Soil organic carbon (SOC) is closely linked to soil fertility, agricultural carbon cycling, and the functioning of cotton field ecosystems, and it provides essential information for sustainable soil management. Rapid and accurate SOC estimation is therefore important for assessing carbon sequestration potential and supporting low-carbon agricultural management. This study focused on cotton fields in northern Shawan City and used optical imagery, Synthetic Aperture Radar (SAR) imagery, and 140 ground-collected SOC samples to estimate SOC content with three machine learning models: Random Forest (RF), Light Gradient Boosting Machine (LightGBM), and Extreme Gradient Boosting (XGBoost). The Kennard-Stone algorithm was applied to partition the 140 SOC samples into training and validation subsets at a 7:3 ratio, ensuring a more representative distribution of samples. Model performance was evaluated using the coefficient of determination (R2) and root mean square error (RMSE), and SHapley Additive exPlanations (SHAP) was used to interpret feature contributions and SOC spatial variability. The results showed that: (1) optical features performed better than SAR features, while fused optical-SAR features achieved the highest accuracy; (2) XGBoost consistently outperformed RF and LightGBM, with the optimal model achieving R2 = 0.726 and RMSE = 1.252% on the validation set; (3) SHAP analysis confirmed the dominant contribution of optical features to SOC estimation; and (4) the predicted SOC distribution showed higher values in the central study area, lower values in the northern and southern parts, and high-value zones mainly along both sides of the Manas River. By comparing optical, SAR, and fused features for SOC estimation in arid-zone cotton fields, this study provides methodological support for rapid SOC monitoring and sustainable soil management, and offers practical guidance for variable-rate fertilization and soil carbon sequestration planning along the Manas River corridor.

Geochemical anomaly detection plays a critical role in mineral exploration, yet conven-tional methods are often limited by compositional effects, sensitivity to outliers, and in-sufficient consideration of spatial relationships. To address these issues, this study pro-poses an integrated analytical framework that combines compositional data analysis and spatial statistics for robust geochemical anomaly identification. The framework incor-porates isometric log-ratio (ILR) transformation to eliminate the closure effect, robust principal component analysis (RPCA) to extract stable geochemical patterns, local indi-cators of spatial association (LISA) to characterize spatial clustering, and compositional balance analysis (CoBA) to enhance anomaly signals. The method is applied to the Barkol Lake area in the Eastern Tianshan, a key metallogenic belt within the Central Asian Orogenic Belt. The results reveal significant geochemical anomalies characterized by Cu-associated element assemblages (e.g., Cu–Ni–Cr), which are spatially correlated with major fault zones and volcanic–intrusive complexes. The identified anomalies show strong consistency with known mineral occurrences and delineate several prospective targets for copper polymetallic mineralization. Compared with conventional approaches, the proposed framework demonstrates improved robustness to outliers, enhanced sensi-tivity to weak anomalies, and better integration of compositional and spatial constraints. These advantages highlight its effectiveness for geochemical anomaly detection and mineral prospectivity mapping in complex geological settings.

Regional ecological risk assessments typically rely on interpolated toxic heavy metal (THM) surfaces derived from sparse field samples. However, this approach is fundamentally constrained by sampling density and interpolation errors, resulting in high spatial uncertainty that masks fine-scale contamination heterogeneity. Hyperspectral remote sensing offers a scalable alternative by enabling continuous spatial characterization of soil properties. This study evaluates an integrated ground-satellite hyperspectral framework to map THMs and ecological risks in the Baixintan (BXT) and Lubei (LB) mining areas (early-stage extraction). We developed a robust inversion workflow combining spectral transformations, band optimization, and machine learning algorithms, which effectively mitigated background noise and enhanced feature extraction. After calibrating satellite spectra with ground-based measurements, estimation accuracy for Cu, Ni, and Cr improved by 31%, 10%, and 34%, respectively, compared to uncorrected baselines. Based on these optimized models, we generated spatially continuous maps for four pollution indices (I_geo, INI, E_i, and RI). Results revealed distinct spatial patterns: while Cu and Ni showed mild-moderate enrichment (I_geo = 0.40-2.01), Cr exhibited severe enrichment (I_geo = 3.24-3.53). Crucially, the high-resolution mapping identified localized high-risk hotspots extending beyond documented mining footprints, likely driven by natural transport mechanisms (e.g., topography and wind) interacting with mining activities. Although the overall ecological risk remained predominantly low to moderate (mean RI of 61.4 in BXT and 82.5 in LB), the detection of these off-site contamination zones demonstrates the superior capability of hyperspectral inversion in capturing spatial nuances missed by traditional monitoring. This framework provides actionable, high-precision data for early-stage risk management and boundary-spanning pollution control.

