Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model

Harshita Guduru; Darshan Joshi; Wisam Bukaita

doi:doi:10.11648/j.ajce.20251305.14

Research Article |

| Peer-Reviewed

Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model

Harshita Guduru^*

, Darshan Joshi

, Wisam Bukaita

Published in American Journal of Civil Engineering (Volume 13, Issue 5)

Received: 24 September 2025 Accepted: 7 October 2025 Published: 30 October 2025

Views: Downloads:

Download PDF

Share This Article

Twitter
Linked In
Facebook

Abstract

This study addresses the critical challenge of earthquake prediction and synthetic seismogram generation through the application of deep learning. A hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) framework is proposed to capture both spatial and temporal characteristics of seismic data, enabling robust forecasting of seismic activity and the generation of realistic waveform simulations. Historical seismographic records and metadata, including magnitude, depth, and epicentral location, were sourced from the United States Geological Survey (USGS). Key predictive features such as amplitude variation, temporal intervals, epicentral distances, and regional spatial attributes were extracted to train and validate the model. Model development and experimentation were conducted in Python using TensorFlow, Keras, NumPy, Pandas, Scikit-learn, and Imbalanced-learn. The CNN component was employed to extract spatial representations from seismograms, while the LSTM component modeled sequential dependencies inherent in seismic waveforms. The final model achieved an accuracy of 84%, with notable improvements across precision, recall, and loss metrics. Statistical evaluations further validated the reliability of the results. The findings demonstrate the potential of hybrid deep learning architectures to enhance early earthquake warning systems and hazard assessment strategies. By integrating prediction with synthetic seismogram generation, this research advances data-driven seismology and provides a scalable foundation for future applications in disaster preparedness and risk mitigation.

Published in	American Journal of Civil Engineering (Volume 13, Issue 5)
DOI	10.11648/j.ajce.20251305.14
Page(s)	284-303
Creative Commons	This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited.
Copyright	Copyright © The Author(s), 2025. Published by Science Publishing Group

Keywords

Earthquake Prediction, Seismic Forecasting, Synthetic Seismogram Generation, Seismology, Seismic Signal Processing, Spatio-Temporal Modeling

