Since the 1950s, geological exploration in Qinghai Province has generated massive geological data resources, which are characterized by fragmentation, lack of structure, multiple sources, heterogeneity, and dispersed storage, as well as large data volumes with insufficient data mining and social service capabilities. In order to adapt to the new situation and demands of geological exploration in the new era and to support a new round of breakthroughs in mineral exploration, the Qinghai Provincial Geological Big Data Integration and Application System has been constructed. Based on the “Qinghai Provincial Geological Big Data Platform Data Standard”, data concerning basic geology, hydrogeology, mineral exploration, urban geology, geological research, mining rights information, project management, and other aspects have been integrated. This integration has achieved unified management of multi-source, heterogeneous, and massive geological data and established the Qinghai Provincial Geological Big Data Resource. Applications such as project duplication analysis, custom-made thematic mapping, comprehensive project application, geological data retrieval management, and three-dimensional maps have been realized. The deep integration of big data, Internet, and database technologies with geological work has been promoted, driving the innovative application of geological big data and improving its service capabilities. In a case study of the Wulonggou gold mining area in Qinghai Province, the geological exploration big data include a 1∶50000 geological and mineral map, a 1∶50000 geochemical anomaly map, ASTER and GF-5 remote sensing interpretation maps, a large-scale elevation map, 1∶1000 exploration borehole profile maps, 1∶1000 three-dimensional solid models of industrial orebodies and their engineering, and a Loop3D geological-geophysical forward and inversion three-dimensional model (vertical depth of 3000 m). The SKUA-GOCAD software has been used to associate the above-mentioned big data maps and multi-parameter models in the same coordinate system, and GeoCube 4.0 software (incorporating linear, nonlinear, machine learning, and deep learning (LSTM) methods) has been independently developed for target optimization and resource evaluation based on multivariate information fusion and integration.

The National Mineral Resources Survey Database integrates geological exploration data formed since the founding of the People’s Republic of China and has the salient characteristics of geological big data such as massive, multi-source, multi-dimensional, multi-disciplinary and multi-spatial-temporal. These massive data elements, which exceed one million levels, bring huge challenges to knowledge services, analysis and mining, quality inspection and information sharing of national survey data. This article adopts a top-down technical route for constructing knowledge graphs based on mineral deposit ontology, and innovatively builds a “multi-source fusion of all survey elements” national survey data standard based on a knowledge-driven model. Based on the data standard system, it enables rapid discovery, positioning, and extraction of mineral deposit knowledge and the correlations among various elements, and supports intelligent services such as machine learning-based mineral deposit knowledge graph construction, spatial data quality inspection and multi-source information consistency verification. These contributions have played a key role in the construction of the National Mineral Resources Survey Database; and by building a big data cluster management and cloud computing platform, it serves geological survey informatization, smart mining and other work, dynamically analyzes the national mineral resources security capabilities, and promotes interdisciplinary integration across Earth science, information science and data science.

Groundwater level changes significantly impact urban construction safety, especially during rapid urbanization. To reveal the complex relationship between groundwater dynamics and urban construction safety, this study takes the Beijing Plain area as a case study and proposes a method for constructing a knowledge graph that integrates multi-source data. This study first reviews the current state of research on knowledge graph technology, groundwater level evolution, and urban safety, and then constructs a unified ontological framework for structuring multi-source data and information extraction. Subsequently, a fine-tuned BERT-BiLSTM-CRF model is employed for named entity recognition, enhancing its accuracy. By integrating with the Neo4j graph database, the knowledge graph enables efficient querying, system scalability, and dynamic representation of causal relationships. The results from the Beijing Plain case study demonstrate that the constructed knowledge graph effectively reveals the potential impacts of groundwater level changes on urban infrastructure, provides real-time warning information, and supports informed decision-making in urban planning and groundwater management. This work offers a novel tool for intelligent and refined urban safety management and demonstrates the potential for applying knowledge graph technology to other complex systems.

The utilization of knowledge graphs for the organization and computation of information within a semantic network, employing a “node-edge” model, represents a pivotal advancement in computational understanding. With profound capabilities in knowledge representation and semantic reasoning, the geological knowledge graph emerges as a foundational infrastructure facilitating the advancement of geological big data and geological artificial intelligence. Addressing prevalent challenges such as disparate data formats, variable quality, scalability limitations, and update latency issues inherent in current geological knowledge graph construction, this study proposes a comprehensive approach. Primarily, from the perspective of geological mechanisms, the paper proposes a unified geological ontology expression model integrating the “semantic conceptual layer-change mechanism layer-feature attribute layer”. Furthermore, it establishes an iterative framework for knowledge graph construction that leverages expert crowd intelligence collaboration to ensure the realization of large-scale, high-quality, and efficient geological knowledge graph development. Detailed methodologies encompassing geological knowledge extraction, alignment, fusion, updating, completion, as well as knowledge storage and visualization within a multi-modal, multi-type, and multi-themed corpus are meticulously expounded. Furthermore, the study delineates various application directions for geological knowledge graphs. These include precise retrieval of geological knowledge and intelligent question-answering systems, three-dimensional fine geological modeling driven by geological knowledge graphs, multi-source data fusion facilitated by geological knowledge graphs, intelligent prediction of mineral occurrences based on knowledge graphs, and knowledge graph-guided monitoring and early warning systems for geological hazards. This paper provides a solid foundation for the advancement and practical application of large-scale, high-quality geological knowledge graphs.

Data-driven 3D mineral potential mapping (MPM) is a key technology for subsurface mineral exploration. Compared with traditional statistical methods and shallow machine learning algorithms, deep learning provides significant advantages in the automatic extraction of deep features and the modeling of complex nonlinear relationships. The Wulong gold district, one of the most important gold districts on the Liaodong Peninsula, was selected as the study area. Conducting subsurface mineral exploration in this district is of great strategic significance for promoting the development of a thousand-ton gold base on the Liaodong Peninsula. Unlike conventional 2D image data commonly used in computer vision, 3D MPM data exist as 3D voxels. To address this issue, the study constructed deep learning prediction frameworks based on 3D Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), focusing on training sample construction, model architecture design, and model training tailored to 3D MPM data. Comparative experiments demonstrated that both CNN and RNN models outperform traditional methods such as Random Forests in terms of 3D mineral prospectivity prediction within the study area. Based on these results, subsurface targets were delineated in the Wulong gold district, further confirming the considerable gold mineralization potential in the subsurface along the Jixingou fault of the Wulong gold deposit.