Optical remote sensing has become an important tool for prospecting in the early stages of geological exploration for rare metal deposits. This analysis enables the identification and mapping of lithium (Li)-bearing minerals using multisource and multiscale remote sensing data. However, the mineral composition of granite pegmatite is complex, and the unique spectral characteristics caused by Li in the mineral structure are still uncertain, which also causes problems of low accuracy and a small range in remote sensing prospecting. Therefore, the purpose of this article is to investigate the scattering properties of granite pegmatite and lay the foundation for the subsequent analysis of grain size and mineral abundance inversion using the Hapke model. First, the relationship between particle size and bidirectional reflectance distribution function response is analyzed. Then, the relationship between the single scattering albedo (SSA) and the particle size and the endmember mineral abundances is extended to the qualitative assessment and quantitative inversion by HySpex imaging hyperspectroscopy. The following results are obtained: 1) a linear relationship between the SSA and the particle size in the characteristic zone is established to evaluate the ore-bearing property, and 2) the abundance of endmember minerals can be estimated by using the photometric parameters obtained from multiangle measurements as the input for the Hapke model. Therefore, the inversion model based on the “multiangle–multisource–multiscale” physicochemical characteristics and mineral abundance of granite pegmatite proposed in this article can provide a new reference for the exploration of Li deposits.

Source edge detection is a crucial task in the analysis and interpretation of potential field data. This paper introduces a novel multi-scale edge detector for potential field data, employing the Laplacian of Gaussian filter. The proposed method was tested on synthetic models and high-precision aeromagnetic data from the Kalatag region in the eastern Tian Shan, NW China. We compared the results with popular edge detection techniques developed by other scholars. Our findings demonstrate that the new detector is able to accurately define the edges of field source bodies, even those characterized by subtle physical differences or larger burial depths. Furthermore, the method exhibits high accuracy in the presence of random noise interference and is adept at identifying field source edges with weak anomalies that may be overlooked by other techniques. The edge identification results of the Kalatag area reveal that the Cu-dominant deposits and mineralization are controlled by two major deep regional faults oriented in the NW-NWW direction. These faults may serve as conduits for Paleozoic magma emplacement or eruption. The borders of magmatic rocks and the intersections of these deep regional faults hold greater potential for mineralization. Overall, this study enhances our understanding of subsurface sources and provides a crucial geophysical foundation for further exploration efforts in the Kalatag area.

Hyperspectral unmixing aims to extract pure spectral signatures (endmembers) and estimate their corresponding abundance fractions from mixed pixels, enabling quantitative analysis of surface material composition. However, in geological mineral exploration, existing unmixing methods often fail to explicitly identify informative spectral bands, lack inter-layer information transfer mechanisms, and overlook the physical constraints intrinsic to the unmixing process. These issues result in limited directionality, sparsity, and interpretability. To address these limitations, this paper proposes a novel model, CResDAE, based on a deep autoencoder architecture. The encoder integrates a channel attention mechanism and deep residual modules to enhance its ability to assign adaptive weights to spectral bands in geological hyperspectral unmixing tasks. The model is evaluated by comparing its performance with traditional and deep learning-based unmixing methods on synthetic datasets, and through a comparative analysis with a nonlinear autoencoder on the Urban hyperspectral scene. Experimental results show that CResDAE consistently outperforms both conventional and deep learning counterparts. Finally, CResDAE is applied to GF-5 hyperspectral imagery from Yunnan Province, China, where it effectively distinguishes surface materials such as Forest, Grassland, Silicate, Carbonate, and Sulfate, offering reliable data support for geological surveys and mineral exploration in covered regions.

