ionopy

Data-Driven Ionosphere Forecasting from sparse GNSS-derived data with Temporal-Fusion Transformers.

Pre-requisites

Install the dependencies mentioned in the environment.yml file, or directly build the environment from that file using conda:

conda env create -f environment.yml
conda activate ionopy

Training

To train a temporal-fusion-transformer model, run the following command:

cd scripts
python train.py

Use -- to pass additional arguments to the training script. For example, to specify a different batch size:

python train.py --batch_size 32

More info and how to cite:

If you would like to learn more about or cite the techniques Ionopy uses, please see the following paper:

• Giacomo Acciarini, Simone Mestici, Halil Kelebek, Linnea Wolniewicz, Michael Vergalla, Madhulika Guhathakurta, Umaa Rebbapragada, Bala Poduval, Atılım Güneş Baydin, Frank Soboczenski. 2025. "Forecasting the Ionosphere from Sparse GNSS Data with Temporal-Fusion Transformers." In Workshop on Machine Learning and the Physical Sciences (NeurIPS 2025), San Diego Convention Center, the United States.

@inproceedings{acciarini2025forecasting,
  title        = {Forecasting the Ionosphere from Sparse GNSS Data with Temporal-Fusion Transformers},
  author       = {Acciarini, Giacomo and Mestici, Simone and Kelebek, Halil and Wolniewicz, Linnea and Vergalla, Michael and Guhathakurta, Madhulika and Rebbapragada, Umaa and Poduval, Bala and Baydin, At{\i}l{\i}m G{\"u}ne{\c{s}} and Soboczenski, Frank},
  year         = {2025},
  booktitle    = {Workshop on Machine Learning and the Physical Sciences (NeurIPS 2025), San Diego Convention Center, US},
  eprint       = {2509.00631}
}

Acknowledgements

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.

Ionosphere Team 2025:

Researchers: Linnea Wolniewicz, Halil Kelebek, Simone Mestici, Michael Vergalla

Faculty: Giacomo Acciarini, Atilim Gunes Baydin, Tom Berger, Frank Soboczencki

Advisors: Bala Poduval, Umaa Rebbapragada, Olga Verkhoglyadova