Proximal multi-scenario SWIR hyperspectral dataset of clay-type lithium deposits from China with mineral abundances and geochemistry

Dataset overview The Proximal Multi-Scenario SWIR Hyperspectral Mineral Abundance Dataset is a proximal shortwave infrared (SWIR) hyperspectral dataset designed for mineralogical analysis, mineral abundance estimation, and cross-scenario spectral studies in geological materials. The dataset covers four complementary observation scenarios: powder, drill cores, hand specimens, and in-situ outcrops. These scenarios together provide a multi-scale and multi-condition framework for linking hyperspectral measurements with mineralogical, geochemical, spatial, and sampling-reference information. This dataset uses clay-type lithium mineralization hosted in carbonate successions as its sample source, with most samples collected from southwest China, mainly Yunnan and Guizhou provinces. These sampling areas are mainly located within the Yangtze Block. The dataset is organized into four main folders: 1-Powder 2-Drill core 3-Hand specimen 4-In-situ outcrop Within these folders, the dataset includes hyperspectral image data, mineralogical and geochemical analytical results, drill-core sampling annotations, white-reference spectra for empirical calibration, and geographic position information. This structure supports integrated studies of spectral response, mineral abundance, sample correspondence, and spatial context across multiple proximal sensing scenarios. Data forms and metadata Most hyperspectral image data in the dataset are provided in ENVI format, typically consisting of paired: .img .hdr files. The associated .hdr header files store the main metadata, including wavelength information and other acquisition-related parameters. For powder, drill cores, and hand specimens, the hyperspectral data are provided as reflectance or relative reflectance products generated during preprocessing. For in-situ outcrop scanning, the hyperspectral images are provided as radiometrically corrected data, and users are expected to further derive reflectance using the corresponding white-reference spectra and an empirical line calibration approach. In addition to hyperspectral image files, the dataset also includes supporting data in CSV, Excel, PNG, compressed archive, and ENVI spectral library formats, depending on data type and usage. Each major data category includes a README.pdf file that describes the data organization, file structure, data content and format information. These README files are provided to help users navigate the data packages and use the files appropriately for further processing and analysis. General folder structure 1. 1-Powder This folder contains SWIR hyperspectral data of powder samples together with their corresponding mineralogical and geochemical analytical results. The hyperspectral images are organized by drill hole, and sample identifiers provide direct links to the mineralogical and geochemical tables. Supporting files include processed QXRD results, raw QXRD data, XRD raw data, geochemical analytical results, and analytical quality-control information. The parent directory also includes representative 5 × 5 pixel average spectra extracted from the image center region in both .txt and .csv formats. 2. 2-Drill core This folder contains continuous drill core hyperspectral scanning data, sampling-location annotations, and drill hole position information. The drill core HSI data are organized by drill hole and stored as compressed archives, with individual cropped core segments named according to drill hole, tray number, and row number. Sampling annotation images indicate the positions of powder-sample collection points on drill core trays and provide the linkage between drill core observations and powder-based mineralogical and geochemical analyses. Geographic coordinates and elevation information for each drill hole are also included (position information.csv). 3. 3-Hand specimen This folder contains reflectance hyperspectral data of hand specimens. The data were acquired and processed using HySpex laboratory software with a fixed white-reference correction procedure. Each hand specimen dataset is stored in ENVI format, and metadata are preserved in the corresponding header files. 4. 4-In-situ outcrop This folder contains in-situ outcrop hyperspectral scanning data, white-board reference spectra, and outcrop position information. Unlike the other scenarios, the outcrop HSI data were processed only to the radiometric-correction stage and are not final reflectance products. White-board spectra measured by an SR-3500 spectroradiometer are therefore provided to support empirical line calibration from radiometrically corrected image data to apparent or relative reflectance. Geographic location and elevation information for each scanned outcrop are also included (position information.csv). Dataset significance By combining proximal SWIR hyperspectral measurements with mineralogical, geochemical, positional, and annotation information across multiple geological sample scenarios, this dataset provides a unified resource for: mineral abundance estimation, hyperspectral unmixing, spectral comparison across sample forms and acquisition conditions, calibration and validation of mineral mapping methods, and reproducible cross-scenario geological spectroscopy research. Sample types and ID coding scheme Sample type ID example Description Drill core DC1111S01 Sample IDs follow the format DCxxxxSxx, where DC denotes drill core, xxxx is the core code, S indicates spectral sample, and xx is the sequence number. Powder PDC1111S01 Sample IDs follow the format Pxxxxxxxxx, like PDCxxxxSxx, where PDC denotes drill-core powder sample. Exploration trench ET1111S01 Sample IDs follow the format ETxxxxSxx, where ET denotes exploration trench, xxxx is the trench code, S indicates spectral, and xx is the sequence number. Stripped outcrop ST1111S01 Sample IDs follow the format STxxxxSxx, where ST denotes stripped site, xxxx is the stripped outcrop code, S indicates spectral, and xx is the sequence number. Hand specimen HS0001R1 Sample IDs follow the format HSxxxxRx, where HS denotes hand specimen, xxxx is the sample code, and Rx indicates the scan number. In-situ outcrop OC0001R1 Sample IDs follow the format OCxxxxRx, where OC denotes outcrop scaning, xxxx is the sample code, and Rx indicates the scan number.