Journal of Earth System Science最新文献

In machine learning, ensemble learning methods (ELM) consist of combining several machine learning algorithms to obtain better quality predictions compared to a single model. The basic idea of this theory is to learn a set of classifiers and allow them to vote. In this paper, to correctly apply the ELM for enhancing of an artificial neural network (ANN) performances, a strategy was devised which is to divide the data to be classified into two categories, ‘easy-to-classify’ category and ‘difficult-to-classify’ category using a main ANN. Hence, reliable ANN and unreliable ANN are created and applied for the classification of ‘easy-to-classify’ data and for the classification of ‘difficult-to-classify’ data, respectively. The AdaBoost algorithm and Bagging algorithm are implemented separately on the unreliable ANN. To increase performance, the AdaBoost results and Bagging results are merged. The developed scheme is applied to remote sensing images from Meteosat Second Generation (MSG). The final results show very interesting performances in the case of the fusion of the results from AdaBoost-ANN and the results from Bagging-ANN (Ada/Bag-ANN). Indeed, the POD, FAR, CSI and Bias pass from 87.2%, 17.4%, 80.8% and 1.3 (ANN) to 96.8%, 06.8%, 92.7% and 1.1 (Ada/Bag-ANN), respectively. The same trend was observed in the case of precipitation estimates. The estimates obtained from the developed model (Ada/Bag-ANN) largely surpass those obtained from the use of ANN without ELM. Compared to ECST (Enhanced Convective Stratiform Technique), EPSAT-SG (Second Generation Satellite Precipitation Estimation), TAMSAT (Tropical Applications of Meteorology using SATellite), and RFE-2.0 (Rain Fall Estimate) which showed correlation coefficients of 87%, 81%, 76% and 71%, respectively, the Ada/Bag-ANN method shows significantly better results with a correlation coefficient of 94%.

Abstract

We report new Sm–Nd whole rock-mineral isochron ages of 2514 ± 13 Ma (MSWD = 0.79) and 2651 ± 95 Ma (MSWD = 7.4) from two east coast dykes (ECD) of Southern Granulite Terrain (SGT), India. The ages from the representative mafic dyke samples correspond to the time of intrusion of ECD into the eastern part of SGT, indicating the presence of an older Archean crust in SGT near the Pondicherry coast. The Sm–Nd ages obtained from the present study, along with geochronological information from Singhbhum Craton, suggest a magmatic linkage between SGT (including southern Dharwar Craton) and Singhbhum Craton during the Neoarchean period. The older ages obtained from the mafic dykes of the present study are comparable with the Sm–Nd ages of older mafic dykes from Nuggihalli green stone belt of Western Dharwar Craton (WDC), Pb–Pb ages of mafic dykes from Singhbhum Craton of India and the U–Pb ages from Pilbara and Kaapvaal cartons. These comparisons unlock a clue to Neoarchean (2.8–2.5 Ga) paleogeographic reconstructions of Pilbara, Kaapvaal, Singhbhum cratons, northern SGT (including southern Dharwar Craton) and also provide an opportunity for wide windows of research to be undertaken considering the dykes from SGT.

Research highlights

• Evidence of Neoarchean magmatism from East coast dykes near Pondicherry coast of Southern Granulite Terrain, India.

• Sm–Nd ages of 2514 ± 13 and 2651 ± 95 Ma represent the time of intrusion of east coast dykes in Southern Granulite Terrain.

• Isotope age indicates the presence of ~2.7 Ga older Archean crust near Pondicherry coast of Southern Granulite Terrain.

• Geochronological studies reveal a magmatic linkage between Southern Granulite Terrain and Singhbhum craton.

• The present study provides clues to the connection between Pilbara, Kaapvaal with SGT and Singhbhum cratons.

In the current study, a seismic inversion based on a hybrid optimisation of genetic algorithm (GA) and pattern search (PS) is carried out. The GA is an approach to global optimisation technique that always converges to the global optimum solution but takes much time to converge. On the other hand, the PS is a local optimisation technique and can converge at local or global optimum solution depending on the starting model. If these two techniques are used together (here termed hybrid optimisation), they can enhance one's benefit and reduce the drawbacks of others. The present study developed a methodology to combine GA and PS in a single flowchart and utilise seismic reflection data exclusively to predict porosity and impedance volume in inter-well regions. The algorithms are initially tested on synthetically created data based on the wedge model, the coal coking model, and the 1D convolution model. The performance of the algorithm is remarkably acceptable, according to the error analysis and statistical analysis between the inverted and the anticipated results. After that, the field post-stack seismic data from the Blackfoot field, Canada, is transformed into impedance and porosity using a developed hybrid optimisation technique. The inverted/predicted sections show very high-resolution subsurface information with impedance varying from 6000 to 14000 m/s×g/cc and porosity varying from 5 to 40% in the region. The error decreases from 1.0 to 0.5 for impedance inversion, whereas it varies from 1.4 to 0.5 for porosity inversion within 3000 iterations, which cannot be achieved by a single optimisation technique. The findings also demonstrated a sand channel (reservoir) anomaly with low impedance (6000–9000 m/s×g/cc) and high porosity (12–20%) in between 1040 and 1060 ms time intervals. This study provides evidence that subsurface parameters like acoustic impedance or porosity may be promptly and affordably determined using seismic inversion based on hybrid optimisation. The developed methodology is very helpful in finding subsurface parameters in a limited time and cost, which cannot be achieved only by global or local optimisation.

