EnsAI: an emulator for atmospheric ammonia ensembles

Sitwell, Michael

doi:10.5194/gmd-19-10283-2026

Articles | Volume 19, issue 20

https://doi.org/10.5194/gmd-19-10283-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-19-10283-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 19, issue 20

Development and technical paper

|

29 Oct 2026

Development and technical paper |

| 29 Oct 2026

EnsAI: an emulator for atmospheric ammonia ensembles

Michael Sitwell

Supplement

https://doi.org/10.5194/gmd-19-10283-2026-supplement

Data sets

GEM-MACH ensemble files for EnsAI training set (2016-01 to 2016-06) Michael Sitwell and GEM-MACH Development Team https://doi.org/10.5281/zenodo.18843197

GEM-MACH ensemble files for EnsAI training set (2016-07 to 2016-12) Michael Sitwell and GEM-MACH Development Team https://doi.org/10.5281/zenodo.18843211

GEM-MACH ensemble files for EnsAI validation set (2014-07) Michael Sitwell and GEM-MACH Development Team https://doi.org/10.5281/zenodo.18820049

GEM-MACH ensemble files for EnsAI test set (2015-01 to 2015-06) Michael Sitwell and GEM-MACH Development Team https://doi.org/10.5281/zenodo.18825231

GEM-MACH ensemble files for EnsAI test set (2015-07 to 2015-12) Michael Sitwell and GEM-MACH Development Team https://doi.org/10.5281/zenodo.18839600

CrIS NH3 retrieval data used in EnsAI evaluation Michael Sitwell and CrIS NH3 Retrieval Team https://doi.org/10.5281/zenodo.18865370

Model code and software

EnsAI: An Emulator for Atmospheric Chemical Ensembles Michael Sitwell https://doi.org/10.5281/zenodo.15342768

GEM-MACHv3.1.0b.2 ECCC https://doi.org/10.5281/zenodo.8346780

Short summary

EnsAI is a newly developed artificial intelligence based program for efficiently generating ensembles of atmospheric surface ammonia concentrations that can be used in emissions inversions systems. Ensemble-based data assimilation methods are widely used for assimilation and emissions inversions, but are usually very computationally demanding. Once trained, EnsAI can run thousands of times faster than the physics-based models when run on graphics processing units.