Arka Roy

The magnetic field from Earth's crust helps us understand its thermal structure by finding the depth to the bottom of magnetic sources, an important indicator of the Crustal thermal properties. This study aims to estimate the depth to the bottom of magnetic sources precisely using the observed magnetic field. Traditional methods, like the spectral peak and centroid techniques, are commonly used to estimate the depth to the bottom of magnetic sources. However, these methods typically require prior knowledge about the magnetization source, derived from empirical relationships of wave‐vectors in the spectral domain, which is challenging to obtain over large regions. We devised an innovative deep learning approach utilizing a convolutional neural network to directly estimate the depth to the bottom of the magnetic sources, eliminating the need for prior knowledge of the fractal magnetization source. Synthetic fractal magnetizations were constructed to train the model, and the performance of the convolutional neural network was compared to the modified centroid approach. Our convolutional neural network methodology was confirmed by utilizing a diverse array of realistic synthetic fractal magnetization, incorporating various window widths and depths to the bottom of the magnetization source. The model is applied to the high‐resolution aeromagnetic data of the southern Indian shield to understand the crustal‐scale thermal structure.

The magnetic field from Earth's crust helps us understand its thermal structure by finding the depth to the bottom of magnetic sources, an essential indicator of the crustal thermal properties. This study aims to estimate the depth to the bottom of magnetic sources precisely using the magnetic field. Traditional methods, like the spectral peak and centroid techniques, are commonly used to estimate the depth to the bottom of magnetic sources. However, these methods typically require prior knowledge about the magnetization source, derived from empirical relationships of wave-vectors in the spectral domain, which is challenging to obtain over large regions. We devised an innovative deep-learning approach utilizing a convolutional neural network to directly estimate the depth to the bottom of the magnetic sources, eliminating the need for prior knowledge of the fractal magnetization source. Synthetic fractal magnetizations were constructed to train the model, and the performance of the convolutional neural network was compared to the modified centroid approach. Our convolutional neural network methodology was confirmed by utilizing a diverse range of realistic synthetic fractal magnetization, incorporating various window widths and depths to the bottom of the magnetization source. The model is applied to the high-resolution aeromagnetic data of the southern Indian shield to understand the crustal-scale thermal structure.

The success of horizontal loop electromagnetic method in determining ore body model parameters is intricately tied to the inversion process. However, geophysical inverse problems often grapple with ill‐posed and non‐unique mathematical difficulties. To handle this situation, we presented two new metaheuristics: the Barnacles Mating Optimizer and the Hunger Games Search algorithm. This pioneering approach is the first in the literature to employ both optimizers simultaneously to invert electromagnetic anomalies. To test the core functionality of these inversion tools, we simplified the horizontal loop electromagnetic problem by using a single frequency. Before carrying out the data inversions, a synthetic anomaly caused by a simple multi‐source model was used to perform some modal analyses. These analyses revealed the composite modality character of the inversion process and highlighted the necessity of an algorithm that effectively balances global exploration (thoroughly searching the model space while avoiding local minima) and local exploitation, which focuses on approximating the global optimum closely. In addition, the findings indicated the importance of conducting post‐inversion analyses to evaluate potential uncertainties in the estimations. Initial experiments with nine challenging benchmark functions revealed that Hunger Games Search outperforms in tackling optimization problems. The performances of both algorithms were then examined using synthetic horizontal loop electromagnetic anomalies and two field data sets from Australia. To achieve a balanced comparison, we fine‐tuned the control parameters of both algorithms, optimizing their performance under comparable conditions. These assessments provided valuable insights into the effectiveness of each algorithm when properly configured. Uncertainty appraisal analyses mostly confirmed the consistency and mathematical consistency of the obtained solutions. In the real data inversion cases, the outputs of Hunger Games Search showed strong agreement with the drilling data. This study illustrates the stronger performance of the Hunger Games Search algorithm relative to Barnacles Mating Optimizer in the presented scenarios and demonstrates its effectiveness in estimating source parameters using horizontal loop electromagnetic anomalies.

