Question. Do the machine-learning-calibrated SHOC parameter sets of Zhang et al. (2026, GRL, doi:10.1029/2025GL120241; "CPE" ranks 1-3, from Peter Bogenschutz) make E3SM/SHOC's BOMEX cumulus-like, i.e. a cloud fraction that peaks near cloud base and decays upward?
Answer: no. In the full-physics EAM-SCM (Brian's setup, radiation off) and in the SHOC standalone driver alike, in almost every one of the 4 dz x 4 dt combinations, the CPE sets have their cloud-fraction maximum in the upper part of the cloud layer and much more liquid water than default SHOC. The exception is dz = 50 m with dt ≤ 60 s, where the CPE cloud switches on and off (large swings in liquid water path) and its time-mean maximum sits low in the layer. Hours 4-6 means; the grey LES curves on the figures are a reference only.
| type | default SHOC | CPE-1 | CPE-2 | CPE-3 | reference |
|---|---|---|---|---|---|
| EAM-SCM (65 p.) | default | cpe1fix | cpe2 | cpe3 | |
| SHOC standalone (59 p.) | default | cpe1 | cpe2 | cpe3 | Matheou (UConn) LES, 21 p. |
cpe1fix = CPE rank 1 including shoc_c_diag_3rd_mom (Peter's C_diag_w3), which Brian's cpe1 runs omitted. The EAM-SCM groups use Brian's template and code (E3SM-fork eam_scm_bugfixes @ 4c07d90b57); CPE values enter only through user_nl_eam. Standalone runs: shoc_in_and_out driver, 12 h, dx = 100 km, same IC files as the SCM.
SHOC+MF results for BOMEX and DYCOMS-RF01, in the standalone driver and in the full-physics SCM, all with Cee = 1 and do_wthv_mf on, are on their own page: SHOCMF.
metrics_fullphysne30*_4-6h.csv, standalone/metrics_4-6h.csv/global/cfs/cdirs/m4359/mchinita/bomex_cpe_archive_20260901/; LES reference files: /global/cfs/cdirs/m4359/mchinita/les_reference/BOMEX_Matheou/main_BOMEXmods, e3sm_scm_scripts/code_test_archive/20260901_e3sm_master_bomex_fullphys_cpe/ (EAM-SCM) and e3sm_scm_scripts/in_and_out_scripts/ (standalone)