The landslide and non-landslide samples are important inputs for machine learning-based landslide susceptibility assessment. Compared to landslide samples, non-landslide samples generally present higher uncertainty due to random sampling. However, most sampling strategies (e.g., the feature space-based) for non-landslides only consider the characteristics of a single factor or the overall characteristics of all factors, which subsequently leads to either excessive artificial concentration of non-landslide samples or sampling information redundancy. To address these issues, a SHapley Additive exPlanations (SHAP) based sampling strategy considering combined characteristics of landslide conditioning factors (LCFs) is proposed. This strategy sorts the importance of LCFs based on SHAP algorithm and generates multiple sampling spaces using different numbers of LCFs in the sense of importance order. The optimal sampling space is selected according to the Bayesian optimization algorithm. Then, random forest (RF) and extreme gradient boosting (XGBoost) models are utilized to assess the susceptibility of Chaling County, Yanling County, and Guidong County, China, based on the proposed strategy and traditional random sampling. The results indicate that, compared with the traditional RF and XGBoost models, the improved models show better performance with an 8.2% and 9.0% increase in the AUC, respectively. Furthermore, the SHAP-based sampling framework demonstrates good adaptability across the study areas with different geological and geomorphic conditions, suggesting its potential transferability to other regions, although local optimization of parameter settings may still be required.

In the context of large-scale infrastructure projects, such as major bridges and docks on shorelines, understanding the behavior of deep piles in saturated sandy soils is crucial. This study employs three-dimensional numerical modeling of vibratory pile driving using Midas GTS NX finite element software and the UBCSAND constitutive model, challenging several common simplifying assumptions found in prior research. The efficacy of the numerical model in predicting pile driving processes and potential liquefaction was rigorously evaluated and validated against experimental data from previous studies. Sensitivity analyses were performed to investigate how pore pressure and liquefaction potential are influenced by various factors, including vibratory pulse counts, pile length-to-diameter ratios, and soil properties. The results from these analyses were utilized to train artificial neural networks and symbolic regression models. The performance of these models was assessed using a range of performance metrics and ROC curves. To enhance interpretability, symbolic regression provided a clear mathematical expression capturing the relationship between key features and soil liquefaction. Furthermore, SHapley Additive exPlanations were employed to offer detailed insights into feature importance and the model’s decision-making process. Design charts were developed based on these models to offer practical guidance for practitioners. Overall, this study underscores the effectiveness of integrating advanced numerical simulations with machine learning techniques, demonstrating significant improvements in understanding and predicting pile driving behavior and liquefaction potential in saturated sandy soils.

Landslide susceptibility mapping (LSM) is an essential tool for the prevention and management of landslide-related disasters. Conventional machine learning-based LSM method faces significant limitations in cross-regional extrapolation. To address this challenge, this study develops a transfer learning (TL) model based on the Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) framework, specifically designed for cross-regional LSM. A total of 11 modelling scenarios is established to compare the cross-regional extrapolation performance of Random Forest (RF), CNN-BiLSTM, and TL models, with Wanzhou District and Wushan County in Chongqing used as case studies. The results indicate that, compared to the strategy of directly expanding training dataset used by RF and CNN-BiLSTM models, the pre-training and fine-tuning strategy employed by the TL model is more suitable for county-scale LSM and its cross-regional extrapolation. Additionally, the cross-regional extrapolation performance of the TL model improves as the volume of source domain data increases. Finally, the SHAP algorithm is used to provide a global interpretation of the TL #3 model, which demonstrates the best performance in cross-regional model extrapolation.

Landslides have different topographic and morphological characteristics due to their different triggering mechanisms. However, the differences in the characteristics of earthquake- and rainstorm-induced landslides remain unclear. In this paper, we collect 12 cases of earthquake- and rainstorm-induced landslides around the world and reveal the differences in characteristics of the two types of landslides. By examining the geometric characteristics and location distribution of the landslides, the results show that earthquake-induced landslides tend to have larger areas, perimeter, lengths, widths, area to perimeter ratios (area/perimeter), major axis (S M), and minor axis (s m) than rainstorm-induced landslides. In addition, earthquake-induced landslides have more complex, rounded, and compact shapes than rainstorm-induced landslides. Earthquake-induced landslides are predominantly clustered near ridges, whereas rainstorm-induced landslides are predominantly clustered near valleys. The results also indicate that earthquake- and rainstorm-induced landslides mostly occur on 30 ° -50 ° and 10 ° -30 ° slopes, respectively, and both types are more likely to occur on sunny slopes. Moreover, the compactness and major axis are negatively logarithmically correlated for earthquake-induced landslides, while they are negatively exponentially correlated for rainstorm-induced landslides. Additional earthquake- and rainstorm-induced landslide events have verified the reliability and extensibility of the research conclusions. This work is beneficial for the management of landslide hazards and the effective implementation of landslide prediction and risk assessment.

