What the study found
The study found that the initial ice crystal number concentration was the main factor controlling ice aggregation rates in persistent supercooled stratiform clouds. The authors also report that the relationship was subquadratic, with a mean exponent of about 0.92, rather than the quadratic dependence expected from theory.
Why the authors say this matters
The authors conclude that their findings provide new insights into the microphysical and environmental controls of ice aggregation. They also say the work establishes a robust methodological foundation for studying aggregation processes in natural clouds.
What the researchers tested
The researchers used targeted glaciogenic seeding experiments called CLOUDLAB to nucleate ice crystals upwind and measure them downwind after a known time in cloud, which let them estimate crystal age. They used a deep-learning detection algorithm, IceDetectNet, to count aggregate monomers and derive the initial ice crystal number concentration, and they examined several possible controls, including initial concentration, temperature, ice crystal size, aspect ratio, and turbulence.
What worked and what didn't
Three independent approaches — causal inference, a physical equation, and machine learning models — all identified initial ice crystal number concentration as the dominant factor. For prediction, CatBoost had the best statistical performance among 11 machine learning models, while the physically based formulation was more robust in sensitivity tests. One possible explanation for the subquadratic behavior is that aggregation may also involve smaller ice crystals, but the abstract says this remains hypothetical.
What to keep in mind
The abstract does not describe detailed limitations beyond noting that the explanation involving smaller ice crystals is hypothetical. The findings are based on in situ measurements in persistent supercooled stratiform clouds, so the scope described in the abstract is limited to that setting.
Key points
- Initial ice crystal number concentration was the dominant control on ice aggregation rates.
- The measured dependence on initial ice crystal number concentration was subquadratic, with a mean exponent of about 0.92.
- Three approaches — causal inference, a physical equation, and machine learning — pointed to the same dominant factor.
- CatBoost performed best statistically among 11 machine learning models.
- The physically based model was more robust in sensitivity tests.
- The abstract says a smaller-ice-crystal explanation is possible but hypothetical.
Disclosure
- Research title:
- Ice crystal concentration strongly controls aggregation rates
- Authors:
- Huiying Zhang, Fabiola Ramelli, Christopher Fuchs, Nadja Omanovic, Anna J. Miller, Robert Spirig, Zhaolong Wu, Yunpei Chu, Xia Li, Ulrike Lohmann, Jan Henneberger
- Institutions:
- ETH Zurich, ETH Zurich, ETH Zurich, ETH Zurich, ETH Zurich, ETH Zurich, ETH Zurich, ETH Zurich, ETH Zurich, ETH Zurich, Leibniz Institute for Tropospheric Research, Peking University
- Publication date:
- 2026-01-28
- OpenAlex record:
- View
Get the weekly research newsletter
Stay current with scholarly research without reading academic papers — one filtered digest, every Friday.