Acta Geophysica最新文献

Sand-sliding slope instability poses difficulties for construction in the Tibetan Plateau and adjacent areas. At present, a simple and reliable method for the engineering evaluation of the water–sand flow susceptibility of such slopes is not available. Therefore, herein, classical machine learning algorithms—gray relation analysis, random forest, and entropy weight method—were employed to evaluate the index weights for the water–sand flow susceptibility of sand-sliding slopes. This analysis was performed on field investigation data from 51 sand-sliding slopes. Building on debris flow susceptibility evaluation methodologies, a dedicated evaluation model for the susceptibility of sand-sliding slopes to water–sand flow was developed. By integrating qualitative field criteria for water–sand flow with a four-tier classification system (extremely susceptible, moderately susceptible, mildly susceptible, and non-susceptible), a quantitative susceptibility evaluation criterion was established through comparative statistical analysis. Our model exhibited an accuracy rate of 92.31% for water–sand flow susceptibility estimation on validation testing with 13 field engineering samples. The practical applicability of the model was further validated on 34 samples under an actual engineering project, achieving an enhanced accuracy of 94.12% with high safety performance. These results confirm the reliability, practical applicability, and generalizability of the proposed methodology.

Volcanic disaster scenes have diverse types and random distribution, displaying complex global features, local information, and sample label ambiguity in remote sensing images. The existing convolutional neural network (CNN)-based classification for remote sensing images is limited by the fixed receptive field of the convolutional kernel, which reduces the modeling ability of local feature and long-term dependencies in remote sensing images. To address this issue, a rough set attribute reduction framework for the Res-Attention_Unet network (RSARNet) used for volcanic disaster scene classification is presented in this paper. In RSARNet, the rough set attribute reduction module uses genetic algorithms to dynamically reduce decision tables and remove redundant attributes so as to better overcome the sensitivity of the network to parameter settings and dependence on sample selection. The Res-Attention_Unet module explores the multi-scale deep features of volcanic disaster scenes by focusing on global contextual information and local details. And then the fully connected layer and classifier are combined to implement the prediction of volcanic disaster scene and output of classification labels. Finally, a volcanic disaster scene (VDS) dataset was used to test the feasibility of the proposed method. Extensive experimental results show that the RSARNet method has the most significant improvement effect on volcanic disaster scene classification, with an overall accuracy of 92.54% compared to traditional machine learning methods. The findings of this paper provide new references for using remote sensing and deep learning for volcanic disaster monitoring and disaster prevention and reduction.

In recent years, especially in geotechnical engineering studies, the relationship between the engineering properties of the soil can be determined through the increasingly frequent use of statistical analysis methods. Standard penetration test results, shear wave velocity (Vs) and electrical resistivity measurements were evaluated during the investigations on different soil types in Şırnak city centre and surrounding settlements. To obtain these correlation equations, correlation and regression analyses were applied to the collected data, investigating linear and nonlinear relationships, and the most suitable relationships and their corresponding correlation coefficients were determined. SPT tests have been conducted in every borehole opened in the field, and N values have been determined. In contrast, the wave speeds were measured at the same points as these drillings. These include SPT-N, shear wave velocity and SPT-N electrical resistivity correlations. The correlations were analysed using three methods: Pearson correlation, simple linear regression and Spearman’s coefficient test analysis. As a result, the regression and correlation relationships obtained are fundamentally based on the geotechnical properties of the examined fields and the amount of data processed.

Accurate estimation of bed shear stress is crucial for understanding sediment transport and morphodynamic processes in fluvial and environmental flows. Shear plates offer a direct measurement approach that avoids the restrictive flow assumptions of many indirect methods. However, their broader use has been limited by misconceptions and concerns about measurement uncertainty. This study addresses these challenges by applying uncertainty analysis based on the Guide to the Expression of Uncertainty in Measurement (GUM), specifically adapted to quantify and evaluate the reliability of shear plate measurements. Experiments were conducted under three hydraulically complex conditions: propeller-induced jets, flow through rigid emergent vegetation, and flexible vegetation over three-dimensional bedforms. Results show that shear plates consistently deliver low relative uncertainties, with over 70% of measurements below 10% and 44% below 5%. The dominant sources of uncertainty were associated with experimental conditions, particularly flow non-uniformity and the need to estimate vegetation drag, while contributions from the measurement instrumentation itself remained minor. Comparisons were made with the traditional gravity method, an indirect approach that estimates bed shear stress based on the energy slope within a control volume. Shear plates provided more accurate and reliable estimates, particularly in vegetated flows where local gradients compromise slope-based calculations. This study demonstrates the practical value of applying the GUM framework in experimental hydraulics, promoting greater confidence in shear plate measurements.

