Earth Science Informatics最新文献

Forest fires pose a serious risk to ecosystems in the Mediterranean region; thus, 2021 fires in the Mediterranean region of Türkiye emphasize the requirement for accurate and interpretable forest fire susceptibility (FFS) mapping. This study presents an innovative approach to FFS mapping for the Mersin, Antalya, and Mugla provinces, integrating machine learning (ML) models with Explainable Artificial Intelligence (XAI). The methodology employs three state-of-the-art ML models: eXtreme Gradient Boosting (XGBoost), Gradient Boosting Machine (GBM), and Light Gradient-Boosting Machine (LightGBM). These models generated FFS maps using 14 fire conditioning factors, including meteorological, topographic, environmental, and anthropogenic factors. LightGBM demonstrated outstanding performance, acquiring the highest accuracy (0.897), outperforming GBM (0.881) and XGBoost (0.851). McNemar’s statistical test demonstrated significant differences in the predictive capabilities between XGBoost and both GBM and LightGBM, whereas no significant difference was found between GBM and LightGBM. Information Gain and SHapley Additive exPlanations (SHAP) analyses were applied to enhance model interpretability and validate feature importance. Both methods agreed that the most influential variables in FFS are soil moisture, Palmer Drought Severity Index (PDSI), and Land Surface Temperature (LST). On the other hand, SHAP plots revealed complex, nonlinear relationships between these factors and fire susceptibility. At the same time, a high increase in LST enhances the risk of fires; higher soil moisture values and the PDSI decrease the possibility of fire risk. This research also contributes to the concept of FFS mapping interpretability and operational utility with the application of XAI, which establishes a transparent basis for identifying fire risk drivers in Mediterranean ecosystems.

To study the stability control scheme of chemical grouting agent for ionic rare earth mine slope. The improved chemical grouting agent comprised lime, sodium silicate, silica micro powder, calcium lignosulfonate and other water solvents. The differences between the enhanced chemical grouting agent and the traditional chemical grouting agent were observed by using indicators such as slope displacement and soil nail tension. The improved chemical grouting agent showed a positive stability effect in both simulation and field experiments. The improved chemical grouting agent is more suitable for the slope stability control scenario of an ionic rare earth mine.

In the present era, scientific computation is gradually becoming a primary research method, with an increasing number of researchers engaging in simulation studies on various high-performance computing platforms. Scientific workflows play a crucial role in organizing these complex research tasks effectively. However, poorly managed scientific workflows can lead to wastage of HPC computational resources and fail to alleviate the operational burden on researchers. The parameter optimization of Earth System Models (ESM) poses specific challenges due to its complexity, exacerbating these issues. To address these challenges, we propose a scientific workflow management framework for surrogate-based ESM parameter optimization. This framework consists of four layers: the resource layer, which gathers current resource information; the service layer, which provides various components to ensure the accurate execution of workflows; the management layer, which monitors the execution status of workflows; and the software environment interaction layer, which serves as the interface for data exchange between users and the framework. We monitored a team engaged in tuning CAM parameters before and after adopting the framework, and the results showed significant improvements in operation numbers, task execution time, and storage resource consumption after deploying the framework. This validates that our proposed scientific workflow management framework effectively addresses the challenges in user operations and resource management during surrogate-based ESM optimization processes, demonstrating the potential of our framework.

This study was carried out within the offshore Niger Delta Basin to generate novel predictive models for estimating rock parameters. MATLAB was employed in obtaining models for four different rock parameter relationships including unconfined compressive strength (UCS) against bulk density, UCS against sonic transit time (STT), shear wave velocity against STT, and permeability against bulk density using multiple ordinary least-squares regression (OLSR) methods. Also, the Adaptive-Neuro Fuzzy Inference System (ANFIS) artificial intelligence network was utilized for modeling and optimization of the data. Statistical tools including the Sum of Squares Total (SST), the Sum of Squares Error (SSE), the Sum of Squares Regression (SSR), and Correlation Coefficient (R-squared) were applied in investigating the prediction performances of the models. Results of OLSR analysis show that only the UCS against bulk density model gave high prediction performance in all the OLSR models with R-squared values of 0.8637, 0.8848, 0.8216, 0.9956, and 0.8108 for linear, quadratic, power, logarithmic, and exponential models respectively. ANN model results revealed that UCS against bulk density, UCS against STT, and shear wave velocity against STT models all gave high prediction performances with respective R-squared values of 0.89635, 0.99365, and 0.52703, while the permeability against bulk density model gave low performance (0.03378). These findings imply that all the OLSR models can be applied for the prediction of rock UCS from bulk density information only, while ANN-generated models can be used in predicting UCS from bulk density and STT, in addition to shear wave velocity from STT in the study area and similar geologic environments.