The Gangdese metallogenic belt in Xizang, a world-class copper polymetallic province, has a poorly understood western segment due to extensive volcanic cover and limited historical exploration. The recent discovery of the Sangmoladong (SMLD) deposit, the first undocumented volcanic-subvolcanic-hosted, tin-dominant polymetallic system in western Gangdese, provides a unique opportunity to investigate collisional metallogeny. Through integrated LA-ICP-MS U-Pb geochronology of zircon and cassiterite, comprehensive whole-rock geochemistry, and Nd-Hf isotopes, this research establishes a genetic link between Paleocene magmatism and Sn-polymetallic mineralization. The mineralization is hosted within granite porphyry stocks and associated rhyolitic tuff breccias of a volcanic dome complex. It comprises two main stages: an early, disseminated cassiterite-sulphide stage with chloritic alteration, followed by later fluorite-cassiterite-tourmaline veins and veinlets. New LA-ICP-MS U-Pb dating constrains the timing of granite porphyry emplacement to 61.3 ± 0.2 Ma and rhyolitic tuff deposition to 61.8 ± 0.5 Ma. Cassiterite mineralization occurred between 61.0 ± 2.2 Ma and 59.3 ± 3.5 Ma, confirming that Sn metallogenesis was coeval with this Paleocene magmatic pulse during the Indo-Asian collision. The ore-forming granitic porphyries are highly evolved, A-type granites, characterized by high SiO₂ (77.50 - 80.40 wt.%), elevated zircon saturation temperatures (831 - 870 °C), and high Ga/Al ratios (10,000 × Ga/Al = 5.75 - 6.51). Their Nd-Hf isotopic signatures (ε_Nd(t) = −6.5 to −6.3; zircon ε_Hf(t) = −5.2 to +2.1) indicate an origin from anatexis of ancient Lhasa terrane metapelites, likely triggered by lithospheric extension during Neo-Tethyan slab rollback. This generated a reduced, fluorine- and boron-rich magmatic-hydrothermal system highly efficient at mobilizing and concentrating tin. On a regional scale, a metallogenic framework is proposed where Fe-Cu mineralization is sourced from hybridized mantle-crust magmas, whereas Pb-Zn and Sn systems derive from similar crustal-dominated sources. This metallogenic divergence of Pb-Zn and Sn reflects contrasting thermal regimes and magma crystallization pathways during a transtensional setting with local extension in pull-apart and uplift structures. The formation of the SMLD tin deposit is attributed to prolonged fractional crystallization of a high-temperature, low fO₂ magma within a subvolcanic dome complex. These conditions suppressed early cassiterite saturation and promoted extreme tin enrichment in the residual melt. This study makes two key contributions: (1) it identifies the first volcanic-hosted tin system in the western Gangdese, challenging traditional exploration models focused on porphyry skarn Cu-Pb-Zn deposits; and (2) it establishes a new tectono-metallogenic model that elucidates the spatiotemporal evolution of Paleocene mineralization (Fe-Cu → Pb-Zn → Sn) during orogenesis. These findings provide crucial insights into metallogeny in continental collision zones and pave the way for new exploration targets for tin resources throughout the central-western Tethyan metallogenic domain, especially in underexplored volcanic terrains with analogous geodynamic histories.

The acquisition of spatiotemporal information for lithological mapping with timeliness, accuracy, and high precision is crucial for mineral resource exploration and geological hazard prevention. However, large-scale lithological mapping remains severely constrained by the limitations of visual interpretation in obtaining representative samples from remote sensing data and error propagation during sample collection based on existing geological maps. To address this, we propose a three-dimensional spatial dual-positioning sample generation methodology (SG-3DSD) using Sentinel-2 (S2) and Landsat 8 (L8) data on the Google Earth Engine (GEE) platform, enabling automated generation of 11 lithological class samples across the Beishan region of Gansu Province, China (covering approximately 6,000 km2). First, boundary association rules were established to reconstruct 1:200,000-scale geological maps, mitigating data acquisition biases and cartographic compilation errors. Second, principal component analysis (PCA) was performed on seven S2 spectral bands, with the first three principal components (capturing > 98% information variance) constituting a 3D feature space for localized clustering. Concurrently, four L8 bands were selected through lithological spectral curve analysis to implement band ratio (BR) transformations for secondary positioning. Finally, a two-step refinement strategy was implemented to filter high-confidence samples across 11 lithological classes, balancing intraclass feature consistency and sample purity. Applying SG-3DSD-derived samples to multiple machine learning models revealed that (1) the Stacking ensemble model demonstrated superior lithological discrimination capability compared to conventional algorithms, achieving peak accuracy of 94.15% and mean F1-score of 93.87%; (2) integrating topographic data (especially Elevation) enhanced lithological positioning accuracy by 4.43% ± 1.13%; (3) PCA and BR transformations effectively enhanced lithological separability, particularly at lithological boundary zones; (4) while SG-3DSD enables efficient large-scale sample generation, it is advisable to avoid using excessively large training samples for regional-scale mapping. This methodology mitigates the weighting dependence on geological maps during sample selection and dilutes inherent cartographic error propagation, providing a novel paradigm for large-scale lithological mapping with broad application potential.

Direct dating of sedimentary successions is a main challenge in geochronology, key for the establishment of chronostratigraphic frameworks for both regional and global events. U-Pb in-situ LA-ICPMS direct dating of carbonate samples is emerging as a promising tool, but complications such as mobility and low U contents hinder most of the attempts on common carbonate rocks. We present new U-Pb in-situ LA-ICPMS data for Ediacaran cap carbonate and related successions from Brazil, China and Canada, along with stable carbon, oxygen, and clumped isotope data for the same samples. The novel dataset reveals that in some instances, especially within calcite-after-aragonite crystal fans and microbialite facies, U is retained from early diagenesis through intermediate to deep burial, resulting in tightly constrained and well-spread linear fits in the Concordia space. Calcite-after-aragonite crystal fan samples from the Guia Fm. (Brazil) and Hayhook Fm. (Canada) caps, sitting immediately above glacial diamictite, yielded 632 ± 14 Ma and 631 ± 6 Ma, respectively, supporting quick deposition and diagenesis following Marinoan deglaciation. Clumped isotope apparent equilibrium temperatures (T D 47) of 79 (+12/ − 11) ° C and 181 (+14/ − 13) ° C (95% confidence level), respectively, indicate that the U-Pb system remained unreset within the crystal fans even through the deep burial realm. In the Sete Lagoas Formation of the Bambuí Group (Brazil), crystal fans are not restricted to the immediate cap carbonate sitting above glacial deposits, but instead occur throughout ca. 400 m of carbonate-dominated facies, in distinct stratigraphic intervals corresponding to the Pedro Leopoldo and Lagoa Santa members. Samples from the basal Pedro Leopoldo member yielded U-Pb ages between 625 Ma and 605 Ma. A crystal-fan bearing sample of the Acauã Formation in the Sergipano Belt (Brazil) yielded similar results, suggesting protracted deposition/diagenesis of the negative d¹³C-bearing limestone above the basal cap dolostone. Crystal fans in the topmost Lagoa Santa member, just below the contact with the mudstone-rich Serra de Santa Helena Formation and 330 m above the contact with the glacials, yielded late Ediacaran ages at ca. 570-550 Ma. All of these yielded T D 47 of around 110-149 ° C. These ages are identical within uncertainty to U-Pb ages obtained in stromatolites at the same stratigraphic level, and from the phosphorite-bearing stromatolites of the Salitre Formation, Una Group, further north in the São Francisco craton, which yielded a lower T D 47 of 91 ± 7 ° C. Finally, both the cap dolostone matrix and isopachous cement filling

