Fingerprinting sediment sources in the outlet reservoir of a hilly cultivated catchment in Tunisia

Purpose Approximately 74 % of agricultural soils in Tunisia are affected by water erosion, leading to the siltation of numerous human-made reservoirs and therefore a loss of water storage capacity. The objective of this study was to propose a methodology for estimating the relative contributions of gully/channel bank erosion and surface topsoil erosion to the sediment accumulated in small reservoirs. Materials and methods We tested an approach based on the sediment fingerprinting technique for sediments collected from a reservoir (which has been in operation since 1994) at the outlet of a catchment (Kamech, 2.63 km 2 ). Sampling concentrated on the soil surface (in both cropland and grassland), gullies and channel banks. A total of 17 sediment cores were collected along a longitudinal transect of the Kamech reservoir to investigate the origin of the sediment throughout the reservoir. Radionuclides (particularly caesium-137, 137 Cs) and nutrients (total phosphorus, total nitrogen and total organic carbon (TOC)) were analysed as potential tracers. Results and discussion The applications of a mixing model with 137 Cs alone or 137 Cs and TOC provided very similar results: The dominant source of sediment was surface erosion, which was responsible for 80 % of the total erosion within the Kamech catchment. Additionally, we showed that the analysis of a single composite core provided information on the sediment origin that was consistent with the analysis of all sediment layers in the core. We demonstrated the importance of the core sampling location within the reservoir for obtaining reliable information regarding sediment sources and the dominant erosion processes. Conclusions The dominance of surface erosion processes indicates that conservation farming practices are required to mitigate erosion in the agricultural Kamech catchment. Based on the results from 17 sediment cores, guidelines regarding the number and location of sampling cores to be collected for sediment fingerprinting are proposed. We showed that the collection of two cores limited the sediment source apportionment uncertainty due to the core sampling scheme to