Located in the eastern segment of the North Qilian Orogenic Belt, the Xihuashan area is a key metallogenic zone in Ningxia Hui Autonomous Region, characterized by favorable geological conditions and densely distributed gold-polymetallic mineralization. A number of small-scale gold-polymetallic deposits, mineralized occurrences, and integrated geophysical-geochemical anomalies have been discovered in this region over previous exploration works. Nevertheless, prior studies have mainly focused on the genetic analysis and basic geological characteristics of individual deposits, failing to systematically summarize regional ore-controlling regularities. In addition, traditional exploration relies on single technical methods rather than integrated approaches, leading to a low degree of overall exploration and the absence of significant prospecting breakthroughs for a long time.

Based on comprehensive collation of historical exploration data and detailed field geological surveys, this study systematically analyzed the dominant ore-controlling factors of local gold-polymetallic deposits. Multiple geoscientific technologies were adopted in this study, including 1∶5000 induced polarization (IP) survey, 1∶50000 stream sediment geochemical survey, and WorldView-3 (WV-3) hyperspectral remote sensing interpretation via principal component analysis (PCA). On this basis, an integrated prospecting prediction model coupled with multi-source geological, geophysical, geochemical, and remote sensing datasets was constructed, and potential prospecting targets were delineated and verified by rock geochemical profile measurements.

The analytical results revealed that regional gold-polymetallic mineralization was jointly controlled by tectonics, stratigraphy, and magmatic activities. A three-tier fault system dominated ore localization. The first-order NW-striking regional faults acted as major ore-transporting channels, the secondary NNW-striking faults controlled ore distribution, and the third-order NW-NWW striking interlayer fractures were the primary ore-hosting spaces. Faults and folds formed synchronously under a regional compressional tectonic regime, and the reactivation of first-order faults in the late stage caused damage to pre-existing ore bodies. Stratigraphically, the Tiandushan Formation served as the optimal host stratum for gold deposits, and the Bojizhang Formation was the favorable horizon for copper mineralization. Caledonian magmatism activity provided abundant ore-forming materials, hydrothermal fluids, and thermal power for the entire mineralization process. IP surveys at a scale of 1∶5 000 in the Liugou area revealed that all anomalies were characterized by high chargeability. The high-resistivity and high-chargeability anomalies were closely associated with shallow mineralized veins and alteration zones, whereas the low-resistivity and high-chargeability anomalies indicated deep-seated sulfide enrichment and active structural-hydrothermal fluid migration. The 1∶50 000 stream sediment geochemical survey delineated two large-scale and high-intensity composite geochemical anomalies: Au-Cu-Ag-As-Mo and Pb-Au-As-Ag. Hyperspectral remote sensing data from WV-3, processed by PCA, successfully identified three typical alteration anomalies including hydroxyl group anomalies, carbonatization anomalies, and iron-staining anomalies, which showed excellent spatial correlation with known deposits and mineralized outcrops in the study area. Integrating all multi-source geological, geophysical, geochemical, and remote sensing information, a total of five prospecting targets are delineated in this study, consisting of three Class Ⅰ high-priority targets and two Class Ⅱ potential targets. The three Class Ⅰ targets are the core areas for prospecting orogenic gold-polymetallic deposits, which correspond to two typical elemental assemblages of Au-Cu-Ag-As-Mo and Pb-Au-As-Ag, respectively. The Class Ⅱ targets also possess favorable metallogenic conditions and considerable prospecting potential. Field verification using rock geochemical profiles demonstrates that geochemical anomalies have a clear and positive correlation with underground and surface mineralization. The findings prove that the delineation of prospecting targets is scientifically sound and reliable.

This integrated research model and target evaluation results can provide solid theoretical support and practical guidance for further regional mineral exploration, deep ore prospecting, and engineering deployment in the Xihuashan area.

Deep- to ultra-deeply buried tight sandstone reservoirs have great potential for oil and gas exploration. The development and spatial distribution patterns of fractures are key factors for the improvement of reservoir performance in such reservoirs.

To clarify fracture genesis and its control on reservoir development in tight sandstones, the Cretaceous Yageliemu Formation in the Kuqa Depression, Tarim Basin, is selected as the study area.

Integrated analyses, including drilling core, thin section, laser confocal microscopy, scanning electron microscopy, detrital zircon geochronology, heavy mineral composition, and carbon and oxygen stable isotopes, were conducted to determine the genetic types and main controlling factors of fractures in deep tight sandstones of the Yageliemu Formation. A fracture-controlled reservoir evolution model was established.

The results showed that the rock types in well area A were mainly lithic sandstone and feldspathic lithic sandstone. The rock fragment were mainly sedimentary rock and metamorphic rock. Rock types in well area B were mainly lithic sandstone, and the rock fragments were mainly magmatic rock. Well area B had higher rock fragment content. Three stages of tectonic fractures were identified in the study area: ① In the first stage, fractures were characterized by relatively wide openings (2-4 mm), high-angle to nearly vertical fractures (70°-90°), straight and smooth surfaces, mainly shear fractures filled with calcite. Fracture filling occurred during 65-45 Ma, corresponding to the slow and shallow burial stage from the late Yanshanian to the early Himalayan period; ② In the second stage, fractures exhibited narrow openings (1-2 mm), medium- to high-angle fractures (40°-60°), slightly curved shapes, mainly tension-shear composite fractures filled with kaolinite cement. Fracture filling occurred during 40-20 Ma, corresponding to the rapid deep burial stage of the middle of the Himalayan period; ③ In the third stage, fractures showed the narrowest openings (0.2-1 mm), low-angle to nearly horizontal fractures (10°-30°), curved shapes, and were mainly tensile fractures filled with ankerite cement. The fracture filling occurred during 10-6 Ma, corresponding to the thrust-adjustment stage of the late Himalayan period. Under a uniform tectonic compression settings, differences in provenance systems and rock composition resulted in different fracture-controlled reservoir evolution models between well areas A and B. Well area A had higher contents of brittle minerals, resulting in significant development of fractures during extensive tectonic compression. This facilitated late-stage acidic dissolution, significantly enhancing porosity and permeability. Overall, reservoir quality in well area A was better than in well area B.

These findings provide a geological basis for the efficient exploration and development of deep tight sandstone reservoirs in the Kuqa Depression.

Dispersed metals including Cd, Ga, In, Ge, Tl, Tc, Se, and Te are classified as national strategic critical mineral resources globally, which are irreplaceable raw materials for new energy equipment, semiconductor manufacturing, and information high-tech industries. Most dispersed elements are characterized by ultra-low crustal abundance and extremely scattered distribution, and they rarely form independent industrial deposits. Instead, they are dominantly hosted in sphalerite within hydrothermal Pb-Zn deposits, making sphalerite an ideal mineral carrier to decode the super-enrichment mechanisms of Cd, Ga, In, and Ge. The Taolin Pb-Zn deposit, located in Linxiang City, northeastern Hunan Province, lies in the central segment of the Jiangnan orogenic belt on the southeast margin of the Yangtze block. It is a large-scale medium-low temperature hydrothermal deposit with total proven Pb+Zn metal reserves of up to 0.98 million tons. Previous research on this deposit mainly focuses on regional tectonic controls, magmatic evolution, and general metallogenic patterns, while systematic microscale constraints on the occurrence and enrichment mechanisms of associated dispersed metals remain lacking, which restricts the comprehensive resource evaluation of co-existing critical metals. To fill this research gap and improve the metallogenic theory of polymetallic deposits in northeastern Hunan, this study takes zoned sphalerite from Taolin deposit as the research object and conducts integrated petrographic and in-situ geochemical analyses.

A total of seven weakly altered drill-core ore samples were collected from multiple mining segments of the deposit. Through hand-specimen observation and polished thin-section microscopic identification, two generations of sphalerite were distinguished: Early-stage SphⅠ formed in early hydrothermal veins and late-stage SphⅡ filling later veinlets that crosscut early mineralized zones. Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) microscale testing was conducted based on a 193 nm ArF excimer laser coupled with an Agilent 7700e mass spectrometer, including 100 quantitative trace element point analyses (69 spots for SphⅠ, 31 spots for SphⅡ) and full elemental mapping for representative sphalerite grains. NIST 610, NIST 612 glass standards and MASS-1 sulfide standard were adopted for signal calibration to ensure data reliability.

