Comparative study using spectroscopic and mineralogical fingerprinting for suspended sediment source apportionment in a river–reservoir system

Abstract

The need to control soil erosion has received increasing attention, but quantitative data on the sources of suspended sediment in many river–reservoir systems is still lacking. The goal of this research was to compare the application of spectroscopic [mid-infrared (MIR)] and mineralogical [X-ray diffraction (XRD)] fingerprints for assessing relative sediment source contributions from different land use groups (agricultural lands, forests and human settlements) in the Konar–Damodar river–reservoir system in India. Source apportionment was estimated using partial least square (PLS) regression for spectroscopic tracers (MIR) and the Bayesian MixSIAR model for mineralogical tracers. Both methods identified differences between the pre- and post-monsoon sediment contributions of forests (overall contribution bounds of ~35–43%). During monsoon seasons, both fingerprinting methods indicated agricultural land use as the primary source of suspended sediment. Although there were some temporal variations in the predicted contributions of the land use sources, the MIR-PLS and mineralogical–MixSIAR methods produced comparable ranges. The respective variations in contributions, using MIR-PLS and mineralogical–MixSIAR, were ~31 to 66% compared with ~36 to 61% for agricultural lands, ~21 to 43% compared with ~15 to 39% for forests and ~16 to 37% compared with ~19 to 32% for human settlements.