Potentially toxic elements (PTE) in soils pose significant environmental risks due to their persistence and bioaccumulation. Integrating hyperspectral remote sensing with machine learning models is a promising approach for quantifying and mapping soil PTE distributions, enabling advanced pollution monitoring. However, comprehensive evaluations of model accuracy remain limited. Here, we conducted a meta-analysis using data from 87 studies across 97 locations, 7 soil elements, 42 spectral transformation methods, 16 band optimization methods, and 34 ML techniques. Our findings indicate that among transformation methods, first derivative (FD), second derivative (SD), wavelet transform (WT), and continuum removal (CR) achieve superior model accuracy. For band optimization methods, principal component correlation (PCC), principal component analysis (PCA), expert knowledge (EK), and the combination (C_2) method effectively enhance predictive performance. In terms of model algorithms, random forest (RF), support vector machine (SVM), artificial neural networks (ANN), extreme learning machine (ELM), and partial least squares regression (PLSR) achieve high accuracy. Furthermore, the selection of FD transformation, PCC method, and RF algorithm yields R² of 79.55 % ± 13.26 %, 82.55 % ± 9.81 %, and 79.55 % ± 13.26 %, respectively. While environmental conditions, sampling design, and covariates influence model accuracy, optimizing preprocessing is key to accurate predictions. Applying scientifically optimized preprocessing methods to field sampling data can significantly enhance model performance by maximizing the utility of available samples. Therefore, we highly recommend that researchers prioritize the FD-PCC-RF strategy combination for exploratory modeling after acquiring soil samples. This study highlights the importance of advanced preprocessing and model integration for soil PTE predictions, enhancing environmental risk assessment and management. Future research should focus on adaptive preprocessing approaches based on element spectral characteristics, hybrid optimization frameworks, advances in predictive algorithms, and multimodal environmental data fusion modeling to enhance model robustness and accuracy.

Accurate terrain perception is essential for safe rover operations and reliable geotechnical interpretation of Martian surfaces. The heterogeneous scales, colors, and textures of Martian terrain present significant challenges for semantic segmentation. We present MarsTerrNet, a dual-backbone segmentation framework that combines Progressive Residual Blocks (PRB) with a Swin Transformer to jointly capture fine-grained local details and global contextual dependencies. To further enhance discrimination among geologically correlated classes, we design a feature-guided loss that aligns representative features across terrain categories and reduces confusion between visually similar but physically distinct types. For comprehensive evaluation, we establish MarsTerr2024, an extended dataset derived from the Curiosity rover, providing diverse geological scenes for terrain understanding. Experimental results show that MarsTerrNet achieves state-of-the-art performance and produces geologically consistent segmentation results, supporting automated mapping and geotechnical assessment for future Mars exploration missions.

The Altyn Tagh Fault plays a critical role in understanding the tectonic evolution of the northern margin of the Tibetan Plateau. However, considerable debate persists regarding its activity and deformation history. This study investigates volcanic rocks from the Beidayao-Jianquanzi-Hanxia-Hongliuxia area in the eastern segment of the fault. By employing zircon U-Pb dating, whole-rock geochemistry, and Sr-Nd isotope analysis, we aim to elucidate their petrogenesis and tectonic setting, thereby providing new insights into the crustal evolution of the eastern Altyn Tagh Fault. Zircon U-Pb dating of the Hongliuxia rhyolite yields a weighted mean 206Pb/238U age of 106.6 ± 0.6 Ma, indicating an Early Cretaceous eruption. Geochemically, the western part of the study area (Beidayao and Jianquanzi) is dominated by basalts that exhibit significant enrichment in large ion lithophile elements and light rare earth elements, together with high Nb concentrations (>20 ppm), as well as high Nb/La (0.64–1.12) and Nb/U (29.8–35.42) ratios, consistent with the characteristics of high-Nb basalt. In contrast, the eastern area (Hanxia and Hongliuxia) is characterized by andesitic rocks that display typical continental arc affinities, marked by enrichment in Th, U, and Pb and depletion in Nb, Ta, and Ti. Isotopically, the basalts show initial 87Sr/86Sr ratios of 0.706–0.707 and εNd (t) values ranging from −3.2 to 0.8, whereas the andesites possess more radiogenic Sr isotopic compositions, with (87Sr/86Sr)i ratios of 0.710–0.717, and more negative εNd (t) values from −11.4 to −1.5, suggesting derivation from an enriched mantle source. Integrating geochemical data with regional geological records, we propose that the eastern part of the Altyn Fault experienced a significant intracontinental extensional setting during the Early Cretaceous, where asthenospheric mantle upwelling played a key role in the generation of the volcanic rocks. This study provides key petrological and geochemical constraints on Early Cretaceous deformation and activity along the Altyn Tagh Fault, and also offers a valuable reference for understanding the evolution of similar fault systems.