The analytical results showed distinct differentiation of dispersed elements in Taolin sphalerite: The mineral was highly enriched in Cd and Ga, moderately enriched in In, and strongly depleted in Ge, while Tl and Te were not detected in all testing points. The mass fraction of Cd ranged from 945.9×10⁻⁶ to 11734.8×10⁻⁶ with an average value of 2978.0×10⁻⁶. Ga varied between 0.7×10⁻⁶ and 3331.7×10⁻⁶, averaging 310.7×10⁻⁶. In had a mean concentration of only 27.1×10⁻⁶, with a maximum of 322.4×10⁻⁶, while Ge averaged merely 6.7×10⁻⁶, and over half of the testing points fell below the limit. Element mapping demonstrated that Cd, Ga, and In were unevenly distributed inside single sphalerite crystals, and Ga displayed an extremely strong positive correlation with Cu (R²=0.93), as did the sum of Ga+In against Cu (R²=0.92). These correlations indicated two isomorphic substitution pathways inside the sphalerite lattice. The coupled charge-balancing reactions were Cu⁺+(Ga+In)³⁺↔2Zn²⁺ and Cd²⁺↔Zn²⁺. Notably, Cd presented a weak positive correlation with Fe in the samples, which contradicted the widely accepted negative correlation between Cd and Fe in high-temperature Fe-rich sphalerite. This discrepancy was interpreted as the limitation of crystal lattice capacity under different formation temperatures. Source tracing indicated that Cd, Ga, and In had mixed material supplies. The primary metal reservoir was late magmatic-hydrothermal fluids derived from the Mufushan Yanshanian biotite monzogranite, whose zircon U-Pb age (136±0.8 Ma) is consistent with the sphalerite Rb-Sr mineralization age (135.4±2.6 Ma). Secondary dispersed metals were extracted by ore-forming fluids when migrating through Neoproterozoic Lengjiaxi Group and Sinian siliceous-carbonaceous slates, containing abundant organic matter and adsorbed trace sulfides. Sulfur isotopic data of ores provided solid evidence for stratigraphic material contribution. During fluid migration, Cd, Ga, and In formed stable fluoride and organic humic complexes for long-distance transportation. As hydrothermal fluids ascended to shallow structural fractures, continuous temperature reduction and progressive mixing of meteoric water increased fluid oxidation intensity. Under high oxidation conditions, Cu⁺ converted to Cu²⁺, eliminating the monovalent cation required for coupled isomorphic substitution, explaining the evidently lower Ga content in late-generation SphⅡ compared with early SphⅠ.

This study proposes a complete multi-source metallogenic model for dispersed elements in medium-low temperature Pb-Zn deposits within the Jiangnan orogenic belt. It innovatively reveals the temperature and redox controls on Cd-Ga-In enrichment in sphalerite, supplements geochemical discrimination criteria for hydrothermal mineralization, and offers significant mineralogical references for prospecting and comprehensive utilization of associated critical metals in analogous deposits across northeastern Hunan and the whole Jiangnan metallogenic belt.

The Fuludi gold deposit is located in the middle segment of the Muping-Rushan gold metallogenic belt, Jiaodong Peninsula, a world-class gold concentration area with substantial gold resources. Previous geological studies on deposits in this belt have mainly focused on gold deposits in its southern and northern sections, while the origin and ore-forming fluid evolution of Fuludi gold deposit have long remained poorly understood. Controversies still exist over the source of ore-forming fluids in the Muping-Rushan belt, with three mainstream viewpoints: Mantle-derived fluid, mixed magmatic fluid and meteoric water, and mixed magmatic water and metamorphic water. Geographically, the Fuludi gold deposit acts as a key link connecting the northern and southern parts of the metallogenic belt. However, previous studies have only carried out basic geological surveys and divided its mineralization stages, and systematic studies on ore-forming fluids have long been lacking. This study aims to clarify the source, spatiotemporal evolution of ore-forming fluids, and gold precipitation mechanism of the Fuludi gold deposit.

Combined with detailed field and microscopic geological characteristics, this study selected quartz samples from four different mineralization stages to conduct a comprehensive analysis, including fluid inclusion microthermometry, laser Raman spectroscopy, and hydrogen-oxygen (H-O) isotope testing. It systematically analyzed the petrographic characteristics of fluid inclusions, physicochemical parameters of ore-forming fluids, fluid compositions, and isotopic compositions, so as to constrain the fluid source and gold metallogenesis.

According to the cross-cutting relationships of quartz veins and paragenetic mineral assemblages, four mineralization stages were divided: Milky quartz stage (Stage Ⅰ), smoky quartz+pyrite early mineralization stage (Stage Ⅱ), smoky quartz+polymetallic sulfide main mineralization stage (Stage Ⅲ), and quartz+calcite late mineralization stage (Stage Ⅳ). Two types of fluid inclusions were identified in quartz: Pure liquid (L-type) aqueous fluid inclusions and gas-liquid two-phase (L+V-type) fluid inclusions. The L+V-type inclusions occurred throughout all four stages, with particle sizes of 3−10 μm and showing elongated, oval, and irregular shapes. Their liquid-phase proportion gradually increased from Stage Ⅰ to Stage Ⅳ, ultimately reaching approximately 80%. The L-type inclusions only developed in Stage Ⅲ, with a particle size of 5−15 μm and mostly oval shapes. Laser Raman analyses revealed that the ore-forming fluid belonged to the CO₂-H₂O-NaCl system, and its components varied significantly at different stages. CH₄ was detected in Stage Ⅰ, no CH₄ existed in Stage Ⅱ, N₂ appeared in Stage Ⅲ, and only CO₂ and H₂O were found in Stage Ⅳ. Microthermometric results showed that the early Stage Ⅰ and Stage Ⅱ had stable physicochemical conditions, with homogenization temperatures ranging from 180.0 °C to 240.0 °C and salinity peaks of 9.0%−17.0%. Fluid boiling occurred at the main Stage Ⅲ, accompanied by obvious decreases in temperature (160.0-200.0 °C) and salinity (11.0%−15.0%). At the late Stage Ⅳ, temperature (120.0-180.0 °C) and salinity (3.0%−11.0%) decreased further. H-O isotope results showed that δD values ranged from −87.0‰ to −72.4‰, and δ¹⁸O values of quartz were 9.0‰-14.0‰. Calculated by oxygen isotope fractionation equation, the δ¹⁸OH₂O of ore-forming fluid was −3.98‰ to 1.82‰.

The ore-forming fluids are dominated by mixed magmatic water and metamorphic water, and are also accompanied by mantle-derived components originating from volatile degassing of enriched mantle. Meteoric water continuously mixed into the fluid system during mineralization. With the gradual decrease of temperature and pressure, combined with fluid boiling and escape of volatile CO₂, the physicochemical properties of fluids change significantly. These processes broke the stability of ${\mathrm{Au}}({\mathrm{HS}})_2^- $ complexes, which are the main transport carrier of gold, and led to the rapid precipitation of gold and polymetallic minerals. Comprehensive geological and geochemical evidence indicates that the Fuludi gold deposit is a typical medium- to low-temperature, low-salinity quartz vein-type hydrothermal gold deposit controlled by NNE-trending faults. This study fills the research gap on ore-forming fluids in Fuludi gold deposit, and provides reliable geological evidence for regional metallogenic theory and further prospecting work.

Although both the Zhaoyuan-Laizhou and Penglai-Qixia metallogenic districts in the Jiaodong Peninsula are located in a Mesozoic compressional-extensional transitional setting, their ore-controlling structural characteristics exhibit significant differences. This study aims to compare the ore-controlling patterns and structural system differences of faults in the two districts and explore the formation mechanisms of these differences.

By comparing the fault attitudes, deformation characteristics, and structural systems of faults at different scales in the two gold-concentrated districts, combined with an analysis of the ore-controlling features and ore body localization patterns along three major faults—Jiaojia, Zhaoping, and Huluxian—the differences in fault-controlled mineralization and structural systems between the two districts were systematically investigated.