Static elastic modulus (E_s) and dynamic elastic modulus (E_d) are two important parameters in rock mechanics that indicate the resistance to deformation under constant and variable loads, respectively. In this study, E_s and E_d were predicted based on physical and textural properties using ML and statistical models. Statistical analysis showed that textural properties have a greater effect than physical features on E_d and E_s. The presence of wackestone and mudstone tends to reduce the elastic properties of rocks, whereas packstone contributes to enhancing these characteristics. The ratio of E_d to E_s for the samples in the present study was found to be equal to 1.14. The relationship between these two parameters, based on the most accurate fit, is a Linear function with a correlation coefficient of 93%. Support vector regression based on radial basis kernel function (SVR-RBF), feedforward multilayer perceptron neural network (FMLPNN), and K-nearest neighbor (KNN), multivariate linear regression (MLR), and random forest (RF) were used to estimate the E_s and E_d. Based on various statistical criteria, the FMLPNN with an R² = 0.99 and RMSE = 0.07 to estimate E_d and an R² = 1.00 and RMSE = 0.01 for estimating E_s demonstrated greater accuracy compared to the other models.

In the present paper, we implement Physics-Informed Neural Networks (PINNs) to study the dispersion of SH-waves in an initially stressed sandy half-space. The traditional numerical methods for simulation studies of seismic wave propagation are computationally expensive. We thus propose the application of PINNs to efficiently solve the governing equations representing SH-wave propagation. In the PINN framework, the governing equation along with the initial and boundary conditions are embedded into the neural network’s loss function which is then minimized using Adam’s technique to optimize the network. The optimal architecture of the PINN framework is obtained by varying the number of hidden layers and neurons. The loss function values are also calculated at various epochs and presented graphically to analyze the convergence of the method. Further, the PINN framework is then utilized to explore how anisotropy, initial stresses, and the presence of sandiness influence the displacement and velocity of SH-waves and are illustrated graphically. Additionally, a three-dimensional graph is generated to illustrate the displacement of the wave as a function of spatial coordinates x and z, as well as time t. The framework is further applied to simulate SH-wave propagation in anisotropic sedimentary basin. The results demonstrate the PINN’s ability to capture anisotropy-induced changes in wave displacement and velocity.

A seismic anisotropy study based on the shear wave splitting technique is conducted in the Valley of Mexico, analyzing 814 seismic events occurred between 1996 and 2023. This analysis provides insight into the region’s tectonic and structural behaviors, as well as the main causes controlling seismic anisotropy. The study reveals a geologically complex setting, with notable spatial and depth-dependent variations across distinct structural regions. Fast polarizations were identified along NW–SE, N–S, E–W, ENE–WSW, and NE–W trends, with anisotropy strength ranging from 0.17 to 21.16 ms km⁻¹, and individual values reaching up to 61 ms km⁻¹. Approximately 52.1% of the mean fast polarization directions correlate with local geological structures. In the central, central–western, and northern sectors of México City, 78.5% of MDP values align with NE–SW and ENE–WSW regional regimes. This percentage decreases in other areas, indicating the combined influence of regional and local stress regimes. The dominant regional anisotropic pattern extends to depths shallower than 15 km. Superimposed on this are localized anisotropic features with NW–SE, N–S, and E–W trends, indicating additional structural controls or ‘dual’ dominant regimes. Anisotropic layers cover much of the Valley, with anisotropy percentages ranging from 2.0 to 9.0%. The highest values are concentrated in the central and central–western sectors, suggesting zones of structural weakness that may facilitate deformation and connectivity with larger fault systems. Anisotropy strength increases with decreasing depths, from 2.5 ms km⁻¹ at 10–12 km to 27.59 ms km⁻¹ in the uppermost 0–2 km layer, indicating pervasive shallow crustal anisotropy. These high anisotropy concentrations may reflect the presence of compliant, self-organized critical systems, contributing to stress-induced and temporal varying anisotropy.