Over the past twenty years, the Middle East has experienced a surge in air pollution and dust, resulting in a range of issues affecting both people and the environment. Monitoring particulate matter (PM₁₀ and PM_2.5) has long been essential in assessing air quality. Thus, creating precise and proficient predictive models to estimate particulate matter concentrations is imperative for effectively managing and reducing air pollution. The estimation of seasonal and intra-annual PM concentrations was conducted in this study through the use of MLR and MLP models. A diverse range of meteorological parameters, including evaporation, temperature, wind speed, visibility, precipitation, and humidity, were employed along with aerosol optical depth (AOD). During autumn, the MLR and MLP models exhibited impressive performances. For PM10, the R values were 0.7 and 0.79, whereas for PM_2.5, they were 0.7 and 0.81, respectively. The MLP’s superior correlation between the observed and estimated seasonal and intra-annual PM concentrations was noteworthy, as it consistently favored PM2.5 and highlighted the superiority of the ANN-MLP model over MLR. The predictive data underscored a correlation between PM concentration and the four seasons, emphasizing the seasonal impact on PM levels. Sensitivity analysis revealed that relative humidity (RH) was the primary factor influencing the intra-annual levels of both PM₁₀ and PM_2.5. This study offers valuable insights into comprehending the formation process, implementing effective control measures, and establishing predictive models for PM, all aimed at proficiently managing air quality.

The Dschang area has substantial mineral and geological exploration potentialities. However, its basement is unclear due to lack of studies on mineral and lithology mapping, and other mineralization indices. The lithological units and potential hydrothermal alteration zones in the Dschang area are investigated here using remote sensing; geographic information systems (GIS); and statistical analysis which are essential method for geological exploration. Landsat 9 OLI, ASTER data using False Color Composites (FCC), Band Ratios (BRs), Principal Component Analysis (PCA), Spectral Angle Mapper (SAM), fuzzy-logic methods, and field observations are used to identify the rocks units and potential mineralization. The integration o these multiple methods allowed the identification of orthogneiss, granites and basalts with iron-oxides, hydroxyl and ferrous bearing as potential mineralization in the Dschang area. The Evaluation of the fuzzy membership of each alteration mineral from Landsat 9 OLI and ASTER data indicates that the highest favorability index varies from 0.8 to 1 indicating a rating index related to iron mineralization. From the statistical analysis of the geochemical data, the calcic, alkaline-calcic, and metaluminous to weakly peraluminous I-type character of the Dschang granites prove their parent magma was fertile for mineralization in Rare Earths, Cu, Sn, Mo, Zn, and Pb. In addition, analysis of lineaments illustrated three structural directions in the area (ENE-WSW to NE-SW, N-S to NNE-SSW, and NW–SE). The innovative aspect of this research is the integration and processing of Landsat 9 OLI, Fuzzy, ASTER, statistical geochemical analysis of previous data, and field investigations, which allows for the identification of rock units and potentially mineralized rock formations and defining exploration targets as well.

The Ba Na region plays a crucial role in deciphering the tectonic evolution of the Indochina terrane. This study addresses the scarcity of geophysical research in the area by utilizing aeromagnetic data to delineate subsurface structures. Various techniques including reduction to the pole (RTP), multi-stage RTP, reduction to the equator (RTE), enhanced analytic signal (EAS), theta map (TM), tilt angle of the horizontal gradient (TAHG), and fast sigmoid-based edge detection (FSED) were examined on synthetic datasets before employing them to analyze the geomagnetic field of the region. The results from the synthetic example show that the use of the RTE filter can provide a more reliable and accurate approach for removing asymmetries caused by non-vertical magnetization. These results also demonstrate the efficacy of applying TAHG and FSED to RTE aeromagnetic data for mapping subsurface structures in the Ba Na area. The findings reveal major magnetic contacts in the approximate ENE-WSW direction and the secondary contacts in the N-S direction, with depths ranging from 200 to 650 m, possibly arising from the collision between the Northern and Southern Vietnam blocks. Additionally, intrusive structures were identified in the region. This study constitutes the initial magnetic interpretation, providing valuable insights into the structural characteristics of the Ba Na area and filling a notable research gap in the understanding of this geologically significant region.