While manganese (Mn) nodules are authigenic metal concretions that form predominantly on deep-sea seafloor, they have also been found along shallow seafloors. The formation environments of these nodules — deep sea vs. shallow water — often result in distinct chemical and morphological characteristics. As Mn is one of the essential components of energy-storing technologies, assessing the proper estimation of Mn and metal contents in both deep- and shallow-water Mn nodules is critical. It has been found that the Mn content of shallow-water nodules is often lower than that from deep-sea environments. Here, we report the discovery of shallow-water Mn nodules with exceptionally high Mn/Fe ratios on the continental slope of the East Siberian Sea, Arctic Ocean. Despite their shallow-water origin, Mn nodules show morphological and chemical characteristics that are typically unique to deep-sea nodules. These distinctive features, including exceptionally high Mn/Fe ratios, may reflect suboxic diagenesis and the preferential remobilization and re-precipitation of Mn from the adjacent continental shelf. The formation of high Mn/Fe nodules may reflect unique ocean circulation patterns that provided oxygenated bottom waters to the study area. Particularly, Pacific Water entering through the Bering Strait, which overlaps with the nodule formation depth (160 - 240 m) and is enriched in dissolved oxygen, could facilitate Mn-rich nodule growth under suboxic diagenetic conditions since the Holocene. Shallow-water Mn nodules with uniquely high Mn/Fe ratios may offer a novel paleo-environmental proxy for reconstructing paleohydrology and biogeochemical evolutions in shallow marine environments.

Natural hydrogen (H₂) generated by the reaction of ultramafic rocks with water is increasingly recognized as a promising low-carbon energy resource with the analysis of rock mineralogy and structural characteristics recognized to play a crucial role in assessing its subsurface generation potential. In this study, micro-computed tomography (micro-CT), micro-X-ray fluorescence spectroscopy (micro-XRF), X-ray diffraction (XRD), and scanning electron microscopy coupled with energy-dispersive X-ray spectroscopy (SEM-EDS) are employed to analyze the density, elemental distribution, mineral composition, and surface spatial relationships of an ultramafic rock sample. In addition, deep learning-based image analysis is employed to achieve high-resolution mineral phase characterization, enabling quantitative analysis of the spatial distribution, co-location, and contact surfaces of the mineral phases. Focusing on a particular sample that was considered a likely initiator of hydrogen generation due to its mineral contents, our results indicate that the sample is primarily composed of Fe-Mg-rich olivine and silicate minerals, with most olivine phases being Mg-rich forsterite or mixtures of forsterite and Fe-rich fayalite. The sample also contains Fe-S sulfides and high-density metal-enriched phases, including Ni-rich phases that may enhance the H₂-generating potential of serpentinization reactions. These findings highlight the mineralogical complexity of the studied ultramafic rock and the value of integrating compositional and spatial data when considering the potential of particular materials for hydrogen generation. The integrated analytical approach proposed in this study provides new insights and practical tools for evaluating the hydrogen generation potential associated with subsurface serpentinization in ultramafic rock.

Precise control of shield tail clearance is a critical factor influencing the safety and quality of shield tunneling construction. Although various methods exist for accurately measuring shield tail clearance, predictive capabilities remain insufficient. This study is based on a shield tunnel project in the karst region of Longgang, Shenzhen, China. By integrating geological parameters obtained from advanced geological prediction with shield construction monitoring data, a predictive calculation method for shield tail clearance is developed, grounded in the spatial relationship between the shield machine and the pipe segments. A knowledge-based data-driven prediction approach is proposed using a Transformer-LSTM deep learning model. Case analysis demonstrates that the proposed Transformer-LSTM model consistently outperformed baseline models such as GRU, LSTM, and pure Transformer. The predicted R² values for the four positions of the shield tail—top, bottom, left, and right—reached 0.990, 0.901, 0.976, and 0.908, respectively, while error indicators (MAE, RMSE, and MAPE) were also minimized. These results confirm that the proposed hybrid approach effectively captures both global dependencies and temporal dynamics, enabling accurate prediction of shield tail clearance and offering practical engineering significance for guiding shield tunneling construction.

