Dépôt numérique

A pan-Canadian calibration of micro-X-ray fluorescence core scanning data for prediction of sediment elemental concentrations.

Zilkey, David R. ORCID logoORCID: https://orcid.org/0000-0002-8841-7665; Baud, Alexandre; Francus, Pierre ORCID logoORCID: https://orcid.org/0000-0001-5465-1966; Antoniades, Dermot et Gregory-Eaves, Irene (2024). A pan-Canadian calibration of micro-X-ray fluorescence core scanning data for prediction of sediment elemental concentrations. Environmental Advances , vol. 15 . p. 100495. DOI: 10.1016/j.envadv.2024.100495.

Ce document n'est pas hébergé sur EspaceINRS.


Sediment geochemistry is one lens through which lake sediments are studied to reconstruct local and regional environmental processes. The measurement of sediment elemental composition has historically relied on expensive and destructive methods that limit the spatial and temporal scale of study. Micro-X-ray fluorescence (µXRF) core scanning offers a non-destructive, high-resolution alternative, but its results (i.e., intensity expressed as counts per second) are considered semi-quantitative and comparison among sites requires calibration. Calibration methods are emerging, although they are not yet widely employed and require further assessment of their efficacy. Using 135 sediment samples from 48 lakes across Canada, we assessed the congruence between µXRF and conventionally measured element compositions with various normalization and calibration techniques. Normalization of µXRF data to common proxies (e.g., Ca, Si, Ti, coherence:incoherence ratio, and total counts per second) often improved correlations between µXRF and conventional data, but increases were modest and not consistent for all elements. Our results suggest that µXRF normalization techniques should be applied cautiously, as no proxy represents a “one-size-fits-all” solution. The performance of multivariate log-ratio calibration (MLC) was more consistent, yielding moderate to strong improvement of the correlations between reference and predicted element concentrations. Random forest regression models outperformed partial least squares regression models for almost all elements. MLC may be applied where knowledge of elemental concentration is of great importance, or when comparing across multiple sites with diverse sediment geochemistries. Overall, our results reinforce uncalibrated µXRF core scanning as a strong investigative tool for measuring sediment geochemistry. Although calibrated µXRF data shows promise, conventional methods for measuring sediment geochemistry are still necessary for comparing element concentrations with sediment quality guidelines.

Type de document: Article
Mots-clés libres: Paleolimnology; µXRF; lake sediments; geochemistry
Centre: Centre Eau Terre Environnement
Date de dépôt: 09 juill. 2024 14:51
Dernière modification: 09 juill. 2024 14:51
URI: https://espace.inrs.ca/id/eprint/15492

Gestion Actions (Identification requise)

Modifier la notice Modifier la notice