Sediment source determination comparing rare earth element tracing and composite fingerprinting approaches on hillslopes

Abstract

Both the rare earth element (REE) tracing and the composite fingerprinting approaches are valuable for sediment source identification. However, few studies have compared the accuracy of sediment source determination based on these two approaches, particularly for coarse-textured soils. This study combined simulated rainfall experiments with artificial mixtures, providing validation data for sediment contribution estimation. Simulated rainfall experiments were conducted using lanthana (La₂O₃), yttria (Y₂O₃), and ceria (CeO₂) separately tagged Acrisols, Chernozems, and Arenosols at 10°, 15°, and 20° slope gradients under a 120 mm h⁻¹ rainfall intensity. Sediment from different soils during 1-h (1 h) erosion process was continuously and separately collected within a 6-min time interval. Then, artificial mixtures were created using sediment from different soils over the same collection time. Sediment contributions were estimated using bulk samples (i.e., <1000 μm) for REE tracing, while they were estimated using a series of particle size ranges (i.e., <10, 10–63, 63–125, 125–250, 250–500, and 500–1000 μm) according to the composite fingerprinting for different source soil groups (i.e., Acrisols–Chernozems, Acrisols–Arenosols, Chernozems–Arenosols, and Acrisols–Chernozems–Arenosols). Here, we also analyzed the impacts of particle correction based on REE enrichment ratio (ER) within fine particles (<10, <63, and 10–63 μm). The results showed that sediment contribution accuracy based on the bulk samples was relatively high for both fine-textured and coarse-textured soils (RMSE<13.4%) on hillslopes, with or without adopting the particle correction factor. Whereas the accuracy of sediment contributions determined using different particle size ranges greatly varied for fine-textured source soils, while all the results presented significant (p < 0.05) differences compared to those obtained using the sediment weighting approach when coarse-textured source soil was included. Moreover, particle correction based on ER values of fine particle size fractions in which REEs were mainly enriched had no obvious effects on decreasing sediment contribution estimation bias. Additionally, particle correction had a high risk of decreasing estimation accuracy of sediment contributions using both REE tracing and the composite fingerprinting approaches. For the bulk samples, ER-corrected sediment contributions were significantly different (p < 0.05) from sediment weighting and uncorrected results when a coarse-textured soil (i.e., Arenosols) was included in the source soils. This was also the case for Acrisols and Chernozems, particularly on 15° hillslopes. Clearly, both REE tracing and composite fingerprinting are useful for sediment source determination, and sediment bulk samples normally provide robust results. Additionally, particle correction is not recommended when sediments have relatively weak sorting effects.