Digital twins of geotechnical structures replicate their physical counterparts, such as underground spaces developed from land reclamations. These spaces often exhibit intricate three-dimensional (3D) stratigraphic distributions, including irregular and interbedded soil layers. Developing a virtual 3D model, such as finite element model (FEM), with complex stratigraphy poses significant computational challenges due to the necessity for numerous stratum voxels, high-resolution meshing, and prohibitive analysis times. Incorporating field settlement data for model updating escalates the computational burden, as repeated evaluations of 3D FEM models are required for each model updating. To address this challenge, this study develops a novel approach for efficiently predicting time-varying 3D settlement from two-dimensional (2D) numerical models with sparsely measured monitoring data. Settlements from 2D FEM analyses, which account for complex stratigraphy, are compiled within a dictionary learning framework and combined with limited monitoring data to estimate time-varying settlements at multiple 2D cross-sections. The 2D settlements are then utilized to reconstruct high-resolution 3D settlements through 3D compressive sampling (3D-CS), eliminating a need for additional numerical model evaluations when integrating new monitoring data. The proposed approach is illustrated using a reclamation project in Hong Kong, China.

Micro-nano fractures serve as the bridge connecting nanopores and macro-fractures. The unclear understanding of their developmental characteristics and controlling factors significantly hinders the large-scale, efficient development of continental shale oil. To address this, this study employs the entropy weight method to establish an evaluation model for fracture development strength that comprehensively considers fracture number, average width, areal density, and areal porosity. Additionally, topology is introduced to evaluate fracture connectivity. The research clarifies the differences in micro-nano fracture developmental characteristics and primary controlling factors among different lithofacies and elucidates the impact of micro-nano fracture development on pore structure and hydrocarbon accumulation in Gulong shale. The results indicate that the HQS (high-organic laminated felsic shale) lithofacies exhibits high micro-nano fracture development strength and connectivity, yielding the highest comprehensive evaluation index. The HCS (high-organic laminated mixed shale) shows high development strength but low connectivity, resulting in a secondary comprehensive evaluation index. Higher organic matter content correlates with greater fracture development strength; clay mineral content controls the characteristics of nano-fracture development; felsic mineral content positively influences fracture connectivity. The development of micro-nano fractures not only enhances macropore content and average pore size but also effectively connects pores of various scales, increasing the effectiveness of the pore-fracture system. Lithofacies with low fracture connectivity (primarily HCS) exhibit more complex pore structures. Shale oil in such lithofacies mainly accumulates via a self-sealing model, making it difficult to form complex fracture networks during hydraulic fracturing and hindering efficient development. Conversely, the HQS lithofacies demonstrates optimal pore-fracture connectivity, favorable oil content, and represents the most favorable lithofacies for Gulong shale oil development. These findings contribute to the optimization of sweet-spot intervals for shale oil exploration in the study area.

Droughts pose escalating threats to global water security, agriculture, and socioeconomic stability amid anthropogenic climate change, with projections indicating an increase in frequency, duration, and severity driven by altered precipitation patterns and amplified evaporative demand. This study introduces a probabilistic framework to quantify drought risk amplification, employing the Risk Ratio (RR) methodology integrated with extreme value theory and non-parametric inference. Utilizing multi-model ensemble (MME) from the Coupled Model Intercomparison Project Phase 6 (CMIP6) under Shared Socioeconomic Pathways (SSP2-4.5 and SSP5-8.5), we evaluate changes in drought characteristics—duration, frequency, and severity — via the Standardized Precipitation Evapotranspiration Index (SPEI) at 3- and 12-month timescales for near-future (NF) and far-future (FF) periods. Our analyses reveal pervasive global intensification, with over 90% of land grids exhibiting positive severity shifts under SSP5-8.5 in the FF, attributed to atmospheric evaporative demand, which accounts for approximately 44% of the trends in SPEI. Threshold-stratified RR assessments reveal nonlinear escalations at higher percentiles (P90 vs. P75), compressing the return periods of extreme events by 20%-30% under high-emission scenarios. Regional hotspots, including the Amazon basin, sub-Saharan Africa, southwestern North America, and Central Asian drylands, exhibit frequency risks that are 4-fold or more amplified, signaling transitions to chronic water stress and potential ecosystem tipping points. These findings underscore the dominance of thermodynamic drivers in drought dynamics, advocating for emissions mitigation to curtail risks by 15%-25% under moderate pathways. By addressing uncertainties in non-stationary regimes, this framework provides adaptive strategies for resilient water management, offering policymakers critical insights to mitigate cascading impacts on global food security and biodiversity in a warming world.

We report here for the first time the detailed spectroscopic investigations on Bhawad meteorite using micro-Raman spectroscopic and high-resolution transmission electron microscopy (HR-TEM) investigation of the Bhawad LL6 ordinary chondrite, focusing on its mineralogical composition and carbonaceous phases. Raman spectroscopy reveals crystalline silicates including olivine, pyroxene, and plagioclase, along with accessory chromite containing ≤ 20% of Al. Carbonaceous material exhibits broad I_D (∼ 1336 cm⁻¹) and I_G (∼ 1587 cm⁻¹) bands with an I_D / I_G ratio of ∼ 1.04, indicative of disordered graphite and nanocrystalline carbon, reflecting shock-induced metamorphism. High-pressure TiO₂ polymorphs are identified by characteristic Raman modes at 146, 394, 446, and 610 cm⁻¹. HR-TEM imaging confirms the presence of nanocrystalline TiO₂ particles embedded within amorphous carbonaceous matrices, demonstrating the coexistence of crystalline and amorphous phases. The Raman spectra of the Bhawad meteorite reveal the presence of high-temperature plagioclase phases, characterized by these distinct vibrational features. This observation indicates possible quenching of the melts having feldspar components, representing the complex thermal and shock metamorphic history of the meteorite. This coexistence of crystalline and amorphous phases highlights the complex thermal and shock history of the Bhawad meteorite, revealing insights into phase transitions and structural order-disorder phase transition induced by impact processes.

