SDO Machine Learning Dataset - Solar Dynamics Observatory

Overview

The SDO Machine Learning Dataset is a curated collection of solar observations from NASA’s Solar Dynamics Observatory (SDO) formatted specifically for machine learning applications. It targets both heliophysicists and ML researchers who want to build models for solar activity and space weather.

Instruments Included

• AIA (Atmospheric Imaging Assembly): Multi-wavelength EUV images of the solar corona.
• HMI (Helioseismic and Magnetic Imager): Vector magnetic field maps and Doppler measurements.
• EVE (Extreme Ultraviolet Variability Experiment): Solar irradiance spectra in the EUV range.

Machine Learning-Ready Design

• Standardized spatial and temporal alignment across instruments.
• Preprocessed data subsets for flare prediction and active-region classification.
• Metadata describing solar events, classes, and time ranges.

Use Cases

• Forecasting solar flares and other eruptive events.
• Studying magnetic field evolution in active regions.
• Developing ML models for space weather applications that impact satellites and power grids.

Access

The dataset is hosted in the AWS Open Data Registry for efficient access via cloud computing, enabling large-scale training without needing to download all raw SDO archives.