Reproducibility Checklist

Reproducibility Checklist

Record these items before launching or comparing a CheMPAS-A run.

Source and build

• CheMPAS-A source commit and branch recorded.
• atmosphere_model was built with MUSICA=true and double precision.
• Compiler, MPI, NetCDF-C, NetCDF-Fortran, PNetCDF, and PIO versions recorded.
• MUSICA-Fortran reports version 0.16.5, source revision 1403e3d22717bc87f3bf9d0aa591caf039c92bbc, MIEM revision 9fdf14a189262eecb677862d877ab72b06c95e21, MIEM ON, and TUV-x ON.
• Executable SHA-256 recorded.

Docker runs

• Source commit used to build the image recorded.
• Local image ID, linux/amd64 platform, Docker version, and resource allocation recorded.
• The fixed eight-rank runner and at least 1 GiB of shared memory were used.
• Host data root and the active or archived case-output directory recorded.
• A passing container run is not treated as validation of an untested mechanism, forcing data set, duration, or MPI decomposition.

Configuration

• Namelist, streams, output list, MICM, TUV-x, MIEM, and static-extension file hashes recorded.
• Mechanism species and photolysis names match case-sensitively.
• config_start_time, initial-condition xtime, streams, and inventory time coverage agree.
• Chemistry cadence, substeps, tolerance, and TUV-x cadence recorded.
• Tracer output units and any species overrides recorded.
• One local MPAS block is used per MPI task.

Mesh and data

• Every raw download matches the byte size and SHA-256 in Data and Provenance.
• Initial state, partition, prescribed field, and inventories match the prepared-artifact hashes.
• Mesh fingerprint is cdd75b787f487302df17b31c2dd7aaf7082a7cb265a711792ef0e5f3ceebda12 for the x1.40962 global example.
• indexToCellID is a permutation of 1 through nCells and packaged inventory rows use ascending global-ID order.
• MIEM uses a Gregorian calendar and never extrapolates in time.
• TUV-x data/ comes from the pinned MUSICA revision.

Global emissions experiment

• No Surface Emissions leaves config_miem_file empty.
• Anthropogenic Emissions stages only the CAMS inventory and global_mvp_cams.yaml.
• Anthropogenic + Fire Emissions stages both inventories and global_mvp_cams_finn.yaml.
• CAMS awb is absent from the applied sums.
• FINN uses direct provider CO, NO, and NO2 and surface injection.
• Lightning NOx is disabled in all three scenarios.
• All scenarios use the same meteorology, chemistry initial state, prescribed O3, decomposition, and output times.

Output checks

• The rank-zero log reports the expected source revision and MICM version and normal completion.
• No chemistry step exceeded the configured consecutive failure limit.
• All expected j_<reaction> and emis_<species> fields exist and are finite.
• Emissions totals close against requested sector/category diagnostics.
• Comparisons use the same xtime, cells, levels, and unit convention.
• Restart comparisons use authoritative dry-air mass mixing ratios, not output-only VMR diagnostics.
• Scientific conclusions retain the examples' reduced-mechanism and idealized-chemistry limitations.