Slope engineering is an uncertain, dynamic, and complex nonlinear spatiotemporal system with time delays. High-fidelity prediction of slope seismic stability has long been a formidable challenge due to the inherent randomness and uncertainty associated with ground motion, geo-material properties, complex topography, etc. Traditional numerical modelling always takes a simplified model by forcedly ignoring those uncertainties, thus failing to replicate precisely the intricate nonlinear interactions between factors that affect slope instability. Notably, the newly emerging deep learning methods have the capability of handling multiple factors with uncertainties. However, these methods heavily rely on extensive and comprehensive sensor data, while arranging sensors at certain important positions is sometimes unachievable. Therefore, we propose a multi-task deep transfer learning (MT-DTL) framework in this study to enhance the prediction accuracy of slope seismic response especially in data-limited conditions. The dynamic response at the locations without sufficient accessible sensor data can be effectively predicted with a newly developed algorithm. To collect the necessary sensor data, we conduct a series of physics experiments with the world’s largest multifunctional shaking table equipment. We demonstrate the efficacy and accuracy of our approach on the shaking-table datasets through comparisons with traditional machine learning (ML) methods. Our findings reveal that the MT-DTL framework can improve the confidence level of prediction results (within 5%) from the highest 86.4% by the optimal traditional ML methods to 92.7%, achieving comparable results with two-thirds fewer data. Additionally, a single response example showed that the trained deep transfer learning model has significantly improved the computational efficiency (0.018 - 0.019 s) compared to the dynamic finite element calculation with GeoStudio (10 min). This highlights its potential for integration into geo-hazards digital twin systems, facilitating rapid risk analysis based on real-time monitoring data.

Stable potassium (K) isotopes are emerging as a novel geochemical tracer for investigating magmatic differentiation and source characteristics. This study presents the K isotopic analyses of Neoarchean-Paleoproterozoic granitoids from the Xing’an Massif, a key microcontinent within the eastern Central Asian Orogenic Belt (CAOB), providing new insights into the granitoid petrogenesis and early crustal evolution of this accretionary orogen. The 2568 Ma peraluminous A-type monzogranite exhibits significantly heavier δ⁴¹K values (− 0.22 ‰ to − 0.05 ‰) compared to the range of the upper continental crust. Subduction zones can effectively transfer heavy K isotopic signature to the mantle wedge through slab-derived fluids/melts. The monzogranite could be formed through co-melting and mixing of previously metasomatized mantle materials and recycled supracrustal metapelites, followed by high degree of fractional crystallization in a post-collisional extensional setting. Although both the 1881 Ma monzogranite and 1843 Ma syenogranite share geochemical affinities with adakites, their markedly different K isotopic compositions and distinct geochemical fingerprints point to substantial heterogeneity within their source regions. The 1881 Ma monzogranite shows more pronounced heavy K isotopic enrichment (δ⁴¹K = − 0.39 ‰ to − 0.18 ‰) and elevated zircon δ¹⁸O values (7.28 ‰ -8.93 ‰). These features demonstrate the incorporation of mantle components metasomatized by melts of altered oceanic crust (with elevated δ⁴¹K values) into the lower crustal source. In contrast, the 1843 Ma syenogranite displays ultrapotassic affinity with lighter K isotopic compositions (δ⁴¹K = − 0.45 ‰ to − 0.38 ‰) and strongly negative zircon εHf (t) values (− 11.5 to − 10.2), indicating a thickened lower crustal source with contributions from ancient supracrustal sediments. Collectively, K isotopic compositions of the ca. 1.8 Ga adakitic granitoids overcome the limitations of traditional geochemical and isotopic proxies in revealing the complex granite petrogenesis, and they potentially provide evidence for a cycle of plate tectonics, from oceanic crust alteration at mid-ocean ridges through slab subduction to continental collision. The onset of plate tectonics promoted remelting of Archean igneous and sedimentary crust, generating abundant peraluminous and potassic granitoids during the late Archean to Paleoproterozoic and driving crustal compositional maturation in this accretionary orogen.

The genesis of bonanza-style gold deposits, characterized by weight-percent-level Au enrichment, challenges conventional models of chemical transport via aqueous complexes. Through high-pressure experiments (0.5-1.5 GPa, 600-1150 °C) combined with thermodynamic modeling and transmission electron microscopy (TEM) analyses, we demonstrate that CO₂-rich fluids generated by metamorphic decarbonization create overpressures exceeding ∼200 MPa. This initiates explosive upward migration of sulfide liquids containing Au-Ag nanoparticles (NPs) into porous peridotite at velocities up to 55.9 ± 12.9 µm/h. High-resolution TEM analyses furthermore confirm the mechanical entrainment of Au-Ag NPs within sulfides. Fractal analysis (FD = 1.55-1.62) of dendritic sulfide networks reveals that viscous fingering dominates fluid dynamics. We propose a unified model where gas-driven filter pressing extracts Au-bearing sulfides from subducted slabs, while viscous fingering further facilitates kilometer-scale transport through lithospheric faults. This novel mechanism bridges mantle-derived carbon fluxes with crustal mineralization, offering new insights into the formation of ultrahigh-grade gold deposits.

