Reproduce Fire VASE

Everything required to inspect, rerun, validate, or extend this research is available here—from source-data contracts and notebooks to frozen claims and figure hashes.

I want the data

Download, inspect, or rebuild

Use the shared v0.1 data lake for the shortest path. Source documentation, schemas, release inventory, checksums, exclusions, and a full rebuild route are kept alongside it.

Data and data lake →

I want to reproduce the paper

Run the canonical v2 workflow

Install the locked environment, materialize the data lake, regenerate the v2 analysis and figures, run scientific validation, and compare output hashes.

Full reproduction vignette →

I want the notebooks

Open all four Jupyter assets

The canonical reproduction notebook and validation notebook sit beside two implementation-focused VASE volume notebooks. Their roles, inputs, outputs, and status are indexed explicitly.

Notebook index →

I want the implementation

Trace code to evidence

Inspect package code, analysis scripts, numbered figure entry points, schemas, tests, configuration, and validation modules.

Code map →

I want the figures

Render or audit every figure

Browse the current main figures, supplementary diagnostics, adversarial checks, and preserved historical generation—with the commands that rebuild them.

Figure workflow →

I want provenance

Follow claims and corrections

Review input/output hashes, correction history, scientific-validation records, AI transparency, citation audit, manuscript sources, and the project migration record.

Provenance map →

Quick start: reproduce the current paper

Requirements: Git, uv, Python 3.9 or newer, and the materialized Fire VASE v0.1 data lake at data_lake/fire-vase-data-lake-v0.1. External validation also requires network access.

git clone https://github.com/CU-ESIIL/fire_vase.git
cd fire_vase
uv sync

PYTHONPATH=src:scripts OPENBLAS_NUM_THREADS=1 \
MPLCONFIGDIR=/tmp/fire-vase-v2-mpl \
.venv/bin/python manuscript_figures/00_run_all.py \
  --generation v2 \
  --data-lake data_lake/fire-vase-data-lake-v0.1

PYTHONPATH=src:scripts MPLBACKEND=Agg \
MPLCONFIGDIR=/tmp/fire-vase-v2-mpl OPENBLAS_NUM_THREADS=1 \
.venv/bin/python scripts/validate_fire_vase_science.py

PYTHONPATH=src:scripts:. .venv/bin/pytest -q

The configuration is config/analysis_v2.json with seed 20260828. Required real inputs must exist; there is no synthetic fallback. After successful statistics generation, --render-only can rebuild figures without recomputing the tables.

Canonical versus historical. manuscript_figures/00_run_all.py --generation v2 is the current paper workflow. Climate-revision scripts and v1 comparison outputs remain available for provenance, but they do not control current quantitative claims.

Reproduction map

Stage	Primary entry point	Inputs	Outputs
Environment	`uv sync`	`pyproject.toml`, `uv.lock`	`.venv/`
Data verification	`scripts/check_reproducibility.py`	v0.1 data lake, checksums	JSON status report
Current analysis + figures	`manuscript_figures/00_run_all.py --generation v2`	data lake, `config/analysis_v2.json`	`analysis/v2/`, current figure assets
Scientific challenge suite	`scripts/validate_fire_vase_science.py`	frozen v2 tables	`analysis/scientific_validation/`
Real-data QA modules	`scripts/run_validation.py`	FIRED/gridMET materialization	module results, plots, suite manifest, PDF
Tests	`pytest -q`	source and fixtures	full repository pass/fail report
Artifact identity	`scripts/check_reproducibility.py`	data, statistics, reference figures	checksums and pixel comparisons

Data, schemas, and licensing

The analysis does not relicense FIRED or gridMET. Upstream reuse and citation requirements remain attached to each source package; access method, exclusions, hashes, and derived-product boundaries are recorded in the data-lake and v2 manifests.