The Kalatongke copper-nickel district is a significant producer of scarce mineral resources. However, with continuous exploitation, it faces a crisis of depleting reserves. Consequently, the discovery of new ore bodies in this district has become a primary objective, and deep exploration presents an opportunity to revitalize such a crisis-hit mining area. This study integrates 3D modeling and machine learning technologies to build district-scale geological and geophysical models for three-dimensional (3D) quantitative mineral resource prediction. The workflow consisted of four main steps: (1) Gravity, magnetic, electromagnetic, and seismic data were used to construct 3D models through geophysical inversion and spatial interpolation, yielding physical property models for density, magnetic susceptibility, resistivity, and seismic wave velocity. (2) Based on metallogenic principles and petrophysical characteristics, a lithological model was constructed using the K-means clustering machine learning method, while a fault model was built by integrating surface fault observations with deep constraints from electromagnetic and seismic sections. (3) The geological and geophysical models were converted into exploration predictor variables. The Bagging-based Positive-Unlabeled Learning (BPUL) algorithm was employed for 3D mineral prospectivity mapping utilizing Random Forest, Support Vector Machine, and XGBoost as base classifiers. Comparison revealed that the BPUL-XGBoost15 model delivered the best performance. This optimal model was used to generate a mineral prospectivity probability map, and initial targets were delineated using the prediction-volume (P-V) plot technique. (4) A risk-return analysis was conducted to evaluate the uncertainty of the optimal model. The initial targets were then refined based on this analysis to identify those with high potential returns and low exploration risk. Finally, these targets were classified according to their metallogenic geological context. The identified key targets are critical for guiding subsequent exploration efforts. The research framework established in this study can serve as a valuable reference for the deep exploration of similar types of mineral deposits.

The rapid development of big data and AI technologies has brought about a new scientific paradigm for studying complex geological phenomena and conducting 3D mineral prospectivity prediction. However, mineral prospectivity prediction is a multi-step process—including model construction, data acquisition, mapping, anomaly detection, variable selection, and information fusion—with each step introducing uncertainties that propagate through the entire workflow. This study proposes an approximately (1+ε) globally optimal truth discovery method to quantitatively evaluate uncertainty in integrated 3D mineral resource prospectivity prediction. The method minimizes the objective function through geometric iterative algorithms, yielding near-optimal truth values and data source weights. Unlike traditional approaches, it provides model-level uncertainty estimates without expert input and requires fewer pre-trained models, thereby significantly reducing computational complexity. Case studies in the Huayuan mineral cluster (Hunan) and the Haoyao’erhudong gold deposit (Inner Mongolia) demonstrated the method’s effectiveness and superiority in multi-scale uncertainty assessment of 3D mineral prospectivity prediction.

A realistic 3D geological model is a vital digital framework in many fields, such as analyzing complex underground structures, conducting engineering surveys, and prospecting mineral resource, etc. However, the accuracy and reliability of 3D geological models are significantly affected by difficulties in data acquisition, data sparsity, and insufficient geological understanding, which limit their in-depth application. Based on the adaptive fully connected deep neural network (AFCDNN) and multi-scale iterative multiple-point statistics (MPS) method, this paper proposes a method for reconstructing 3D geological structures, for which two-dimensional geological cross-sections are used as the modeling data source. Two AFCDNNs are built: one to generate geological surfaces and the other to predict geological attributes. Geological relationships are then rectified in a post-processing manner. An initial model is subsequently constructed. To optimize its local features, the multi-scale iterative MPS method is applied. A concrete example of 3D model construction for a station site of a Guangzhou metro line illustrates that the constructed model correctly reproduces the spatial relationship between faults and strata, achieving a maximum accuracy of 89.3% when validated against actual borehole data. Furthermore, the AFCDNN can generate global features of the geological structure from small sample datasets, and the multi-scale iterative strategy fully leverages the advantages of local optimization. The proposed method exhibits good feasibility, accuracy, and reliability, and can provide an important reference for reconstructing 3D geological structures.

Accurate characterization of three-dimensional (3D) geological structures is crucial in geological modeling, yet effectively characterizing structures with multi-scale spatial features remains a significant challenge. The rapid advancement of deep learning technology offers new avenues for characterizing 3D multi-scale subsurface structures. To enhance the extraction and reconstruction of multi-scale spatial features, we propose a multi-scale 3D geological modeling method based on Laplacian Pyramid-Self-Attention Generative Adversarial Network (Lap-SAGAN). To efficiently extract features at multiple spatial scales, Lap-SAGAN employs a 3D Gaussian filter sampling method to hierarchically represent spatial features. In addition, we introduce a spatial context loss function based on residual computation to ensure consistency between the reconstructed geological models and the input conditioning data. The performance of Lap-SAGAN is evaluated using multiple 3D datasets. Experimental results demonstrate that Lap-SAGAN can effectively characterize 3D geological structures with multi-scale spatial features from 2D cross-section data, confirming its feasibility and practicality for 3D geological model reconstruction.

Driven by ecological water replenishment, precipitation, and reduced groundwater extraction, the groundwater level in the Beijing Plain has been continuously rising. While this trend has improved the aquatic ecological environment and promoted restoration, it also poses risks and challenges to urban construction and the safety of subsurface infrastructure. Therefore, accurate prediction of groundwater levels is crucial for managing groundwater resources and ensuring the safety of surface and subsurface structures in Beijing. Specifically, using the area surrounding Shougang Park on the eastern bank of the Yongding River as a case study, this research enhances the traditional LSTM model by incorporating both spatial and temporal characteristics of groundwater dynamics. A spatio-temporal attention mechanism-based LSTM (STA-LSTM) model was developed for short-term groundwater level forecasting, achieving accurate dynamic predictions of regional groundwater levels. The results show that: (1) The STA-LSTM model demonstrated superior prediction performance across both temporal and spatial scales compared to the LSTM, SA-LSTM, and TA-LSTM models, achieving optimal metrics: MAE of 0.08 m, RMSE of 0.11 m, and NSE of 0.98; (2) The model effectively captured groundwater level dynamics during intense precipitation and ecological water replenishment events, achieving simulation accuracy greater than 99% and 97%, respectively; (3) Prediction accuracy decreased over time due to cumulative errors, with MAE values of 0.10 m, 0.18 m, and 0.40 m for the 7th, 14th, and 28th days, respectively; (4) Predictions of groundwater dynamics under five designed precipitation scenarios and two ecological water replenishment scenarios revealed a strong correlation between groundwater level fluctuations and the intensity and duration of these events. This study demonstrates that the STA-LSTM model effectively addresses the complexity and uncertainty inherent in groundwater dynamics and provides reliable technical support for the short-term forecasting of urban shallow groundwater levels that are significantly influenced by precipitation and ecological replenishment.