The South China Block (SCB) is recognized as one of the most significant uranium deposit clusters in the world, characterized by its complex genetic types and geodynamic drives. Based on host rocks, uranium deposits in the SCB can be categorized into three primary types, exhibiting a trend from black shale-related deposits in the west, to granite-related, and ultimately to volcanic-related deposits toward the eastern margin of the SCB. We identify that three types of deposits are primarily distributed within or along margins of ancient crustal domains. Geochronological data reveals large-scale uranium mineralization occurred predominantly during Cretaceous and Paleogene periods. Uranium mineralization was mainly controlled by structures in the extensional setting, developed particularly at subsidiary faults, lithological (unconformity, intrusion contacts) and physicochemical interfaces. Uranium mineralization is dominantly characterized by medium to low ore-forming temperature with pitchblende as the main industrial mineral, and with silicification, carbonatization, hematitization, fluoritization and chloritization as common alteration. Isotopic studies show that sulfur sourced from host rocks, while carbon isotopes distinguish mantle-derived signatures in granite- and volcanic-related deposits from primarily sedimentary organic matter sources in black shale-related deposit. Uranium was mainly contributed by host rocks which are relatively U-fertile geological formations. Magmatic and/or mantle-derived mineralizing agents promote the activation and migration of uranium in host rocks, and accelerate the accumulation of U in ore-forming fluids. Our study suggests that the coupling of shallow and deep-seated energy and conduit system within a crustal extension setting, together with the pre-enrichment of uranium in basement and host rocks, controlled the formation of uranium deposits in the SCB.

Landslides trigger high loss of life, damage to property and infrastructure, particularly in sensitive terrains like Kerala, India. Real-time monitoring and forecasting remain difficult due to rugged topography and low connectivity in remote terrain. The current work depicts a low-power, long-range IoT framework for monitoring applications utilizing LoRaWAN for data transmission and machine learning for forecasting. Soil moisture, accelerometer-gyroscope (MPU6050), humidity (DHT22), and simulated piezometer sensor nodes periodically store important slope-stability parameters. The sensed data are transmitted across LoRa to a base hub where the site-specific machine learning program analyzes the data in real time. Experimental results reveal soil moisture increasing from 2% to 10%, humidity from 89.8% to 91.5%, pore water pressure from 0.2 kPa to 0.5 kPa, and fluctuating accelerometer during simulated slope failure—variables closely related to landslide initiating factors. Machine learning outcomes reveal the ExtraTrees Classifier obtained 87.0% accuracy and gave the best results relative to different algorithms. The system provides automatic SOS messages to the Geological Survey of India (GSI) and executes site-based alarms for communities at risk. In comparison with the current GSM or satellite-based systems, the presented method provides longer-range communications and reduced energy consumption, along with quicker responses. The work presents a field-applicable and scalable solution for landslide risk management and disaster preparedness applications.

An innovative framework for correlating physical-mechanical properties of deep-sea sediments is established through a comprehensive database integrating microstructural, mineralogical, and geotechnical data from over 300 samples. Advanced cold field emission SEM analyses reveal unique flocculated-laminated microstructures dominated by organic components and smectite-rich clay minerals. Microstructural parameters and relationships between macroscopic and microscopic characteristics are further examined, which enhances the fundamental understanding of the correlations between physical and mechanical properties. Statistical analyses demonstrate strong interdependencies among water content, buoyant unit weight, and void ratio, confirming their equivalence as physical descriptors. Crucially, conventional terrestrial soil models show limited applicability for predicting undrained shear strength in deep-sea environments, particularly underestimating strength parameters by neglecting sediment sensitivity and liquidity index. Through multiple nonlinear regression and the construction of multivariate distribution, predictive models are developed incorporating buoyant unit weight, liquidity index, and sensitivity as key governing factors, achieving superior accuracy compared to existing methods. This investigation advances the understanding of physical-mechanical properties of deep-sea sediments, thus providing critical insights for assessing subsea geo-hazards.

Increasingly frequent extreme climate events have intensified urban flood risks, underscoring the urgent need for accurate, interpretable assessment methodologies. This study establishes an explainable artificial intelligence (XAI) framework for flood risk assessment in the Guangdong-Hong Kong-Macao Greater Bay Area (GBA), integrating the LISFLOOD-FP hydrodynamic model with Gradient Boosting Decision Tree (GBDT). To resolve model opacity, Local Interpretable Model-agnostic Explanations (LIME) quantifies the contributions of critical disaster-inducing indicators. The framework achieves over 91% predictive accuracy, revealing a 1.33% expansion of very high-risk zones and a 3.80% increase in high-risk areas under the 100-year flood scenario, with the most affected cities including Guangzhou, Shenzhen, Zhuhai, and Foshan. LIME-based interpretability analysis under this scenario underscores the dominant influence of hydrological and topographic variables, with FD (flood depth), SD (submerge duration), and DEM (Digital Elevation Model) collectively contributing over 60% of the total explanatory contribution. This XAI approach significantly enhances flood risk prediction precision, delivering actionable insights for evidence-based resilience planning across the GBA.

Inherited structures in rifted continental margins strongly influence the architecture and evolution of collisional orogens. The northern Dora-Maira Massif in the Western Alps (NW Italy) preserves records of such inheritances, capturing the transition from Gondwana inheritance to Alpine convergence. New lithostratigraphic and structural data, together with U-Pb zircon dating, reveal a long-lasting tectonostratigraphic and/or magmatic evolution during (i) pre-Permian, (ii) Permian, (iii) Triassic and (iv) Jurassic time intervals. The heterogeneous Paleozoic basement consists of pre-Variscan micaschist and metabasite, and was intruded by Permian igneous bodies now corresponding to the Borgone metagranite and Luserna augen gneiss. The basement was later overlain by a Mesozoic cover made up of Lower Triassic siliciclastic sediments, a Middle to Upper Triassic carbonate platform and Lower to Middle Jurassic syn-rift deposits linked to the opening of the Ligurian-Piedmont Ocean Basin. Our results highlight that the Dora-Maira Massif was located within a transitional paleogeographic domain, emphasizing the role of pre-rift architecture in governing margin segmentation. Successive cycles of sedimentation, magmatism, and rifting created structural and rheological heterogeneities that may have localized strain during the Cenozoic Alpine-related overprinting. The Dora-Maira case illustrates that deep-time inherited structures and tectonostratigraphic settings continue to influence rifting, subduction, and collision, offering a broader framework for understanding the dynamics of orogens worldwide.

