Aerosol-Driven Parameterization of Ice Nucleation and Secondary Ice Processes in EC-Earth3: Evaluation and Climate Impacts

Clouds remain a major source of uncertainty in climate projections, particularly due to complexities in aerosol-cloud interactions. To improve the representation of mixed-phase clouds in EC-Earth3, the model's heterogeneous ice nucleation scheme has been updated. The previous temperature-based parameterization has been replaced with aerosol- and temperature-sensitive immersion freezing schemes for mixed-phase clouds that consider ice-active desert dust minerals (K-feldspar and quartz) and marine organic aerosols, both explicitly tracked in EC-Earth3. Additionally, a secondary ice production scheme based on a random forest regressor further enhances the ice crystal concentrations.The updated model is evaluated against an extensive observational dataset of ice-nucleating particle (INP) concentrations, satellite observations of cloud properties (MODIS and CALIPSO), and both Top of the Atmosphere (TOA) and surface radiative Cloud Radiative Effect (CRE) flux components from CERES-EBAF. The impact of the updates is analysed relative to the previous temperature dependent parameterization.Results from 12-year (2009-2020) nudged simulations show improved agreement with INP observations using the updated aerosol-aware scheme compared to the earlier approach. The ice nucleation parameterization clearly links simulated ice crystal number concentrations with aerosol emission sources and transported pathways. Despite remaining biases largely attributed to other processes, this update improves consistency with MODIS and CALIPSO retrieved data, including total cloud cover, low/mid/high cloud area percentages, liquid and ice cloud fractions, and water paths. Sensitivity analyses reveal that the new scheme impacts global cloud cover, liquid and ice water content, temperature, and radiative balances. Evaluation with CERES-EBAF indicates that the new parameterization reduces surface net CRE bias at mid-to-high latitudes while slightly increasing bias at low latitudes, despite no specific model tuning for this configuration.Our approach offers potential enhancements in future climate projections using EC-Earth3-AerChem and future generations of the model.