Quantifying Sediment Provenance Using Multiple Composite Fingerprints in a Small Watershed in Oklahoma.

Abstract

Quantitative information on sediment provenance is needed for improved calibration and validation of process-based soil erosion models. However, sediment source data are often limited due to difficulties in directly measuring source contributions at a watershed scale. Our objectives in this study were to estimate sediment source contributions in a 15-km watershed using analytical solutions to a three end-member mixing model using multiple composite fingerprints and to compare the results with those estimated with a single radionuclide, Cs. Surface soil samples were collected from 23 croplands, 19 rangelands, and 26 gully banks in the watershed, and 31 geochemical elements were analyzed for each sample. The elements served as tracers and were screened using statistical tests and range checks. The mean concentrations of all the nonconflict tracer pairs were used in the mixing model to calculate source contributions for the three sources. Results showed that although source contributions were strongly influenced by topography and land use, gully or subsoil erosion was found to be the main source of fine sediment in most subwatersheds. This study demonstrates that estimated source contributions may vary substantially among different composite fingerprints and that the use of multiple composite fingerprints greatly improves accuracy while reducing uncertainty. The source contributions estimated using multiple composite fingerprints agreed well with those estimated with Cs, with a correlation coefficient of 0.69 for gully contributions. This good agreement increases our confidence in using the multiple composite fingerprint method to identify sediment provenance in relatively small watersheds.