International Journal of Applied Earth Observations and Geoinformation (Dec 2021)
Quantitative estimation of rare earth element abundances in compositionally distinct carbonatites: Implications for proximal remote-sensing prospection of critical elements
Abstract
Rare earth elements (REE) became a strategic raw material in the 21st century and carbonatite-related deposits frequently carry potentially economic levels of these critical metals. To determine concentrations of major and trace elements (including REE) conventional analysis of bulk rock geochemical samples by means of inductively coupled plasma optical emission spectrometry (ICP-OES) and inductively coupled plasma mass spectrometry (ICP-MS) is mostly applied. However, these analytical methods are labor-intensive, costly and time-consuming. In contrast, modern Remote Sensing (RS) – e.g. proximal remote sensing or imaging spectroscopy – has become a novel tool for detecting and quantifying geological materials as they offer an efficient, non-destructive and cost-effective way not only to identify different minerals and/or chemical elements but also model their abundances. In this study, we investigated whether Partial Least Squares Regression (PLSR) models linking the laboratory sample reflectance to mineral/geochemical property could be employed for quantitative predictions of bulk LREE and HREE concentrations. The suite included samples from four different geographic regions in India, East Africa and the USA. Our results showed that for such samples the quantitative approaches have some limitations.First, our models required to exclude samples with high modal abundance of hematite and samples with too low REE contents (LREE < 500 μg/g, HREE < 100 μg/g). Furthermore, it was revealed that the sample mineralogy has a significant impact on the PLSR predictions. To achieve reliable models the sample suite had to be divided into two datasets following their mineralogy and geochemistry. The first dataset comprised rocks with increased amounts of strontium (Sr, DatasetSr), where REE were predominantly bound in Sr-carbonates. On the other hand, REE mainly bound in Ca–REE-carbonates characterized the second dataset (DatasetOther). Following this division, we were able to construct valid prediction models for bulk LREE and HREE concentrations using PLSR. However, the detected spectral assignments associated with the REE presence indicated that the predictions were mainly indirect, based on the present mineral phases rather than direct absorptions related to LREE and HREE. This illustrates the difficulty and limitations for further model generalization and the ability to be further transferred to other lithologically diverse carbonatite sites.We posit that this topic requires future systematic investigations using the extended carbonatite datasets – samples having a wide range of REE abundances – collected from lithologically diverse regions. This would enable further validation and re-calibration of the constructed prediction models.