Abstract

The method of target classification known as hyperspectral imaging (HSI) is commonly used in the field of remote sensing. However, recent research has shown that categorizing HSI can be problematic due to the limited availability of labelled data. There is significant interest in applying this technique to hyperspectral data. Previous graph neural network (GNN)-based methodologies often used a graph filter to obtain HSI properties, but the potential advantages of various graph neural networks and graph filters have not been fully exploited. GNNs often operate under the assumption that a node’s neighbours are independent of each other, neglecting potential interactions among them. To overcome these limitations, graph attention neural network-based remote target classification (GANN-RTC) has been proposed. It has the ability to handle both the labelled and unlabelled datasets. To evaluate the performance of GANN-RTC, we compared it with existing methods using performance measures such as individual class accuracy, overall accuracy, and the Kappa coefficient. The findings indicate that the GANN-RTC yields enhancements in OA, ICA, and KC by 2.32, 7.89, and 2.47% for the Cuprite dataset and 4.79, 11.85, and 2.82% for the Pavia University dataset.

Research highlights

• The research focuses on remote target classification in hyperspectral imaging using a Graph Attention Neural Network.

• Previous methods in this field have not fully utilized the potential advantages of graph filters and graph neural networks.

• The proposed approach overcomes limitations by considering interactions between neighbouring nodes and can handle both labelled and unlabelled datasets.

• Performance evaluation shows significant improvements in overall accuracy, individual class accuracy, and the Kappa coefficient compared to existing state-of-the-art methods.

In this study, the focus is on predicting the properties of rocks beneath the Earth’s surface using global optimisation techniques such as genetic algorithms (GA), simulated annealing (SA) and particle swarm optimisation (PSO). The goal is to minimise the difference (error) between actual seismic data and synthetic (computed) seismic traces. Global optimisation is an approach that is independent of the initial model and aims to identify the global minimum of an objective function. In contrast, local optimisation relies on the accuracy of the initial model, and if an accurate initial model is not provided, it may become trapped in a local minimum, leading to an inaccurate representation of the subsurface model. What makes global optimisation powerful is that it does not get stuck in local minima (suboptimal solutions), but seeks the absolute best solution in the entire search space. This property is crucial in seismic inversion, where finding the most accurate representation of subsurface properties is of utmost importance for geophysical applications. The study includes one synthetic example and one real dataset, with a specific emphasis on evaluating acoustic impedance rock properties. While acoustic impedance is characteristic of rock layers, seismic data represents properties at the interfaces between these layers. Consequently, seismic data is highly valuable for gaining detailed insights into the subsurface. The results of the optimisation process provide exceptionally detailed views of the subsurface, aiding in the interpretation of seismic data. GA, SA and PSO algorithms perform well, both with synthetic data and real data. The inversion process identifies a zone with low acoustic impedance, corresponding to a prominent seismic anomaly. The evaluation of the inverted outcomes reveals that the impedance within the area ranges from 4300 to 4700 m/s*g/cc, situated within a specific time range of 900–950 ms in the seismic data of F3-block, Netherland.

Rock magnetic analyses of easternmost Indus Molasses of Nyoma–Rhongo section, Ladakh Himalaya have been performed. The thermomagnetic curve gives Curie point (Tc) information on the constituent magnetic minerals, which vary between 574° and 592°C. The hysteresis loops are harmonious and the resultant remanence ratio (Mrs/Ms) ranges between 0.10 and 0.19 and the coercivity ratio (Bcr/Bc) between 1.91 and 3.07. The domain states of magnetic grains majorly belong to the pseudo-single domain (PSD) state. The isothermal remanent magnetization (IRM) acquisition curves show saturation ranges between 250 and 300 mT, and the coercivity spectra show coercive force ranging between 24 and 41 mT. These investigations indicate that magnetic mineralogy in samples is predominantly controlled by fine to medium-sized PSD state magnetite and accessorily Ti-poor magnetite, pyrrhotite, and greigite. This magnetic mineralogy seemed homogenous and was not considerably affected by weathering, lithogenesis and geotectonic events, suggesting their deposition over well-developed palaeogeography with small-scale tectonic modulations. The best possible source for Indus Molasses, as identified in the current study is the Ladakh batholith having an affinity to the Eurasian Plate and these interpretations are in line with the literature.

