FDL-X Heliolab 2024: Spectral Irradiance of the 3D Sun (SPI3S) on Mars

This project estimates the Sun's full EUV spectral irradiance at arbitrary heliospheric viewpoints — including Mars — by coupling two models:

1. SuNeRF-DT — a Neural Radiance Field adapted to solar physics. Reconstructs a time-evolving 3D model of the corona (electron density, temperature) from multi-viewpoint EUV observations (SDO/AIA + STEREO/EUVI). Synthesizes AIA-like images from any vantage point, including Mars.
2. MEGS-AI (KANDEM-Spectrum) — maps a stack of 7 AIA EUV images to the full MEGS-A/MEGS-B spectral irradiance, acting as a "Virtual EVE" that works on real or SuNeRF-synthesized imagery.

End-to-end: multi-viewpoint solar images → SuNeRF-DT → 3D Sun → virtual Mars viewpoint → MEGS-AI → spectral irradiance at Mars, validated against MAVEN/EUVM.

This is the umbrella repository. The two model codebases are git submodules:

.
├── spi3s/                  # integration glue, EVE data loaders, showcase renders
├── sunerf/                 # submodule → 2024-HL-SPI3S-SuNeRF
├── megsai/                 # submodule → 2024-HL-SPI3S-MEGS-AI
├── notebooks/              # MAVEN download, EVE download, prototyping
└── scripts/                # release packaging (create_subset.py)

Clone with submodules:

git clone --recurse-submodules https://github.com/FrontierDevelopmentLab/2024-HL-SPI3S.git

Note: the submodule pins point to feature branches (sunerf → spie3s_integration, megsai → dem) rather than each submodule's main. This is intentional and reflects the FDL24-canonical state.

A public release of trained model checkpoints and processed training data is available on AWS S3:

• Models bucket: s3://nasa-radiant-data/helioai-datasets/hl-spect/models/
• Dataset bucket: s3://nasa-radiant-data/helioai-datasets/hl-spect/processed_data/

Start with the Colab inference notebook (source in public/inference_demo.ipynb) for a self-contained demo that downloads a checkpoint and a small data subset, then predicts the EUV spectrum on both Earth-view and Mars-view AIA stacks.

• 2024-HL-SPI3S-SuNeRF — SuNeRF / SuNeRF-DT implementation
• 2024-HL-SPI3S-MEGS-AI — MEGS-AI irradiance model

This work is the research product of FDL-X Heliolab a public/private partnership between NASA, Trillium Technologies Inc (trillium.tech) and commercial AI partners Google Cloud, NVIDIA and Pasteur Labs & ISI, developing open science for all Humankind.