MMS Data Analysis Project Documentation

Project Overview

This project implements a machine learning pipeline for analyzing Magnetospheric Multiscale (MMS) spacecraft data using Gaussian Mixture Models (GMM). The pipeline processes ion spectrogram data to identify and classify different plasma regions in the magnetosphere.

Pipeline Components

1. Data Loading (read_cdf.py)

• Reads MMS spacecraft data from CDF (Common Data Format) files
• Extracts ion spectrogram data and timestamps
• Resamples the data to 1-minute intervals for analysis
• Uses spacepy.pycdf for handling CDF files

2. Feature Engineering (feature_engineering.py)

Processes ion spectrogram data to extract three key features:

Features:

• ratio_max_width: Measures peak characteristics in the energy spectrum
• ratio_high_low: Compares high-energy (>4000 eV) to low-energy (<100 eV) intensities
• norm_Bt: Normalizes the total magnetic field to 50 nT

Pseudofeature Implementation:

• Zeros all features if:
  • High-energy spectrum shows strong linear correlation (r > 0.7)
  • No valid peaks are detected
• Saves processed features to CSV format

3. Visualization (vis_v3.py)

Creates visual representations of the GMM clustering results:

• Loads processed features and trained GMM model
• Generates 2D scatter plots of the first two features
• Visualizes GMM clusters using:
  • Color-coded cluster assignments
  • Ellipses showing Gaussian distributions (2σ)
  • Clear labels and legends

Data Processing Flow

1. Raw CDF data → read_cdf.py
2. Resampled time series → feature_engineering.py
3. Extracted features → GMM model
4. Cluster assignments → Visualization

Dependencies

• Python 3.9
• spacepy
• numpy
• pandas
• scikit-learn
• matplotlib
• joblib

Project Structure

gaussian_mixed_model_mms/
├── data/
│   ├── processed/
│   └── raw/
├── models/
│   └── gmm_v1.py
├── scripts/
│   ├── feature_engineering.py
│   └── read_cdf.py
├── visualizations/
│   ├── vis_v1.py
│   ├── vis_v2.py
│   └── vis_v3.py
├── requirements.txt
├── Dockerfile
└── README.md

Setup and Installation

1. Clone the repository:

   git clone https://github.com/YOUR_USERNAME/gaussian_mixed_model_mms.git
   cd gaussian_mixed_model_mms

2. Using Docker:

   docker build -t gaussian_mixed_model_mms .
   docker run -v $(pwd)/data:/gaussian_mixed_model_mms/data gaussian_mixed_model_mms

Usage

1. Place raw CDF files in the data/raw directory
2. Run the scripts in order:

   python scripts/read_cdf.py
   python scripts/feature_engineering.py
   python visualization/vis_v3.py

Output

• Processed features are saved in data/processed/features.csv
• GMM model is saved in models/gmm_model.pkl
• Visualization plots show cluster assignments and Gaussian components

Notes

• The project is designed for MMS ion spectrogram analysis
• Features are carefully chosen to identify plasma regions
• Visualization helps in understanding the clustering results