Accuracy verification of optical fingerprinting methods in sediment tracing study

Abstract

AbstractSediment fingerprints have been widely used in source identification studies. Optical features have become a powerful substitute for traditional fingerprints in recent years due to their fast measurement and low analysis costs. However, the accuracy of optical fingerprinting methods has received little attention. Here, artificial mixing and indoor scouring experiments were carried out to compare and assess the accuracy of optical fingerprinting results using three spectroscopic ranges [visible (VIS), near‐infrared (NIR), and mid‐infrared (MIR) spectroscopy] in multivariate models and using 19 colour parameters [such as red (R), green (G) and blue (B) in an RGB system; virtual component X (X), brightness (Y) and virtual component Z (Z) in a CIE XYZ system and so on] coupled with the conventional method. Furthermore, we examined how sediment sorting (particle sorting and organic matter enrichment) affects the accuracy of source apportionments. The results showed that VIS, NIR and MIR spectroscopic tracers presented high accuracy in scouring and artificial mixtures, with mean absolute error (MAE) values of 4.98% and 5.91%. In contrast, the colour parameters had weak performances in two experiments (MAE = 16.83% and 15.05%). Additionally, similar fingerprinting results of the scouring mixtures (MAE = 7.95%) and artificial mixtures (MAE = 8.20%) indicated that slight particle sorting and organic matter enrichment have little effect on the accuracy of optical fingerprinting results. Our study shows that sediment fingerprinting based on optical features, especially three spectroscopic ranges, has good applicability in sediment source identification.