The production of geochemical data serves diverse purposes, and a variety of analytical methods are utilized for analyzing geochemical element content. However, due to limitations in project funds, censored or missing values are common in geochemical data. This scarcity of data becomes more pronounced when dealing with large datasets. Regrettably, numerous data analysis techniques are unable to process datasets containing missing values, which presents a significant hurdle for researchers who depend on geochemical data. To address this issue, here we employed a random forest model to simulate the geochemical elements of rocks and stream sediments. By comparing and analyzing the effects of model parameters and feature variable selection on the simulation results of major and trace elements, the study found that with appropriate model parameters and variable selection, the simulation results for many elements are reliable, and the generalization performance of the random forest model is satisfactory. This research sheds light on the inherent correlations among various elements in nature, offers solutions to the challenges posed by missing values in geochemical data, and provides valuable technical support for disciplines such as geology, environmental science and soil science.

The Baixintan deposit is one of newly discovered magmatic Ni-Cu sulfide deposit in eastern Tianshan nickel belts, NW China. As exploration progresses, the understanding of the whole subsurface structure and unexposed deposits in this area remains insufficient to meet the challenges of deep exploration. In this study, a cost-effective and high-precision ground magnetic surveying was employed to investigate the underground structures of the mining area, primarily due to the frequent presence of ferromagnetic minerals. Utilizing the high-precision magnetic data, the main fault structures were delineated based on robust edge detectors such as horizontal derivative, tilt angle and analytic signal. Additionally, the extensions of the primary mineralization zones and the burial depths of the ore-bearing rock bodies were determined through 3D physical property inversion technology. The target area for ore exploration was defined based on the available geological and drilling information. Our findings suggest that the primary magnetic anomalies in the study area are attributed to shallow geological bodies, specifically those less than 400 m in depth, and their morphology agrees well with the mafic-ultramafic rocks exposed on the surface. The predominant magnetic structures are oriented in the NE-SW and NW-SE directions, of which the NE direction is closely related to mineralization. The Ni-Cu mineralization bearing mafic-ultramafic rocks extends eastward along the NE-trending fault. This study not only enhances mineral exploration efforts in the Baixintan mining area but also serves as a valuable reference for exploring similar deposits covered by the Gobi Desert.

Remote sensing technology has significant technical advantages over traditional geological methods in geological mapping and mineral resource exploration, especially in high-altitude and steep topography areas. Geochemical sampling and geological mapping methods in these areas are difficult to use directly in mountainous regions such as West Kunlun. Therefore, in the face of Li-Be-Nb-Ta mineralization of the Dahongliutan rare-metal pegmatite deposit in West Kunlun, remote sensing has become an effective means to identify areas of interest for exploration in the early stage of the exploration campaigns. Several methods have been developed to detect pegmatites. Still, in this study, this methodology is based on spectral analysis to select bands of the ASTER and Landsat-8 OLI satellites, and methods, such as principal component analysis (PCA) and mixture tuned matched filtering (MTMF), to delineate the prospective areas of pegmatite. The results proved that PCA could map the hydrothermal alteration and structure information for pegmatites. To define new locations of interest for exploration, we introduced the spectra of spodumene-bearing pegmatites and tourmaline-bearing pegmatites as endmembers for the MTMF approach. The results indicate that the location of pegmatite areas on the ASTER and Landsat-8 OLI images overlaps with the ore deposits, and the location of potential ore-bearing pegmatites is delineated using remote sensing and geological sampling. Although this does not guarantee that all prospective areas have the mining value of ore-bearing pegmatites, it can provide basic data and technical references for early exploration of Li.

Geochemical data is crucial for reflecting geological features and is extensively applied in mineral exploration, environmental impact assessment and geological research. However, the high economic cost of geochemical data analysis hinders large-scale studies, leading to low spatial resolution, especially in remote areas. Although remote sensing data provides rich surface spectral information and shows a strong correlation with geological features, its accuracy in large-scale geochemical data inversion is insufficient. Therefore, we improve the accuracy and reliability by fusing multi-source geoscience data. Vegetation information, digital elevation models (DEMs), and aeromagnetic data, among other geoscience data, offer new perspectives for geochemical data analysis. This paper proposes a novel Multi-modal Spatial-Spectral Fusion Model with Swin Transformer and Convolutional Networks for Regression (MSSF-SCR). This model extracts spatial features from multi-source geoscience data using a multi-branch Swin Transformer and dynamically adjusts feature weights with the MMHCA module. The Swin Transformer unifies spatial features, addressing semantic disparities among diverse data sources. Spectral features from remote sensing data are then fused with spatial features through two-dimensional convolutional regression, producing 15-meter-resolution geochemical maps. Experiments conducted in the Dananhu-Tousuquan Island Arc region of East Tianshan demonstrate that MSSF-SCR achieves superior performance in terms of R 2 , R, MAE, and RMSE indices for five elements (Al 2 O 3 , Fe 2 O 3 , K 2 O, MgO, and SiO 2 ).

