nasa-ibm-ai4science/surya-bench-coronal-extrapolation

Coronal Field Extrapolation Dataset

Dataset Summary

This dataset contains spherical harmonic coefficients of the coronal magnetic potential generated by emulating the physics-based ADAPT-WSA PFSS (Potential Field Source Surface) code, driven by SDO/HMI solar magnetogram observations. The target spherical harmonic coefficients represent the magnetic potential between the photosphere and the source surface (set to 2.51 Rs).
Each file also contains additional variables from the WSA model output, such as:

• Photospheric polarity maps
• Open/closed field line masks

Because ADAPT-WSA is an ensemble model, data is included for all 12 realizations (R000 – R011) in the filenames. Spherical harmonic data are stored in FITS HDU index 3, formatted as a (2, 91, 91) array. Each (91, 91) array is triangular, with unused elements left as zero. The dataset covers May 2010 to December 2024 and includes daily simulations, with splits designed for consistent training, validation, and testing across years.

Supported Tasks and Applications

• Surrogate modeling: Training ML models to emulate ADAPT-WSA PFSS outputs
• Coronal field reconstruction: Predicting global magnetic field structure from photospheric inputs
• Space weather forecasting: Estimating open/closed field lines and solar wind connectivity
• Physics emulation: Foundation models replacing computationally expensive PFSS runs

Data Structure

Data Files

• train.csv: Instances from Feb 15 to Dec 31 in each year between 2010–2019
• validation.csv: Instances from Jan 15–28 of each year between 2010–2019
• test.csv: All instances from each year between 2020–2024
• leaky_validation.csv: Instances from Jan 1–14 and Jan 29–Feb 11 of each year between 2010–2019

Compressed simulation model outputs containing spherical harmonics are available in data.tar.gz. Individual files are stored in FITS format. There are 12 ensemble realizations per one timestamp. Each file includes:

• Spherical harmonic coefficients (HDU index 3)
• Photospheric polarity information
• Open/closed field masks
• The paths to realizations (R000–R011) are provided in above mentioned csv files.

Features

• Spherical Harmonics: (2, 91, 91) triangular arrays
  • Order: up to 90
  • Normalization: Schmidt

Dataset Details

Field	Description
Temporal Coverage	May 13, 2010 – Dec 31, 2024
Data Format	FITS
Data Shape	(2, 91, 91) per instance
Data Size	5,347 instances * 12 realizations
Cadence	1 day
Total File Size	~36.2 GB (compressed)
Size per file	~763 KB

Example Code (Python)

from astropy.io import fits
import numpy as np
import torch

fits_file = fits.open("wsa_201005010800R000_ahmi.fits")
sph_data = fits_file[3].data.copy()
fits_file.close()

coefficients = torch.from_numpy(
    np.array(
        [
            sph_data[0, :, :][np.triu_indices(sph_data.shape[1])],
            sph_data[1, :, :][np.triu_indices(sph_data.shape[1])],
        ]
    )
)

Contact

Daniel da Silva, [email protected]