The results showed that the Jiaojia and Zhaoping fault zones in the Zhaoyuan-Laizhou district were dominated by low-angle listric extensional mechanisms, exhibiting multi-stage extensional shear deformation with moderate to gentle dips. The mineralization is primarily altered-rock type, with ore bodies occurring within the main fault zones at small pitch angles. A series of steeply dipping secondary ore-controlling structures were developed in the footwall of the main fault zones, characterized by steep or nearly vertical attitudes. These structures had quartz-vein and altered-rock type mineralization, with ore bodies displaying larger pitch angles, indicating that the deformation mechanism of the secondary structures was dominated by strike-slip movement. In the Penglai-Qixia district, represented by the Huluxian fault, high-angle brittle faults were developed. The subsidiary faults on both sides were steeply dipping and controlled quartz-vein type mineralization, with ore bodies showing negligible pitch angles, indicating a predominantly strike-slip mechanism. The structural differences between the two districts reflect a transition in the tectonic regime from extension in the west to strike-slip in the east across the northwestern Jiaodong Peninsula. These differences may be controlled by changes in the regional stress field during the Mesozoic compressional-extensional transition.

The research findings can provide a structural theoretical basis for deep mineral exploration in the Zhaoyuan-Laizhou and Penglai-Qixia metallogenic districts in the Jiaodong Peninsula.

The South China Sea is rich in deep-water oil and gas resources, but dee-pwater low-temperature environments significantly delay the early strength development of oil well cement, which restricts the safety, quality, and efficiency of cementing operations and increases engineering costs.

To solve the problem of insufficient early strength of oil well cement under deep-water low-temperature conditions and promote the application of green low-carbon building materials in petroleum engineering, rice husk ash (RHA), as an eco-friendly supplementary cementitious material, was incorporated into Class G oil well cement in this study. A systematic experimental investigation was conducted at a low temperature of 10℃ to reveal the effects of RHA particle size (four grades: 11.4-56.9 μm) and dosage (5%, 10%, 15%) on the early hydration characteristics, compressive strength development, hydration heat release behavior, hydration product evolution, and microstructure formation of oil well cement pastes. A series of characterization methods were adopted, including compressive strength test, isothermal calorimetry, thermogravimetric and derivative thermogravimetry (TG) analysis, and scanning electron microscopy equipped with energy-dispersive X-ray spectroscopy (SEM-EDS).

The results showed that RHA dosage and particle size synergistically regulated cement hydration kinetics, microstructure evolution, and the generation of hydration products, thereby dominating the mechanical performance of hardened cement pastes. With the increase of RHA dosage, the compressive strength increased first and then decreased. The 1 d hydration age strength reached the maximum at 5% RHA dosage, while the 3 d and 7 d strengths reached their peaks at 10% RHA dosage. As RHA particles were refined, the 1 d and 3 d strengths increased continuously, whereas the 7 d strength rose first and then declined, with the T1RHA group exhibiting the optimal overall performance. Low-dosage RHA enhanced cement strength through three mechanisms: Pozzolanic reaction, nucleation site effect, and spatial filling effect. In contrast, high-dosage RHA caused performance degradation due to the dilution effect and particle agglomeration. Finer RHA possessed higher pozzolanic activity and more intense hydration heat release. Nevertheless, excessively fine RHA accelerated the early formation of C-S-H gels, wrapping unhydrated cement particles and hindering subsequent hydration. The TG results verified that RHA consumed calcium hydroxide (CH) via pozzolanic reaction to generate additional C-S-H gels, optimizing the composition and microstructure of hydration products. SEM-EDS observations showed that RHA refined the pore structure, converted amorphous C-S-H into fibrous and ribbon-like morphologies, and lowered the Ca/Si molar ratio, contributing to a denser microstructure.

This study clarifies the coordinated regulation mechanism of RHA particle size and dosage on the early hydration of oil well cement under low-temperature conditions, and provides a theoretical basis and technical support for the design and application of green and low-carbon cementing systems suitable for deep-water low-temperature environments.

Landslides, as one of the most prevalent geological hazards in China, are widely distributed and have also extended into the western regions. The Qinghai region is characterized by complex geomorphic units and clustered mountain systems, which provide favorable geological conditions for the initiation and development of landslides. A comprehensive investigation into the formation mechanisms and influencing factors of representative landslides in this region can provide essential theoretical support for landslide prevention, mitigation, and hazard forecasting, thereby reducing casualties and economic losses.

This study focused on the accumulation landslide group of Hanjiacun, Qutan Town, Ledu District, Qinghai Province. Based on field investigations and monitoring data, the macroscopic deformation characteristics and formation mechanisms of the landslide group were systematically analyzed. Furthermore, the correlation between rainfall, temperature, and the deformation time series was examined using wavelet coherence analysis. Temperature and rainfall were selected as the principal external variables. A linear regression ensemble model was employed, in which the predicted displacements from individual models were combined through a weighted summation approach to estimate the displacement at the GNSS2 monitoring point of the landslide.

The results indicated that the Hanjiacun landslide group exhibited an average annual deformation rate of approximately 8.5 mm, classifying it as a typical creep-type landslide. Its displacement demonstrated a step-like deformation pattern under the influence of both rainfall and temperature. Specifically, rainfall showed a positive correlation with cumulative displacement, with abrupt increases observed during periods of concentrated summer rainfall, followed by stabilization after the rainy season, while a lag effect was also evident. Temperature, in contrast, was negatively correlated with cumulative displacement. As temperatures decreased in winter, frost heave was induced by the freezing and volumetric expansion of pore water within the soil matrix of the slope, resulting in an increase in landslide deformation. With the onset of spring, thaw settlement caused a rebound phenomenon in the accumulation layer. The ensemble model achieved a goodness-of-fit of 0.990 for displacement prediction at the GNSS2 monitoring point and can accurately predict the landslide deformation of the landslide group.

The established multiple linear regression ensemble model shows excellent prediction accuracy and can be applied to short-term displacement forecasting of similar creep-type colluvial landslides in alpine cold regions.

The stability coefficient of unstable rocks is a critical metric for assessing rockfall hazards. Traditional two-dimensional (2D) cross-sectional models, which fail to account for three-dimensional (3D) geometric characteristics and the synergistic effects of multiple fractures, often result in substantial errors in the calculation of stability coefficients.

In this study, a 3D stability calculation model for sliding unstable rocks with steeply inclined fractures at the rear edge was developed based on the theory of limit equilibrium. Additionally, a calculation method for the water pressure acting on unstable rocks under the influence of multiple groups of rear-edge fractures in 3D space, along with a calculation method for the uplift force on the sliding surface of unstable rocks in 3D spatial configurations, was proposed. The model was applied to the Dazhaokou unstable rocks in Fuling District, Chongqing, and the differences between the 3D and 2D model calculations were compared and analyzed.

The results indicate that the 3D model could accurately characterize the irregular geometry of the unstable rocks and the hydro-mechanical coupling effects of multiple fractures. Under heavy rainfall conditions, the stability coefficient calculated for both fractures filled with water (case ⑦) was 5.04% lower than that for single fracture filled with water (case ③). Numerical simulation validation demonstrated that the discrepancy between the 3D limit equilibrium method and the strength reduction method was about 0.4%. The shape of unstable rocks significantly influences stability. Except for regular cubic shapes, 3D analysis methods are required in most cases to ensure assessment accuracy.

This research provides theoretical and technical support for accurate stability assessment of sliding unstable rocks under complex conditions.

The southern region of Dengfeng City in Henan Province lies in the transitional zone between the Songshan Mountains and the Middle and Lower reaches of the Yellow River plain. Complex topographic and geological conditions in this region lead to frequent landslide hazards, posing serious threats to regional production safety and residents' lives. Conducting early identification of landslide hazards and high-precision susceptibility assessment is practically significant for the prevention and control of regional geological hazards.

This study applied optical remote sensing and small baseline subset interferometric synthetic aperture radar (SBAS-InSAR) for the early identification of landslide hazards. Twelve key evaluation factors, including elevation, slope, and lithology, were selected. Landslide susceptibility assessment was carried out based on the information value model and machine learning methods (artificial neural network, random forest, and stacking ensemble model). Additionally, slope units were used to optimize the output of evaluation results.

The results showed that: (1) A total of 33 landslide hazard sites were identified through multi-source remote sensing interpretation and field verification. They were mainly distributed in the central, southwestern, and southeastern parts of the study area, and their spatial distribution was significantly correlated with terrain slope, weak lithological layers, and human engineering activities. (2) Landslide susceptibility in the study area showed a distribution pattern of lower susceptibility in the north and higher susceptibility in the south. The stacking ensemble model achieved the highest prediction accuracy with an area under the curve (AUC) value of 0.96, which was significantly better than single models and the traditional information value model. The high susceptibility zone accounted for only 8.60% of the total area but captured 87.72% of the landslide samples.