Precise determination of Moho topography holds paramount importance in advancing our comprehension of Earth's structural characteristics, geodynamic phenomena, and the exploration of resources. This study introduces an innovative methodology employing conditional Generative Adversarial Networks (cGAN) to unveil Moho topographies from observed gravity anomalies. To address the scarcity of real Moho datasets for training the cGAN model, we meticulously generated a comprehensive set of quasi-realistic synthetic training data using the FFT filtering technique. The forward estimation of gravity anomalies, arising from synthetic Moho topographies, was assessed through spherical prism-based gravity modeling. These estimated anomalies served as input data for the training of the cGAN model. For evaluating the efficacy of our developed cGAN algorithm in deriving Moho architecture, we conducted a comparative analysis against a conventional inversion scheme. This assessment utilized various synthetic datasets and a real case study in Southern Peninsular India, renowned for its geological diversity and ancient continental tectonic blocks. The established Bott's inversion scheme was employed as a benchmark to validate the Moho surface estimation obtained through the Deep Learning approach. To mitigate the impact of diverse factors such as topography, bathymetry, sediments, crustal and mantle heterogeneities, observed gravity anomalies underwent meticulous corrections using spherical prism-based forward gravity modeling for real case studies. The gravity contribution exclusively associated with the pure Moho was subsequently inverted using both the cGAN and traditional Bott's inversion schemes. Crucial hyperparameters, including the mean Moho depth and density contrast between the crust and mantle, were determined by utilizing seismic constraints. Our results underscore the potential of the cGAN and spherical prism-based gravity modeling approach in accurately predicting Moho topography. This study provides valuable insights into high-resolution Earth's Moho architecture and contributes to advancing our understanding of geodynamic processes, facilitating resource exploration endeavours with reduced computational demands.

Sedimentary basins, integral to Earth's geological history and energy resource exploration, undergo complex changes driven by sedimentation, subsidence and geological processes. Gravity anomaly inversion is a crucial technique offering insights into subsurface structures and density variations. Our study addresses the challenge of complex subsurface structure assessment by leveraging deep neural networks to invert observed gravity anomalies. Optimization approaches traditionally incorporate known density distributions obtained from borehole data or geological logging for inverting basement depth in sedimentary basins using observed gravity anomalies. Our study explores the application of deep neural networks in accurate architectural assessment of sedimentary basins and demonstrates their significance in mineral and hydrocarbon exploration. Recent years have witnessed a surge in the use of machine learning in geophysics, with deep learning models playing a pivotal role. Integrating deep neural networks, such as the feedforward neural networks, has revolutionized subsurface density distribution and basement depth estimation. This study introduces a deep neural network specifically tailored for inverting observed gravity anomalies to estimate two‐dimensional basement relief topographies in sedimentary basins. To enhance computational efficiency, a one‐dimensional discrete cosine transform based discretization approach is employed. Synthetic data, generated using non‐Gaussian fractals, compensates for the scarcity of true datasets for training the deep neural network model. The algorithm's robustness is validated through noise introduction with comparisons against an efficient and traditional global optimization‐based approach. Gravity anomalies of real sedimentary basins further validate the algorithm's efficacy, establishing it as a promising methodology for accurate and efficient subsurface imaging in geological exploration.

This paper is well-written in its current form. The following reasons can explain why this paper is suitable to be published in Qeios: This paper mainly focuses on comparing constantly applied mathematical models to examine and predict the populatio...