Hyperspectral remote sensing has been widely used for geological mapping and for identifying alteration minerals during mineral exploration. With the development of machine learning, the random forest algorithm has been widely applied to extract alteration mineral information in complex metallogenic settings and across diverse mineral types; however, the quality of training samples often affects recognition accuracy. Therefore, we first quantitatively compared different extraction methods and selected the SID-SGA_tan algorithm as the most effective method for extracting reliable and representative training samples from hyperspectral remote sensing imagery. Subsequently, we conducted hyperspectral alteration mineral identification experiments using random forest classifiers to further validate the applicability of remotely sensed alteration information in geological prospecting. The results indicate that: (1) the SID-SGA_tan algorithm extracts training samples that are more accurate and reliable than those produced by alternative methods; and (2) the overall accuracy of the random forest classifier based on SID-SGA_tan reaches 78.46% (77.09% after dimensionality reduction), with a Kappa coefficient of 0.7053 (0.6863 after dimensionality reduction). These metrics represent an improvement in recognition accuracy over results obtained using SAM and other benchmark methods. The method examined in this study demonstrates strong potential for high-precision mineral identification in hyperspectral remote sensing imagery characterized by rich information content and complex backgrounds.

Understanding the distribution of rock glaciers is crucial for studying the hydrogeology and climate change in cold and arid regions. Apart from traditional methods such as field surveys and visual interpretation, deep learning (DL) applied to high-resolution natural color (RGB) remote sensing imagery has been used to compile the Tibetan Plateau Rock Glacier Inventory (TPRoGI). However, although the near-infrared (NIR) band provides sharp edge features and rich spectral information that are useful for identifying rock glaciers, it has rarely been utilized in DL-based recognition models due to the three-band input restriction of typical DL networks. In this study, an improved DeepLabV3+ network (IDNet) was designed to simultaneously extract and fuse features from both RGB and NIR bands. The IDNet was trained using Sentinel-2 imagery and the rock glacier labels from the TPRoGI in the Qilian Mountains (QLMs), yielding a well-trained model that achieved an accuracy of 0.7830, a precision of 0.7830, a recall of 0.7840, a specificity of 0.7835, and a mean Intersection over Union (mIoU) of 0.6916. The IDNet model further identified 459 rock glaciers in the QLMs that were missed by the TPRoGI, demonstrating the feasibility and effectiveness of combining the IDNet with RGB and NIR bands for rock glacier recognition. This approach effectively enhances the efficiency and accuracy of rock glacier inventory compilation.

In the era of Industry 4.0 in the 21st century, big data and artificial intelligence technology have promoted the development of large-scale mining from digital mining and intelligent mining to the intelligent mining integrating “geology, mining, mineral processing and smelting”. Large and very large metal mines at home and abroad usually adopt the “resource-environment-economy” integrated management and control mode, and the artificial intelligence technology methods of “digitalization, information technology, visualization, quantification and intelligence” are being applied more and more widely, especially the expansion and application of stereo observation and 5G+ real-time communication technology. So far, a new scientific paradigm, deep knowledge discovery and four-dimensional control of wisdom have been born. By combing the data sets of geology, orebody, survey, mining and mineral processing of Yanshan iron mine in Hebei Province, this paper carries out multi-temporal remote sensing (spectrum and radar) UAV image acquisition of the exposed mining site and its explosive pile, and uses artificial intelligence technology and methods to dig deep geological information of the mine to serve the intelligent management and control of the real-time mining industry. Specific research contents and achievements are summarized as follows: (1) Construct high-precision 3D geological and ore body and engineering models by using mineral exploration and mining borehole data set, mining meter blasting borehole data set and centimeter engineering survey data set; (2) Using “UAV” (high) spectrum, radar point cloud and ground spectrum, and in-situ analysis of rock and ore microzones, the multi-parameter (identification of mineral, grade, radiation) accurate information model of 3D ore body is established; (3) Using the intelligent pattern recognition of sub-meter images of UAV, the geological factors affecting mine production are identified, showing that the difference between the formation lithology and the low-grade ore in the periphery of the ore body is not significant in terms of tone, structure and texture, and the inclusion of lens-shaped intrusive dike and fault structure are in the ore-rich section; (4) The 3D ore body model, mathematical statistics and geostatistics mining were used to identify the hyperspectral band information of the ore body and construct the empirical mathematical model formula; (5) Develop the workflow to identify geological and environmental factors (variables and parameters) affecting ore body mining by integrating multi-parameter three-dimensional models of geology, minerals, spectra, radar point cloud, magnetic method, etc., and to identify the visible (visual) and near-infrared bands of surrounding rock formation, lean ore, ore significant difference information, and carry out real-time dynamic mining intelligent spatial decision-making in open pit mines. It serves the integrated control of geology, mining and mineral processing of smart mining, and improves the recovery rate of mining and mineral processing.

The development and utilization of underground space is an important approach to optimizing urban spatial structure and enhancing urban functions. The data generated in this process involve multiple elements such as geological bodies, underground structures, and utility pipelines. These diverse elements result in complex characteristics of underground space data, including multi-scale, multi-domain, multi-format, multi-semantic, and multi-temporal features. This complexity leads to several challenges in large-scale application scenarios, such as difficulties in data storage and integration, as well as low loading efficiency, which significantly hinder data utilization. To address these issues, this paper systematically reviews the elements and characteristics of underground space data and investigates technical challenges in multi-source data integration and visualization. The study focuses on three aspects: multi-source data management, multi-source data integration, and data lightweighting and visualization, proposing a practical set of technical methods for urban-level underground space multi-source data integration and visualization. For multi-source data management, a data classification method that progressively refines from business domains to data entities is employed to organize, manage, update, and enable the publication and sharing of data services. For multi-source data integration, a unified spatiotemporal benchmark is adopted to achieve the integration of entity models with attribute information, unstructured data, and real-time streaming data through spatial and semantic methods. Additionally, different precision control measures are proposed for various types of data integration processes and methods to enhance integration outcomes. Regarding data lightweighting and visualization, a multi-level LOD-based lightweighting algorithm for underground structure models, as well as lightweighting methods for geological models and detailed models, are proposed. The study also explores technical pathways for the efficient rendering and display of urban-level, multi-scale, high-precision 3D model data on web platforms, thereby providing technical support for the application of multi-source underground space data.