The technical framework of multi-source remote sensing identification combined with ensemble learning evaluation constructed in this study provides high-precision data support and a technical paradigm for accurate prevention and control of landslide risks in the southern region of Dengfeng. It also verifies the significant advantages of ensemble learning for landslide susceptibility assessment in complex terrain areas.

Haibei Tibetan Autonomous Prefecture is located in the central part of the Qilian Mountains with widespread permafrost coverage. Accelerated global warming has aggravated permafrost degradation and triggered frequent thermal thawing geohazards, posing a serious threat to local residents' lives and property safety. The dominant controlling factors of regional thermal thawing geohazards remain unclear, and the conventional direct elimination of redundant conditioning factors tends to cause the loss of valid information. To address these two key scientific bottlenecks, this study carries out susceptibility assessment to provide technical support for regional sustainable infrastructure construction and geohazard early warning.

Haibei Tibetan Autonomous Prefecture was taken as the study area, and 16 preliminary conditioning factors covering topography, meteorology, geological setting, and human engineering interference were selected in this study. Pearson correlation analysis and Kruskal-Wallis test were sequentially adopted to quantify multicollinearity and factor importance. Principal component analysis (PCA)-based dimensionality reduction was applied to integrate highly correlated redundant variables for indicator optimization. Three machine learning algorithms-logistic regression (LR), support vector machine (SVM), and random forest (RF)-were used for susceptibility modeling and analysis. The frequency ratio method was also applied. Model performance was verified via ROC-AUC metric and five-fold cross-validation.

The results showed that: ① Kruskal-Wallis testing identified distance to roads, multi-year average temperature, thawing index, elevation, and annual snow cover days as the dominant controlling factors. ② Different from common rainfall-triggered landslides, precipitation did not play a dominant role in the formation of thaw-related geohazards. Therefore, relevant indicators characterizing the thermal state of permafrost and the intensity of engineering disturbances should be prioritized in assessment. ③ PCA-based dimensionality reduction was applied to process redundant factors. It not only effectively eliminated strong inter-factor correlation but also preserved the main information of the original data. The importance of the generated dimensionality-reduced fusion factors was significantly improved, thereby optimizing the evaluation indicators. ④ The prediction performance of all three machine learning models was improved after optimization of the evaluation indicators using dimensionality reduction techniques, and the optimized LR model achieved the best overall performance with an average AUC of 0.875 via five-fold cross-validation.

The indicator optimization framework combining Kruskal-Wallis significance test and PCA-based dimensionality reduction possesses high reliability, and the optimized LR coupling model is applicable to thaw hazard susceptibility mapping in the study area. The proposed research framework can serve as a technical reference for geohazard assessment in analogous permafrost regions across the Qinghai-Xizang Plateau.

Loess distributed across the Loess Plateau is characterized by prominent water sensitivity, collapsibility, and well-developed vertical joints, making regional landslides frequently triggered by rainfall infiltration, freeze-thaw cycles, and intensive human engineering activities. With the large-scale construction of ultra-high-voltage power transmission infrastructure in mountainous loess terrain, refined landslide susceptibility assessment has become an essential prerequisite for engineering safety management. However, remote hilly loess regions generally suffer from incomplete historical landslide inventories. Conventional sampling strategies obtain positive samples merely from archived landslide records and extract negative samples randomly across the entire study area. Such sampling patterns lead to insufficient positive samples and contaminated negative samples mixed with ambiguous non-landslide grids that share similar geological settings, which seriously degrades the prediction performance of machine learning-based susceptibility models. To solve this technical bottleneck, this study proposes a collaborative optimization strategy for positive and negative landslide samples by integrating time-series small baseline subset interferometric synthetic aperture radar (SBAS-InSAR) remote sensing interpretation and the information value method.

The study area was located at the southern foot of Lyuliang Mountain in Linfen, Shanxi Province, covering a total area of 189.62 km² with typical loess ridge-gully geomorphology. Nine assessment factors closely related to loess landslide initiation were selected for susceptibility modeling: Elevation, slope gradient, slope aspect, plan curvature, profile curvature, topographic wetness index (TWI), normalized difference vegetation index (NDVI), gully density, and distance to gullies. Based on Sentinel-1A ascending SAR images collected from March 2023 to June 2024, SBAS-InSAR deformation inversion was implemented, and grid cells with slope-parallel annual average deformation rate ≤−15 mm/a were preliminarily defined as potential unstable landslide zones. Combined with visual interpretation of typical geomorphic features such as cirque-shaped scarps from high-resolution optical remote sensing images, dual verification was conducted to screen reliable positive samples. Specifically, 230 raster positive samples were expanded from 10 historically recorded landslides, and another 920 supplementary raster samples were identified from 31 newly detected hidden landslides, forming a final positive dataset consisting of 1150 grid cells. Subsequently, the information value model was adopted to classify the entire study area into five susceptibility grades via the natural breaks algorithm, and qualified negative samples were randomly selected only from extremely low and low susceptibility zones with a fixed 1∶1 positive-to-negative sample ratio. Four comparative sampling schemes were constructed for quantitative comparison, and all datasets were randomly split into training and testing subsets at a ratio of 7∶3. Random forest (RF) and back propagation neural network (BP) were employed to establish landslide susceptibility models, and the area under the receiver operating characteristic curve (ROC-AUC) was adopted as the quantitative assessment indicator of model accuracy.

The modeling results revealed an obvious hierarchical improvement effect. Optimizing only positive samples greatly improved model accuracy, with AUC values reaching 0.87608 (RF) and 0.77174 (BP), while independent negative sample optimization brought limited accuracy improvement, with AUC values of 0.59124 (RF) and 0.58785 (BP). The collaborative optimization scheme combining remote-sensing-derived positive samples and information-value-filtered negative samples achieved optimal performance, with RF-AUC=0.91812 and BP-AUC=0.81937, representing accuracy improvements of 60.27% and 47.43%, respectively, compared with the traditional sampling scheme (RF-AUC=0.57285, BP-AUC=0.55577).

This study verifies that the proposed hybrid sample optimization framework can significantly improve the reliability of loess landslide susceptibility assessment. The core technical idea can be extended to other data-deficient regions such as red-bed hilly terrains and alpine canyon areas, providing solid technical support for geological disaster prevention and safe operation of major power transmission projects on the Loess Plateau.

To address the challenges in preventing and controlling mass landslides triggered by typhoon-induced heavy rainfall in granitic regions, this study focuses on Beiliu City, Guangxi Province as the study area, and systematically investigated the rainfall response mechanism and early warning technology of typhoon rainstorm-triggered landslides.

By analyzing the spatiotemporal distribution characteristics of rainfall and the mechanisms triggering landslides during typhoon-induced heavy rainfall events, the study quantified the rainfall kurtosis, skewness, peak location coefficient and classified typhoon-induced heavy rainfall into three types: Post-peak, pre-peak, and concentrated. A regional slope stability evaluation method under heavy rainfall conditions was developed using the TRIGRS-Scoops3D coupled model. The method was validated using the "6・26" rainfall event in 2023 as an example.

The results indicated that over 50% of the study area experienced stability degradation under rainfall conditions, with extremely unstable zones accounting for 5.73%. High-risk areas were concentrated in the northern, eastern, and southwestern steep slope terrain units. All landslide points induced by heavy rainfall were located within the warning zones delineated based on stability evaluation results.

The results indicate that the typhoon-induced shallow landslide early warning method based on "rainfall pattern recognition-quantitative stability assessment-dynamic delineation of risk areas" possesses high reliability and applicability. The research findings provide scientific support for shallow landslide risk prevention and control, monitoring and early warning system development, and emergency management of geological disasters in typhoon-prone granitic regions.

At 8:40 am on July 17, 2024, the Wanshuitian landslide in Jiajiadian Village, Guizhou Town, Zigui County, Yichang City, Hubei Province, underwent instability and failure, with a total displaced volume of approximately 800000 m³. This landslide damaged 1200 m of village-level roads and 60 mu (around 9.88 acres) of citrus farmland and forest. This study aims to systematically reveal the development characteristics, movement process, and instability mechanism of the landslide.

