Research



The Mathematical Geophysics and Geoinformatics
(MGG) Research Group

Mathematical Geophysics and Geoinformatics (MGG) Research Group at the Institute of Geophysics and Engeneering Seismology after A. Nazarov! The MGG studies geophysical processes in various layers of the Earth, seismic hazard assessment, eartquake rediction etc. are basic, fundamental, and applied problems for the territory of the Republic of Armenia. It will allow us to develop a physical foundation for the creation of technologies and systems to prevent natural and man-made extreme and emergency situations. To solve these problems, new knowledge in the field of Earth physics is required, as well as a transition to the use of modern methods and approaches of mathematical geophysics (incluing discrete mathematical analysis, modern methods, and algorithms of geoinformatics, new mathematical tools for processing large volumes of geophysical information). In the proposed project, fundamentally new physical and mathematical models will be built and tested, mathematical algorithms for operational forecasting of earthquakes will be developed using patern recognition methods and modem artificial intelligence tools. To this end, within the framework of the project, a group of scientists from IGES NAS RA will be supplemented by geophysics, mathematics, information technology, GIS technologies, iding machine learning), etc.

Research Topics

Earthquake clustering in modern seismicity



Group Lead

Jon Karapetyan

Group

Ani Gevorgyan,
Shahen Shahiyan,
Sevada Hovhannisyan,
Lilya Hovhannisyan

Collaborators

Li Li,
Yongzhe Wang,
Anatoly Soloviev,
Boris Dzeboev

Funded By

Mathematical Geophysics and Geoinformatics (MGG) Research Group
Institute of Geophysics and Engeneering Seismology


About

Sensitive instruments have been able to record ever smaller earthquakes in different parts of the world, especially since the 1950s. As the data on these small earthquakes accumulated, two fundamental questions emerged: (i) what are their characteristics and physical properties? and (ii) how do they relate to larger shocks? One of the prime characteristics of small earthquakes is their clustering in space and time, for example, as sequences of foreshocks, aftershocks, and earthquake swarms (Richter 1958; Mei 1960; Bak et al.2002; Baiesi & Paczuski 2004; Zaliapin & Ben-Zion 2013; Gu et al.2013; Moradpour et al.2014; Zhang & Shearer 2016). Aftershocks occur almost universally, as a transient sequence of smaller events occurring after a main shock, at a rate that decays over a few months or years to the background level in the form of an inverse power law known as the Omori–Utsu law (Utsu et al.1995). Swarms are clusters of events in space and time with no clear main shock.

We examine the case for persistent clusters for the case of seismicity around Caucasus. It is a highly densely populated and an important economic region in an intraplate continental setting that has been associated with large historical earthquakes. The region has a rich historical archive that is still being evaluated and updated.


Earthquake Time-series Forecasts



Group Lead

Jon Karapetyan

Group

Ani Gevorgyan,
Shahen Shahiyan,
Sevada Hovhannisyan,
Lilya Hovhannisyan

Collaborators

Li Li,
Yongzhe Wang,
Anatoly Soloviev,
Boris Dzeboev

Funded By

Mathematical Geophysics and Geoinformatics (MGG) Research Group
Institute of Geophysics and Engeneering Seismology


About

Over the last few years, many works have been done in earthquake prediction using different techniques and precursors in order to warn of earthquake damages and save human lives. Plenty of works have failed to sufficiently predict earthquakes, because of the complexity and the unpredictable nature of this task. Therefore, we use the powerful deep learning technique. The results show that learning decomposed datasets gives more wellfunctioning predictions since it exploits the nature of each type of seismic events.


Earthquake Early Warning System



Group Lead

Jon Karapetyan

Group

Ani Gevorgyan,
Shahen Shahiyan,
Sevada Hovhannisyan,
Lilya Hovhannisyan

Collaborators

Li Li,
Yongzhe Wang,
Anatoly Soloviev,
Boris Dzeboev

Funded By

Mathematical Geophysics and Geoinformatics (MGG) Research Group
Institute of Geophysics and Engeneering Seismology


About

We explored the feasibility of predicting the structural drift from the first seconds of P-wave signals for On-site Earthquake Early Warning (EEW) applications. To this purpose, we investigated the performance of both linear least square regression (LSR) and four non-linear machine learning (ML) models: Random Forest, Gradient Boosting, Support Vector Machines and K-Nearest Neighbors. Furthermore, we also explore the applicability of the models calibrated for a region to another one. Earthquakes release energy that travels through the Earth as seismic waves. Seismic sensors detect the first energy to radiate from an earthquake, the P-wave, which rarely causes damage. The sensors transmit this information to data centres where a computer calculates the earthquake's location and magnitude, and the expected ground shaking across the region. This method can provide warning before the arrival of secondary S-waves, which bring the strong shaking that can cause most of the damage.



