Predicting peak ground acceleration using the ConvMixer network

IF 1.2 4区地球科学 Q3 Earth and Planetary Sciences

Earthquake Science Pub Date : 2025-02-27 DOI:10.1016/j.eqs.2024.11.005

Mona Mohammed , Omar M. Saad , Arabi Keshk , Hatem M. Ahmed

{"title":"Predicting peak ground acceleration using the ConvMixer network","authors":"Mona Mohammed , Omar M. Saad , Arabi Keshk , Hatem M. Ahmed","doi":"10.1016/j.eqs.2024.11.005","DOIUrl":null,"url":null,"abstract":"<div><div>The level of ground shaking, as determined by the peak ground acceleration (PGA), can be used to analyze seismic hazard at a certain location and is crucial for constructing earthquake-resistant structures. Predicting the PGA immediately after an earthquake occurs allows for the issuing of a warning by an earthquake early warning system. In this study, we propose a deep learning model, ConvMixer, to predict the PGA recorded by weak-motion velocity seismometers in Japan. We use 5-s three-component seismograms, from 2 s before until 3 s after the P-wave arrival time of the earthquake. Our dataset comprised more than 50,000 single-station waveforms recorded by 10 seismic stations in the K-NET, Kiki-NET, and Hi-Net networks between 2004 and 2023. The proposed ConvMixer is a patch-based model that extracts global features from input seismic data and predicts the PGA of an earthquake by combining depth and pointwise convolutions. The proposed ConvMixer network had a mean absolute error of 2.143 when applied to the test set and outperformed benchmark deep learning models. In addition, the proposed ConvMixer demonstrated the ability to predict the PGA at the corresponding station site based on 1-second waveforms obtained immediately after the arrival time of the P-wave.</div></div>","PeriodicalId":46333,"journal":{"name":"Earthquake Science","volume":"38 2","pages":"Pages 126-135"},"PeriodicalIF":1.2000,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Earthquake Science","FirstCategoryId":"89","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1674451924001150","RegionNum":4,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"Earth and Planetary Sciences","Score":null,"Total":0}

引用次数: 0

Abstract

The level of ground shaking, as determined by the peak ground acceleration (PGA), can be used to analyze seismic hazard at a certain location and is crucial for constructing earthquake-resistant structures. Predicting the PGA immediately after an earthquake occurs allows for the issuing of a warning by an earthquake early warning system. In this study, we propose a deep learning model, ConvMixer, to predict the PGA recorded by weak-motion velocity seismometers in Japan. We use 5-s three-component seismograms, from 2 s before until 3 s after the P-wave arrival time of the earthquake. Our dataset comprised more than 50,000 single-station waveforms recorded by 10 seismic stations in the K-NET, Kiki-NET, and Hi-Net networks between 2004 and 2023. The proposed ConvMixer is a patch-based model that extracts global features from input seismic data and predicts the PGA of an earthquake by combining depth and pointwise convolutions. The proposed ConvMixer network had a mean absolute error of 2.143 when applied to the test set and outperformed benchmark deep learning models. In addition, the proposed ConvMixer demonstrated the ability to predict the PGA at the corresponding station site based on 1-second waveforms obtained immediately after the arrival time of the P-wave.

查看原文本刊更多论文

求助全文

约1分钟内获得全文求助全文

来源期刊

Earthquake Science GEOCHEMISTRY & GEOPHYSICS-

CiteScore

1.10

自引率

8.30%

发文量

审稿时长

3 months

期刊介绍： Earthquake Science (EQS) aims to publish high-quality, original, peer-reviewed articles on earthquake-related research subjects. It is an English international journal sponsored by the Seismological Society of China and the Institute of Geophysics, China Earthquake Administration. The topics include, but not limited to, the following ● Seismic sources of all kinds. ● Earth structure at all scales. ● Seismotectonics. ● New methods and theoretical seismology. ● Strong ground motion. ● Seismic phenomena of all kinds. ● Seismic hazards, earthquake forecasting and prediction. ● Seismic instrumentation. ● Significant recent or past seismic events. ● Documentation of recent seismic events or important observations. ● Descriptions of field deployments, new methods, and available software tools. The types of manuscripts include the following. There is no length requirement, except for the Short Notes. 【Articles】 Original contributions that have not been published elsewhere. 【Short Notes】 Short papers of recent events or topics that warrant rapid peer reviews and publications. Limited to 4 publication pages. 【Rapid Communications】 Significant contributions that warrant rapid peer reviews and publications. 【Review Articles】Review articles are by invitation only. Please contact the editorial office and editors for possible proposals. 【Toolboxes】 Descriptions of novel numerical methods and associated computer codes. 【Data Products】 Documentation of datasets of various kinds that are interested to the community and available for open access (field data, processed data, synthetic data, or models). 【Opinions】Views on important topics and future directions in earthquake science. 【Comments and Replies】Commentaries on a recently published EQS paper is welcome. The authors of the paper commented will be invited to reply. Both the Comment and the Reply are subject to peer review.