Based on the analysis of movement characteristics of the Wanshuitian landslide, detailed field geological surveys, UAV aerial photography, and monitoring data analysis were carried out. Combined with the Geo-Studio finite element simulation software, the internal seepage characteristics and the evolution process of slope stability under heavy rainfall conditions were calculated, thereby revealing its genetic mechanisms and failure modes.

The results showed that the Wanshuitian landslide was a high-speed rock landslide, which could be divided into five subzones based on movement characteristics: The initiation zone, secondary disintegration zone, main accumulation zone, right scattering zone, and left scattering zone. The interbedded lithology of sandstone and mudstone, the micro-geomorphology of alternating troughs and ridges, and the jointed rock mass structure were the internal factors for the occurrence of the Wanshuitian landslide. The main external factor was two rounds of continuous heavy rainfall with a cumulative rainfall of 253.8 mm over 17 days before sliding. Continuous heavy rainfall led to a sustained increase in pore water pressure within the slope, and the landslide safety coefficient decreased from 1.2 to 0.97, significantly reducing slope stability. After two rounds of heavy rainfall, the pore water pressure in the rock and soil mass of the landslide body and sliding zone increased significantly, reaching a maximum of 75.4 kPa and a maximum incremental increase of 303.9 kPa. The sudden increase in pore water pressure ultimately triggered slope instability and failure. As the sliding movement was blocked along the layer dip, the slope slid along the free surface. The main sliding direction of 10° formed an angle of 88° with the rock dip direction of 282°, representing a special failure mode of sliding nearly along the rock strike. The mode was significantly different from the instability mechanism of consequent bedding landslides and was characterized by high concealment and strong suddenness.

The research results have important theoretical and practical significance for disaster prevention and mitigation, monitoring and early warning, and engineering prevention at potential disaster sites with similar geological conditions in mountainous areas of China.

Soil-rock accumulations are widely distributed in hilly and mountainous regions across China. As loose geomaterials with widely graded particles and mixed soil-rock components, they are highly prone to internal erosion induced by rainfall infiltration and subsurface seepage. During internal erosion, fine soil particles detach and migrate through intergranular pores, which simultaneously deteriorates the bearing capacity of the soil skeleton and alters the internal permeability of the accumulations. This coupled deterioration is the dominant triggering mechanism for rainfall-induced landslides, posing major threats to geotechnical facilities and geological disaster prevention. Therefore, developing an accurate method to predict the evolution and magnitude of fine particle erosion is of great theoretical significance and practical value for ensuring the safe operation of geotechnical projects and mitigating landslide risks.

Traditional internal erosion prediction models fail to consider the influence of complex in-situ stress states, limiting their application in practical engineering. To address this research gap, this study first regarded saturated soil-rock accumulations as a five-phase mixed medium and established a set of coupled governing equations for seepage, erosion, and stress fields based on mass conservation, momentum balance, and the effective stress principle. The finite element numerical model was compiled and solved on the COMSOL Multiphysics platform, and Voronoi diagrams were adopted to reconstruct geometric models of accumulations with different rock contents. A series of triaxial erosion-shear tests under three deviatoric stress conditions with a constant confining pressure of 50 kPa were carried out to verify the reliability and calculation accuracy of the proposed numerical method. On the basis of massive numerical simulation results, this study took volumetric strain, rock content, average seepage velocity, and erosion time as four input parameters, and adopted the least squares method for regression fitting. A novel evolution characterization model for internal erosion was further established, realizing the full quantitative characterization of internal erosion under complex stress conditions. Additionally, the intrinsic mechanisms of how rock content and volumetric strain affect internal erosion behaviors were systematically explored.

The combined results of physical tests and numerical simulations demonstrated that the proposed characterization model could accurately predict the evolution of eroded fine particles in soil-rock accumulations with different stress levels and rock contents, provided that no erosion-induced instability failure occurs in the soil skeleton. The increase of rock content could extend the seepage path of pore water and reduce the average seepage velocity inside the medium, thereby effectively restraining the development of internal erosion. In contrast, when the accumulation was subjected to deviatoric stress, shear dilatancy occurred, leading to a continuous rise in volumetric strain. The enlarged volumetric strain further increased porosity and permeability, which was the essential internal factor aggravating the degree of internal erosion. This newly developed internal erosion evolution model can quantitatively describe the dynamic evolution of permeability in soil-rock accumulations under the combined action of seepage, internal erosion, and complex stress fields. The major innovations of this study lie in two aspects. First, it introduces volumetric strain to quantitatively characterize the stress effect on internal erosion, compensating for the inherent limitations of traditional models that neglect stress influence. Second, it integrates multiple key parameters to establish a multi-factor coupling model, which greatly expands the applicability of classical erosion equations.

This study not only enriches the theoretical system of coupled seepage-erosion-stress effects for wide-graded soils, but also provides reliable theoretical support and technical references for stability evaluation of accumulation slopes and foundations, as well as the prevention and control of rainfall-triggered landslides in mountainous areas.

This study aims to examine the influence of groundwater level at different temporal scales (specifically hourly, daily, monthly, and annual average water levels) on the accuracy of annual shallow groundwater storage variation calculations.

The shallow groundwater system of the Handan Plain at 2019 was selected as the study object. The grid method and the Thiessen polygon method were applied to calculate groundwater storage variations using water level data at different temporal scales, and the results were compared to evaluate differences in calculation accuracy.

The results indicated that, for the same temporal scale, groundwater storage variation estimates obtained using the grid method and the Thiessen polygon method were generally consistent, with a maximum difference of 0.0114 billion m³. At different temporal scales, both methods showed that the results calculated using monthly average water levels deviated the most from those calculated using hourly water levels, which were considered more accurate in theory. The deviations were 0.0727 billion m³ for the grid method and 0.0611 billion m³ for the Thiessen polygon method, with no consistent directional bias. In contrast, estimates using annual average water levels exhibited relatively small discrepancies compared to those calculated using hourly water levels, with a difference of 0.0015 billion m³. However, the degree of agreement also exhibited randomness. For the grid method, the estimated results based on annual average water levels did not change significantly with grid size, with a maximum difference of 0.0011 billion m³. At non-annual temporal scales, calculation accuracy improved as the grid resolution became finer. The grid method yielded more accurate results than the Thiessen polygon method when the grid resolution was finer than 1 km. However, the Thiessen polygon method demonstrated superior accuracy when the grid cell size of regular partition approached the average area of Thiessen polygons.

These findings provide theoretical and methodological support for the rational selection of water level data at different temporal scales, thereby improving the accuracy of groundwater storage variation calculations, which is essential for evaluating the effectiveness of groundwater overexploitation control strategies.

Groundwater flow exchanges heat with the surrounding rock and alters geothermal field distribution, playing a key controlling role in the occurrence and exploration of geothermal resources. This study aims to reveal the influence mechanism of groundwater flow on geothermal field in the Yuncheng Basin, Shanxi Graben.

Taking the Yuncheng fault basin as the study area, a two-dimensional geological profile model was constructed. Finite element numerical simulations were conducted under three scenarios: Pure heat conduction; gravity-driven heat conduction-convection; Gravity and buoyancy-driven heat conduction-convection. The controlling effects of groundwater flow on deep geothermal field were comparatively analyzed.

Under the pure heat conduction model, the geothermal field exhibited a North-South symmetrical distribution with alternating high and low temperatures. High temperature zones were concentrated in the Fenhe and Sushui depressions, controlled by basement undulation and caprock thickness. Under gravity-driven conditions, groundwater flowed along high-permeability strata and fault zones, causing cooling in recharge areas and heating in discharge areas. When buoyancy effects caused by temperature differences was superimposed, the flow velocity and direction were changed within deep major fault zones, resulting in local positive temperature anomalies at the northern and southern marginal faults of the Emei Platform and in the deep part of the Zhongtiaoshan Fault. Borehole temperature comparisons indicated that the heat transfer in the Yuncheng Basin was dominated by a combined heat conduction-convection mode, with the permeability of deep major faults being approximately 1.0×10⁻¹² m².

Groundwater flow significantly controls the geothermal field distribution and heat redistribution in the Yuncheng Basin. The coupled heat transfer mode is the dominant mechanism of the geothermal system in this area. The results provide a scientific basis for the exploration and prediction of geothermal resources in the Yuncheng Basin and similar areas within the Shanxi Graben.

