Welcome to our project repository for the analysis of seismic tremor signals from the 2021 Geldingadalir eruption in Iceland. Using unsupervised machine learning, we aim to uncover hidden patterns in continuous seismic data to enhance understanding of volcanic activity.
In this project, we replicate and expand upon the findings from the paper Tremor clustering reveals pre-eruptive signals and evolution of the 2021 Geldingadalir eruption of the Fagradalsfjall Fires, Iceland. The paper demonstrates the use of machine learning to analyze tremor signals and detect pre-eruptive activity.
We apply Deep Embedded Clustering (DEC), an unsupervised machine learning technique, to seismic data collected during the eruption. DEC allows us to automatically detect various eruption phases by clustering the continuous tremor signals. Notably, this analysis led to the identification of a previously unknown eruption phase by the authors, highlighting the potential for unsupervised learning in volcanic monitoring.
- Understand and reproduce the results from our analysis of the 2021 Geldingadalir eruption.
- Expand the application of DEC to new seismic datasets from different volcanic events.
- Explore the potential for real-time monitoring of volcanic activity using similar techniques.
- Step through the code: We will carefully review the workflow to ensure full comprehension of the methods and results.
- Reproduce results: We will run the provided code to verify that the clustering and phase identification techniques produce consistent results.
- Apply to new data: Once verified, we will test DEC on new seismic datasets to uncover additional insights.
- Investigate real-time potential: Ultimately, we hope to explore the possibility of implementing these methods for real-time seismic event monitoring, using insights from the original paper.
These instructions are for a *nix system.
git clone https://github.com/UW-MLGEO/MLGEO2024_Geldingadalir
conda env create -f environment.yml
conda activate mlgeo
pip install -r requirements.txt.
├── ClusTremor-main # Original model
├── LICENSE # MIT License information
├── README.md
├── docs # Supporting documentation
│ └── test.txt
├── images
│ └── fagradalsfjall_eruption.png # Image of the eruption for README.md
├── notebooks
│ ├── AutoML_Hyperparameter_Tuning.ipynb # Hyperparameter Tuning notebook with CML - Alaska
│ ├── Clustering_Analysis.ipynb # Clustering Analysis notebook - Alaska
│ ├── Computational_Time_Analysis.ipynb #
│ ├── Data_Cleaning.ipynb # Data Cleaning notebook - Alaska
│ ├── Dimensionality_Reduction.ipynb # Dimensionality Reduction notebook - Alaska
│ ├── Download_Data.ipynb # Data Download notebook - Alaska
│ ├── EDA.ipynb # EDA notebook - Alaska
│ ├── Model_Training_Assessment.ipynb # Assessment on CML clustering - Alaska
│ ├── Prepare_AI_Ready_Data.ipynb # Data Preparation notebook - Alaska
│ └── data # Data - Alaska
│ ├── ai_ready
│ ├── clean
│ └── raw
│ └── HOPS_analysis # Data processing results from 9F.HOPS of the Iceland Reykjanes experiment 2021
│ └── NUPH_analysis # Data processing results from 9F.NUPH of the Iceland Reykjanes experiment 2021
│ └── HVUWB_analysis # Data processing results from HV.UWB of the Hawaiian Volcano Observatory's network
│ └── AVPVV_analysis/data # Data processing results from AV.PVV
| └── StHelens_analysis # Data processing results from UW.ELK
│ └── Classic_ML # Classic Machine Learning notebook for alternate analysis
│ includes CML results for select DEC stations
├── tests # Staging for future test modules
│ └── test_modules.py
└── Geldingadlir: [ˈkɛltiŋaˌdalɪr̥]
Fagradalsfjall: [ˈfaɣraˌtalsˌfjatl̥]