TRACER – DP-SCREAM-matched periodic case

cases/tracer_dp_scream.jl runs a doubly periodic Breeze LES with the forcing of the published DP-SCREAM TRACER experiment (Oware et al. 2025, JGR Atmospheres, doi 10.1029/2025JD044113). It is a separate experiment identity from ENA (Covert, LASSO) and from regional TRACER-MIP; it is not the ENA-SCREAM assignment. Constructor: tracer_dp_scream(; ...) in src/tracer_dp_scream.jl; inputs pinned in cases/tracer_dp_scream/inputs.toml (fetch with data_wrangling/fetch_manifest.jl).

Sources (pinned 2026-10-07)

What DP-SCREAM actually did (from the script, the IOP file and the E3SM source at 1d551ea2b0)

• 200 km × 200 km, 20 × 20 spectral elements (3.33 km), 128 hybrid levels to ~3 hPa, physics step 100 s, dynamics 8.33 s, RRTMGP every 3 steps (300 s), SHOC + P3, prescribed aerosol. The 0.5 km run used 15 s / 1.25 s and started cold on 2022-08-05 00 UTC (Table 1 of the paper; the archive's first records show the spin-up). The 3 km August run started on 2022-08-01 00 UTC; its days 5–15 are the paper's comparison window.
• Forcing: hourly VARANAL profiles on 40 pressure levels. EAMxx forms divT3d = divT + vertdivT and divq3d = divq + vertdivq when the vertical terms exist and adds only those to T and qv (advance_iop_forcing); iop_dosubsidence = false, so the omega profile is unused. Data are read piecewise-constant over each hourly interval.
• iop_srf_prop = true: sensible heat flux, evaporation (lhflx/Lv) and the radiative surface temperature Tg come from the file; the momentum stress comes from the land/ocean coupler (not reproducible from public inputs).
• Planar HOMME sets fcor = 0; iop_coriolis = false: no Coriolis force.
• Wind nudging: the current script sets iop_nudge_uv = true and the paper states the winds are nudged, but the archived code revision predates the DP-EAMxx nudging port (E3SM 3f7eee0053, 2024-04-16; the 1d551ea2b0 namelist has no iop_nudge_* option). The archived runs therefore ran without wind nudging; wind_nudging_timescale = nothing is the default and 10800 reproduces the script's intent as a sensitivity.
• Archived outputs: 30-minute domain means (PRECL, TMQ, TGCLDLWP, TGCLDIWP; TOTCLOUDFRAC, Z3, T, RELHUM, W_SEC, Q on 128 levels), time axis labelled in local time (CDT, UTC−5). No 3-D fields, no winds, no surface fluxes; PRECL has no units attribute (mm/day inferred).

Breeze implementation

iop_sam_inputs maps the IOP window onto the SAM-record pipeline (build_case): initial sounding from the start record (θ from T, specific humidity as mass fraction), tls/qls = the 3-D tendencies, wls = 0, no geostrophic columns, sfc = (Tg, shflx, lhflx), every record duplicated 1 s before the next so the linear interpolation holds the hourly value. The large-scale transport therefore enters exactly once (vertical_advection = nothing, geostrophic = false, coriolis = nothing, upper_boundary_relaxation = false).

Intentional differences (recorded in the run label and provenance): Smagorinsky–Lilly LES at 200 m on 51.2 km (160 levels, 50 m near the surface, top 22 km, sponge above ~16.5 km) versus SHOC at 3.33 km on 200 km; P3 with a prescribed droplet number (200 cm⁻³, assumption) versus prognostic droplet number with prescribed CCN; constant-coefficient bulk drag for a 0.1 m roughness length versus the coupler stress; IOP-implied albedo 0.15 and emissivity 0.98; anelastic dynamics; no aerosol radiative effects.

Checks

Pkg.test() covers the IOP reader against reference values, the adapter (3-D transport once, hold expansion, basis), the archive reader (local time → UTC) and a 4-step CPU case with RRTMGP, outputs and a checkpoint pickup, all on excerpts in test/fixtures/. The GPU script writes budget.toml with column water and (approximate) static-energy closure residuals and plots water paths, precipitation and time–height cloud fraction against the archived 3 km and 0.5 km runs. Agreement with DP-SCREAM is not observational validation; the ARM products used in the paper (ARMBECLDRAD, VARANAL precipitation, MRMS) are not staged here.