Karman - Data Driven Thermospheric Density Nowcast and Forecast with Machine Learning
Karman
Explore the docs »
Report bug
·
Request feature
Karman is a machine learning (ML) Python package for thermospheric density modelling. It was developed as a collaboration between the Heliophysics Division of NASA and Trillium Technologies.
Its main goal is to improve our understanding and modeling of thermospheric density variations due to the Sun's influence. Furthermore, it also has the objective to provide a shared framework where people can design, train and benchmark their own thermospheric density models.
- Release of ML models to nowcast & forecast thermospheric density changes from solar and geomagnetic data, trained on precise orbit determination (POD)-derived thremospheric density measurements.
- A framework to support ingestion of solar irradiance, geomagnetic and thermospheric density data from different sources. Currently, it supports data from: solar indexes (F10.7, M10.7, S10.7, Y10.7), geomagnetic indixes (Dst, Ap), SOHO EUV irradiance data, OMNIWeb high-resolution data, GOES EUV irradiance missions, POD-derived thermospheric density data.
- Framework to enable comparison between ML and empirical models at different geomagnetic storm conditions, altitude, times, etc.
Currently, the Karman software supports several input data from different sources:
- High frequency (i.e., every 30 or 10 seconds) POD-derived thermospheric density data from five space missions (CHAMP, GOCE, GRACE, SWARM-A, SWARM-B), from TU Delft thermosphere data.
- NRLMSISE-00 thermospheric density empirical models inputs (and corresponding predicted density).
- Geomagnetic input data from NASA's OMNIWeb high resolution data.
- Solar irradiance proxies commonly used in empirical thermospheric density models (i.e., F10.7, S10.7, M10.7, Y10.7).
- EUV Irradiance data from SOHO mission (from LASP website)
- EUV Irradiance data from GOES mission (from NOAA website)
We have dedicated tutorials and scripts to download and prepare each of these data sources for ML ingestion.
https://spaceml-org.github.io/karman/
Currently active contributors are:
Many other people contributed to Karman since 2021 over the course of several FDLs. For a list of previous and current contributors see the "credits" page in the doc.
If you would like to learn more about or cite the techniques Karman uses, please see the following papers:
- Giacomo Acciarini, Edward Brown, Thomas E. Berger, Madhulika Guhathakurta, James Parr, Christopher Bridges, Atılım Güneş Baydin. 2024. "Improving Thermospheric Density Predictions in Low-Earth Orbit with Machine Learning". In Space Weather Journal.
@article{acciarini2024improving,
title={Improving Thermospheric Density Predictions in Low-Earth Orbit With Machine Learning},
author={Acciarini, Giacomo and Brown, Edward and Berger, Tom and Guhathakurta, Madhulika and Parr, James and Bridges, Christopher and Baydin, At{\i}l{\i}m G{\"u}ne{\c{s}}},
journal={Space Weather},
volume={22},
number={2},
pages={e2023SW003652},
year={2024},
publisher={Wiley Online Library}
}- Giacomo Acciarini, Edward Brown, Christopher Bridges, Atılım Güneş Baydin, Thomas E. Berger, Madhulika Guhathakurta. 2023. "Karman - a Machine Learning Software Package for Benchmarking Thermospheric Density Models". In Advanced Maui Optical and Space Surveillance Technologies Conference (AMOS) – www.amostech.com.
@inproceedings{karman-amos-acciarini,
author = {Giacomo Acciarini and Edward Brown and Christopher Bridges and Atılım Güneş Baydin and Thomas E. Berger and Madhulika Guhathakurta},
title = {Karman - a Machine Learning Software Package for Benchmarking Thermospheric Density Models},
booktitle = {Proceedings of the Advanced Maui Optical and Space Surveillance Technologies Conference (AMOS)},
year = {2023},
month = {September},
day = {19},
}- Shreshth Malik, James Walsh, Giacomo Acciarini, Thomas Berger, Atılım Güneş Baydin. 2023. "High-Cadence Thermospheric Density Estimation enabled by Machine Learning on Solar Imagery". In NeurIPS, Machine Learning and the Physical Sciences Workshop.
@article{malik2023high,
title={High-Cadence Thermospheric Density Estimation enabled by Machine Learning on Solar Imagery},
author={Malik, Shreshth A and Walsh, James and Acciarini, Giacomo and Berger, Thomas E and Baydin, At{\i}l{\i}m G{\"u}ne{\c{s}}},
booktitle={NeurIPS, Machine Learning and the Physical Sciences Workshop},
year={2023},
month={December}
}The authors would like to ackowledge the support and precious feedback of several people: Dr. Madhulika Guhathakurta, and all the NASA-FDL 2021, 2023, 2024 reviewers.