Seismic profiling and gravity inversion are the main geophysical methods for studying the Moho depth. Previous studies calculated the residual gravity anomalies caused by Moho undulations and then used seismic Moho depth data to constrain the parameters for frequency-domain gravity inversion, thereby generating a high-resolution Moho depth distribution map of the South China Sea. However, gravity inversion involves two key parameters—the density contrast across the Moho and the regional reference Moho depth—so the Moho depth derived from conventional analytical formulas inevitably contains certain errors. This paper adopts a machine learning approach to directly establish a correlation model between residual gravity anomalies and seismic Moho depths. Using the residual gravity anomalies from previous studies, seismic Moho depth points, and the machine learning model, we predict the Moho depth. The results show that the machine learning model outperforms traditional methods in Moho depth inversion, with a higher R² and lower NRMSE, MAE, and RMSE values. Furthermore, this paper recalculates the residual gravity anomalies of the South China Sea Basin and its peripheral regions with updated data, and compiles and collects recent seismic Moho depth data in this region. The machine learning model is then used to re-predict the Moho depth, and statistical metrics indicate that the new predictions are more accurate.

The Xiajiadian gold deposit, located on the northern margin of the Qinling Orogenic Belt, is a representative deposit in the Zhashui-Shanyang ore concentration area. In recent years, significant breakthroughs have been achieved in mineral exploration, indicating considerable potential for mineral discovery in deep and peripheral areas. However, exploration has only reached depths of up to 500 meters below the surface. The search for deep and concealed ore deposits has become the most pressing issue in mineral exploration in this region. Guided by the theory of ore prospecting prediction based on ore-forming geological bodies, this paper summarizes the geological background, geophysical and geochemical characteristics of the Xiajiadian gold deposit. It constructs a geological model for mineral exploration prediction, focusing on the characteristics of ore-forming geological bodies, ore-controlling structures, ore-controlling structural planes, and mineralization features, providing an example for gold and polymetallic mineral exploration prediction in the area. The Xiajiadian gold deposit is a low-temperature magmatic hydrothermal gold deposit, with mineralization occurring during the Late Jurassic to Early Cretaceous. The ore-forming geological body is a concealed intermediate-acidic intrusive body, and the ore-controlling structures are the Xiajiadian-Yaolinghe regional overturned anticline and the Zhen’an-Banyanzhen major fault complex system. The ore-controlling structural planes include fault structures, lithological interfaces, and unconformities. The alteration types most closely related to gold mineralization are silicification, sericitization, calcitization, and fluoritization. Orpiment and realgar are relatively developed in the upper part and periphery of the mining area, while minerals such as pyrite and arsenopyrite are more developed in and below the ore body. The electrical characteristics show “high polarizability and low resistivity”, and the ore-forming elements exhibit an association of Au, As, Hg and Sb. The primary halo zoning generally shows As-Hg-Sb as the front halo, Au-Ag-Zn as the ore halo, and W-Sn-Mo as the tail halo. Integrating geological, geophysical, and geochemical information, a comprehensive information-based mineral exploration prediction model is established, and three mineral exploration target areas are delineated. Engineering verification has led to the discovery of a thick and extensive gold orebody within the Zhen’an-Banyanzhen major fault.

Since the discovery of skarn-type iron ore deposits in the Qihe-Yucheng area in 2015, a variety of geophysical exploration methods, including ground gravity, magnetic, and electromagnetic surveys, have been carried out. However, due to thick overburden, deep burial of orebodies, and weak mineralization signals, data interpretation suffered from severe non-uniqueness. Seismic exploration, advantageous for its depth penetration and high resolution, was employed for deep iron ore prospecting in this region. By adopting a wide-line geometry and focusing on enhancing the signal-to-noise ratio of deep reflections via advanced processing, high-quality seismic profiles were obtained. Analysis of seismic wavegroups across different strata precisely delineated the base of the Cenozoic and Carboniferous-Permian sequences and mapped the distribution of intrusive bodies. These results provided accurate horizon and structural constraints for joint gravity-magnetic inversion. Thereby, this integrated approach differentiated gravity-magnetic anomalies caused by depth and morphology variations of intrusions from ore-induced anomalies, substantially reduced inversion non-uniqueness, and enhanced orebody targeting. This integrated geophysical methodology guided the planning of borehole ZK11, which intersected high-grade iron ore. Drilling results confirmed a strong correspondence between predicted stratigraphy and depths, demonstrating the efficacy and considerable potential of seismic methods for exploring deep-buried iron ores in covered areas.

High-grade iron deposits associated with dioritic intrusions in the North China Craton (NCC) are a current research focus. Understanding the magmatism linked to these deposits and its metallogenic response remains a significant challenge. By combining in-situ zircon trace element analysis with established petrogenetic and metallogenic theories, this study investigates this magmatism-mineralization relationship, using the western Shandong high-grade iron metallogenic belt (within the eastern NCC) as a case study. The findings aim to provide insights for future research and mineral exploration. Zircon U-Pb dating reveals two distinct age groups: Mesozoic (~130 Ma) and Neoarchean to Paleoproterozoic (2500-2300 Ma). Both groups consist of non-metamorphic magmatic zircons that crystallized within a narrow temperature range. Notably, the Precambrian zircons are temporally coincident with the Great Oxidation Event (GOE) at the Archean-Proterozoic boundary. Significant systematic differences in zircon trace elements are observed among different diorite plutons, including: (1) the degree of divergence in chondrite-normalized REE patterns, (2) the variation ranges of ∑LREE, ∑HREE, and ∑REE, and (3) the variation range of δCe values. These systematic differences can therefore serve as potential geochemical markers for mineral exploration. Trends in zircon δCe, δEu, and Hf concentrations suggest that the terminal stages of magma evolution may have been unfavorable for large-scale mineralization. Calculated oxygen fugacity (f(O₂)) indicates that samples plotting above the FMQ buffer (or even the MH buffer) constitute a necessary favorable condition for mineralization, highlighting the importance of high oxygen fugacity. Based on these findings and by integrating the complex mineralization characteristics of the western Shandong region with anatectic granite petrogenesis theory, we propose a comprehensive magmatism-mineralization response model.