The eastern Henan Plain is a typical agricultural irrigation area in the lower reaches of the Yellow River, where high-arsenic groundwater is widely distributed, posing a severe threat to drinking water safety. Revealing the biogeochemical mechanisms of arsenic migration and transformation in groundwater in alluvial plain aquifers can provide a scientific basis for prevention and control of endemic arsenic contamination.

In this study, 200 groundwater samples were collected from three geomorphic units, including Yellow River alluvial plain, crevasse splays, and interriver depressions, to identify the distribution of high-arsenic groundwater. Hydrogeochemical analysis, three-dimensional excitation-emission matrix (3D-EEM) fluorescence spectroscopy, and parallel factor analysis (PARAFAC) were applied to clarify its spatial differentiation pattern and the arsenic activation mechanism mediated by dissolved organic matter (DOM).

High-arsenic groundwater (ρ(As)>10 μg/L) was mainly distributed in shallow aquifers at depths of 20-50 m. Its spatial distribution was controlled by sedimentary systems of modern Yellow River channel and paleochannels, forming enrichment zones at the fronts of crevasse splays and interriver depressions. DOM in high-arsenic groundwater was characterized by high aromaticity and strong humification, dominated by low-molecular-weight humic-like (C1, 54%) and fulvic-like (C3, 29%) components, revealing a synergistic input mechanism of terrestrial and microbial sources. Correlation analysis indicated that arsenic concentration in groundwater was significantly positively correlated with Fe(Ⅱ), NH₄⁺-N, and DOM components C1 and C3 F_max (maximum fluorescence intensity).

In weakly reducing to reducing sedimentary environments, arsenic activation is jointly controlled by two pathways: Microbially mediated reductive dissolution of Fe (hydr)oxides driven by organic matter, and desorption of humic-Fe-As complexes. Anaerobic degradation of tryptophan-like component (C2) enhances microbial metabolic activity and accelerates secondary release of arsenic from sediments. The results provide theoretical support for risk management and safe utilization of high-arsenic groundwater in the alluvial fan in the lower reaches of the Yellow River.

With the global promotion of carbon neutrality and the rapid development of renewable energy, hydrogen has become one of the most promising clean energy carriers due to its high energy density and pollution-free characteristics. Underground hydrogen storage (UHS) is regarded as an effective solution to the large-scale and long-term storage of hydrogen, which has been widely studied in energy and geological engineering fields in recent years. The diffusion coefficient of H₂ in water under high-temperature and high-pressure (HTHP) conditions is a key parameter for quantifying hydrogen migration behavior in reservoir pores, simulating diffusion fluxes, and evaluating the leakage risk of hydrogen through caprocks. However, previous studies are mostly limited to ambient temperature and pressure, and experimental data under real UHS-suitable HTHP conditions are still insufficient, with obvious discrepancies among different reported results.

To fill this data gap, this study conducted in-situ quantitative observations of the dissolution and diffusion processes of H₂ in aqueous solutions using micro-laser Raman spectroscopy in transparent high-pressure quartz capillaries. A series of diffusion experiments was carried out at pressures of 10-30 MPa and temperatures of 298.15-393.15 K, and the diffusion coefficients of H₂ in pure water were accurately obtained.

The results showed that temperature imposed a dominant effect on the diffusion coefficient of H₂ in water. As temperature rose, the diffusion coefficient increased significantly. At 20 MPa, when the temperature increased from 298.15 K to 363.15 K, the diffusion coefficient increased by approximately 211%. The relationship between the diffusion coefficient and temperature could be well fitted by the Speedy-Angell power-law equation: D=23.572×10⁻⁹[(T/213.54)−1]^2.021, with an average absolute deviation (AAD) of only 1.8%. In contrast, pressure had a weak effect on the diffusion coefficient. With increasing pressure, the diffusion coefficient showed a slight decreasing trend. At 363.15 K, when pressure increased from 10 MPa to 30 MPa, the diffusion coefficient decreased by only about 4.8%, which was consistent with the low compressibility of liquid water. Furthermore, combined with the Bruggeman empirical formula and actual geological parameters of the Underground Sun Storage project in Austria, the effective diffusion coefficient and leakage flux of H₂ through caprocks were calculated. It revealed that the total cumulative diffusion mass of H₂ decreased obviously with increasing caprock thickness. Thicker caprocks significantly slowed down the diffusion rate and prolonged the time required for H₂ to migrate outward, thus greatly reducing the loss risk of stored hydrogen.

Therefore, deep geological structures with relatively low temperature and thick caprocks are strongly recommended for practical UHS engineering. This study provides systematic and reliable HTHP diffusion coefficient data of H₂ in pure water, which supplements the basic parameter database for underground hydrogen storage. The findings support the quantitative characterization of hydrogen migration, the calculation of diffusion fluxes, and the assessment of confinement security, and they offer an important scientific basis for site selection, scheme design, and risk management in underground hydrogen storage projects worldwide.

Soil moisture content acts as a fundamental physical parameter to characterize multi-phase media consisting of soil solids, gas, liquid water, and ice, and it dominates freeze-thaw phase transition processes in frozen ground. Accurate quantification of unfrozen water and ice contents is essential for hydrological cycle simulation, farmland irrigation regulation, ecological environment assessment, and stability evaluation of geotechnical infrastructures such as frozen soil subgrades, slopes, and landslides. Existing review papers on soil moisture monitoring mostly focus on single measurement technology or single spatial scale, while few studies systematically compare the applicability of various techniques for differentiating ice and liquid water under freeze-thaw conditions. This research gap restricts the precise assessment of frost heave and thaw settlement risks in cold-region engineering, which necessitates a comprehensive systematic review to clarify the full technical system and key bottlenecks of ice-water differentiation.

Based on bibliometric analysis of literature published from 2005 to 2026, this paper retrieves 766 valid Chinese core papers from CNKI and 6 025 international articles from the Web of Science Core Collection. Statistical results of annual publication number, core research institutions, and keyword bursts reveal that the research hotspot has shifted from simple single-point soil moisture monitoring to multi-source data fusion and artificial intelligence inversion over recent decades. All prevailing soil moisture measurement technologies are systematically classified into contact measurement and non-contact measurement categories. The contact category includes reference oven-drying method, in-situ dielectric sensors (TDR, FDR), thermal response probes, nuclear magnetic resonance (NMR), and actively heated-fiber Bragg grating (AH-FBG) sensing. The non-contact category covers shallow geophysical methods (GPR, ERT, electromagnetic induction, shallow seismic) and multi-type remote sensing inversion. Each technique is comprehensively evaluated from four dimensions: Working principle, applicable spatial scale, capacity of unfrozen water-ice differentiation, and inherent error sources. Furthermore, this review elaborates on the distortion mechanism of monitoring signals triggered by phase transition: Frozen soil exhibits unique thermal, dielectric, NMR, and elastic wave discrepancies between liquid water and ice, and the coexistence of bound water, capillary water, and ice crystals further aggravates the non-uniqueness of sensor response. Three major categories of interference factors affecting measurement precision are summarized, including soil physicochemical properties (texture, salinity, organic matter, bulk density), external environmental conditions (temperature fluctuation, vegetation coverage, freeze-thaw cycles) and inherent limitations of monitoring equipment, with targeted calibration and error correction strategies proposed correspondingly. In addition, this paper compares three mainstream inversion frameworks: Pure empirical physical models, data-driven machine learning algorithms, and physics-data hybrid inversion constrained by hydrothermal coupling theories.

The analytical results demonstrate distinct complementary characteristics among different monitoring technologies. The oven-drying method can only serve as a calibration benchmark and fails to realize long-term continuous field monitoring. In-situ sensors enable real-time point monitoring but are highly susceptible to soil-sensor contact state and soil physicochemical properties. Shallow geophysics and satellite remote sensing expand monitoring coverage but suffer from ambiguous physical response and scale mismatch problems. A single monitoring method cannot reliably distinguish between unfrozen water and ice contents in frozen soils, and cross-validation combining laboratory tests, field sensing, geophysical prospecting, and remote sensing data is indispensable to improve phase identification accuracy. Among all inversion frameworks, the physics-data hybrid model balances physical interpretability and prediction accuracy and outperforms single-model methods. In future research, integrated space-air-ground collaborative observation networks, multi-sensor fusion algorithms, and physics-informed intelligent inversion models will become core technical approaches for achieving cross-scale, simultaneous high-precision measurement of soil moisture and ice content, providing theoretical support for hazard prevention in cold-region geotechnical engineering.

