Predicting mine water inrush accidents based on water level anomalies of borehole groups using long short-term memory and isolation forest

Journal of Hydrology, Journal Year: 2022, Volume and Issue: 616, P. 128813 - 128813

Published: Nov. 26, 2022

Language: Английский

---

Fluid geochemistry and geothermal anomaly along the Yushu-Ganzi-Xianshuihe fault system, eastern Tibetan Plateau: Implications for regional seismic activity

Journal of Hydrology, Journal Year: 2022, Volume and Issue: 607, P. 127554 - 127554

Published: Feb. 5, 2022

Language: Английский

---

Meteorological Anomalies During Earthquake Preparation: A Case Study for the 1995 Kobe Earthquake (M = 7.3) Based on Statistical and Machine Learning-Based Analyses

Atmosphere, Journal Year: 2025, Volume and Issue: 16(1), P. 88 - 88

Published: Jan. 15, 2025

The purpose of this paper is to discuss the effect earthquake (EQ) preparation on changes in meteorological parameters. two physical quantities temperature (T)/relative humidity (Hum) and atmospheric chemical potential (ACP) have been investigated with use Japanese “open” data AMeDAS (Automated Meteorological Data Acquisition System), which a very dense “ground-based” network stations higher temporal spatial resolutions than satellite remote sensing open data. In order obtain clearer identification any seismogenic effect, we used station at local midnight (LT = 01 h) our initial target EQ was chosen be famous 1995 Kobe 17 January (M 7.3). Initially, performed conventional statistical analysis confidence bounds it found that (very close epicenter) exhibited conspicuous anomalies both parameters 10 1995, just one week before EQ, exceeding m (mean) + 3σ (standard deviation) T/Hum well above 2σ ACP within short-term window month weeks after an EQ. When looking whole period over year including day case only detected three additional extreme anomalies, except winter, but unknown origins. On other hand, anomalous peak largest for ACP. Further, distributions anomaly intensity presented using about 40 provide further support relationship has compared recent machine/deep learning methods. We utilized combinational NARX (Nonlinear Autoregressive model eXogenous inputs) Long Short-Term Memory (LSTM) models, successful objectively re-confirming same prior combination these results elucidates are considered notable precursor Finally, suggest joint examination their real prediction, as future lithosphere–atmosphere–ionosphere coupling (LAIC) studies information from bottom part LAIC.

Language: Английский

---

Multi-step modeling of well logging data combining unsupervised and deep learning algorithms for enhanced characterization of the Quaternary aquifer system in Debrecen area, Hungary

Modeling Earth Systems and Environment, Journal Year: 2024, Volume and Issue: 10(3), P. 3693 - 3709

Published: March 22, 2024

Abstract In this research, a multi-step modeling approach is followed using unsupervised and deep learning algorithms to interpret the geophysical well-logging data for improved characterization of Quaternary aquifer system in Debrecen area, Hungary. The Most Frequent Value-Assisted Cluster Analysis (MFV-CA) used map lithological variations within system. Additionally, Csókás method discern both vertical horizontal fluctuations hydraulic conductivity. MFV-CA introduced cope with limitation conventional Euclidean distance-based k-means clustering known its low resistance outlying values, resulting deformed cluster formation. However, computational time demands are evident, making them costly time-consuming. As result, Deep Learning (DL) methods suggested provide fast groundwater aquifers. These include Multi-Layer Perceptron Neural Networks (MLPNN), Convolutional (CNN), Recurrent (RNN), Long Short-Term Memory (LSTM), which implemented classification regression. categorized inputs into three distinct lithologies trained initially by results MFV-CA. At same time, regression model offered continuous estimations conductivity model. demonstrated significant compatibility between outcomes derived from approaches DL algorithms. Accordingly, lithofacies across main hydrostratigraphical units mapped. This integration enhanced understanding system, offering promising development management.