Climate change can have adverse effects on various ecosystems on the globe, with the cryosphere being affected to a significant extent. Of the cryosphere, mountain or alpine glaciers are essential resources for freshwater and various ecosystem services. Glacial ablation is the process of removal of snow and ice from a glacier, which includes melting, evaporation, and erosion. The increase in temperature on the Earth due to climate changes is causing rapid glacial abrasion. The rapid global decline in alpine glaciers makes it necessary to identify the key drivers responsible for a glacial retreat to understand the eventual modifications to the surroundings and the Earth’s ecosystem. This study attempts to understand the influence of different driving factors leading to glacier retreat using Machine Learning (ML) and Remote Sensing (RS) techniques. Three models have been developed to estimate the glacial retreat: Feedforward Artificial Neural Network (ANN), Recurrent Neural Network (RNN) and Long-Short Term Memory (LSTM). The RNN performed the best with an average training and validation accuracy of 0.9. The overall shift of the area estimate has been identified over 10 years. The model thus generated can lead to a better understanding of the region and can provide a baseline for policy and mitigation strategies in the future.

Malshej Ghat is one of the busiest transport corridors in Maharashtra. It is highly prone to landslides, as the region receives heavy precipitation and steep slopes along the road section. This manuscript focuses on the applicability and reliability of block size and shape classification techniques in the assessment of the stability of the road-cut rock slopes. Structurally weak zones extending along and across the road section are identified as a lineament, and accordingly, slopes are chosen for in-depth analysis. Kinematic analysis signifies that wedge failure is observed to be the most common type of failure, while planar and toppling failures are observed at places. The rock blocks which are produced as a result of the intersection of different joint sets are mainly cubic, cubic-elongated and elongated in shape, whereas platy, platy-cubic and elongated-platy blocks are very less in proportion. Volume and surface area of rock blocks play a vital role in the movement of the blocks. The rock mass is observed to be nearly poor to fair in quality; however, the numerical simulation shows that the slopes MRS-1, MRS-5, and MRS-7 are critically stable.

A group of algorithms for estimating the current intensity (CI) of typhoons, which use infrared and microwave sensor-based images as the input of the algorithm because it is more skilled than each algorithm separately, are used to create a technique to estimate the typhoon intensity which is known as SATCON. In the current study, an effort was undertaken to assess how well the SATCON approach performed for estimating typhoon intensity throughout the West Pacific basin from year 2017 to 2021. To do this, 26 typhoons over the West Pacific basin were analysed using the SATCON-based technique, and the estimates were compared to the best track parameters provided by the Regional Specialized Meteorological Centre (RSMC), Tokyo. The maximum sustained surface winds (V_max) and estimated central pressures (ECP) for various ‘T’ numbers and types of storm throughout the entire year, as well as during the pre-monsoon (March–July) and post-monsoon (July–February) seasons, have been compared. When compared to weaker and very strong typhoons, the ability of the SATCON algorithm to estimate intensity is determined to be rather excellent for mid-range typhoons. We demonstrate that SATCON is more effective in the post-monsoon across the West Pacific basin than in the pre-monsoon by comparing the algorithm results.

A hazard is a natural occurrence that might harm humans, animals or the environment. It may cause loss of life, illness or other health consequences, property damage, social and economic crisis or environmental degradation. Various regions around the world are vulnerable to one or more types of disasters. Flooding is one of the worst environmental catastrophes that impacts both civilisation and the environment globally. Various datasets and methods, such as meteorological data, satellite images and GIS, were used to create the hazard assessment map. For a particular region, flood hazards can be developed by integrating an assessment map for several parameter categories. The aim of the study was to evaluate the hazard of flooding and map the areas that will be flooded in Gujarat. This study develops and tests flood-hazard maps to visualise the spatial variation of hazards in Gujarat, India. The parameters for flood-hazard assessment are mainly considered as elevation, slope, aspect, curvature, lithology, soil, land use/cover, drainage density and distance from the river, and rainfall to create a map in the context of a GIS. The acquired data was evaluated using ArcGIS and fuzzy-logic techniques to build a flood hazard map. Five categories have been assigned to the computed flood hazard map: very low, low, moderate, high, and very high. Engineers, planners and local governments may find this study useful in the future when it comes to land use planning and the control of hazards. Flood hazard potential mapping is necessary to manage and mitigate flooding.