The Diedao iron deposit, newly discovered in 2023, is located on the northwestern margin of the Luxi Uplift, North China Craton. It is a high-grade skarn-type iron deposit. The high-grade iron orebodies exhibit sharp contacts with the surrounding rocks and are stratified within the coal-bearing strata adjacent to the contact zone between the intrusive rock mass and the Carboniferous-Permian strata. The ore has a total iron content of approximately 55% and is dominated by magnetite. The mineralization is predominantly iron, with notably low concentrations of Cu, Co, and other elements. Cobalt and sulfur show consistent variation trends. The characteristics of the ore bodies identify it as a typical “Yucheng-type” high-grade skarn iron deposit. The ore-forming intrusive rocks are peraluminous and characterized by high Mg and Na but low Fe and Ti contents. Strong albitization is widespread in the adjacent rock mass, and this alteration process may have contributed some iron to the mineralization. The relatively high emplacement level of the intrusion facilitated the long-distance migration and subsequent mineralization of the ore-bearing hydrothermal fluids. The deposit is associated with a secondary, low-amplitude, and gentle magnetic anomaly. Integrated geophysical exploration, combining gravity and magnetic profile inversion with either 2D seismic or wide-field electromagnetic methods, can effectively distinguish magnetic anomalies caused by ore bodies and intrusive rock mass protrusions, thereby improving prospecting success rates. While the orebodies in the southeastern and eastern parts of the Diedao anomaly may have been eroded, the western area offers broader prospecting potential.

The Qihe-Yucheng area of Shandong Province hosts a newly discovered, high-grade iron deposit, where diorite constitutes the primary ore-hosting intrusion. Studying its main rock-forming minerals can elucidate physicochemical changes during magma evolution and mineralization. Using electron probe microanalysis (EPMA), we systematically analyzed the chemical composition of biotite and hornblende in diorites from different mining areas, based on detailed petrographic observations.Hornblende is categorized into two types by color: early brown-yellow and late yellow-green hornblende, the latter coexisting with biotite. Biotite is magmatic magnesian biotite, and hornblende is predominantly magnesio-hornblende, both exhibiting geochemical characteristics consistent with a mixed crust-mantle source. The brown-yellow hornblende has low Si and high Al, while the yellow-green type has high Si and low Al. Major-element mapping reveals that brown-yellow hornblende is unzoned but shows fluid metasomatism along cleavages, whereas yellow-green hornblende displays clear compositional zoning, correlating with point analysis trends.Thermobarometry indicates brown-yellow hornblende crystallized at 847-914 ℃ and 144-269 MPa (depth: 5.4-10.1 km), with oxygen fugacity (f_O2) from ΔNNO+0.59 to+1.39 and melt water content from 4.29% to 5.40%. Yellow-green hornblende crystallized at 687-790 ℃ and 33-94 MPa (depth: 1.23-3.55 km), with f_O2 from ΔNNO+1.64 to+2.89 and water content from 2.92% to 4.74%. The significant decrease in crystallization temperature and pressure from the brown-yellow to the yellow-green hornblende delineates a clear magma ascent and cooling path, evidencing continuous fractional crystallization. Temperature, pressure, and high f_O2 estimates for coexisting biotite corroborate the hornblende data.Integrating these findings, we propose the following model: The iron-rich dioritic magma originated from partial melting of the lithospheric mantle with crustal contamination. It experienced brief stagnation at ~7.64 km depth, crystallizing brown-yellow high-Al hornblende, before final emplacement at ~2.05 km depth, where yellow-green low-Al hornblende and biotite formed. The evolving physicochemical conditions (decreasing t, p, H₂O_melt, and increasing f_O2) during ascent promoted iron enrichment in the residual melt and subsequent exsolution of iron-rich hydrothermal fluids. These fluids facilitated mineralization upon interaction with the carbonate country rocks.

Iron skarn deposits in the Middle-Lower Yangtze River metallogenic belt (MLYRB) and the Hanxing area are known to involve evaporite rocks in their mineralization. However, the specific interaction between evaporites and the ore-forming magma remains unclear. This study investigates the textures, in situ compositions, and Sr isotopes of igneous apatite from the Chengchao Fe skarn deposit (Edong district) using electron probe microanalysis (EPMA) and laser ablation multi-collector ICP-MS (LA-MC-ICP-MS). Results show that igneous apatite at Chengchao exhibits a wide range of Sr contents and isotopic ratios. The higher and lower Sr isotope values correlate with those of sedimentary/hydrothermal anhydrite/gypsum and local mafic rocks, respectively. This indicates a mixed Sr source and incorporation of evaporite components into the magma, although the quantity was limited, as indicated by the incomplete resetting of the bulk-rock Sr isotope signature. The REE content of primary magmatic apatite increases progressively from quartz diorite through diorite to granite. Apatite REE patterns mirror whole-rock patterns, confirming a close genetic link among these rock types. The addition of limited evaporite components did not significantly alter the magmatic differentiation trend (dominated by plagioclase). Nevertheless, evaporite involvement is interpreted to have modified magma composition and evolution, likely promoting the partitioning of Cl and Fe into an exsolved fluid phase—a process key to iron enrichment during early mineralization. This study highlights that when intermediate-felsic magma intrudes evaporite-bearing strata, it can generate an intrusion-centered system comprising Fe skarn and hydrothermal anhydrite/gypsum mineralization, alongside pre-existing sedimentary gypsum/anhydrite. If the intrusion occurs at the edge of a volcanic basin, iron oxide-apatite (IOA) mineralization may also develop at the system’s apex. Thus, these four mineralization types can form an intrusion-centered system and serve as mutually indicative prospecting targets.

To resolve the debate on the genesis of the Bayan Obo Fe-Nb-REE deposit, this study conducted in situ carbon isotopic and trace-element analyses on its carbonate minerals. Calcites from the carbonatite of the No.1 dyke and the Boluotou area exhibit δ¹³C_V-PDB values ranging from -9.24‰ to -7.47‰ (avg. -8.42‰) and from -5.43‰ to -4.14‰ (avg. -4.82‰), respectively, showing carbon isotopic compositions similar to those of igneous carbonatites or magmatic-hydrothermal fluids. Dolomites from the Main, East, and West open pits have δ¹³C_V-PDB values ranging from -5.53‰ to -2.47‰ (avg. -4.23‰), -5.26‰ to -2.46‰ (avg. -3.76‰), and -3.91‰ to 1.45‰ (avg. -2.97‰), respectively, falling between the isotopic fields of mantle and sedimentary carbonates. In contrast, siderites from the West open pit yield δ¹³C_V-PDB values from -0.61‰ to 1.54‰ (avg. 0.58‰), within the field of sedimentary carbonate rocks. Calcite, dolomite, and siderite are also distinguished by their trace-element compositions. Calcites display significant REE enrichment, with ΣREE contents ranging from 963.44×10^-6 to 5242.62×10^-6 (avg. 1945.83×10^-6). Dolomites have ΣREE contents ranging from 17.29×10^-6 to 321.15×10^-6 (avg. 89.79×10^-6; one anomalous sample at 3527.84×10^-6 is excluded), an order of magnitude lower than calcites, yet they remain significantly enriched. Siderites, however, show ΣREE contents from 0.28×10^-6 to 6.10×10^-6 (avg. 1.54×10^-6), three orders of magnitude lower than calcites, indicating no REE enrichment. Combined with geological and petrographic evidence, we conclude that the West open pit siderites are of sedimentary origin and unrelated to REE mineralization. The carbonatites with epigenetic features in the No.1 dyke and Boluotou area are likely magmatic or hydrothermal in origin and derived from the same source as the REE ore-forming materials. The widespread dolomite formed via magmatic-hydrothermal replacement of original sedimentary carbonate rocks, driven by Ca-, REE-, and F-rich fluids. Furthermore, magnetite mineralization in the Bayan Obo deposit may be derived from the transformation of sedimentary siderite. While a genetic link exists between magnetite and REE mineralization, their respective ore-forming materials were derived from distinct sources.