Language: Английский

---

Anomaly Detection Using Machine Learning in Hydrochemical Data From Hot Springs: Implications for Earthquake Prediction

Water Resources Research, Journal Year: 2024, Volume and Issue: 60(6)

Published: June 1, 2024

Abstract This study explores the potential of machine learning algorithms for earthquake prediction, utilizing fluid chemical anomaly data from hot springs. Six springs, located within an active fault zone along southeastern coast China, were carefully chosen as hydrochemical monitoring sites extended period two and a half years. Using this data, prediction model integrating six was developed to forecast M ≥ 5 earthquakes in Taiwan. The model's performance validated against recorded events, factors influencing its predictive capability analyzed. Our comprehensive analysis conclusively demonstrates superiority over traditional statistical methods prediction. Additionally, including sampling time sets significantly improves performance. However, it is important note that varies across different spring indicators type, highlighting importance identifying optimal specific scenarios. parameters, detection rate (P) response threshold (M), impact capabilities. Therefore, adjustments are needed optimize practical use. Despite limitations such inability differentiate pre‐earthquake anomalies post‐earthquake pinpoint precise location earthquakes, successfully showcases paving way further research improved methods.

Language: Английский

---

Groundwater Radon Precursor Anomalies Identification by EMD-LSTM Model

Water, Journal Year: 2022, Volume and Issue: 14(1), P. 69 - 69

Published: Jan. 1, 2022

Groundwater radon concentrations can reflect the changes of crustal stress and strain. Scholars scientific institutions have also recorded groundwater precursor anomalies before earthquakes. Therefore, monitoring is an effective means predicting seismic activities. However, variation within not only affected by structural factors, but environmental such as air pressure, temperature, rainfall. This causes difficulty in identifying possible anomalies. EMD-LSTM model proposed to identify study investigated time series data from well #32 located Sichuan province. Three models (including LSTM (Long Short-Term Memory) with auxiliary data, (Empirical Mode Decomposition Long without data) were developed order predict variations. The results indicated that prediction accuracy was much higher than model, obtain ideal result. Furthermore, different durations activities T (T = ±10, ±30, ±50, ±100) comparing identification results. rate highest when ±30. identified five among seven selected Taking example, we provided a promising method, detect It suggested be used future studies.

Language: Английский

---

Application of the extreme gradient boosting method to quantitatively analyze the mechanism of radon anomalous change in Banglazhang hot spring before the Lijiang Mw 7.0 earthquake

Journal of Hydrology, Journal Year: 2022, Volume and Issue: 612, P. 128249 - 128249

Published: July 27, 2022

Language: Английский

---

Unbalanced network attack traffic detection based on feature extraction and GFDA-WGAN

Computer Networks, Journal Year: 2022, Volume and Issue: 216, P. 109283 - 109283

Published: Aug. 17, 2022

Language: Английский

---

Importance of Machine Learning and Deep Learning Algorithms in Earthquake Prediction: A Review

Published: May 11, 2023

Earthquakes are the leading natural disasters that have caused loss of life and property since formation world. Machine learning deep frequently used in studies for earthquake prediction. This article consists a compilation using machine algorithms. In article, on topics such as magnitude estimation, signal discrimination, electron density estimations ionosphere, examination radon gas anomalies algorithms included. The this paper show Deep Learning more forecasting. It is expected will provide successful results future due to its ability work with larger data sets compared improve itself from errors.

Language: Английский

---

Electromagnetic and Radon Earthquake Precursors

Published: June 27, 2024

Earthquake forecasting is arguably one of the most challenging tasks in Earth sciences owing to high complexity earthquake process. Over past 40 years, there has been a plethora work on finding credible, consistent and accurate precursors. This paper cumulative survey precursor research, arranged into two broad categories: electromagnetic precursors radon In first category, methods related measuring radiation wide frequency range, i.e. from few hz several MHz, are presented. Precursors based optical radar imaging acquired by space borne sensors also considered, sense, as electromagnetic. second concentration measurements gas found soil air, or even ground water after being dissolved, form basis activity Well-established mathematical techniques for analysing data derived described with an emphasis fractal methods. Finally, physical models generation propagation aiming at interpreting foundation aforementioned seismic precursors, investigated

Language: Английский

---