Provenance discrimination of fine sediments by mid‐infrared spectroscopy: Calibration and application to fluvial palaeo‐environmental reconstruction

Abstract

Determining sediment provenance allows a better understanding of fluvial palaeo-dynamics, and identifying involved watersheds, at broad spatio-temporal scales. Conventional approaches for source identification are usually based on the physical, mineralogical, geochemical, magnetic or isotopic properties of sediments. Rapid, non-destructive and, in well-established contexts, highly accurate, mid-infrared spectroscopy is an alternative method for investigating sediment sources. The present research objectives are: (i) to use the mid-infrared spectroscopy method to discriminate the provenance of fine sediments, by applying discriminant analysis on a large set of reference samples from three different watersheds in the Upper Rhine area (associated with the Rhine, Ill and Vosges tributaries); (ii) to clarify whether the provenance spectra signatures are influenced by riverine depositional contexts (bars versus banks) and, to some extent, by grain size and/or high organic matter content; and (iii) to apply the mid-infrared spectroscopydiscriminant analysis method to a study of fluvial palaeo-dynamics and determine the provenance of palaeo-channel infillings. Three main sedimentary sources, divided into eight sub-categories, have been characterized by 196 modern reference samples from 78 collecting sites. Discriminant analysis displayed a strong separating power by classifying correctly the origin of samples without any inter-group overlap, independently from the geomorphological context (bar or bank) and associated slight changes in organic matter contents or grain size. Mid-infrared spectroscopydiscriminant analysis investigations of the palaeo-channel infill, complemented by radiocarbon dates and mineralogical data, allowed reconstructing general trends for the local morpho-sedimentary dynamics over the last ca 12 millennia.

Improving the quantification of sediment source contributions using different mathematical models and spectral preprocessing techniques for individual or combined spectra of ultraviolet–visible, near- and middle-infrared spectroscopy

“…Possibly, the SGD2 technique may be eliminating features (predictor variables) of the spectra that are important for the prediction of sediment sources. Nevertheless, several studies have used the 2nd derivative technique to treat UV-Vis, NIR and MIR spectra (Tiecher et al, 2015(Tiecher et al, , 2016(Tiecher et al, , 2017 and MSC (Ni et al, 2019) and SNV (Chapkanski et al, 2019) techniques to treat MIR spectra, with the objective to derive the characteristics of interest along the spectrum and to improve the predictions of the models of sediment source contributions. However, as observed in the current research, these pre-processing methods do not appear to be the most promising.…”

“…They include studies in the ultraviolet-visible (UV-Vis) and near-infrared (NIR) ranges in the French Alps (Legout et al, 2013) and the Southern France (Uber et al, 2019), Luxemburg (Martínez-Carreras et al, 2010c, 2010b, 2010a, Spain (Brosinsky et al, 2014a(Brosinsky et al, , 2014b, Ethiopia (Verheyen et al, 2014), South Africa (Pulley et al, 2018;Pulley and Rowntree, 2016), Argentina (Batistelli et al, 2018), the United Kingdom (Collins et al, 2014), Canada (Barthod et al, 2015;Boudreault et al, 2018;Liu et al, 2017), Brazil (Tiecher et al, 2016(Tiecher et al, , 2015Valente et al, 2020) and Iran (Nosrati et al, 2020). In the middle infrared region (MIR), sediment fingerprinting studies were carried out in France (Poulenard et al, 2012(Poulenard et al, , 2009, Mexico (Evrard et al, 2013), the United Kingdom (Vercruysse and Grabowski, 2018), China (Liu et al, 2019), Brazil (Tiecher et al, 2017) and in a transnational river catchment covering part of Switzerland, France and Germany (Chapkanski et al, 2019). Many of these studies have shown a good agreement between the results obtained with the spectroscopic method and those provided by the conventional approach based on geochemical and / or radionuclide properties Legout et al, 2013;Martínez-Carreras et al, 2010c;Tiecher et al, 2015Tiecher et al, , 2016Tiecher et al, , 2017Verheyen et al, 2014).…”

“…In addition to the more qualitative attempts that combine the use of discriminant analysis with spectroscopy (Chapkanski et al, 2019), previous studies conducted to trace sediment source contributions using spectroscopy (Table 1) can be divided into three main groups, according to the way they used the spectroscopic information. The first group uses color parameters extracted from the visible range (Barthod et al, 2015;Martínez-Carreras et al, 2010c, 2010b, 2010aPulley et al, 2018;Pulley and Rowntree, 2016) and other spectral characteristics (spectral features and overtones) (Brosinsky et al, 2014b(Brosinsky et al, , 2014aCollins et al, 2013Collins et al, , 2014 in an optimized mixed linear model, separately or in combination with conventional geochemical tracers (Tiecher et al, 2015).…”

“…However, so far, very few fingerprinting sediment tracing studies have reported whether they used any type of spectral pre-processing. Most of them used only pre-processing (Brosinsky et al, 2014b(Brosinsky et al, , 2014aChapkanski et al, 2019;Tiecher et al, 2017Tiecher et al, , 2016Tiecher et al, , 2015, except for Ni et al (2019), who used four pre-processing techniques, although they did not compare them to each other.…”

Analyses des matériaux de construction bruts de Portus par la Spectroscopie Moyen InfraRouge (SMIR)

“…Après l'acquisition spectrale, la matrice de données a été standardisée et alignée à sa base en utilisant le logiciel Unscrambler (Camo Unscrambler 10.3). Des précisions techniques sont détaillées par S. Chapkanski et collaborateurs 8 . Le spectre de quartz est fourni comme un exemple.…”