Robust Principal Component Analysis (RPCA) is a dimensionality reduction method widely used in machine learning and pattern recognition for unsupervised feature learning in multivariate statistical analysis. In response to increasing challenges in mineral exploration, this study employed an RPCA mathematical model based on the Inexact Augmented Lagrange Multiplier (IALM) algorithm to enhance the capability of airborne geophysical exploration for the rapid identification and extraction of weak anomaly information. This approach is not only robust but also well-suited for processing large volumes of data. Specifically, it was applied to separate the regional background field from local anomalies within aeromagnetic anomaly data. Theoretical model experiments demonstrated that this spatial-domain information extraction method effectively separates regional and local magnetic fields. The results are consistent with the model’s forward modeling outcomes, and the method achieves high computational efficiency. For measured data, the computation time per dataset was significantly shorter than that of the Exact Augmented Lagrange Multiplier (EALM) algorithm, representing a speed improvement of approximately two orders of magnitude (i.e., about 100 times faster). Finally, the algorithm was applied to process and interpret aeromagnetic anomalies in the Anshan-Benxi area, a sedimentary-metamorphic iron ore metallogenic district. The extracted local high magnetic anomalies exhibit strong correlations with the spatial distribution of known magnetite deposits, successfully delineating prospective exploration targets. Significant exploration potential remains within the local high magnetic anomaly areas near Anshan, Benxi, Gongchangling, and Dengta, where hidden magnetite deposits are likely to be discovered.

With the progression of geological prospecting, a distinct shift has emerged: a transition from shallow to deep-seated exploration targets and from singular methods to integrated technical systems. Over nearly seven decades of geological practice, China has established critical energy and mineral resource supply bases through systematic, multi-phase geological surveys. Yet, persistent declines in prospecting efficiency demand innovative solutions. This study leverages big data frameworks to advance quantitative prospecting target prediction, overcoming limitations of traditional approaches by pioneering a shift from integrated to classification-based modeling. Focusing on target accuracy, we employ statistical correlation analysis to integrate multi-source heterogeneous data (geological, geophysical, geochemical, and mining-related), constructing association models between multivariate parameters and known deposits to reveal latent correlations. Applied to the Jinchuan Cu-Ni deposit and its periphery, our integrated model delineated 37 targets (0.11-21 km², 91.89% of which are <2 km²), drastically narrowing the search scope. Field validation of five high-priority targets confirmed that: (1) there was an 86% overlap between Jinchuan’s mineralization and predicted Cu-Ni zones; (2) four targets fully coincided with tailings ponds and ash storage sites; and (3) three were validated as high-confidence exploration zones. Conversely, the M-15 magnetic anomaly (Area IV’s eastern extension) showed negligible Cu-Ni potential. This research demonstrates big data’s capacity to transform geological prospecting through empirical validation and iterative refinement, thereby setting a benchmark for analogous resource exploration.

Intelligent mineral prediction is a cutting-edge field in digital geology, where numerous machine learning methods have been widely applied to construct mineral prediction models. Current approaches typically convert various geological data into raster images, segment them into training units, and then learn to predict based on the presence or absence of known mineral occurrences as labels within each unit. However, these methods suffer from problems such as insufficient training data, complex model architectures, and an inability to clearly distinguish false anomalies. This study aims to construct a multi-source mineral prediction model by integrating regional geochemical exploration data and geological mineralization elements. Using gold deposits in southern Sichuan as a case study, we apply convolutional neural networks (CNNs) to directly process raw geological data, using distance intervals from known deposits as labels for supervised learning. The trained model is then applied to delineate prospective areas in frontier regions far from any known mineralization. The model ultimately identified seven prospective areas, demonstrating that CNNs can effectively capture spatial features around known gold deposits and extract meaningful patterns from raw input data. This study provides a new solution for exploration in blank areas, improves the accuracy of gold prospectivity mapping, and offers a novel experimental approach for applying machine learning to mineral exploration.

This study presents, for the first time, a high-precision analytical method for Mg isotopes using the Nu Sapphire MC-ICP-MS. The chemical purification procedure for Mg isotopes is described, and the effects of residual matrix elements (Al, Ca, Fe, K, Mn, Na, and Ti) in the purified solutions on the accuracy of Mg isotope measurements are systematically evaluated. The impacts of mismatches in Mg concentration and acidity between samples and standards are also assessed. Sixteen geological reference materials with diverse chemical compositions were analyzed, and the measured Mg isotope compositions are consistent with previously published data within analytical uncertainties. Furthermore, this study reports the Mg isotope values of four low-Mg carbonate reference materials: GBW07108, GBW07120, GBW03107A, and NIST SRM-1d. Their δ²⁶ Mg values are -1.66‰±0.05‰, -2.02‰±0.02‰, -1.73‰±0.03‰, and -3.41‰±0.01‰, and their δ²⁵ Mg values are -0.90‰±0.04‰, -1.07‰±0.02‰, -0.90‰±0.02‰, and -1.79‰±0.03‰, respectively, providing reference values for inter-laboratory comparison. Finally, twelve geological samples with varying MgO, CaO, and MnO contents, previously measured at the University of Washington Isotope Laboratory, were reanalyzed. The results obtained in this study are consistent with the original data within error, further validating the reliability of the established chemical purification and instrumental analysis methods.

