Sediment sources in a small agricultural catchment: A composite fingerprinting approach based on the selection of potential sources

Fingerprinting techniques have been widely used as a reasonable and reliable means for investigating sediment sources, especially in relatively large catchments in which there are significant differences in surface materials. However, the discrimination power of fingerprint properties for small catchments, in which the surface materials are relatively homogeneous and human interference is marked, may be affected by fragmentary or confused source information. Using fingerprinting techniques can be difficult, and there is still a need for further studies to verify the effectiveness of such techniques in these small catchments. A composite fingerprinting approach was used in this study to investigate the main sources of sediment output, as well as their relative contributions, from a small catchment (30 km(2)) with high levels of farming and mining activities. The impact of the selection of different potential sediment sources on the derivation of composite fingerprints and its discrimination power were also investigated by comparing the results from different combinations of potential source types. The initial source types and several samples that could cause confusion were adjusted. These adjustments improved the discrimination power of the composite fingerprints. The results showed that the composite fingerprinting approach used in this study had a discriminatory efficiency of 89.2% for different sediment sources and that the model had a mean goodness of fit of 0.90. Cultivated lands were the main sediment source. The sediment contribution of the studied cultivated lands ranged from 39.9% to 87.8%, with a mean of 76.6%, for multiple deposited sediment samples. The mean contribution of woodlands was 21.7%. Overall, the sediment contribution from mining and road areas was relatively low. The selection of potential sources is an important factor in the application of fingerprinting techniques and warrants more attention in future studies, as is the case with other uncertainty factors. (C) 2016 Elsevier B.V. All rights reserved.