vegetative strip of 6 meters long, situated between a vineyard plot and the river was submitted to runoff events, either coming from the upslope field or simulated. Even for events corresponding to a two years return period, all the incoming water was infiltrated, on the condition that the soil was not initially too wet. Moreover, four lysimeters, situated under the strip, 50 cm deep, showed that only 30 to 52 % of the applied pesticides moved below 50 cm, depending on their adsorption capacity (Koc) (Gouy et al., 2007). An other example is given by the western site of La Jaillière, near Angers ; there, soils are hydromorphic, due to the presence of a perched table in winter. In these conditions, field evidence and numerical simulations show that the infiltration capacity is very limited so that vegetative strip efficiency is poor during the tile drainage season, that is to say when pesticides exportation is significant. In such an example, it seems necessary to use other kind of buffer zones, as humid woods or artificial wetlands.