Achieving coordinated development among social equity (SE), economic development (ED), and ecosystem health (EH) is central to resolving the sustainability trilemma. This study investigated the spatiotemporal evolution and driving forces of SE–ED–EH coordinated development in Hebei Province, China, from 2005 to 2020 using a 1 km grid dataset. A comprehensive analytical framework integrating the Coupling Coordination Degree (CCD) model, fuzzy C-means clustering, and interpretable machine learning (XGBoost–SHAP) was developed to quantify changes in coupling and coordination (CC) levels and reveal nonlinear threshold effects. Results show pronounced spatial heterogeneity: urban cores exhibit “high coupling degree (C)–high coordination degree (T)–high CC level,” southeastern plains show “high C–low T–medium CC level,” and northwestern mountainous areas present “low C–medium/high T–low CC level.” Six dominant temporal evolution types were identified. XGBoost–SHAP reveals that nighttime lights (NL), population density (POP), and elevation (DEM) are the dominant drivers, with clear threshold ranges (NL 500–1500 nits; POP threshold near 40 persons km−2 with diminishing returns beyond 100 persons km−2; DEM constraint at 1000–1250 m) and strong interaction effects. The results suggest that Hebei is entering a quality- and structure-oriented rebalancing stage, where threshold-based management is critical for avoiding marginal loss of coordinated development. This study demonstrates that interpretable machine learning provides a transferable paradigm for threshold calibration, spatial zoning, and policy optimization aligned with SDGs, particularly applicable for resource-constrained regions undergoing late industrial transition.

To promote the transformation of remote sensing (RS) data into geoscience knowledge, it is necessary to provide better data discovery capabilities, especially when large amounts of RS data have been accumulated. Spatio-temporal range query is one of them to enable data discovery on RS images by the spatio-temporal range. Although existing RS image indexing methods are suitable for data discovery based on spatio-temporal range queries, they do not fully consider the problem of low data retrieval efficiency when the temporal and spatial range scales are unbalanced. To address this problem, we propose a multi-scale spatio-temporal grid index model (MSTGI). First, we divide the time dimension into three levels according to the granularity of year, month and day, and build an independent grid index structure for the object at different levels. Second, we design an adaptive hierarchical indexing strategy to perform parallel retrieval in an appropriate combination of partitions. MSTGI linearizes the grid obtained after global geospatial subdivision using Hilbert curves. Our experimental results, obtained on the Landsat series satellite dataset, reveal that the proposed method improves query efficiency levels by approximately 12.582% and 87.754% compared with GeoSOT-ST and AGMC at various spatial scales, respectively, and reaches 100% recall.

The inversion of Cu contents in rocks based on hyperspectral remote sensing is indicative of geochemical exploration as well as mineral resource exploration. Due to the strong redundancy of band information in hyperspectral data, characteristic band selection is the key to improving the accuracy of metal element inversion models in hyperspectral data. In this study, we proposed a framework that innovatively focuses on secondary screening method (SSM) as the core, allowing users to add different data preprocessing methods and machine learning algorithms to the framework based on expert experience, to obtain high-precision metal element inversion models suitable for sampling data. we first preprocessed the measured spectral curves using the first order derivative, second order derivative, continuum removal, multiple scattering correction and variance analysis to highlight band absorption characteristics. In addition, we performed SSM to obtain Cu-related characteristic bands, which were further used as input data to train and compare the performance of Partial Least Square Regression (PLSR), Support Vector Regression (SVR), Multi-Layer Perceptron (MLP), Random Forest (RF) and Gradient Boosting Regression Tree (GBRT). The results show that both the spectral pretreatment and modeling method could potentially impact the inversion accuracy. The accuracy of the ensemble method applied in fresh faces data (i.e., a CR-SSM-RF with R2 values of 0.77 and CR-SSM-GBRT with R2 values of 0.81) performed better than other models with an accuracy improvement of up to 80%. Compared to the full band model, the accuracy of characteristic bands selection model increased 50%. This also illustrates that the CR-SSM-RF or CRSSM-GBRT model significantly improves the accuracy of the inversion of Cu contents in rocks compared to the original spectral reflectance. Our study provides a new idea for selecting different spectral preprocessing methods as the input values for the future model and will also offer an executable framework for the inversion of Cu contents in rocks of different lithology on a large scale in the future.

