Combining visible-based-color parameters and geochemical tracers to improve sediment source discrimination and apportionment
Résumé
Parameter selection in fingerprinting studies are often time-consuming and costly because successful fingerprint properties are generally highly site-specific. Recently, spectroscopy has been applied to trace sediment origin as a rapid, less expensive, non-destructive and straightforward alternative. We show in this study the first attempt to combine both geochemical tracers and color parameters derived from the visible (VIS) spectrum in a single estimate of sediment source contribution. Moreover, we compared the discrimination power and source apportionment using VIS-based-color parameters and using the whole ultra-violet-visible (UV-VIS) spectrum in partial last square regression (PLSR) models. This study was carried out in a small (1.19 km(2)) rural catchment from southern Brazil. The sediment sources evaluated were crop fields, unpaved roads, and stream channels. Color parameters were only able to discriminate unpaved roads from the other sources, disabling its use to fingerprint sediment sources itself. Nonetheless, there was a great improvement in source discrimination combining geochemical tracers and color parameters. Unlike VIS-based-color parameters, the distances between sediment sources were always significantly different using the whole UV-VIS-spectrum. It indicates a loss of information and, consequently, loss of discriminating power when using VIS-based-color parameters instead of the whole UV-VIS spectrum. Overall, there was good agreement in source ascription obtained with geochemical tracers alone, geochemical tracers coupled with color parameters, and UV-VIS-PLSR models, and all of them indicate clearly that the main sediment source was the crop fields, corresponding to 57 +/- 14, 48 +/- 13, and 62 +/- 18%, respectively. Prediction errors for UV-VIS-PLSR models (6.6 +/- 1.1%) were very similar to those generated in a mixed linear model using geochemical tracers alone (6.4 +/- 3.6%), but the combination of color parameters and geochemical tracers decreases the prediction error (5.4 +/- 2.0%). Therefore, the use of VIS-based-color parameters combined to geochemical tracers can be a rapid and inexpensive way to improve source discrimination and precision of sediment source apportionment.