Acta Geophysica最新文献

Lithology identification is a fundamental task in seismic reservoir characterization. However, existing studies have primarily focused on optimizing single algorithms, with limited systematic comparisons of ensemble models and hyperparameter optimization strategies. To address this issue, this study, based on well seismic data from the North Sea F3 block, integrates recursive feature elimination with cross-validation (RFECV), the Near-Miss Synthetic Minority Oversampling Technique (NM-SMOTE) sampling strategy, and four mainstream hyperparameter optimization methods to evaluate the performance of random forest, XGBoost, LightGBM, CatBoost, and stacking ensemble method. NM-SMOTE (Near-Miss SMOTE) effectively alleviates the class imbalance problem by synthesizing minority sandstone samples and retaining key mudstone samples that are closest to the minority class (while reducing the majority class size), thereby improving the reliability of minority class recognition. Fivefold cross-validation was employed, using well log lithology interpretation as ground truth for validation. The results indicate that Optuna achieves the best balance between efficiency and accuracy, outperforming Bayesian optimization and grid search in terms of test accuracy, training time, and model stability. CatBoost achieves the highest prediction accuracy (area under the receiver operating characteristic curve, AUC = 0.91), demonstrating clear sandstone–mudstone boundaries and superior continuity in predictions. These findings provide a reliable basis and methodological support for the selection and optimization of intelligent lithology identification models under complex geological conditions.

An experimental investigation was undertaken to examine the temporal evolution of turbulence characteristics and higher-order flow correlations for a discharge of 0.0176 m³/s in a rigid bank, mobile bed channel. Three-dimensional velocity components were captured using an Acoustic Doppler Vectrino Profiler to facilitate detailed turbulence analysis. Turbulent kinetic energy (TKE) exhibits its minimum and maximum values close to the outer and inner bend, respectively, at the upstream and apex regions, while at the downstream, the higher TKE occurs at the center. Near the bed, the TKE flux shows downstream-downward flux transport toward the inner bend. The TKE budget indicated that, in the near bed region, the TKE production rate and diffusion rate exhibited both positive and negative tendencies, whereas the TKE dissipation rate was positive, and the pressure energy diffusion rate showed a negative tendency. The eddy size, computed using the Taylor microscale with in the inertial subrange, increases at the inner and center points at the upstream. At the apex and downstream locations, it is located at the center points of the bend. The turbulence indicator states that the level of local turbulence is more dominant at the outer bend points. The second order streamwise–vertical and streamwise–lateral correlations indicate a sign reversal, confirming the existence of secondary flow. Third order correlations can be helpful to confirm the streamwise–downward flux, which shows the characteristics of sweep events responsible for the bed movement. The fourth order correlations indicated strong characteristics of turbulence intermittency behavior. The results of the current study are applicable to the selected case of the modeled river reach and can be extended to natural sinuous rivers by considering the limitations. The current study provides significant understanding of flow turbulence and can be helpful to hydraulic engineers for designing structures in an asymmetric sinuous channel qualitatively.

Due to the increasingly negative impacts of drought, which affects water resources, agricultural activities, and all sectors, there is a need to analyze the drought trend to understand its effects and mitigate them comprehensively. This research aims to spatiotemporally analyze the drought trend based on the well-known Standardized Precipitation Index (SPI) at different timescales using classical and innovative trend analysis methodologies, including Mann–Kendall (MK), Sen’s slope (SS), and newly proposed Frequency Innovative Trend Analysis method (F-ITA) over the Antalya basin, Türkiye with monthly precipitation data from 1969 to 2022 for the first time in the literature. Also, the research calculates the slope and actual trends using SS and ITA methodologies along with the drought classifications and frequencies, providing a deeper understanding of the drought patterns and their variability. This research generally indicated an increasing trend in drought events for SPI-3 and SPI-6 based on classical methods and F-ITA graphs for specific stations. However, F-ITA for SPI-12 showed a significant drought trend across all stations, with increased drought frequencies; for example, Alanya station exhibited a monotonic increasing trend, with the frequencies of MD, SD, ED, and EXD approximately doubling over the study period. The spatial distribution of slopes computed by SS and ITA and their respective actual trends exhibited significant parallels, resulting in more frequent drought events and an increased drought trend in the southern parts of the region near the coast. The southern and southeastern parts of the study area exhibited the highest trends and slopes. In summary, analyzing drought trends and their spatio-temporal distribution provides critical and crucial insights for sustainable water resources management and agriculture, and guides policymakers in developing effective adaptation and mitigation strategies.

