Seismic Event Classification in the Pacific Northwest

📌 Created by: @Akash Kharita (PhD Candidate, University of Washington)

This repository provides a framework for automated seismic event classification in the Pacific Northwest. We train multiple machine learning (ML) and deep learning (DL) models on a dataset spanning 2001–2023, containing 200K+ events across four classes:

1. Earthquakes
2. Explosions
3. Noise
4. Surface Events

More information about these events can be found in this paper

The primary objective is to evaluate and compare various ML and DL approaches to improve surface event classification while balancing accuracy, interpretability, and efficiency. The key differences between ML and DL techniques are illustrated below:

---

🚀 Installation

For cloud-based execution, refer to HPSBook.

1️⃣ Set up environment

sudo yum install -y git  
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh  
chmod +x Miniconda3-latest-Linux-x86_64.sh  
./Miniconda3-latest-Linux-x86_64.sh -b -p $HOME/miniconda  
./miniconda/bin/conda init bash  
bash  
sudo yum groupinstall -y "Development Tools"

2️⃣ Clone the repository & install dependencies

git clone https://github.com/Akashkharita/PNW_Seismic_Event_Classification.git  
cd PNW_Seismic_Event_Classification  
conda create -y -n surface python=3.9.5  
conda activate surface  
pip install -r requirements.txt  
pip install jupyter  
conda install -y ipykernel  
python -m ipykernel install --user --name=surface  

3️⃣ Launch Jupyter Notebook

• Cloud

  bash
  jupyter notebook --ip 0.0.0.0 --allow-root
  
  

• Local

  bash
  jupyter notebook
  

🛠️ Usage

📌 Classic Machine Learning

Feature extraction is a critical step in ML-based seismic classification. This repository includes various approaches:

• Physical Features: Notebook
• TSFEL Features: Notebook
• ScatNet Features: Notebook
• Combined Feature Sets( Physical + Tsfel): Notebook
• Algorithm Comparison: Notebook
• Sampling Rate & Window Size Study: Notebook
• Full ML Workflow for SCOPED: Notebook

📌 Feature Extraction Scripts

Scripts for extracting different types of features:

• Physical Features: Feature extraction and Feature calculation script

• TSFEL Features: Feature extraction

• ScatNet Features: Feature extraction

📌 Deep Learning

Deep learning models enable automatic feature extraction:

• Neural Network Architectures: Script
• Training & Validation: Notebook
• Testing on Common Dataset: Notebook

🤝 Contributing

Contributions to code, visualization, and methodology are welcome! Contact me at ak287@uw.edu for collaborations.

📜 License

This repository is licensed under the MIT License.