References

[1]	Zechar, J. Douglas, and Thomas H. Jordan. “Testing Alarm-Based Earthquake Predictions.” Geophysical Journal International 172, no. 2 (2008): 715–24. https://doi.org/10.1111/j.1365-246X.2007.03676.x
[2]	Zhu Lijun, Zhigang Peng, James McClellan, Chenyu Li, Dongdong Yao, Zefeng Li, and Lihua Fang. “Deep Learning for Seismic Phase Detection and Picking in the Aftershock Zone of 2008 7.9 Wenchuan Earthquake.” Physics of the Earth and Planetary Interiors 293 (2019). https://doi.org/10.1016/j.pepi.2019.05.004
[3]	Tapia-Hernández, Edgar, Elizabeth A. Reddy, and Laura J. Oros-Avilés. “Earthquake Predictions and Scientific Forecast: Dangers and Opportunities for a Technical and Anthropological Perspective.” Earth Sciences Research Journal 23, no. 4 (2019): 309-15. https://doi.org/10.15446/esrj.v23n4.77206
[4]	Ratiranjan Jena, Biswajeet Pradhan, Abdullah Al-Amri, Chang Wook Lee, and Hyuck-jin Park. “Earthquake Probability Assessment for the Indian Subcontinent Using Deep Learning.” Sensors 20, no. 16 (2020): 4369. https://doi.org/10.3390/s20164369
[5]	Jain Rachna, Nayyar Anand, Arora Simrann, and Akash Gupta. “A Comprehensive Analysis and Prediction of Earthquake Magnitude Based on Position and Depth Parameters Using Machine and Deep Learning Models.” Multimedia Tools and Applications 80, no. 18 (2021): 28419-38. https://doi.org/10.1007/s11042-021-11001-z
[6]	Bhardwaj, A., L. Sam, and F. J. Martin-Torres. “The Challenges and Possibilities of Earthquake Predictions Using Non-Seismic Precursors.” The European Physical Journal Special Topics 230, no. 1 (2021): 367-80. https://doi.org/10.1140/epjst/e2020-000257-3
[7]	Meier, Men-Andrin, Jean-Paul Ampuero, Elizabeth Cochran, and Morgan Page. “Apparent Earthquake Rupture Predictability.” Geophysical Journal International 225, no. 1 (2021): 657-63. https://doi.org/10.1093/gji/ggaa610
[8]	Wang, Yanwei, Xiaojun Li, Zifa Wang, and Juan Liu. “Deep Learning for Magnitude Prediction in Earthquake Early Warning.” Gondwana Research 123 (2023): 164-73. https://doi.org/10.1016/j.gr.2022.06.009
[9]	Zhu, Jingbao, Shanyou Li, Qiang Ma, and Jindong Song. “Onsite Intensity Prediction for Earthquake Early Warning with Multimodal Deep Learning.” Soil Dynamics and Earthquake Engineering 195 (2025). https://doi.org/10.1016/j.soildyn.2025.109430
[10]	Kaushal, Arush, Ashok Kumar Gupta, and Vivek Kumar Sehgal. “Earthquake Prediction Optimization Using Deep Learning Hybrid RNN-LSTM Model for Seismicity Analysis.” Soil Dynamics and Earthquake Engineering 195 (2025). https://doi.org/10.1016/j.soildyn.2025.109432
[11]	Gao, Zhaoqi, Sichao Hu, Chuang Li, Hongling Chen, Xiudi Jiang, Zhibin Pan, Jinghuai Gao, and Zongben Xu. “A Deep-Learning-Based Generalized Convolutional Model for Seismic Data and Its Application in Seismic Deconvolution.” IEEE Transactions on Geoscience and Remote Sensing 60 (2022). https://doi.org/10.1109/TGRS.2021.3076991
[12]	Elsayed, Hagar S., Omar M. Saad, M. Sami Soliman, Yangkang Chen, and Hassan A. Youness. “Attention-Based Fully Convolutional DenseNet for Earthquake Detection.” IEEE Transactions on Geoscience and Remote Sensing 60 (2022). https://doi.org/10.1109/TGRS.2022.3194196
[13]	Jiang, Wenbin, Chengpeng Xi, Wenchuang Wang, and Youyi Ruan. “Time Window Selection of Seismic Signals for Waveform Inversion Based on Deep Learning.” IEEE Transactions on Geoscience and Remote Sensing 60 (2022). https://doi.org/10.1109/TGRS.2021.3076531
[14]	Irwandi, Irwandi, and Yunita Indris. “Realistic Shakemap M6.5 Pidie Jaya Earthquake 7 December 2016 Based on Modal Summation Technique.” IOP Conference Series. Earth and Environmental Science 318, no. 1 (2019). https://doi.org/10.1088/1755-1315/318/1/012005
[15]	REDDY, RAMAKRUSHNA, and RAJESH R NAIR. “The Efficacy of Support Vector Machines (SVM) in Robust Determination of Earthquake Early Warning Magnitudes in Central Japan.” Journal of Earth System Science 122, no. 5 (2013): 1423-34. https://doi.org/10.1007/s12040-013-0346-3
[16]	Zhang Jie, Zhu Huiyu, Siwei, and Ma Jianwei. “Constructing the Seismograms of Future Earthquakes in Yunnan, China, Using Compressed Sensing.” Seismological Research Letters, 2020. https://doi.org/10.1785/0220190382

Cite This Article

Plain Text BibTeX RIS

APA Style

Guduru, H., Joshi, D., Bukaita, W. (2025). Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model. American Journal of Civil Engineering, 13(5), 284-303. https://doi.org/10.11648/j.ajce.20251305.14

Copy | Download

ACS Style

Guduru, H.; Joshi, D.; Bukaita, W. Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model. Am. J. Civ. Eng. 2025, 13(5), 284-303. doi: 10.11648/j.ajce.20251305.14

Copy | Download

AMA Style

Guduru H, Joshi D, Bukaita W. Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model. Am J Civ Eng. 2025;13(5):284-303. doi: 10.11648/j.ajce.20251305.14