The analysis of apparent earth pressure (AEP) in braced excavations in soft clay environments demands advanced methodologies to address complex soil-structure interactions and nonlinear parameter inter-dependencies. Traditional empirical approaches often oversimplify these critical factors, compromising design reliability. This study introduces a data-driven framework that merges machine learning (ML) techniques with finite element (FE) modeling to enhance AEP prediction and interpretation. A novel Dynamic Time Warping (DTW)-based KMeans clustering algorithm is employed to classify AEP distributions, validated against FE simulations and field-monitored data. By integrating FE modeling with data-driven clustering, the framework generates refined apparent pressure diagrams (APDs) tailored to T_sc-specific conditions, outperforming conventional Terzaghi-Peck and CIRIA diagrams. Results demonstrate that ML models reduce prediction errors compared to empirical approaches. This work underscores the transformative potential of ML in advancing geotechnical engineering, offering a paradigm for robust excavation design in heterogeneous soil strata.

Natural resource exploitation—particularly the extraction of minerals and related primary commodities—continues to shape patterns of economic expansion, structural transformation, and environmental strain across developing regions. Understanding how these resource dynamics interact with broader economic structures and institutional conditions is crucial for designing sustainable development pathways. In this context, productive capacity, economic policy uncertainty, and ecological pressure emerge as central dimensions through which the environmental consequences of development can be assessed. This study investigates the impact of the productive capacity index and economic policy uncertainty on the ecological footprint of 33 Asian developing countries from 2000 to 2022, explicitly considering mineral resource dependence, foreign direct investment, and economic growth as control variables. Using advanced econometric techniques—including slope heterogeneity diagnostics, the Westerlund cointegration test, Moment Quantile Regression (MMQR), and Kernel-Based Regularized Least Squares (KRLS)—the analysis reveals that productive capacity, policy uncertainty, and natural resources (including minerals) are negatively associated with the ecological footprint, suggesting that stronger institutional and productive structures mitigate environmental pressures. By contrast, economic growth and foreign direct investment are positively related to ecological footprint, highlighting the environmental trade-offs of rapid expansion and external capital flows. The findings underscore the need for sustainable mineral resource management and integrated policy frameworks that align productive capacity with environmental stewardship. The study concludes that resource-rich economies must balance mineral exploitation with long-term energy and environmental strategies, ensuring that productivity gains do not come at the cost of ecological degradation.

The dynamic interactions between mantle plumes and continental collision zones are still inadequately defined or poorly understood. Focusing on the Early Permian Tarim LIP and the adjacent Central Asian Orogenic Belt (CAOB), this study employs a Random Forest-based tectonic affinity prediction model (98% accuracy) to quantitatively evaluate the relative contributions of distinct mantle components (subduction-modified, asthenospheric, and plume-related) in 461 basalt samples. Combined with lithospheric thickness reconstruction via Y/Yb ratios, we demonstrate that: (1) the influence of the Tarim mantle plume extended northward into the CAOB, but was deflected into an east-west trajectory upon encountering the thick lithosphere (>70 km) of the Yili Block; (2) within the orogen, ocean island basalt (OIB)-affinity anomalies (e.g., East Tianshan, Junggar) are spatially consistent with thin lithosphere zones (55-65 km), and clusters of Ni-Cu sulfide deposits; and (3) major trans-lithospheric faults served as preferential conduits for plume upwelling. These findings provide a ‘‘channel-barrier” model where lithospheric thickness variations control plume spreading asymmetry, with preexisting structural weaknesses regulating spatial distribution. This study establishes a methodological framework for plume identification and Ni-Cu sulfide exploration in analogous settings, with broad implications for deep Earth material cycles and lithosphere-mineralization interactions.

During the Ediacaran Period (635 - 538.8 Ma), the photosynthetic activity due to cyanobacterial communities and early photosynthetic eukaryotes prompted the wide oxygenation of the terrestrial atmosphere. Biogeochemical evidence of this type of communities and activity in different terrestrial environments is very scarce. In this work, we search for lipid biomarkers and their carbon specific isotopic composition in stromatolites from an Ediacaran volcanic alkaline lake in the Anti-Atlas Mountains, in Morocco. Molecular analysis reveals the presence of n-alkanes, isoprenoids, hopanes and steranes in the Amane-n’Tourhart stromatolites, with compound-specific d13C values for n-alkanes and isoprenoids within the range of autotrophic organisms using the Calvin-Benson-Bassham cycle. Results from contamination controls and laboratory tests attest for the indigeneity and syngenicity of the detected biomarkers. In addition, molecular and isotopic analysis of hydrocarbons allows for the recognition of phototrophic activity from the prokaryotic and eukaryotic communities developed in this extreme alkaline lake in anoxic conditions. These unique results shed light on a key Period in the evolution of life on Earth in the particular region of Amane-n’Tourhart. The set of molecular and isotopic biomarkers detected in the Amane-n’Tourhart stromatolites supports the presence of some of the first complex organisms (i.e. fungi and early animals) and the relevance of the most prominent metabolism in present day biology (i.e. Calvin cycle), and expands the catalog of biomarkers preserved from that geological Period to reconstruct its paleobiology.

