Skip to main content
ARS Home » Research » Publications at this Location » Publication #313497
Title: A new method for fingerprinting sediments source contributions using distances from discriminant function analysis

Author
item LIU, BING - China Institute Of Water Resources
item STORM, DANIEL - Oklahoma State University
item Zhang, Xunchang
item CAO, WENHONG - China Institute Of Water Resources
item DUAN, XINGWU - Beijing Normal University

Submitted to: Catena
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/27/2016
Publication Date: 7/5/2016
Citation: Liu, B., Storm, D.E., Zhang, X.J., Cao, W., Duan, X. 2016. A new method for fingerprinting sediments source contributions using distances from discriminant function analysis. Catena. 147:32-39.

Interpretive Summary: Soil erosion varies over landscapes, and information on soil loss contributions at specific locations or landform units is essential for laying out precision conservation measures. Sediment tracers and mixing numerical models have been used to predict sediment source contributions. The objective of this study is to develop and evaluate a new method using Discriminant Function Analysis (DFA) to fingerprint sediment source contributions. We hypothesized that the information given by the DFA can be potentially used to directly predict source contributions without using the mixing models. Results showed that both mixing model and DFA method were useful for predicting sediment source contributions. The mixing model provided useful information about the sediment source contributions and performed best with two sources. However, when the number of sources was greater than two, the DFA method may be a better method since it avoids spurious numerical solutions. The sediment ascription method is useful to erosion scientists and soil and water conservationists for identifying major sediment source types or areas for better conservation planning.

Technical Abstract: Mixing models have been used to predict sediment source contributions. The inherent problem of the mixing models limited the number of sediment sources. The objective of this study is to develop and evaluate a new method using Discriminant Function Analysis (DFA) to fingerprint sediment source contributions. We hypothesized that the information given by the DFA can be potentially used to directly predict source contributions without using the mixing models. The hypothesis of using the DFA method to replace mixing models was tested at the Bull Creek watershed in Oklahoma State. Both mixing model and DFA method were used to predict source relative contributions at the subwatershed to watershed scales. The results showed that the mixing model provided useful information about the sediment source contributions and performed best with two sources. For the Bull Creek study area, the mixing model predicted the source contributions with the lowest model errors for two sources; however, the results were not significantly different from those of the DFA method. When the number of sources was greater than two, the DFA method may be a better method since it avoids spurious numerical solutions.