Soil salinization is a prominent environmental issue in arid regions, and rapid and accurate monitoring of soil salinity is critical for regional ecological conservation and sustainable agricultural development.

To improve the accuracy of remote sensing inversion of soil salinity in arid regions, the Kongque River irrigation area of the Xinjiang Uygur Autonomous Region was selected as the study area. Soil samples were collected from field survey points. Based on measured soil hyperspectral data acquired with the ASD FieldSpec 4 spectroradiometer, Landsat 8 satellite remote sensing data were calibrated, and the calibrated spectral indices were then used to construct a soil salinity remote sensing inversion model with the random forest algorithm to estimate surface soil salinity in the Kongque River irrigation area.

The results showed that soil spectral reflectance increased with increasing salinization degree. After ASD hyperspectral calibration, the correlations between some salinity indices and soil salinity were significantly improved. The random forest was used to construct a remote sensing inversion model for soil salinity. The R² was 0.847 for the training set and 0.713 for the validation set, both significantly higher than the R² values of the training and validation sets obtained from the original data. Soil salinization was most severe in the western part of the Kongque River irrigation area and gradually decreased from Southwest to Northeast. Slightly and moderately saline soils were mainly distributed in the southern part, while non-saline soils were mainly distributed in the central and northern parts.

The hyperspectrally calibrated random forest inversion model shows good accuracy and can provide reliable technical support for dynamic monitoring of soil salinization in the Kongque River irrigation area and similar arid regions.

Images acquired by a single remote sensing sensor are inherently constrained by hardware and physical limitations, making it difficult to simultaneously achieve high spatial resolution and high spectral resolution. Hyperspectral images provide rich spectral information but typically suffer from low spatial resolution, whereas multispectral images contain finer spatial details at the cost of reduced spectral fidelity. To address this trade-off, this study proposes a local-global collaborative multi-scale feature augmentation method for hyperspectral and multispectral image fusion. The objective is to fully exploit the complementary spatial and spectral characteristics of heterogeneous data sources, thereby generating fused images that preserve spectral consistency while significantly enhancing spatial detail expression.

The proposed fusion framework consisted of four cooperative modules: Feature extraction, feature fusion, feature augmentation, and image reconstruction. First, the feature extraction module independently encoded the hyperspectral and multispectral inputs using dedicated convolutional layers to obtain hierarchical spectral and spatial feature representations. Second, the feature fusion module integrated the extracted features into a shared latent space, enabling cross-modal interaction and alignment. The core component was the feature augmentation module, which was divided into local and global sub-modules. The local feature augmentation sub-module employed multiple convolutional blocks with different receptive fields to strengthen fine-grained spatial details such as edges, textures, and local structures. The global feature augmentation sub-module introduced a spectral-spatial fusion Transformer architecture combined with multi-scale convolutions to model long-range dependencies and enhance global contextual information as well as spectral consistency. Finally, the image reconstruction module mapped the augmented fusion features back to the image domain to produce the final high-resolution hyperspectral image.

Extensive experiments were conducted on several benchmark hyperspectral and multispectral datasets, including both quantitative evaluations and qualitative visual comparisons. The proposed method consistently outperformed state-of-the-art fusion methods across multiple evaluation indicators. In terms of spatial detail preservation, the fused images exhibited sharper edges and clearer textures with significantly improved spatial resolution. Regarding spectral fidelity, the proposed method achieved low spectral distortion, with SAM (spectral angle mapper) values comparable to those of the best-performing competitors. Comprehensive evaluation indicators such as PSNR, SSIM, and ERGAS also demonstrated superior performance. For example, on the widely used CAVE and Harvard datasets, the proposed method achieved average PSNR improvements of 1.5-2.5 dB over the best baseline methods. Visual comparisons further confirmed that the proposed method effectively avoided common artifacts such as blurring and spectral aliasing. Moreover, the method showed robust performance across different scenes and varying degradation conditions.

The proposed local-global collaborative multi-scale feature augmentation method effectively mitigates the inherent spatial-spectral trade-off in single-sensor imaging systems. By jointly enhancing local fine-grained details and global contextual dependencies, the method generates fused images with both high spatial resolution and high spectral fidelity. Experimental results demonstrate its superiority over existing approaches in terms of accuracy, robustness, and visual quality. The proposed framework provides a powerful and versatile solution for hyperspectral and multispectral image fusion, with strong potential for practical applications in remote sensing, environmental monitoring, and beyond.

Shear wave velocity is a critical parameter that characterizes the physical and mechanical properties of subsurface media, and it plays an indispensable role in the exploration and development of oil and gas resources. In carbonate fracture-cavity reservoirs, complex lithologic assemblages and strong reservoir heterogeneity bring great challenges to shear wave velocity acquisition. Traditional rock physics models and empirical formulas are difficult to adapt to such complex geological conditions, resulting in low prediction accuracy and poor applicability.

Taking the ultra-deep fracture-cavity reservoirs in Tahe Oilfield, Xinjiang Uygur Autonomou Region as the research target, this study proposed a shear wave velocity prediction method using long short-term memory (LSTM) neural networks based on dimensionality reduction and reservoir classification. Firstly, the distorted logging curves were corrected by using valid undistorted logging data to guarantee the reliability of input datasets. Secondly, principal component analysis (PCA) was adopted to reduce the dimensionality of 11 logging parameters including acoustic slowness, density logging, and neutron logging, and five principal components were extracted to eliminate data redundancy. On the basis of imaging logging and electrical logging characteristics, support vector machine (SVM) was applied to divide reservoirs into six categories: Dissolved pores, fractures, intact bedrock, unfilled caves, sand-mud filled caves, and breccia-filled caves. Then, targeted LSTM deep learning models were established to realize classified shear wave velocity prediction for different reservoir types.

The application results showed that the correlation between predicted results and measured values of the proposed method reached 91%, representing a significant improvement over conventional empirical formulas and rock physics methods. Independent verification using blind wells further proved that the maximum correlation coefficient between predicted and measured shear wave velocity was up to 0.9694. The predicted curves were highly consistent with measured data.

The proposed PCA-SVM-LSTM combined method can well describe the strong heterogeneity of ultra-deep fracture-cavity reservoirs in Tahe Oilfield, and the prediction results show good agreement with measured data. This method avoids the complicated rock physics modeling process and has the advantages of simple workflow and high computational efficiency. It provides an efficient and feasible technical reference for shear wave velocity prediction of similar carbonate fracture-cavity reservoirs.

As a key iron ore base in China, southeastern Hubei Province hosts numerous large and medium-sized skarn-type iron deposits. The Chengchao iron deposit, one of the most representative large skarn-type iron mines in this region, has been explored intensively by drilling projects. The deepest existing engineering control has reached an elevation of −1300 m, and the ore body has not been fully delineated, indicating great exploration potential in its deep and peripheral areas. High-grade iron ore is a strategic mineral resource in China, and gravity and magnetic prospecting have become efficient geophysical techniques for magnetite exploration. Nevertheless, traditional potential field separation methods are highly dependent on manual parameter selection and have limited capability in extracting weak deep-seated anomaly signals. Therefore, it is urgent to develop an effective technical means to accurately identify deep mineralization information from gravity-magnetic anomaly data for deep mineral exploration in the study area.

This study took the Chengchao iron deposit area and its surroundings as the research object. Based on the GMS gravity-magnetic exploration software, the DB4 wavelet basis was adopted to conduct wavelet multi-scale decomposition on collected gravity and magnetic data. The power spectrum analysis method was further applied to quantitatively calculate the apparent source depths corresponding to each order of wavelet detail anomalies. From planar and profile perspectives, this study systematically analyzed the intensity, scale, positive-negative anomaly combination, and gradient characteristics of magnetic and gravity detail anomalies at different depths. Combined with physical property test data, drilling records, and magnetotelluric sounding results, this study established the spatial correlation between geophysical anomalies and known geological bodies as well as iron ore bodies.

