Green roofs may induce many benefits as well at the building scale (thermal and acoustical comfort of the building) as at the city scale (reduction of urban runoff, mitigation of city climate, landscape aesthetics (Wong et al., 2003; Rowe, 2010, Susca et al. 2011). However there is less consensus among the authors about other ecosystemic services like retention of pollutants from atmospheric fallout or carbon sequestration. Several short studies on run-off quality report different degrees of nutrients release and trace metal mobility (Hunt et al., 2006; Teemusk and Mander, 2007, Czermiel-Berndtsson, 2010, Beck et al., 2011, Seidl et al 2012), but rare are those worked out over long periods or handling organic pollutants. The presentation proposes therefore to discuss the incidence of vegetated roofs on the concentration of (micro)pollutant and their loads emitted from roofs into the runoff, during one year of follow up of experimental green roofs in the suburbs of Paris. In order to test effects of the green roof composition on the runoff water quality, an existing flat roof of 300 m², was transformed to 6 green roofs and 2 reference roofs combining 2 types of plants, 2 types of substrates, 2 different substrate depths and 2 types of drainage layers. The two reference roofs consisted of a flat bituminous roof with and without gravel cover. During 2012 a dozen of rain events have been followed, monitoring in detail the hydraulic fluxes and using whole event samplers for the chemical parameters. The results can be divided in three pollutant groups: the global ones (nutrients), the metals (iron, copper, zinc) and the organic pollutants (PAH, alkylphenols, nonylphenol, MCPP). The emitted contaminant loads are dependant of the hydrologic behaviour of the roof and thus on soil moisture condition and rain depth. The behaviour of vegetated roofs towards pollutant depends moreover on the characteristics of the considered pollutant like origin and molecular properties. The green roofs studied appear as source of solids accompanied by carbon, phosphorus and in some extend nitrogen. As the roofs have about six month at the beginning of sampling, the loss of solid and nutrients issued from the green roof support is not very surprising. This behaviour is usual for new roofs and should diminish with the time. This release trend is also followed by major metals like iron and manganese. However some trace metals like zinc and in less extend copper are retained (Figure 1). The organic micro-pollutants give a similar picture. Some, like PAH are retained and others, like nonylphenol are released. On one hand these results suggest a release of components originating from the roof growth media like nutrients or some metals, and release from the roof structure like nonylphenol or MCCP. On the other hand, retention can be observed for substances coming with atmospheric fall out like zinc and PAH. The influence of roof types and the effect of meteorological conditions will be discussed in detail for each pollutant group.