<p>Sediment provenance is an important factor in understanding soil erosion processes or assessing the ecological effects of soil and water conservation measures. Sediment fingerprinting is an effective technique used globally for identifying sediment sources. Few studies have examined sediment sources at different spatial scales. In this study, sediment fingerprinting was used with a Bayesian mixing model to quantify the relative contributions of different sediments to streambeds in the Hebei catchment (ca. 28.0 km<sup>2</sup>) and its sub-catchment (ca. 3.5 km<sup>2</sup>) in the black soil region of Northeast China. Three potential sediment sources were identified: cultivated topsoil, uncultivated topsoil, and gullies. A similar number of sediment samples were collected for each source in both catchments: 71 and 69 sediment samples from the sub-catchment and Hebei catchment, respectively. Five uniformly distributed streambed sediment samples were collected from each catchment. The results showed a significant difference in the spatial variability of fingerprinting properties between the two catchments (p < 0.01). The spatial variability in fingerprint properties of cultivated topsoil and gully soil was more sensitive to scale than that of uncultivated topsoil. The optimum composite fingerprint that was used to discriminate potential sediment sources differed between the sub-catchment and Hebei catchment. Cultivated topsoil and gully soil were the main sediment sources, comprising more than 95% of the streambed sediment. There were significant differences (p < 0.01) in the contributions of cultivated topsoil and gully soil at different spatial scales. Cultivated topsoil contributed 47.8% and 42.0% in the sub-catchment and Hebei catchment, respectively, whereas gully soil contributed 49.6% and 55.3% (mean absolute fit >0.95). The upper stream segment mainly received sediment from the gullies (>60%) and the contribution from cultivated topsoil gradually increased downstream.</p>