Ultra-deep sandstone reservoirs are characterized by poor petrophysical properties. Identifying effective reservoir rocks and evaluating reservoir quality are key but challenging aspects in the exploration and development of ultra-deep hydrocarbon reservoirs. Adopting the Cretaceous Bashijiqike Formation of the Keshen gas field in the Tarim Basin with burial depths exceeding 8000 m as an example, we evaluated the quality of this ultra-deep tight sandstone reservoir by classifying petrofacies and analyzing the diagenetic evolution of different petrofacies. We revealed that although the petrophysical properties of ultra-deep reservoirs are poor, effective reservoir rocks with relatively high porosities and permeabilities can still develop locally. According to the detrital mineralogy and texture, diagenesis, and pore system, we classified sandstone into effective petrofacies (ductile lithic-lean sandstone) and tight petrofacies (ductile lithic-rich sandstone and tightly carbonate-cemented sandstone), which underwent differential diagenetic evolution processes. Such processes significantly influence the quality of ultra-deep tight sandstone reservoirs. High contents of ductile grains and carbonate cement explained the low reservoir quality. The ductile lithic-rich sandstone was relatively fine-grained and contained a high content of ductile grains, which, owing to mechanical compaction during early burial, were compacted and largely occupied the pore space, yielding a low reservoir quality. The carbonate-cemented sandstone pores were filled with large amounts of carbonate cements during early diagenesis, resulting in a low reservoir quality. The ductile lithic-lean sandstone was relatively coarse-grained, contained a high content of rigid grains, and exhibited moderate compaction, with relatively well-developed primary pores and secondary dissolution pores. This sandstone exhibited the highest reservoir quality and represents an effective reservoir rock type in ultra-deep tight sandstone reservoirs. This study provides new insights for the evaluation of the effective properties of ultra-deep tight sandstone reservoirs.

The lunar surface element distribution obtained from Chang’e-2 gamma-ray spectrometer has provided new insights into the thermal activity and element migration of the Moon. To further investigate lunar thermal evolution and geological activities, the heat production rate (HPR) distribution was selected as a breakthrough. An optimized inversion method for Chang’e-2 gamma-ray spectrum data, based on multivariate statistical analysis, was developed to effectively reduce the influence of time-varying factors by improving the background estimation and subtraction process. The results validated the utility of HPR for lunar research. The global HPR distribution maps not only provide a reference for assessing the thermal state of the lunar surface, demonstrating that radiogenic heat production can be reliably studied at a global scale, but also enable detailed investigations of regional geological processes. In the Imbrium Basin, HPR clearly reflects the effects of large-scale impact events and subsequent mare volcanic activity. High-HPR materials associated with impact ejecta can be distinguished from the lower-HPR mare basalts. Furthermore, by integrating HPR data with additional geological information, it is possible to assess and partially subdivide the structure of the Imbrium Basin, providing new quantitative insights into its evolution and compositional heterogeneity.

Accurate sea surface temperature (SST) forecasting across multiple timescales remains challenging. Daily forecasting frequently relies on autoregressive models prone to instability and over-smoothing, whereas monthly forecasting suffers from sparse data and the complex dynamics of ocean systems. Existing deep learning methods struggle to address these diverse challenges simultaneously. We introduce SSTFormer, a novel physics-guided deep learning framework that achieves leading results, with root mean squared error of 0.17 °C for daily forecasts and 0.60 °C for monthly forecasts, yielding lower bias and improved spatial coherence. The model’s core innovation is its unified and flexible architecture. For multi-step daily forecasts (1-15 days), it deploys as a ‘‘two-phase sequential ensemble” that replaces conventional autoregression and uses ocean current to solve instability and mitigate error accumulation. For single-step monthly forecasts, it is used in a direct forecasting configuration, proving effective at handling ‘‘sparse data” and ‘‘complex ocean dynamics.” SSTFormer demonstrates how a single architecture, through flexible deployment, can address the unique challenges of multi-scale SST forecasting, highlighting its potential as a unified and robust framework.

This study presents an integrated approach, combining machine learning clustering with gravity data analysis, to characterize the region’s aquifer systems. K-means and Self-Organizing Maps (SOM) were applied to well log data, including Gamma Ray (GR), Spontaneous Potential (SP), and resistivity (R), to delineate lithofacies. Three distinct units were identified: clean sand, shaly sand, and clay-rich facies. The SOM algorithm outperformed the clustering of K-means in accurately estimating layer thickness and resolving lithological transitions. A 3D lithofacies model revealed spatial heterogeneity within the NSSA, highlighting clean sand layers as primary groundwater extraction zones. Gravity data analysis using upward continuation and edge-filtering techniques identified dominant NE-SW, NW-SE, and E-W lineaments controlling groundwater flow dynamics. The 3D gravity inversion model revealed density contrasts associated with structural features, providing insights into potential groundwater flow between aquifers. Spatial analysis reveals lower groundwater salinity in the southern part of the Oasis, coinciding with areas of reduced structural complexity. Higher salinity zones in central and northeastern regions show spatial correlation with gravity-derived structural systems, though causal relationships require additional validation through hydrochemical studies. This integrated approach provides critical insights for sustainable groundwater management in structurally complex arid environments. Groundwater salinization in arid oasis environments poses significant challenges for sustainable water resource management. In Siwa Oasis, Egypt, the deep Nubian Sandstone Aquifer System (NSSA) and the shallow Tertiary Carbonate Aquifer (TCA) interact through fault systems. At the same time, the potential leakage from hypersaline surface lakes creates complex hydrogeological conditions that require comprehensive characterization. Despite the critical importance of understanding aquifer connectivity and salinization processes, there remains a significant knowledge gap in the quantitative integration of multiple geophysical datasets for objective aquifer characterization and structural control identification. Traditional methods lack the spatial resolution and objective framework necessary to map lithofacies distributions and identify structural pathways controlling groundwater flow in complex multi-aquifer systems.