The Huainan and Huaibei coalfields, as two of the most important coal-producing areas in China, have long attracted significant attention regarding coal resource mining and utilization. As an important environmental resource, coal requires a thorough understanding of the content, occurrence modes, and enrichment factors of its trace elements, which is crucial for geochemical characterization, efficient utilization, and environmental protection. Based on a comprehensive review of domestic and international literature and previously published laboratory data, we analyzed the coal quality and conducted a geochemical study of trace elements in the Huainan and Huaibei coalfields, focusing on: (i) distribution and enrichment characteristics of trace elements in different coal seams; (ii) comparison of trace elements between the two coalfields; (iii) modes of occurrence; and (iv) controlling factors of enrichment. Our findings show that As, B, Be, Cd, Sb, Se, and Sn are slightly enriched in Huainan coal, whereas Be, Cd, Cr, Cu, Hg, Ni, Pb, Sb, Se, Sn, U, and V are slightly enriched in Huaibei coal, with considerable variation among seams. Huainan coal is enriched in light and medium rare earth elements, whereas Huaibei coal is enriched in heavy rare earth elements. Vertically, trace elements in the coal-bearing strata exhibit three overall trends from bottom to top: decreasing, increasing, and increasing then decreasing. Elemental associations exhibit both commonalities and differences across seams. Trace elements in these coalfields occur in various forms, including silicate-bound, sulfide-bound, carbonate-bound, organically bound, ion-exchangeable, and water-soluble forms. Mineralogical transformations in coal are also closely related to trace element redistribution. The degree of trace element enrichment is controlled by multiple factors, including parent rocks in the terrigenous area, sedimentary environment, structure, and magmatic-hydrothermal activity. A deeper understanding of these trace element geochemical characteristics is essential for efficient coal utilization, environmental protection, and advancing green coal chemistry.

Seismic technology is a primary method for coalbed methane (CBM) exploration. By analyzing the seismic response characteristics of CBM reservoirs, their prediction and evaluation can be achieved. This article presents a forward modeling study based on the wave equation to simulate seismic responses. We constructed geological models with varying coal seam thicknesses and coal structures to analyze their influence on post-stack seismic attributes. Additionally, pre-stack AVO response characteristics were investigated using models with different coal seam thicknesses, roof/floor lithologies, and gas contents. Experimental results demonstrate that increasing coal seam thickness and transforming coal structure from undeformed to mylonitized coal both significantly enhance the amplitude of seismic reflections from the coal seam. This relationship provides a theoretical basis for using post-stack seismic amplitude to predict coal seam thickness and coal structure. For coal seams of various thicknesses, the AVO intercept remains negative, and the absolute reflection coefficient decreases with increasing incident angle. Regarding the surrounding lithology, seismic reflections exhibit stronger amplitudes when coal seams are bounded by hard sandstone roofs and floors, and weaker amplitudes when bounded by soft mudstone. Moreover, variations in floor lithology have a weaker influence on seismic reflections than changes in roof lithology. As gas content increases, the corresponding seismic reflection intensity also rises, indicating higher gas saturation. These findings provide a theoretical basis for seismic exploration of CBM reservoirs and help optimize CBM exploration strategies.

The theory of the whole petroleum system (WPS) can explain the orderly coexistence mechanism of conventional and unconventional resources. The Badaowan Formation in the Sishilicheng area of the Yanqi Basin contains three types of resources: shale oil, tight oil, and conventional oil, exhibiting characteristics of a whole petroleum system. Based on the theory of the whole petroleum system, a static element matching analysis and a dynamic element evolution analysis of reservoir formation were conducted for the conventional and unconventional resources in the Badaowan Formation within the area. This analysis clarified the reservoir formation characteristics of each type of resource and established a reservoir formation model for the whole petroleum system. The research indicates that in the study area, the Badaowan Formation features a sequential coexistence of structural oil reservoirs, tight oil reservoirs, and shale oil reservoirs, with significant differences in the reservoir formation characteristics of various types of oil and gas. The whole petroleum system in the Badaowan Formation of the study area has favorable reservoir formation conditions, with source rocks supplying sufficient oil and gas, multiple types of reservoirs providing favorable storage space, strong sealing performance of regional cap rocks, stratigraphic occurrence and faults controlling oil and gas migration and entrapment, and time matching of reservoir formation elements for conventional and unconventional resources. In the low-lying sag area of the study area, unconventional oil reservoirs exhibit characteristics of retention and accumulation, close proximity of source and reservoir, and overpressure driving. In the high-lying northwest edge of the slope zone, conventional oil reservoirs exhibit characteristics of source-reservoir separation, long-distance migration, and buoyancy driving.

Focusing on global plate drift, this study addresses the open question of whether plate drift velocity influences the development of hydrocarbon source rocks. Taking the Paleozoic as an example, we reconstructed global plate paleogeography and statistically analyzed source rock data from major plates worldwide. On this basis, we calculated plate drift velocities and examined their relationship with source rock development. Our results show that the average drift velocities of global plates during the Paleozoic can be categorized into three levels: low (0.10-4.00 cm/a), medium (4.00-8.00 cm/a), and high (8.00-25.76 cm/a). Among these, low drift velocities were the most favorable for source rock development, followed by medium velocities, while high velocities were unfavorable. Furthermore, when low drift velocities occurred at low latitudes, they significantly promoted source rock development, showing distinct advantages in both the quantity of source rocks and the occurrence of high-TOC source rocks. These findings provide a new perspective and theoretical basis for global hydrocarbon exploration and for advancing petroleum geology research.

The Paleozoic strata in the Ordos Basin are important natural gas-rich and high-yielding intervals. Natural gas from the Carboniferous Benxi Formation to the Permian Taiyuan Formation, which includes bauxite-bearing strata, is generally enriched in helium; however, helium contents vary significantly across different areas. The key controlling factors and distribution patterns of helium enrichment in these bauxite strata require further investigation.Both the Proterozoic-Archean basement rocks and the Carboniferous-Permian sedimentary rocks—including organic-rich mudstone, coal, and bauxite—contain high concentrations of uranium and thorium, making them effective helium source rocks. Helium contents in the Carboniferous-Permian bauxite strata mostly range from 0.02% to 0.15%, with ³He/⁴He ratios generally below 0.05 R_a and a maximum not exceeding 0.1 R_a, indicating a predominantly crustal origin for the helium. In addition to basement volcanic and metamorphic rocks, sedimentary rocks such as mudstone, shale, coal, and bauxite also serve as effective helium source rocks for the bauxite formations. Helium content varies considerably among gas fields. Overall, it exhibits a pattern of high values in the northern and southern parts, low values in the central region, and moderate values on the eastern and western sides. Specifically, the Dongsheng and Qingyang gas fields in the north and south have helium contents reaching 0.3%; the Shixi and Linxing gas reservoirs in the east generally exceed 0.1%; whereas the Sulige, Yulin, and Shenmu gas fields in the central region typically have helium contents below 0.1%. Based on the relationship between helium content and the distribution of basement rocks, faults, uplifts, and sedimentary strata in different areas of the Ordos Basin, the enrichment pattern of helium in bauxite strata is revealed: a dual helium supply from both the basement and bauxite sedimentary layers, with faults, uplifts, and gypsum-salt strata controlling the accumulation. Influenced by uplifts, basement faults, and the dual helium sources, the Qingyang and Dongsheng gas fields have relatively high helium content. In contrast, due to the blockage of basement-derived helium by Cambrian-Ordovician carbonates and evaporites in the central basin, gas fields such as Sulige, Yulin, and Jingbian generally exhibit lower helium content. Favorable helium-rich zones in the bauxite strata include Hangjinqi in the north, Qingyang-Longdong in the south, Etuokeqianqi-Dingbian-Wuqi in the west, Fuxian-Yichuan in the southeast, and Linxian-Shilou in the east.