The research results showed that the intensity of the first- to fourth-order wavelet detail anomalies increased continuously with the growth of apparent source depth, and the fourth-order anomalies reached the maximum amplitude with a corresponding apparent source depth of approximately 1587.8 m. Although the intensity of the fifth-order detail anomalies decreased slightly, their distribution range expanded significantly. At present, the magnetite ore bodies controlled by existing engineering works are mainly distributed in the shallow and middle zones corresponding to the first- to third-order detail anomalies. The strong fourth-order anomalies, however, have not been verified by deep drilling, indicating significant geophysical responses of high-density and high-magnetic geological bodies in the deep part of the mining area.

This study verifies that wavelet multi-scale decomposition can effectively separate local gravity-magnetic anomalies at different depths and realize multi-dimensional refined interpretation. A deep verification borehole and logging data have uncovered industrial magnetite ore at depths below 1600 m, setting a new record of the deepest ore occurrence in the mining area. The proposed integrated technical workflow proves reliable for detecting concealed ore bodies and interpreting deep geological structures in old mines. It can also provide a solid technical reference for deep exploration of similar skarn-type iron deposits worldwide.

Acoustic logging curves, particularly compressional wave slowness (DTC) and shear wave slowness (DTS), serve as fundamental data for petrophysical analysis, synthetic seismogram generation, and refined reservoir characterization. However, during actual drilling operations, these curves are prone to distortion or gaps due to factors such as borehole conditions and complex environmental measurement noise, which constrains their practical application. Traditional empirical formulas and statistical regression methods struggle to capture the complex nonlinear relationships between logging curves. Although machine learning and deep learning methods introduced in recent years have improved reconstruction accuracy to some extent, they still exhibit limitations in comprehensively representing the non-stationary features, local variations, and long-range geological dependencies of logging signals under complex borehole conditions.

To address these issues, this study proposed an acoustic logging curve reconstruction method based on a fusion architecture combining variational mode decomposition (VMD) and temporal convolutional network (TCN)-Transformer. The method first employed VMD to perform multi-scale decomposition of the original logging signals, preserving the effective formation signals to the greatest extent while effectively filtering out high-frequency environmental noise. Subsequently, TCN was introduced to characterize the local variation features of the logging curves, while the Transformer's multi-head self-attention mechanism was employed to extract long-range dependencies within the logging sequences, enabling holistic modeling of complex sedimentary cyclicity. Based on measured logging data from a block in Shanxi, comparative model analysis, ablation experiments, curve reconstruction experiments under conditions of severe borehole enlargement, and blind-well prediction validation were conducted.

The results demonstrated that the proposed method performed well in terms of accuracy and stability for acoustic logging curve reconstruction. The coefficients of determination (R²) for DTC and DTS predictions in the test intervals reached 0.9142 and 0.9165, respectively. Both the VMD signal decomposition and the TCN-Transformer hybrid architecture contributed significantly to the model’s performance. In intervals with significant borehole enlargement, the model effectively suppressed environmental noise interference, producing reconstructed curves with continuous and geologically reasonable morphology. The synthetic seismograms generated from the blind-well prediction results showed good consistency with the measured seismic profile in terms of wavelet characteristics and phase features.

The proposed method exhibits strong adaptability and practicality under complex borehole conditions. It can provide reliable foundational data for the correction and completion of low-quality logging data, as well as for subsequent seismic inversion and refined reservoir characterization.

The Quaternary strata are widely developed in urban areas. They exhibit complex sedimentary characteristics, including frequent sedimentary cycles, multiple interbedded lens bodies, and laterally disordered distribution. Traditional 3D geological modeling methods rely heavily on fixed stratigraphic sequences and struggle to handle discontinuous lens bodies, disordered layer connections, and locally inverted strata, leading to distorted interfaces, illogical connectivity, and low modeling accuracy. These limitations severely restrict the digital management of urban underground space and intelligent early warning of geological hazards.

To tackle these key technical bottlenecks, this study proposed an improved 3D geological modeling method for Quaternary strata based on stratigraphic penetration-driven layer correlation. Driven by borehole data, this method first automatically identified three types of lens bodies, including simple, nested, and top/bottom lens bodies, and conducted spatial clustering under the constraints of relative elevation difference and lens body thickness to eliminate local discontinuity interference. Guided by expert geological knowledge, a "major layer-sub-layer-sub-sub-layer" hierarchical system was constructed. With stratigraphic penetration as the core index, strata with high penetration were prioritized for standardized coding to realize the unification of stratigraphic sequences including those with inverted structures. On this basis, the stratigraphic pinch-out boundary was calculated using the angular unconformity pinch-out coefficient, and a stratigraphic partition model was constructed. Finally, a smooth and topologically consistent 3D geological grid model was established via thin-plate spline interpolation, and clustered lens bodies were embedded into the framework model to restore real sedimentary structures.

A case study was conducted using 102 engineering boreholes in the Zhongguancun area of Beijing to verify the method. The results showed that profiles extracted from the established 3D model were highly consistent with manual geological profiles. All lens body structures were automatically and accurately identified. The stratigraphic connection error rate decreased by 67%, and the geological interface agreement rate increased to 92%. The method effectively avoided unreasonable layer connections and redundant zero-thickness layers caused by loose Quaternary sediments.

This approach can intelligently identify lens bodies and accurately unify stratigraphic sequences, significantly improving the accuracy and rationality of 3D modeling for complex Quaternary strata. It provides reliable and high-precision geological model support for urban underground space development, intelligent early warning of geological disasters, and engineering survey and design, and has important theoretical value and broad application prospects for international urban geological digitalization.

Granite residual soil is widely distributed in humid and hot regions of southern China, and its highly variable engineering properties bring great challenges to slope stability evaluation and foundation design. The shear strength indices, including cohesion and internal friction angle, are the most critical mechanical parameters for analyzing the stability of geotechnical structures. Traditional laboratory tests for obtaining shear strength parameters are time-consuming, costly and labor-intensive, and cannot meet the demand for rapid parameter acquisition in disaster early warning. In addition, conventional single machine learning models often suffer from limited generalization performance when dealing with the strong nonlinear relationship between soil physical properties and shear strength. To solve the above practical problems, this study develops an innovative prediction framework based on the Stacking ensemble learning algorithm to realize the high-accuracy prediction of cohesion and internal friction angle, and further reveal the dominant influencing mechanism of physical indices on soil shear strength.

In this study, a comprehensive dataset was compiled from published literature and field geotechnical investigation data, and data screening and normalization preprocessing were conducted to unify data quality and eliminate the interference of dimensional differences. A two-layer Stacking ensemble learning architecture was established. Three typical heterogeneous machine learning models—random forest (RF), support vector machine (SVM), and back propagation neural network (BPNN)—were adopted as base learners, and a 5-fold cross-validation strategy was applied to complete model training and avoid overfitting. Ridge regression was employed as the meta-learner to synthesize the prediction outputs from three base learners. Six common geotechnical indices, namely fines content, void ratio, natural water content, liquid limit, plastic limit, and specific gravity, were set as model inputs, while cohesion and internal friction angle were defined as model outputs. Furthermore, the SHapley Additive exPlanations (SHAP) method was introduced to interpret the black-box model and quantitatively analyze the contribution degree of each input parameter.

The results demonstrated that the determination coefficient (R²) of the proposed Stacking model reached 0.88 for cohesion and 0.90 for internal friction angle on the validation set, with corresponding root mean square error (RMSE) values of 3.60 kPa and 2.46°, respectively. Compared with the best-performing single base learner, the R² values increased by 1% and 5%, respectively. Verified by independent engineering test samples collected from Zixing City, Hunan Province, the absolute prediction deviation of cohesion ranged from 0 kPa to 1.91 kPa, and that of internal friction angle varied from 0° to 0.67°. The ensemble model exhibited obviously better prediction capability and robustness than individual models. SHAP interpretation results indicated that cohesion was mainly controlled by water content, liquid limit, and fines content, whereas fines content, void ratio, and water content served as the primary factors affecting internal friction angle. The variation characteristics of all parameters were well consistent with classical soil mechanics theories.

The study proves that the Stacking ensemble learning strategy can effectively combine the respective strengths of different single machine learning models and overcome their inherent defects. The proposed method greatly improves the prediction accuracy and generalization ability for shear strength parameters of granite residual soil. It provides an efficient, low-cost, and reliable technical solution for rapid parameter determination, and has good application prospects in slope stability assessment and landslide disaster prevention in areas covered by granite residual soil.