Feasibility of Deep Learning in Shear Wave Splitting analysis using Synthetic-Data Training and Waveform Deconvolution
Teleseismic shear-wave splitting analyses are often performed by reversing the splitting process through the application of frequency- or time-domain operations aimed at minimizing the transverse-component energy of waveforms. These operations yield two splitting parameters, ɸ (fast-axis orientation...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
McGill University
2024-03-01
|
| Series: | Seismica |
| Subjects: | |
| Online Access: | https://seismica.library.mcgill.ca/article/view/1124 |
| _version_ | 1797213908714389504 |
|---|---|
| author | Megha Chakraborty Georg Rümpker Wei Li Johannes Faber Nishtha Srivastava Frederik Link |
| author_facet | Megha Chakraborty Georg Rümpker Wei Li Johannes Faber Nishtha Srivastava Frederik Link |
| author_sort | Megha Chakraborty |
| collection | DOAJ |
| description | Teleseismic shear-wave splitting analyses are often performed by reversing the splitting process through the application of frequency- or time-domain operations aimed at minimizing the transverse-component energy of waveforms. These operations yield two splitting parameters, ɸ (fast-axis orientation) and δt (delay time). In this study, we investigate the applicability of a baseline recurrent neural network, SWSNet, for determining the splitting parameters from pre-selected waveform windows. Due to the scarcity of sufficiently labelled real waveform data, we generate our own synthetic dataset to train the model. The model is capable of determining ɸ and δt with a root mean squared error (RMSE) of 9.7° and 0.14 s on a noisy synthetic test data. The application to real data involves a deconvolution step to homogenize the waveforms. When applied to data from the USArray dataset, the results exhibit similar patterns to those found in previous studies with mean absolute differences of 9.6° and 0.16 s in the calculation of ɸ and δt respectively. |
| first_indexed | 2024-04-24T11:05:45Z |
| format | Article |
| id | doaj.art-7fd14808616f4b2bb0fbe9dd464bf7e2 |
| institution | Directory Open Access Journal |
| issn | 2816-9387 |
| language | English |
| last_indexed | 2024-04-24T11:05:45Z |
| publishDate | 2024-03-01 |
| publisher | McGill University |
| record_format | Article |
| series | Seismica |
| spelling | doaj.art-7fd14808616f4b2bb0fbe9dd464bf7e22024-04-11T16:55:18ZengMcGill UniversitySeismica2816-93872024-03-013110.26443/seismica.v3i1.11241118Feasibility of Deep Learning in Shear Wave Splitting analysis using Synthetic-Data Training and Waveform DeconvolutionMegha Chakrabortyhttps://orcid.org/0000-0001-7319-6137Georg Rümpkerhttps://orcid.org/0000-0002-5348-9888Wei Lihttps://orcid.org/0000-0003-1637-1386Johannes FaberNishtha Srivastavahttps://orcid.org/0000-0003-0328-0311Frederik Linkhttps://orcid.org/0000-0003-1639-0093Teleseismic shear-wave splitting analyses are often performed by reversing the splitting process through the application of frequency- or time-domain operations aimed at minimizing the transverse-component energy of waveforms. These operations yield two splitting parameters, ɸ (fast-axis orientation) and δt (delay time). In this study, we investigate the applicability of a baseline recurrent neural network, SWSNet, for determining the splitting parameters from pre-selected waveform windows. Due to the scarcity of sufficiently labelled real waveform data, we generate our own synthetic dataset to train the model. The model is capable of determining ɸ and δt with a root mean squared error (RMSE) of 9.7° and 0.14 s on a noisy synthetic test data. The application to real data involves a deconvolution step to homogenize the waveforms. When applied to data from the USArray dataset, the results exhibit similar patterns to those found in previous studies with mean absolute differences of 9.6° and 0.16 s in the calculation of ɸ and δt respectively.https://seismica.library.mcgill.ca/article/view/1124shear-wave splittingseismic anisotropywaveform analysis |
| spellingShingle | Megha Chakraborty Georg Rümpker Wei Li Johannes Faber Nishtha Srivastava Frederik Link Feasibility of Deep Learning in Shear Wave Splitting analysis using Synthetic-Data Training and Waveform Deconvolution Seismica shear-wave splitting seismic anisotropy waveform analysis |
| title | Feasibility of Deep Learning in Shear Wave Splitting analysis using Synthetic-Data Training and Waveform Deconvolution |
| title_full | Feasibility of Deep Learning in Shear Wave Splitting analysis using Synthetic-Data Training and Waveform Deconvolution |
| title_fullStr | Feasibility of Deep Learning in Shear Wave Splitting analysis using Synthetic-Data Training and Waveform Deconvolution |
| title_full_unstemmed | Feasibility of Deep Learning in Shear Wave Splitting analysis using Synthetic-Data Training and Waveform Deconvolution |
| title_short | Feasibility of Deep Learning in Shear Wave Splitting analysis using Synthetic-Data Training and Waveform Deconvolution |
| title_sort | feasibility of deep learning in shear wave splitting analysis using synthetic data training and waveform deconvolution |
| topic | shear-wave splitting seismic anisotropy waveform analysis |
| url | https://seismica.library.mcgill.ca/article/view/1124 |
| work_keys_str_mv | AT meghachakraborty feasibilityofdeeplearninginshearwavesplittinganalysisusingsyntheticdatatrainingandwaveformdeconvolution AT georgrumpker feasibilityofdeeplearninginshearwavesplittinganalysisusingsyntheticdatatrainingandwaveformdeconvolution AT weili feasibilityofdeeplearninginshearwavesplittinganalysisusingsyntheticdatatrainingandwaveformdeconvolution AT johannesfaber feasibilityofdeeplearninginshearwavesplittinganalysisusingsyntheticdatatrainingandwaveformdeconvolution AT nishthasrivastava feasibilityofdeeplearninginshearwavesplittinganalysisusingsyntheticdatatrainingandwaveformdeconvolution AT frederiklink feasibilityofdeeplearninginshearwavesplittinganalysisusingsyntheticdatatrainingandwaveformdeconvolution |