Solar Physics · nasa-ibm-ai4science
core-sdo
Multi-modal Solar Dynamics Observatory dataset for solar foundation models and space weather applications.
What core-sdo is
core-sdo is a machine learning-ready solar physics dataset created by nasa-ibm-ai4science, compiling multi-modal observations from NASA's Solar Dynamics Observatory (SDO). Spanning more than fourteen years from 13 May 2010 to 31 December 2024, the dataset aggregates full-disk observations across a wide range of solar atmospheric layers.
The dataset is curated to support large-scale machine learning in heliophysics, including foundation model pre-training, unsupervised representation learning, and operational space weather forecasting. It integrates high-cadence extreme ultraviolet imagery with photospheric vector magnetograms and Doppler velocity measurements into a single unified schema.
How it was made
The dataset is compiled from Level-1.5 processed observations taken by two SDO instruments: the Atmospheric Imaging Assembly (AIA) and the Helioseismic and Magnetic Imager (HMI). Every 12 minutes, observational channels are co-aligned and packaged into netCDF (.nc) files storing 32-bit floating-point arrays of dimension [13, 4096, 4096].
Each file contains 13 physical channels: eight AIA extreme ultraviolet and ultraviolet bands (94 Å, 131 Å, 171 Å, 193 Å, 211 Å, 304 Å, 335 Å, and 1600 Å) alongside five HMI variables consisting of line-of-sight magnetograms, vector magnetic field components in three dimensions (bx, by, and bz), and Doppler velocity measurements.
Key facts
- Temporal coverage
- 13 May 2010 to 31 December 2024
- Temporal granularity
- 12 minutes
- Data shape
- [13, 4096, 4096] per file
- Data format
- netCDF (.nc), float32
- Total channels
- 13 (8 AIA EUV/UV + 5 HMI magnetic and velocity)
- File size
- Approximately 570 MB per file
- Total dataset volume
- Approximately 360 TB
- DOI
- 10.57967/hf/7273
What you can use it for
- Solar flare forecasting
- Binary classification systems can be trained to predict the occurrence of solar flares within a 24-hour lead time.
- Solar wind speed prediction
- Regression models can predict ambient solar wind speeds with a forecast horizon of up to four days.
- Active region segmentation
- Computer vision models can perform pixel-wise semantic segmentation of active solar regions across multi-wavelength solar disk images.
- Coronal magnetic field extrapolation
- Models can reconstruct and extrapolate 3D magnetic field configurations from photospheric measurements up into the solar corona.
- Active region emergence forecasting
- Predictive models can forecast the emergence of new active regions based on historical magnetic field and velocity features.
Explore the data
The schema, real rows — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from nasa-ibm-ai4science/core-sdo at the moment you asked. Nothing is cached or stored — every row above came from that request.
Strengths and limits
Good at
- Combines 13 coordinated channels including multi-wavelength extreme ultraviolet imagery, vector magnetic fields, and Doppler velocity.
- Prepared in a consistent 4096 by 4096 float32 netCDF structure suited for large-scale vision and multi-modal neural networks.
Watch out for
- The dataset demands vast storage infrastructure, totalling approximately 360 TB of data.
- Individual files are large at approximately 570 MB each, requiring substantial memory and data transfer bandwidth.
Questions
Can I look inside core-sdo without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser. Nothing is downloaded to your machine.
What licence is core-sdo under?
Its authors declare cc-by-4.0. Read the terms on the original page before reusing the data.
Which instruments aboard the SDO provide data for core-sdo?
The dataset incorporates Level-1.5 observations from the Atmospheric Imaging Assembly (AIA) and the Helioseismic and Magnetic Imager (HMI).
Which solar atmospheric channels are captured by the AIA instrument?
The AIA instrument provides eight channels: seven extreme ultraviolet passbands (94 Å, 131 Å, 171 Å, 193 Å, 211 Å, 304 Å, and 335 Å) and one ultraviolet continuum channel (1600 Å).
What measurements are provided by the HMI instrument?
HMI provides line-of-sight magnetograms, three components of the vector magnetic field (bx, by, and bz), and Doppler velocity data.
What is the resolution and file format of the dataset?
Each file has a shape of [13, 4096, 4096] stored as float32 values within the netCDF (.nc) file format.
Written from core-sdo's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.