Model simulation data and analysis scripts used in "Mixed-phase cloud response to dust INP perturbations strongly depends on ice nucleation scheme choice" (Vella et al., 2026)

This deposit contains the data and analysis code that reproduce every figure and table in Mixed-phase cloud response to dust INP perturbations strongly depends on ice nucleation scheme choice (Vella et al., 2026).

It provides two levels of data and the code connecting them, so the published results can be re-plotted in seconds or recomputed from the model fields with different choices.

Contents

code/ One Python script per figure and table, plus a shared configuration module. Running run_all.sh regenerates every published figure and table from data/derived/ in a few seconds.

data/derived/ Eight small netCDF files, one per figure or table, each holding exactly the reduced quantity plotted.

data/native/ Model fields on the native ICON R2B4 grid at daily resolution (about 37 GB), from which the reductions were computed, so the analysis can be repeated with different bin edges, thresholds or regions rather than only re-plotted.

reference_output/ The figures and tables as published, for comparison.

tools/ The two reduction stages, for provenance and re-analysis. build_derived_data.py rebuilds data/derived/ from data/native/ and runs on this package alone, with every threshold, bin edge and mask exposed at the top. extract_native_data.py documents how data/native/ was extracted from the raw model output.

environment.yml / requirements.txt: The Python environment used.

The simulations

ICON-A coupled to the aerosol module HAM2.3, on the R2B4 icosahedral grid (about 160 km) with 47 vertical levels and a 600 s time step, in an AMIP configuration with prescribed sea surface temperatures and horizontal winds nudged to ERA5 on a six-hour relaxation time scale. The experiment is a full factorial design of 24 simulations:

Heterogeneous ice nucleation scheme: LD06 (Lohmann and Diehl, 2006), Niemand (Niemand et al., 2012), CNT (classical nucleation theory).

Wegener-Bergeron-Findeisen treatment: wbf1 (explicit conversion of cloud liquid to ice active), wbf0 (inactive).

Dust ice-nucleating particle scaling: ctrl (unscaled), x10, x100, x1000, applied to the immersion freezing pathways.

Two analysis windows are used, and both are already applied in the data: 2009 to 2013 for the CALIOP cloud-phase evaluation (control simulations only), and 2009 to 2010 for every dust-sensitivity diagnostic (all 24 simulations). October to December 2008 is discarded as spin-up.

The CALIOP bulk supercooled liquid fraction (in the CALIOP data file and reproduced in the Figure 1 derived file) is taken from the retrieval of Hofer et al. (2024). Please cite that work when using it.