Diffuse and point sources of dissolved organic matter (DOM) in streams influence its composition, interactions and fate in the aquatic ecosystem. These inputs can be very numerous at the scale of a watershed, and their identification remains a challenge, especially for diffuse sources related to land use. The complexity of the transfer mechanisms and the reactivity of DOM throughout the soil-water column continuum raise questions about the sampling of diffuse sources in watercourses. To answer this issue, we compared the characteristics of soil-extracted DOM influenced by a particular land use (homogenous sub-catchment of forest and vineyard) and DOM collected from the watercourse adjacent to the soil samples. A 28-day incubation experiment of soil extracts was designed to remove the labile fraction of DOM. During the first 3 days, between 40 and 70 % of the DOC mass was lost for both types of soils. A set of optical indicators (UV–Visible, EEM fluorescence and HPSEC/UV-fluorescence) showed that the labile fraction was mostly composed by low (<1 kDa) and high (>10 kDa) protein-like molecules. At the end of the incubation, soil-extracted DOM was mainly composed of medium molecules (1–10 kDa) associated to terrigenous humic-like compounds. Its optical and size molecular signature tended towards that in the adjacent watercourses and was specific to land use. However, the characteristics of DOM in watercourses was also influenced by the hydrological conditions, probably due to a transfer of top soil DOM during high water periods and both deep soil and autochthonous DOM during low water periods. These results were obtained by a set of indicators, including novel ones derived from HPSEC/UV-fluorescence. Finally, this study demonstrated that it is possible to sample the DOM representative of a land use directly in the river downstream of the homogeneous sub-basin by multiplying the samples during contrasting hydrological conditions.