The self-potential method (SP) has been used extensively to reveal some model parameters of various ore deposits. However, estimating these parameters can be challenging due to the mathematical nature of the inversion process. To address this issue, we propose here a novel global optimizer called the Modified Barnacles Mating Optimizer (MBMO). We improved upon the original approach by incorporating a variable genital length strategy, a novel barnacle offspring evolving method, and an out-of-bounds correction approach. The MBMO has not been previously applied to geophysical anomalies. Prior to inversion of real data sets, modal and sensitivity Analyzes were conducted using a theoretical model with multiple sources. The Analyzes revealed that the problem is modal in nature, model parameters have varying levels of sensitivity, and an algorithm that can well balance global exploration with local exploitation is required to solve this problem. The MBMO was tested on theoretical SP anomalies and four real datasets from Türkiye, Canada, India, and Germany. Its performance was compared to the original version under equal conditions. Uncertainty determination studies were carried out to comprehend the reliability of the solutions obtained via both algorithms. The findings indicated clearly that the MBMO outperformed its original version in estimating the model parameters from SP anomalies. The modifications presented here improved its ability to search for the global minimum effectively. In addition to geophysical datasets, experiments with 11 challenging benchmark functions demonstrated the advantages of MBMO in optimization problems. Theoretical and field data applications showed that the proposed algorithm can be used effectively in model parameter estimations from SP anomalies of ore deposits with the help of total gradient anomalies.

The need for adequate data on the spatial and temporal variability of freshwater resources is a significant challenge to the water managers of the world in water resource planning and management. The problems will be acute in the coming years because of the increase in frequency and intensity of hydrologic extremes due to climate change. Therefore, streamflow prediction has become an important area of research because of its importance in flood mitigation, reservoir operation, and water resource management. In this paper, we have tested four Machine Learning models (ML models): Support Vector Machines (SVM), Random Forest (RF), Long Short-Term Memory (LSTM), and Multivariate Adaptive Regression Splines (MARS) for streamflow prediction at daily and monthly time scales in three rivers draining in the different climatic and geological setting. The SVM, RF, LSTM, and MARS models have been trained and tested in the Suvarna, Aghanashini, and Kunderu River Basins in peninsular India. Model intercomparison was made to identify the best suitable model for streamflow prediction. The RF outperforms other models for daily streamflow, and MARS outperforms other models for monthly streamflow prediction in the Suvarna river with Nash-Sutcliffe efficiency (NSE) values of 0.676 and 0.924, respectively. SVM (NSE = 0.741) and RF (NSE = 0.826) are found to be the best models for daily and monthly streamflow prediction in the Aghanashini river. MARS outperformed other models in the case of high, severe, and extreme flow simulation with NSE values of 0.481, 0.374, and 0.455, respectively, in the Aghanashini river. Other hydrological variables (groundwater level data, antecedent soil moisture, potential evapotranspiration data) and a better spatial resolution of rainfall data can be used to develop more accurate machine-learning models for streamflow predictions.

This study presents a high‐resolution 3D Moho structure beneath southern India and its surrounding regions from observed gravity anomalies. The global gravity disturbance model (XGM2019e) with a grid resolution of 0.1° is considered for this study. The extended Bott's inversion algorithm and Gauss‐Fast Fourier Transform based forward model are adopted to invert for the Moho undulations beneath the Indian peninsula. The inversion algorithm is tested for a synthetic model having a predefined density contrast and mean Moho depth. The robustness of this inversion algorithm is further tested for noise‐incorporated gravity data. The control points are required for estimating two hyper‐parameters, viz. density contrast, and reference depth, which play a crucial role in the precise estimation of Moho depth. In real case study, the inverted Moho depth of Southern India and its surrounding regions by seismic constraint receiver function‐driven control points show a very complex architecture of Moho topography. The observed average crustal thickness in the study region is 35.35 km, corroborating with the previously reported Moho depths. The maximum crustal thickness is 53.04 km beneath the southern part of Archean Western Dharwar Craton and west of Salem block, around 44–47 km Moho depth is observed at the south of Salem block into Madurai block till Achankovil Shear Zone, which suggests the possible continuation of the Achaean crust of Palghat‐Cauvery Shear Zone System. The lowest crustal thickness values are observed along the eastern margin of the Cuddapah basin, which overlaps with the Proterozoic Krishna basin of the Eastern Ghats Mobile belt.