☀️ Project Helios — Space Weather Prediction Engine

A real-time space weather prediction engine that uses Gaussian Process Regression to forecast geomagnetic storm severity (Kp index) from NASA DONKI coronal mass ejection data, with an interactive dashboard showing Earth impact predictions.

Architecture

┌─────────────────────────────────────────────────┐
│              DATA INGESTION LAYER               │
│  NASA DONKI (CME, GST, FLR) + NOAA OMNI        │
└──────────────────────┬──────────────────────────┘
                       │ ETL + Feature Engineering
┌──────────────────────▼──────────────────────────┐
│            BAYESIAN ML PIPELINE                 │
│  GP Classifier (ADF) → Storm/Quiet              │
│  GP Regressor → Kp with uncertainty bands       │
│  CME Arrival Estimator → Transit time           │
│  Impact Mapper → GPS, aurora, satellites        │
└──────────────────────┬──────────────────────────┘
                       │ JSON API
┌──────────────────────▼──────────────────────────┐
│          API + REAL-TIME LAYER                  │
│  FastAPI + Redis Cache + PostgreSQL + WebSocket  │
└──────────────────────┬──────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────┐
│           FRONTEND (Next.js + TS)               │
│  Live CME Tracker (Three.js)                    │
│  Kp Prediction with GP Posterior Bands          │
│  Aurora Visibility Map (D3 Azimuthal)           │
│  Earth Impact Dashboard                         │
│  Interactive GP Explainer                       │
└─────────────────────────────────────────────────┘

Tech Stack

Backend

• Python 3.11+ with FastAPI
• GPyTorch + PyTorch for Gaussian Process models
• scikit-learn for preprocessing and baselines
• PostgreSQL for historical event storage
• Redis for DONKI API response caching
• APScheduler for periodic data ingestion
• WebSockets for real-time prediction streaming

Frontend

• Next.js 14 with TypeScript
• Three.js / React Three Fiber for 3D solar system viz
• D3.js for aurora visibility map
• Recharts for GP posterior visualization
• Tailwind CSS for styling
• Framer Motion for animations

Deployment

• Vercel (frontend)
• Railway or Render (backend + PostgreSQL + Redis)

Key Features

1. Real-time CME tracking — 3D visualization of CME propagation from Sun to Earth
2. Probabilistic Kp prediction — GP posterior mean + 95% credible intervals
3. Storm classification — GP classifier with ADF (storm vs quiet)
4. Earth impact assessment — GPS degradation, HF radio blackout zones, satellite risk
5. Aurora forecast map — Kp-to-latitude aurora oval on azimuthal projection
6. Interactive GP explainer — Drag points, change kernels, watch posteriors reshape
7. Historical model validation — Compare GP predictions vs NASA's WSA-ENLIL

Data Sources

Source	Endpoint	Data
DONKI CME	/DONKI/CME	CME speed, half-angle, source location
DONKI CME Analysis	/DONKI/CMEAnalysis	WSA-ENLIL simulation results
DONKI GST	/DONKI/GST	Geomagnetic storm Kp values
DONKI FLR	/DONKI/FLR	Solar flare class, X-ray flux
DONKI IPS	/DONKI/IPS	Interplanetary shock arrivals
NOAA OMNI	omniweb	Solar wind Bz, density, velocity

Scientific References

• Chakraborty & Morley (2020). "Probabilistic prediction of geomagnetic storms and the Kp index." J. Space Weather Space Clim.
• Rasmussen & Williams (2006). Gaussian Processes for Machine Learning.
• Newell et al. (2007). "A nearly universal solar wind-magnetosphere coupling function."

Setup

# Backend
cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env  # Add your NASA API key
python -m app.main

# Frontend
cd frontend
npm install
npm run dev

Environment Variables

NASA_API_KEY=your_key_here
DATABASE_URL=postgresql://...
REDIS_URL=redis://localhost:6379

License

MIT