Exploring ice nucleation particle concentrations in a boreal environment: limits of machine-learning-assisted variable screening

Wu, Yusheng; Brasseur, Zoé; Castarède, Dimitri; Heikkilä, Paavo; Keskinen, Jorma; Möhler, Ottmar; Kulmala, Markku; Petäjä, Tuukka; Thomson, Erik S.; Duplissy, Jonathan

doi:10.5194/ar-4-293-2026

Articles | Volume 4, issue 2

https://doi.org/10.5194/ar-4-293-2026

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

https://doi.org/10.5194/ar-4-293-2026

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

Articles | Volume 4, issue 2

Research article

|

09 Jul 2026

Research article |

| 09 Jul 2026

Exploring ice nucleation particle concentrations in a boreal environment: limits of machine-learning-assisted variable screening

Yusheng Wu, Zoé Brasseur, Dimitri Castarède, Paavo Heikkilä, Jorma Keskinen, Ottmar Möhler, Markku Kulmala, Tuukka Petäjä, Erik S. Thomson, and Jonathan Duplissy

Data sets

Datasets to: Measurement Report: Introduction to the HyICE-2018 Campaign for Measurements of Ice-Nucleating Particles and Instrument Inter-Comparison in the Hyytiälä Boreal Forest Zoé Brasseur et al. https://doi.org/10.5281/zenodo.10469663

PINC ice-nucleating particle dataset from the HyICE-2018 campaign at SMEAR-II Mikhail Paramonov et al. https://doi.org/10.3929/ethz-b-000397022

Model code and software

Title HyICE-2018 INP analysis code: machine-learning variable screening, multi-seed random-forest sensitivity, and dust-episode screen (ar-2026-4) Yusheng Wu et al. https://doi.org/10.5281/zenodo.20367476

Short summary

Clouds in cold regions affect climate and precipitation, but their behavior depends on rare airborne particles that help ice form. We measured these particles over several months in a Finnish forest and compared them with many environmental observations. We found that ice formation in winter was largely unpredictable, while in spring and summer it was more strongly linked to particle amount and composition. This shows that local conditions are needed to better represent clouds in climate models.