SuNeRF

Solar Neural Radiance Fields for 3D Tomography

SuNeRF is a Python framework for 3D tomographic reconstructions of the solar atmosphere and heliosphere using Neural Radiance Fields (NeRFs). It performs forward modeling of optically thin observations with explicit spacecraft geometry and image-formation physics.

SuNeRF supports:

• EUV emission tomography of the solar corona
• White-light CME tomography using Thomson scattering

---

Installation

conda create -n sunerf python=3.10 -y  
conda activate sunerf  
pip install -r requirements.txt  
pip install -e .

---

EUV Emission Tomography

Reconstructs a 3D EUV emissivity field from multi-viewpoint EUV observations (e.g., AIA + EUVI).

Data preparation

python -m sunerf.data.prep.sdo  
  --sdo_file_path "/path/to/aia/*.fits"  
  --output_path "workspace/data/euv/aia_193"  
  --resolution 512  

python -m sunerf.data.prep.stereo  
  --stereo_file_path "/path/to/euvi/*.fts"  
  --output_path "workspace/data/euv/euvi_195"  
  --resolution 512  

Training

python -m sunerf.run_emission  
  --config "config/emission/emission_2012_08-193.yaml"

---

CME Tomography (White-Light)

Reconstructs the 3D electron density of CMEs using a full Thomson-scattering forward model.

CME Challenge Dataset (HAO / PUNCH)

Dataset:
https://download.hao.ucar.edu/pub/punch/cme_challenge_v2/

Download and place the data locally, for example:

mkdir -p data/cme_challenge_v2

---

Data preparation

python -m sunerf.data.prep.prep_hao_cme  
  --resolution 512  
  --hao_path "data/cme_challenge_v2/**/*.fits"  
  --output_path "workspace/data/cme/prepared"  
  --check_matching  

---

Training / reconstruction

python -m sunerf.run_thomson  
  --config "config/cme/hao_2view.yaml"

---

Evaluation

python -m sunerf.evaluation.cme.evaluate_cme_parameters  
  --sunerf_path "workspace/outputs/cme/save_state.snf"  
  --data_path "data/cme_challenge_v2/**/*.sav"

---

Configuration

All experiments are configured via YAML files in:

config/  
  cme/
  emission/  

---