Water bodies are an important geographical feature in freshwater security, irrigation, climate regulation, and flood risk management. Thus, monitoring and extracting water bodies are widespread uses in remote sensing. This is the artificial intelligence (AI) age, demonstrated by a variety of AI models developed for a wide range of applications. Additionally, there is an increasing remote sensing data that can be used for AI models’ inputs. The high-quality Segment Anything model (HQ-SAM), a newly improved version of the SAM, is proposed to accurately enable the segmentation of a broad range of objects while maintaining the promptable architecture, efficiency, and zero-shot generalizability of the original SAM. We applied the HQ-SAM, and water indices (NDWI, MNDWI, SWI, AWEI) in the Otsu method for lake/reservoir extractions using optical and Synthetic Aperture Radar (SAR) remote sensing imagery, including Sentinel-1, 2, ALOS-2/PALSAR-2, RadarSAT, Landsat 5 and 8, and Google-based satellite images (Leafmap) for selected lakes in South Korea. The HQ-SAM model is evaluated as working well, exhibiting excellent accuracy (above 95%) of water body masks compared to the measured boundary of the lake. The HQ-SAM results surpassed Otsu’s results applied to four common water indices. Both approaches revealed advantages and disadvantages, where the HQ-SAM worked well with larger, complex lakes but had some mis-segmented small, thin parts of lakes. Nevertheless, the Otsu method did not separate surface water bodies from the snow and ice on the mountains. The HQ-SAM revealed an accurate and promising potential model for water body extraction using remote sensing imagery.

The study analyzed thermal and precipitation conditions during the thermal growing season (TGS) in Central and Northern Europe over the period 1950–2022. The mean season length was 189 days, with substantial spatial variability ranging from 76 days in northern Scandinavia to 293 days in the southwestern part of Germany and southern Netherlands. A statistically significant increase in season length was observed over the study period. On average, the season commenced on April 24 and ended on October 30, with its onset and termination shifting toward earlier and later dates, respectively. The mean air temperature during the TGS was 12.1 °C, increasing at a rate of 0.13 °C/10 years, while the sum of temperatures rose on average by 53 °C/10 years. The highest rates of change were recorded in the southern part of Central Europe. Precipitation totals during the growing season exhibited pronounced spatial and seasonal variability, with a mean value of 390 mm and a weak decreasing trend (− 1.1 mm/10 years). The number of days with precipitation averaged 73, while values of the hydrothermal coefficient of Selyaninov (HTC) ranged from 0.5 to over 3.0, with a mean of 1.39, corresponding to optimal conditions for plant development. HTC trends were regionally differentiated but statistically insignificant for the study area as a whole. The results indicate a systematic warming of the TGS across the entire study area, whereas precipitation exhibits both strongly varied trends and spatial variability, thereby significantly altering the region’s thermal and moisture conditions.

Predicting scour parameters downstream of sluice gates due to the high velocity of a jet and submerged hydraulic jumps seems to be essential with accurate physical models. The present study investigated the scour parameters downstream of a gate via laboratory approach in three cases for the sedimentary bed, including uniform (σ_g = 1.25), semi-uniform (σ_g = 1.35), and nonuniform (σ_g = 1.45). Four scour parameters were measured as the maximum depth of scour hole (d_se), its longitudinal distance from the end of the apron (x_se), the maximum dune height (h_d), and its horizontal location from the beginning of the sediment bed (x_d), and became dimensionless by dividing by the gate opening (b₀). The comparison of velocity profiles indicated acceptable accuracy of the results with the previous empirical equation. The difference between the scour proposed equations in the present study and the previous formulas is adding parameter σ_g, which could provide better reality of the physical features of the variation in the uniformity of the sediment bed grains. Statistical analysis revealed that multiple nonlinear regression analysis (MNLRA) could calculate d_se/b₀ more accurately than multiple linear regression analysis (MLRA) with R² = 0.957, RMSE = 0.263. Furthermore, proposed equation for x_se/b₀, h_d/b₀, and x_d/b₀ has better performance in MNLRA compared to MLRA.

Machine learning (ML) techniques offer major improvements for ground motion prediction in India's high seismic hazard zones—specifically Seismic Zones IV and V, encompassing the Himalayas, Indo-Gangetic Plain, and Kachchh. This study harnesses a dataset of 564 three-component acceleration records from 145 earthquakes (M_w 2.3–7.9) and 95 strong-motion stations to develop and benchmark XGBoost (eXtreme Gradient Boosting), LightGBM (Light Gradient Boosting Machine), and artificial neural network (ANN) models. The XGBoost model, trained with rigorous cross-validation strategies and explicit regularization, achieves excellent generalization (test R² = 0.96, Pearson’s correlation coefficient ρ = 0.998), outperforming established ground motion prediction equations (GMPEs) and ANNs while capturing regional and site-specific variability. Model robustness and uncertainties are analyzed using RMSE, MAE, F1-Score, Bayesian Information Criterion (BIC), and comprehensive residual checks. The Bayesian Information Criterion (BIC) values obtained for the training and full datasets are −1710.24 and -2251.49, respectively. The substantial negative BIC values demonstrate that our XGBoost regression model achieves excellent predictive performance by balancing fit and simplicity effectively. The XGBoost approach demonstrates robust physical consistency but reveals elevated uncertainties for long-period/distant events, highlighting data-driven limitations and motivating further research. This ML-based framework offers substantial advances for seismic hazard assessment and resilient structural design tailored to India's most hazardous regions.

