Abstract:Lithium, a rare metal of strategic importance, has garnered heightened global attention. This investigation delves into the laboratory visible-near infrared and short-wavelength infrared reflectance (VNIR-SWIR 350 nm-2500 nm) spectral properties of lithium-rich rocks and stream sediments, aiming to elucidate their quantitative relationship with lithium concentration. This research seeks to pave new avenues and furnish innovative technical solutions for probing sedimentary lithium reserves. Conducted in the Tuanjie Peak region of Western Kunlun, Xinjiang, China, this study analyzed 614 stream sediments and 222 rock specimens. Initial steps included laboratory VNIR-SWIR spectral reflectance measurements and lithium quantification. Following the preprocessing of spectral data via Savitzky-Golay (SG) smoothing and continuum removal (CR), the absorption positions (Pos2210nm, Pos1910nm) and depths (Depth2210, Depth1910) in the rock spectra, as well as the Illite Spectral Maturity (ISM) of the rock samples, were extracted. Employing both the Successive Projections Algorithm (SPA) and genetic algorithm (GA), wavelengths indicative of lithium content were identified. Integrating the lithium-sensitive wavelengths identified by these feature selection methods, A quantitative predictive regression model for lithium content in rock and stream sediments was developed using partial least squares regression (PLSR), support vector regression (SVR), and convolutional neural network (CNN). Spectral analysis indicated that lithium is predominantly found in montmorillonite and illite, with its content positively correlating with the spectral maturity of illite and closely related to Al-OH absorption depth (Depth2210) and clay content. The SPA algorithm was more effective than GA in extracting lithium-sensitive bands. The optimal regression model for quantitative prediction of lithium content in rock samples was SG-SPA-CNN, with a correlation coefficient prediction (Rp) of 0.924 and root-mean-square error prediction (RMSEP) of 0.112. The optimal model for the prediction of lithium content in stream sediment was SG-SPA-CNN, with an Rp and RMSEP of 0.881 and 0.296, respectively. The higher prediction accuracy for lithium content in rocks compared to sediments indicates that rocks are a more suitable medium for predicting lithium content. Compared to the PLSR and SVR models, the CNN model performs better in both sample types. Despite the limitations, this study highlights the effectiveness of hyperspectral technology in exploring the potential of clay-type lithium resources in the Tuanjie Peak area, offering new perspectives and approaches for further exploration.

Estimating Lithium Concentrations in Groundwater Used as Drinking Water for the Conterminous United States

Evaluation of potential human health risks associated with Li and their relationship with Na, K, Mg, and Ca in Romania’s nationwide drinking water

Lithium in Drinking Water and its Potential Impact on Mental Health: A Review

Towards an environmental quality standard (EQS) for lithium in fresh water

Concentration-dependent effects of lithium on Daphnia magna : Life-history profiles and integrated biomarker response implementation

Evaluation of potential human health risks associated with Li and their relationship with Na, K, Mg, and Ca in Romania's nationwide drinking water

Estimates of lithium mass yields from produced water sourced from the Devonian-aged Marcellus Shale

Predictions of groundwater PFAS occurrence at drinking water supply depths in the United States

Predicting the performance of lithium adsorption and recovery from unconventional water sources with machine learning

Geological controls on lithium production from basinal brines across North America

Lithium an emerging contaminant: bioavailability, effects on protein expression, and homeostasis disruption in short-term exposure of rainbow trout

The potential water quality impacts of hard-rock lithium mining: Insights from a legacy pegmatite mine in North Carolina, USA

Perfluoroalkyl and Polyfluoroalkyl Substances in Groundwater Used as a Source of Drinking Water in the Eastern United States

Predictions of Arsenic in Domestic Well Water Sourced from Alluvial Aquifers of the Western Great Basin, USA

Influences of Lithium on Soil Properties and Enzyme Activities

Reflectance spectroscopy analysis and lithium content estimation in lithium-rich rocks and stream sediments: Insights from Tuanjie Peak, Western Kunlun, China

Lithium and its isotopes as tracers of groundwater salinization: A study in the southern coastal plain of Laizhou Bay, China

Evaluation of the lithium resource in the Smackover Formation brines of southern Arkansas using machine learning

Reimagining safe lithium applications in the living environment and its impacts on human, animal, and plant system

Behaviors of lithium and its isotopes in groundwater with different concentrations of dissolved CO2

Predicting Serum Levels of Lithium-Treated Patients: A Supervised Machine Learning Approach