Hemispheric contrasts in recorded volcanic explosivity are sensitive to temporal adjustment and missing VEI
Bulletin of Volcanology
Rongzhen Dai
Correspondence: dairongzhen3@gmail.com

Contents
run_frozen_portable.py: executed regression and sensitivity code, preserved byte-for-byte.
reference_results.json: fresh execution on the supplied tables with relative input paths.
diagnostic_encoding_2x2.json: original fixed-sample coding diagnostic.
reproduce_support_tables.py: regional MH/RBG calculation and SI support tables S4-S6.
support_reference_results.json: execution of that additional entry point.
verified_environment.json: environment actually used for this package's verification.
requirements.txt: required scientific packages; installation is handled by the reader.
verify_source_snapshot.py: source worksheet/position-to-derived membership check.
source_snapshot_verification.json: actual source-membership verification output.

Layout
Place the extracted code files together. Extract the data archive into data/ alongside
them, or supply the actual CSV paths to the commands below.

Commands
python run_frozen_portable.py --ng data/eruptions_NG.csv --full data/eruptions_full.csv --out reanalysis_new.json
python reproduce_support_tables.py --ng data/eruptions_NG.csv --out support_new.json
python verify_source_snapshot.py --data data --out source_check_new.json
Use new output filenames to preserve the supplied reference files.

The regression entry reproduces the fixed-sample models, two cluster calculations,
region-ambiguity exclusion, leave-region-out results and 25 missing-VEI scenarios.
The support entry supplies the four-stratum MH/RBG benchmark and support summaries.
Standardization averages over retained episodes. The missing-VEI grid assigns expected
outcomes among included records under stated odds multipliers. See manuscript/SI for
the estimands, empirical targets, model definitions and uncertainty calculations.