The increasing generation of data in different areas of life, such as the environment, highlights the need to explore new techniques for processing and exploiting data for useful purposes. In this context, artificial intelligence techniques, especially through deep learning models, are key tools to be used on the large amount of data that can be obtained, for example, from weather radars. In many cases, the information collected by these radars is not open, or belongs to different institutions, thus needing to deal with the distributed nature of this data. In this work, the applicability of a personalized federated learning architecture, which has been called adapFL, on distributed weather radar images is addressed. To this end, given a single available radar covering 400 km in diameter, the captured images are divided in such a way that they are disjointly distributed into four different federated clients. The results obtained with adapFL are analyzed in each zone, as well as in a central area covering part of the surface of each of the previously distributed areas. The ultimate goal of this work is to study the generalization capability of this type of learning technique for its extrapolation to use cases in which a representative number of radars is available, whose data can not be centralized due to technical, legal or administrative concerns. The results of this preliminary study indicate that the performance obtained in each zone with the adapFL approach allows improving the results of the federated learning approach, the individual deep learning models and the classical Continuity Tracking Radar Echoes by Correlation approach.

The technology for classifying and recognizing underwater targets is crucial for supporting underwater acoustic information countermeasures. The research focus is on the extraction and classification of features of underwater targets. Researchers have conducted an in-depth study from various perspectives. Due to the influence of ambient noise and various operating conditions of different targets, the signal-to-noise ratio of underwater acoustic signals is generally meager. Additionally, the components of these signals are complex, often requiring specific signal pre-processing techniques such as signal enhancement and decomposition. In current methods, there is a primary focus on extracting and classifying features of underwater acoustic signals after multi-step preprocessing. However, these methods do not effectively integrate feature extraction and classification. To address these limitations, we propose a new model called Depthwise Separable Convolutional Attention (DSCA) and use multiple instances of DSCA to construct a neural network, which we call DSCANet. The DSCANet integrates feature extraction and target classification for underwater acoustic targets. The ’target’ in our work should be mentioned as it refers to underwater sources of sound. The structure of DSCANet is unified and simple, and no specific pre-processing of the underwater acoustic signal is necessary. The DSCANet is trained and validated on ShipsEars, an open dataset. It achieves a classification accuracy of 93%, which is the highest in the contrast experiment.

The Taihang Mountains in China span numerous cities, where landslide disasters occur frequently in the mountainous areas, jeopardizing the lives and properties of residents. Consequently, it is of great significance to focus on prevention and control of landslide disasters in the region. Currently, a single model is commonly employed to analyze landslide susceptibility mapping (LSM), but the accuracy of the results fails to meet the demands of early warning, prevention, and control. This paper focuses on the Taihang Mountain area as the research area, organizes the collection of landslide disaster potential points and related influence factor data, and employs the information quantity method to derive a composite machine learning model by coupling with Random Forest (RF) and Extreme Gradient Boosting (XGB), subsequently utilizing the Genetic Optimization Algorithm (GA) to optimize the model. The performance of the composite model is enhanced using the Genetic Algorithm (GA), employing accuracy, regression rate, precision, F1 score, AUC value, and Taylor diagram to evaluate the comprehensive accuracy of the model results, with a susceptibility map generated for comparative analysis. The results demonstrate that the IV-GA-RF model performs optimally (accuracy = 0.956, precision = 0.96, recall = 0.953, F1 score = 0.957, AUC = 0.946 for the testing set, AUC = 0.929 for the training set), with all-around improvement in performance metrics compared to the unoptimized composite model, with metric values improving by 0.044, 0.051, 0.046, 0.044, 0.021 and 0.020, respectively. The IV-GA-RF model exhibits a significant advantage over the IV-GA-XGB algorithm, also optimized using the GA algorithm. The accuracy of the susceptibility map produced by the IV-GA-RF model is superior, as assessed by the Seed Cell Area Index (SCAI) method. The four factors of slope, rainfall, seismicity, and stratigraphic lithology are crucial in determining the occurrence of landslides in the study area. In summary, the IV-GA-RF model can be utilized as an effective model for analyzing landslide disasters, providing a reference for research in this field and contributing scientific insights to disaster prevention and control efforts in the study area; simultaneously, the concept of the composite optimization model introduces new perspectives into this field.