The hydrate reservoir in the Shenhu Sea area is a complex stratigraphic system characterized by the coexistence of gas hydrate, free gas, and aqueous phase. The presence of underlying free gas significantly influences fluid migration and heat transfer dynamics, thereby affecting overall production behavior. To elucidate the dissociation mechanisms of stratigraphic hydrate reservoirs under depressurization and investigate the role of pre-existing free gas, we developed a multi-layer physical model based on the W17 hydrate site, explicitly incorporating hydrate, free gas, and water layers. Using the TOUGH+HYDRATE numerical simulator, we simulated long-term depressurization-induced gas production and comprehensively analyzed the co-production of methane from both hydrate dissociation and the underlying free gas zone. Results show that hydrate dissociation rates progressively decline over time, leading to a diminishing contribution of hydrate-derived gas to total wellhead output. After ten years of continuous operation, over 50% of the produced gas originates from dissolved methane and initial free gas originally present in the formation. The presence of free gas alters pressure propagation through the sedimentary column, thereby creating localized conditions that favor hydrate reformation at layer boundaries and near the dissociation front. Although free gas contributes to higher wellhead energy yields by supplementing gas supply, it simultaneously impedes efficient hydrate dissociation in the reservoir matrix, resulting in an approximately 30%-42% reduction in cumulative hydrate decomposition after ten years. Furthermore, pore water from adjacent caprocks is continuously drawn into the reservoir and flows into the wellbore under the combined influence of gravity and pressure gradients. In hydrate reservoirs containing free gas, the external water influx is slightly greater than the wellhead water production. This discrepancy may be attributed to substantial free gas production and water consumption caused by hydrate reformation during extraction.

The bioavailability of coal-sourced organic matter (CSOM) is crucial for the sustainable utilization of coal resources and environmental protection. This review summarizes the definition, origin, microscopic composition (macerals), and chemical structure of CSOM, as well as the effect of coalification. It focuses on the mechanisms of microbial (bacteria, fungi, and archaea) degradation of CSOM, including the overall process, the role of key enzyme systems (e.g., oxidases, hydrolases), major microbial degraders, and degradation products. Three major categories of factors influencing CSOM bioavailability are analyzed: physicochemical properties of coal (coal rank, macerals, structure, etc.), environmental factors (temperature, pH, redox potential, etc.), and microbial factors (community structure, functional genes, syntrophic interactions, etc.). Key applications based on CSOM biodegradation mechanisms are discussed, including microbial enhanced coalbed methane recovery (ME-CBM), coal bioliquefaction for chemical production, bio-humic acid production and soil amendment, and environmental remediation practices such as acid mine drainage treatment and resource recovery from coal-based solid wastes. The review indicates that a deeper understanding and effective regulation of CSOM bioavailability hold promise for the development of novel biotechnological pathways for the clean and efficient conversion of coal.

The background concentrations of groundwater components are crucial for evaluating groundwater pollution and assessing the genesis of natural poor-quality water. However, research on the background concentrations of arsenic in natural high-arsenic groundwater has been relatively limited. This study selected the Datong Basin as the research area and comprehensively employed the cumulative frequency method, component separation method, and machine learning random forest model to investigate arsenic enrichment processes, quantify the relative contributions of various environmental factors, and preliminarily predict the background arsenic concentration in groundwater. Results showed that arsenic concentrations in groundwater ranged from 0.03 to 1081 μg/L, with high-arsenic groundwater mainly distributed in runoff-discharge areas. The cumulative frequency method indicated that the upper limit of the background arsenic concentration in the study area was 7.59 μg/L; for the recharge area and the runoff-discharge area, the respective upper limits were 4.47 μg/L and 10.9 μg/L. The component separation method yielded a background value of 13.3 μg/L for the study area, with values of 6.16 μg/L and 33.4 μg/L for the recharge area and runoff-discharge area, respectively. The random forest model predicted that the formation of high-arsenic groundwater was the result of multiple interacting environmental factors, with varying relative importance among them. The concentrations of SO₄^2- and Fe²⁺ were found to be highly important for predicting the arsenic background value, indicating that high-arsenic groundwater in the Datong Basin was mainly influenced by coupled sulfur-iron cycling. These findings improve the understanding of high-arsenic groundwater formation mechanisms and provide a scientific basis for ensuring local drinking water safety.

To address the challenges of global climate change, the Paris Agreement, signed by nearly 200 countries, sets the goal of achieving “carbon neutrality” in the latter half of this century. China has established targets to peak carbon emissions before 2030 and achieve carbon neutrality by 2060, referred to as the carbon peak and carbon neutrality goals. These objectives not only reshape the global climate governance landscape but also drive a profound transformation of territorial spatial development patterns toward green and low-carbon directions. Currently, China's existing underground space exceeds 3.2 billion square meters and continues to grow rapidly, constituting strategic territorial space and resources that must be considered for achieving the carbon peak and carbon neutrality goals. This paper primarily considers four resources (underground space, water, geotechnical materials, and geothermal energy) as urban underground multiple resources, and analyzes the synergistic utilization relationships among them. It proposes three main pathways for urban underground multiple resources utilization in response to the carbon peak and carbon neutrality goals: artificial carbon sinks, ecological carbon sinks, and carbon source reduction. Furthermore, it identifies four key research and development directions to facilitate the carbon peak and carbon neutrality goals: (1) prioritized development of “underground-aboveground ecological transition”; (2) timely and site-specific underground artificial carbon sinks; (3) novel low-carbon underground transportation systems; and (4) underground multi-energy complementary system based on wind-photovoltaic-hydrogen storage.