Machine Learning Approach to Investigating the Relative Importance of Meteorological and Aerosol-Related Parameters in Determining Cloud Microphysical Properties

Aug 5, 2026

Aerosol effects on cloud properties are hard to separate from meteorological variability. A machine learning model is trained using reanalysis data and satellite retrievals to predict cloud microphysical properties and quantify the relative roles of meteorology and aerosol. Cloud droplet effective radius is predicted with useful skill from meteorology alone, outperforming a reference linear regression and a climatological-mean model. Gradient boosting matches a neural network, while adding aerosol information yields only limited, region-dependent improvement and inconsistent aerosol-burden influence patterns.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Abstract

Introduction

Study goal and key research questions

Data sources and model inputs

Machine Learning Approach to Investigating the Relative Importance of Meteorological and Aerosol-Related Parameters in Determining Cloud Microphysical Properties FRIDA A. -M. BENDER VERENA JUNG ANNA STAFFANSDOTTER TOBIAS LORD SABINE UNDORF *Author affiliations can be found in the back matter of this article ABSTRACT Aerosol effects on cloud properties are notoriously difficult to disentangle from variations driven by meteorological factors. Here, a machine learning model is trained on reanalysis data and satellite retrievals to predict cloud microphysical properties, asa way to illustrate the relative importance of meteorology and aerosol, respectively, on cloud properties. It is found that cloud droplet effective radius can be predicted with some skill from only meteorological information, including estimated air mass origin and cloud top height. For ten geographical regions the mean coefficient of determination is 0.41 and normalised root-mean square error 24% . The machine learning model thereby performs better than a reference linear regression model, and a model predicting the climatological mean. A gradient boosting regression performs on par with a neural network regression model. Adding aerosol information as input to the model improves its skill somewhat, but the difference is small and the direction of the influence of changing aerosol burden on cloud droplet effective radius is not consistent across regions, and thereby also not always consistent with what is expected from cloud brightening. ORIGINAL RESEARCH PAPER CORRESPONDING AUTHOR: Frida A.-M. Bender Department of Meteorology, and Bolin Centre for Climate Research, Stockholm University, SE [frida@misu.su.se](frida@misu.su.se) KEYWORDS: Aerosol-cloud interaction; Cloud brightening; Machine learning; Gradient boosting regression; Reanalysis; Remote sensing TO CITE THIS ARTICLE: Bender, FA-M, Jung, V, Staffansdotter, A, Lord, T and Undorf, S. 2024. Machine Learning Approach to Investigating the Relative Importance of Meteorological and Aerosol-Related Parameters in Determining Cloud Microphysical Properties. Tellus B: Chemical and Physical Meteorology, 76(1): 1–18. DOI: [https://doi.org/10.16993/](https://doi.org/10.16993/)[ ](https://doi.org/10.16993/)[tellusb.1868](tellusb.1868) 1 INTRODUCTION Aerosol-cloud interactions and their effects on Earth’s radiation balance remain one of the main uncertainties in future climate projection (Bellouin et al. 2020, Forster et al. 2021, Bender 2020) . This is not only because future changes in aerosol loading are not known but also because the sensitivity in clouds to aerosol changes is uncertain. Investigation of aerosol-cloud interactionson large scale relies on more or less complex versions of correlation analysis, making it difficult to isolate and assess causality of aerosol effects, and to distinguish any potential signals from variation due to meteorological factors, that often co-vary with aerosols, and aerosol influence on cloud (Mauger and Norris 2007, Engströmand Ekman 2010, Koren et al. 2010, Zhang et al. 2022) . Attempts to account for varying meteorology are continuously made, [e.g. by](e.g. by) segregating analysis based on meteorological regime or time interval (Gryspeerdt and Stier 2012, Chen et al. 2014, Oreopoulos et al. 2017, Malavelle et al. 2017, Oreopoulos et al. 2019, Douglas and L’Ecuyer 2019, Chen et al. 2022) or like Gryspeerdt et al. (2014) rather investigating how the occurrence of and transition between different regimes varies with aerosol. Still, the challenge to separate meteorological variation from aerosol influence remains unsolved. Methods from data science provide new ways of studying aerosol-cloud interaction, and here we apply machine learning techniques on large sets of reanalysis and remote sensing data to investigate the relative importance of meteorological and aerosolrelated parameters in determining cloud microphysical propertie

FAQs

Why is separating aerosol effects from meteorology on cloud properties difficult?

Meteorological factors and aerosols often co-vary, making it challenging to isolate causality and distinguish aerosol-driven signals from meteorology-driven variations.

What does the machine learning model predict in this study?

The model predicts cloud droplet effective radius (reff) for each geographical region using inputs derived from reanalysis, satellite retrievals, and trajectory-based air mass origin.

How does adding aerosol information affect prediction performance?

Including aerosol variables improves model skill only somewhat, and the inferred influence of aerosol burden on droplet effective radius is not consistent across regions, limiting agreement with expectations such as cloud brightening.

Aug 5, 2026

Upload by Emma Wilson