Hyperspectral remote sensing is a fast and non-destructive technology for identifying geological information, and many successful cases have been achieved in mineral identification and estimation of soil heavy metal content. However, there have been fewer studies on the application of this technology to rare metals, especially the detection of lithium (Li) resources. Whether the hyperspectral process can effectively identify Li anomalies is significant for expanding the exploration of Li resources. To this end, this study explores the potential of hyperspectral techniques for Li elemental content estimation by collecting rock debris samples in the field and extracting spectral feature coefficients using a Gaussian Mixture Model (GMM). The results show that (1) the feature parameter extraction technique based on the GMM can quickly and accurately extract the spectral absorption feature parameter. (2) Compared with the spectral reflectance, the feature coefficients can improve the correlation with Li content, and the constructed model is more effective. (3) The estimation model based on the full-width at half maximum (FWHM) at 1.93 mu m is the most effective, with the coefficients of determination (R2), relative root mean squared error (RRMSE), and the ratio of the performance to deviation (RPD) of 0.61, 0.516 and 1.601, respectively, which are significantly better than that of the spectral reflectance model. The above results show that the use of hyperspectral technology can effectively estimate the Li content in rock debris, which provides a technical reference for the identification of regional Li anomalies by airborne hyperspectral remote sensing and technical support for improving the efficiency of lithium resources exploration and narrowing the geological focus of investigation.

Lithium (Li) is growing in importance and demand in several industrial applications, such as portable electric devices, electric vehicles, and hybrid electric vehicles. Li-rich pegmatites are one of the main sources of Li production in the world, so low-cost and high-efficiency exploration using remote sensing has become an important means for promoting the discovery of Li resources. Although imaging spectroscopy has great potential for Li-rich pegmatites identification and regional delineation, due to the limitations of resolution and data acquisition, the exploration of the Earth’s surface at various scales on a variety of platforms requires further research. The purpose of this study is to use a ground-based HySpex imaging hyperspectrometer to map pegmatite zones directly, developing a new approach for mineralized pegmatite exploration. To achieve this, we present the minimum wavelength mapper (MWM), which combines the position and depth information of the deepest absorption feature, to give a per-pixel overview map of the Jingerquan Li–Be–Nb–Ta pegmatite deposit pit and heap. The results show that images between 2100 and 2450 nm provide useful information about mineral assemblages that display strong spectral features, such as Li-rich pegmatites, Li-poor pegmatites and alteration, and their spatial distribution at the surface in areas covered by the imagery. The wavelength position and the depth of Al–OH can provide an overview of the mineral assemblages and abundances, to rapidly classify possible regions of interest for further analysis.

To reduce the high redundancy of band information in hyperspectral data, various band optimization methods have been adopted, which could be divided into two types namely band extraction (e.g., principal component analysis, PCA) and band selection (e.g., Spearman correlation coefficient, SRC). However, the applicability and effectiveness of different band optimization methods were rarely reported in the literature. Therefore, based on the rock sample data of the Baixintan deposit, we compare the performance of two band optimization algorithms (principal component analysis−based band extraction and SRC-based band selection) in inverting metal elements (Cu, Fe, Ni, Cr, Mg) using the adaptive genetic algorithms-gradient boosting regression tree (AGA-GBRT) algorithm. Two band optimization methods have shown different effects in improving the accuracy of target metal elements. The five models with the highest accuracy in metal elements include LT-R-Cu, LT-PCA-Fe, ORI-PCA-Ni, SDT-PCA-Cr, and LT-PCAMg. In the established model, the inversion accuracy of the Cu element is the lowest, possibly due to the high variability of the data itself (coefficient of variation is 3.55). Fe and Ni highly correlated with Cu elements were used to indirectly invert Cu element. Compared with the direct inversion model, the accuracy of the indirect inversion model has increased by 11%. Overall, PCA is more effective than SRC in predicting the content of metal elements in rocks. The conclusion presented in this article provides a ground experimental basis and technical support for future optimization of hyperspectral bands and inversion methods of metal element content in rocks.