Peatlands play a critical role in water storage and as reservoirs of organic carbon, with their degradation contributing to carbon dioxide emissions. However, the study of peatlands is challenging due to the variable thickness of peat and muck layers and limited accessibility to research sites. Traditional sensing methods, while effective for surface layer data collection, often fail to provide a comprehensive understanding of peatland dynamics. To address these limitations, this study employs electromagnetic geophysical methods, specifically the ground conductivity meter (GCM), to investigate several peatland sites located on river terraces. The GCM method, based on the interaction between transmitter and receiver coils to generate and measure electromagnetic fields, offers insights into subsurface conductivity, which is influenced by the composition and moisture content of the peat. Data collected from the study areas were processed using 2D inversion techniques, revealing distinct boundaries between low-resistivity peat zones and higher-resistivity sandy soil areas. The results include resistivity distribution maps along the profiles of various peatlands, highlighting sandy zones where peat accumulates on river terraces. These findings demonstrate the effectiveness of the GCM method in estimating peat thickness, assessing moisture content, and detecting significant changes in peat wetness. Furthermore, this study lays the groundwork for long-term monitoring, as potential future changes in peat resistivity could indicate drying processes, such as mucking, and the associated release of greenhouse gases. This research underscores the utility of electromagnetic methods in advancing peatland conservation and management strategies.

Reservoir heterogeneity in geologically complex settings poses significant challenges for accurate characterization and predictive modeling, especially in hydrocarbon-rich regions such as the eastern Sirte Basin, Libya. This study develops a robust workflow that combines Artificial Neural Networks (ANNs) and Self-Organizing Maps (SOMs) to enhance the prediction of porosity and permeability while integrating computational outputs with geological insights. The methodology utilizes an extensive dataset from twenty-nine wells, comprising 3,417 core plugs, 2,945 core descriptions, wireline logs, and detailed chemostratigraphic data, addressing the limitations of traditional regression models constrained by linear assumptions. Traditional regression methods, limited by their inability to model nonlinear relationships, yielded correlation coefficients R² of 0.33 and 0.24 for porosity and permeability predictions, respectively. ANNs demonstrated significantly superior predictive performance, achieving R² values of 0.89 for porosity and 0.85 for permeability, coupled with minimal bias and robust error distributions. Complementing this, SOM clustering delineated depositional facies and stratigraphic controls, effectively linking machine learning outputs with practical geological interpretations. This integrated approach bridges computational precision and geological understanding, offering a scalable framework applicable to diverse geological settings worldwide. The study’s findings underscore the potential of combining advanced machine learning techniques with core and log data to optimize hydrocarbon recovery strategies and address reservoir heterogeneity. By leveraging these methodologies, this workflow establishes a new standard for reservoir characterization and resource management in geologically complex basins.

The change in duration, amplitude, and frequency content of the earthquake ground motion as it passes through the rock and ground environment is referred to as the local ground effect. Impedance differences between bedrock and soil, as well as the dynamic behavior of soils, can amplify this effect. The geometry of both dense and loose soil layers must be known to accurately define soil–structure interaction and properly assess how soil behavior affects a structure during an earthquake. Local ground effects are known to play a significant role in structural damage during earthquakes. In basin-like environments, however, studies based on foundation and sub-base depth often lack sufficient information, making it difficult to identify problems associated with basin effects. It is not appropriate to provide construction-related information, especially in environments with a basin structure like Sivas, without determining the bedrock or solid ground conditions. This study aimed at determining the bedrock/seismic foundation depth for the central settlement of Sivas and defining the basin structure, involved large-scale microgravity measurements. The study area was modeled in three dimensions using the gravity data obtained. Long-term microtremor measurements were also conducted, and one-dimensional depth–shear-wave (Vs) velocity models were generated using the Rayleigh ellipticity method. The bedrock/seismic foundation structure of the study area was defined using two different methods, and these definitions were combined into two-dimensional sections. A depth map of the study area was created, revealing that the thickness of the loose basin unit is approximately 90 m. Ambient noise was recorded at 35 points with a velocity seismometer, and S velocity (Vs) profiles were obtained from joint inversion of Rayleigh ellipticity data and dispersion curves from MASW and ReMi data. Furthermore, the Vs-depth structure of the basin was defined along the profiles cutting the basin in NW–SE and S–N directions, based on the Vs velocities in the 2D gravity model. The frequency range along these profiles was found to be 0.6 Hz in the deep parts of the basin and 2.5 Hz in the shallow parts.