Copy | Download

@article{10.11648/j.ajce.20251305.14,
  author = {Harshita Guduru and Darshan Joshi and Wisam Bukaita},
  title = {Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model
},
  journal = {American Journal of Civil Engineering},
  volume = {13},
  number = {5},
  pages = {284-303},
  doi = {10.11648/j.ajce.20251305.14},
  url = {https://doi.org/10.11648/j.ajce.20251305.14},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ajce.20251305.14},
  abstract = {This study addresses the critical challenge of earthquake prediction and synthetic seismogram generation through the application of deep learning. A hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) framework is proposed to capture both spatial and temporal characteristics of seismic data, enabling robust forecasting of seismic activity and the generation of realistic waveform simulations. Historical seismographic records and metadata, including magnitude, depth, and epicentral location, were sourced from the United States Geological Survey (USGS). Key predictive features such as amplitude variation, temporal intervals, epicentral distances, and regional spatial attributes were extracted to train and validate the model. Model development and experimentation were conducted in Python using TensorFlow, Keras, NumPy, Pandas, Scikit-learn, and Imbalanced-learn. The CNN component was employed to extract spatial representations from seismograms, while the LSTM component modeled sequential dependencies inherent in seismic waveforms. The final model achieved an accuracy of 84%, with notable improvements across precision, recall, and loss metrics. Statistical evaluations further validated the reliability of the results. The findings demonstrate the potential of hybrid deep learning architectures to enhance early earthquake warning systems and hazard assessment strategies. By integrating prediction with synthetic seismogram generation, this research advances data-driven seismology and provides a scalable foundation for future applications in disaster preparedness and risk mitigation.
},
 year = {2025}
}

Copy | Download

TY  - JOUR
T1  - Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model

AU  - Harshita Guduru
AU  - Darshan Joshi
AU  - Wisam Bukaita
Y1  - 2025/10/30
PY  - 2025
N1  - https://doi.org/10.11648/j.ajce.20251305.14
DO  - 10.11648/j.ajce.20251305.14
T2  - American Journal of Civil Engineering
JF  - American Journal of Civil Engineering
JO  - American Journal of Civil Engineering
SP  - 284
EP  - 303
PB  - Science Publishing Group
SN  - 2330-8737
UR  - https://doi.org/10.11648/j.ajce.20251305.14
AB  - This study addresses the critical challenge of earthquake prediction and synthetic seismogram generation through the application of deep learning. A hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) framework is proposed to capture both spatial and temporal characteristics of seismic data, enabling robust forecasting of seismic activity and the generation of realistic waveform simulations. Historical seismographic records and metadata, including magnitude, depth, and epicentral location, were sourced from the United States Geological Survey (USGS). Key predictive features such as amplitude variation, temporal intervals, epicentral distances, and regional spatial attributes were extracted to train and validate the model. Model development and experimentation were conducted in Python using TensorFlow, Keras, NumPy, Pandas, Scikit-learn, and Imbalanced-learn. The CNN component was employed to extract spatial representations from seismograms, while the LSTM component modeled sequential dependencies inherent in seismic waveforms. The final model achieved an accuracy of 84%, with notable improvements across precision, recall, and loss metrics. Statistical evaluations further validated the reliability of the results. The findings demonstrate the potential of hybrid deep learning architectures to enhance early earthquake warning systems and hazard assessment strategies. By integrating prediction with synthetic seismogram generation, this research advances data-driven seismology and provides a scalable foundation for future applications in disaster preparedness and risk mitigation.

VL  - 13
IS  - 5
ER  -

Copy | Download

Author Information

Harshita Guduru

Math and Computer Sciences, Lawrence Technological University, Southfield, USA

Contact Email

http://orcid.org/0009-0003-4149-0472
Darshan Joshi

Math and Computer Sciences, Lawrence Technological University, Southfield, USA

Contact Email

http://orcid.org/0009-0000-7652-9240
Wisam Bukaita

Math and Computer Sciences, Lawrence Technological University, Southfield, USA

Contact Email

http://orcid.org/0000-0001-6255-3848

Download PDF

Submit an Article

Sections

Plain Text BibTeX RIS

APA Style

Guduru, H., Joshi, D., Bukaita, W. (2025). Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model. American Journal of Civil Engineering, 13(5), 284-303. https://doi.org/10.11648/j.ajce.20251305.14

