High-resolution greenhouse gas flux inversions using a machine learning surrogate model for atmospheric transport

Dadheech, Nikhil; He, Tai-Long; Turner, Alexander J.

doi:https://doi.org/10.5194/acp-25-5159-2025

Articles | Volume 25, issue 10

https://doi.org/10.5194/acp-25-5159-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/acp-25-5159-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 25, issue 10

Research article

|

21 May 2025

Research article |

| 21 May 2025

High-resolution greenhouse gas flux inversions using a machine learning surrogate model for atmospheric transport

Nikhil Dadheech, Tai-Long He, and Alexander J. Turner

Data sets

High-resolution greenhouse gas flux inversions using a machine learning surrogate model Nikhil Dadheech https://doi.org/10.5281/zenodo.13750963

STILT footprints data set 2 T.-L. He https://doi.org/10.5281/zenodo.12803736

Model code and software

High-resolution greenhouse gas flux inversions using a machine learning surrogate model Nikhil Dadheech https://doi.org/10.5281/zenodo.13750963

Short summary

We developed an efficient GHG (greenhouse gas) flux inversion framework using a machine-learning emulator (FootNet) as a surrogate for an atmospheric transport model, resulting in a 650 × speedup. Paradoxically, the flux inversion using the ML (machine-learning) model outperforms the full-physics model in our case study. We attribute this to the ML model mitigating transport errors in the GHG flux inversion.