Seismic Hazard and Risk analysis using Deep Learning



Group Lead

Jon Karapetyan

Group

Ani Gevorgyan,
Shahen Shahiyan,
Sevada Hovhannisyan,
Lilya Hovhannisyan

Collaborators

Li Li,
Yongzhe Wang,
Anatoly Soloviev,
Boris Dzeboev

Funded By

Mathematical Geophysics and Geoinformatics (MGG) Research Group
Institute of Geophysics and Engeneering Seismology


About

Earthquake prediction is currently the most crucial task required for the probability, hazard, risk mapping, and mitigation purposes. Earthquake prediction attracts the researchers' attention from both academia and industries. Traditionally, the risk assessment approaches have used various traditional and machine learning models. However, deep learning techniques have been rarely tested for earthquake probability mapping. Therefore, this study develops a convolutional neural network (CNN) model for earthquake probability assessment in Caucasus.


Seismic Tomography



Group Lead

Jon Karapetyan

Group

Ani Gevorgyan,
Shahen Shahiyan,
Sevada Hovhannisyan,
Lilya Hovhannisyan

Collaborators

Li Li,
Yongzhe Wang,
Anatoly Soloviev,
Boris Dzeboev

Funded By

Mathematical Geophysics and Geoinformatics (MGG) Research Group
Institute of Geophysics and Engeneering Seismology


About

Seismic tomography using transmission data is widely used to image the Earth's interior from global to local scales. In seismic imaging, it is used to obtain velocity models for subsequent depth-migration or full-waveform inversion. In addition, cross-hole tomography has been successfully applied for a variety of applications, including mineral exploration, reservoir monitoring, and CO2 injection and sequestration. Conventional tomography techniques suffer from a number of limitations, including the use of a smoothing regularizer that is agnostic to the physics of wave propagation. Here, we propose a novel tomography method to address these challenges using developments in the field of scientific machine learning. Using seismic traveltimes observed at seismic stations covering part of the computational model, we train neural networks to approximate the traveltime factor.



Artificial Intelligence and Machine Learning in Earth science



Group Lead

Jon Karapetyan

Group

Ani Gevorgyan,
Shahen Shahiyan,
Sevada Hovhannisyan,
Lilya Hovhannisyan

Collaborators

Li Li,
Yongzhe Wang,
Anatoly Soloviev,
Boris Dzeboev

Funded By

Mathematical Geophysics and Geoinformatics (MGG) Research Group
Institute of Geophysics and Engeneering Seismology


About

Earthquake prediction is of significance to risk assessment, prevention and safe design of major structures. However, it is typically challengeable to characterize earthquake response and unveil features from continu- ously detected, massive noisy data. To address those severe challenges in seismology, the AI techniques have been used as powerful statistical tools to address the data-related issues. Along the way, AI has demonstrated advantages for mass adoption, which is emerged into seismology to open a promising direction for AI-enhanced seismic analysis. This is in contrast to the traditional approaches dominated in the field of earthquake until now. The exciting debut of ML and its robust branches such as DL in the last decade, combining with the advents of cluster computing environment and more powerful personal computers, immediately offers a potential solution to the fields requested to address massive seismology data. Aiming at developing seismic prediction models that identify seismic response from noisy data (i.e., effective seismic data) while reveal unseen patterns and features from detected seismic data (i.e., undetected earth- quake), preliminary efforts have been dedicated to deploying DL to earthquake analysis. Here, we move a step forward to envision the future development trends of the DL-enhanced seismology in IoT platform. Other than the fact that DL seismic analysis is still in its infancy, IoT has just reached the peak at the Gartner’s hype cycle. Integrating the cutting-edge technologies of DL and IoT techniques and applying them to seismic data can lead us to the great-leap-forward development of seismology.