Copy | Download

ACS Style

Guduru, H.; Joshi, D.; Bukaita, W. Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model. Am. J. Civ. Eng. 2025, 13(5), 284-303. doi: 10.11648/j.ajce.20251305.14

Copy | Download

AMA Style

Guduru H, Joshi D, Bukaita W. Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model. Am J Civ Eng. 2025;13(5):284-303. doi: 10.11648/j.ajce.20251305.14

Copy | Download

@article{10.11648/j.ajce.20251305.14,
  author = {Harshita Guduru and Darshan Joshi and Wisam Bukaita},
  title = {Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model
},
  journal = {American Journal of Civil Engineering},
  volume = {13},
  number = {5},
  pages = {284-303},
  doi = {10.11648/j.ajce.20251305.14},
  url = {https://doi.org/10.11648/j.ajce.20251305.14},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ajce.20251305.14},
  abstract = {This study addresses the critical challenge of earthquake prediction and synthetic seismogram generation through the application of deep learning. A hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) framework is proposed to capture both spatial and temporal characteristics of seismic data, enabling robust forecasting of seismic activity and the generation of realistic waveform simulations. Historical seismographic records and metadata, including magnitude, depth, and epicentral location, were sourced from the United States Geological Survey (USGS). Key predictive features such as amplitude variation, temporal intervals, epicentral distances, and regional spatial attributes were extracted to train and validate the model. Model development and experimentation were conducted in Python using TensorFlow, Keras, NumPy, Pandas, Scikit-learn, and Imbalanced-learn. The CNN component was employed to extract spatial representations from seismograms, while the LSTM component modeled sequential dependencies inherent in seismic waveforms. The final model achieved an accuracy of 84%, with notable improvements across precision, recall, and loss metrics. Statistical evaluations further validated the reliability of the results. The findings demonstrate the potential of hybrid deep learning architectures to enhance early earthquake warning systems and hazard assessment strategies. By integrating prediction with synthetic seismogram generation, this research advances data-driven seismology and provides a scalable foundation for future applications in disaster preparedness and risk mitigation.
},
 year = {2025}
}

Copy | Download

TY  - JOUR
T1  - Earthquake Prediction and Synthetic Seismogram Generation Using Hybrid CNN – LSTM Model

AU  - Harshita Guduru
AU  - Darshan Joshi
AU  - Wisam Bukaita
Y1  - 2025/10/30
PY  - 2025
N1  - https://doi.org/10.11648/j.ajce.20251305.14
DO  - 10.11648/j.ajce.20251305.14
T2  - American Journal of Civil Engineering
JF  - American Journal of Civil Engineering
JO  - American Journal of Civil Engineering
SP  - 284
EP  - 303
PB  - Science Publishing Group
SN  - 2330-8737
UR  - https://doi.org/10.11648/j.ajce.20251305.14
AB  - This study addresses the critical challenge of earthquake prediction and synthetic seismogram generation through the application of deep learning. A hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) framework is proposed to capture both spatial and temporal characteristics of seismic data, enabling robust forecasting of seismic activity and the generation of realistic waveform simulations. Historical seismographic records and metadata, including magnitude, depth, and epicentral location, were sourced from the United States Geological Survey (USGS). Key predictive features such as amplitude variation, temporal intervals, epicentral distances, and regional spatial attributes were extracted to train and validate the model. Model development and experimentation were conducted in Python using TensorFlow, Keras, NumPy, Pandas, Scikit-learn, and Imbalanced-learn. The CNN component was employed to extract spatial representations from seismograms, while the LSTM component modeled sequential dependencies inherent in seismic waveforms. The final model achieved an accuracy of 84%, with notable improvements across precision, recall, and loss metrics. Statistical evaluations further validated the reliability of the results. The findings demonstrate the potential of hybrid deep learning architectures to enhance early earthquake warning systems and hazard assessment strategies. By integrating prediction with synthetic seismogram generation, this research advances data-driven seismology and provides a scalable foundation for future applications in disaster preparedness and risk mitigation.

VL  - 13
IS  - 5
ER  -

Copy | Download