01 / NASA Space Apps Challenge

Solar Predictor

NASA Space Apps Challenge2023Global Nominee

Solar storms driven by coronal mass ejections can disrupt power grids, satellites, and communications. Predicting their severity requires understanding the interplanetary magnetic field's Bz component — specifically when it flips southward and triggers magnetic reconnection at Earth's magnetopause.

My Role

  • Built a custom graph reader pipeline to extract Bz time-series data from NOAA imagery
  • Implemented the Sweet-Parker magnetic reconnection model for severity classification
  • Developed the 8-day positive-to-negative Bz flip detection algorithm
  • Created the full-stack web application with real-time forecasting

Approach

Rather than using black-box ML, we modeled the physics directly. The pipeline reads Bz graphs, detects the critical southward flip pattern, and applies the Sweet-Parker reconnection rate equation to estimate storm severity. This physics-first approach gave us interpretable predictions grounded in magnetohydrodynamics.

PythonTensorFlowFlaskOpenCVNumPy

Graph Reader Pipeline

-40-2002040Day 0Day 5Day 10Day 15Day 20Day 25Day 30Bz (nT)

Click on the graph to extract data points. Extract points on both sides of a zero-crossing to detect the Bz flip.

Magnetic Reconnection — Sweet-Parker Model

Links

Next Project

Safe CrossingSafe Crossing