Geomagnetic storms are complex phenomena during which highly energetic solar wind interact with geomagnetic field and transfer energy to earth’s magnetosphere, auroral atmosphere and ionosphere, through magnetic reconnection process. This can result in geomagnetic substorms, ring current enhancement and other magnetic and ionospheric disturbances. Thirty-two intense geomagnetic storm time substorms of Solar Cycle 23 has been considered for energy budget analysis. Solar wind energy incident onto the magnetosphere gets disbursed through different energy sinks and a part of it may also get stored in the magnetosphere. The share of coupled energy to different energy sinks has been analyzed using different empirical methods involving geomagnetic indices and from different magnetosphere models runs of Community Coordinated Modeling Centre, NASA. The major dissipation happened through ionosphere joule heating and ring current enhancement during all the substorm events. The coupling efficiencies of the events revealed loading–unloading process (CE > 100%) as well as driven processes (CE < 100%) during energy transfer. Higher the substorm intensity larger is the amount of energy transferred affecting the satellite technologies and power grids on earth. The highest amount of energy (30.1PJ) transferred to magnetosphere was during the most intense substorm during the superstorm of November 20, 2003, with peak AL of − 4141 nT. Energy budget analysis is important in understanding more of space weather hazards and its impacts on electronics and technology, thereby useful in minimizing economic losses and enhancing the knowledge base of underlying dynamics of such geomagnetic phenomena.

Marine simultaneous source seismic acquisition can optimize efficiency and increase data density; however, the recorded data frequently experience signal interference among sources. Effective deblending is essential to suppress blending noise and recover useful signals, ensuring high-quality data processing and analysis. Generally, we use sparse transform techniques to eliminate blending noise. This paper employs the seislet transform, whose prediction operator relies on the local slope of seismic events. The interference from blending noise makes it difficult to forecast the local slope accurately in blended seismic data. We establish an effective two-stage deblending workflow utilizing the seislet transform to resolve this problem. Leveraging the random time delays in simultaneous source acquisition, some segments of the blended seismic data remain unblended. We use this prior information to improve the deblending outcomes. In the first stage, pre-deblending is conducted on the blended data. We subsequently employ the pre-deblended data to inform the prediction of local slopes and construct a mute operator. In the second stage, an iterative deblending framework is utilized within the seislet transform domain, where the seismic event slopes are based on the results predicted in the first stage. The blended seismic data are subjected to the mute operator for each iteration, which separates it into blended and unblended components. Meanwhile, based on the local similarity between recovered useful signals and removed blending noise, we propose an iterative stopping criterion that avoids unnecessary iterations. Tests on model and field data confirm the effectiveness and extensibility of the proposed workflow.

Distributed acoustic sensing (DAS) technology has gained widespread use in vertical seismic profiling (VSP) data acquisition due to its efficiency. However, high-energy noise introduced by complex geological conditions significantly degrades data quality, posing challenges for traditional denoising methods. While deep learning offers new approaches for seismic denoising, its reliance on large-scale training data and high computational resources remains a limitation. To address this, we propose a Weight Fusion Adversarial Network based on a Self-Collaborative Strategy (SC-WFAN). This network dynamically fuses features from different processing stages, incorporating a weight fusion (WF) module between the encoder and decoder to preserve contextual information and enhance detail recognition. Additionally, the denoising network replaces the generator in generative adversarial networks (GANs), optimizing the process through adversarial training, while the self-collaborative strategy further improves training efficiency. A training dataset comprising 483 pairs of field DAS-VSP records covering four dominant noise types (random background, fading, horizontal, and optical noises) was constructed. Experimental results demonstrate that SC-WFAN excels in suppressing strong noise and recovering weak signals from thin and deep layers, requiring only 66.56G floating-point operations (FLOPs) and 1.87 M parameters, outperforming traditional methods and mainstream deep learning models (e.g., DnCNN, AttU-Net). Its efficiency and robustness provide an innovative solution for processing complex DAS-VSP seismic records, particularly suited for high-precision data processing in unconventional oil and gas resource exploration.