SuryaBench: Benchmark Dataset for Advancing Machine Learning in Heliophysics and Space Weather Prediction.

Area of Science:

• Heliophysics and Space Weather
• Solar Physics
• Machine Learning Applications

Background:

• Solar Dynamics Observatory (SDO) data is crucial for understanding solar phenomena.
• Existing datasets often require extensive preprocessing for machine learning (ML) tasks.
• Advancing space weather forecasting necessitates accessible, high-resolution solar data.

Purpose of the Study:

• Introduce a high-resolution, ML-ready heliophysics dataset from SDO.
• Facilitate ML applications in solar physics and space weather prediction.
• Provide benchmark datasets for key heliophysics and space weather tasks.

Main Methods:

• Utilized imagery from SDO's Atmospheric Imaging Assembly (AIA) and Helioseismic and Magnetic Imager (HMI).
• Processed data spanning a full solar cycle (May 2010 - December 2024).
• Applied preprocessing steps including angle correction, normalization, and degradation compensation.

Main Results:

• Developed a unified, standardized dataset suitable for ML.
• Created auxiliary benchmark datasets for tasks like active region segmentation and solar flare prediction.
• Ensured data quality through rigorous preprocessing for ML readiness.

Conclusions:

• The dataset accelerates the development of AI-driven models for space weather.
• Enhances reproducibility and benchmarking in solar physics research.
• Bridges the gap between solar physics, ML, and operational forecasting.