Getting started¶

Installation¶

git clone https://github.com/zerotonin/thermostrife.git
cd thermostrife
pip install -e ".[climate,dev]"

Python ≥ 3.11 is required (meteostat 2.x dropped Python 3.10).

Conda environment (recommended)¶

conda env create -f environment.yml
conda activate thermostrife
pip install -e ".[dev]"

Copernicus CDS API key (needed for the ERA5 fallback tier)¶

url: https://cds.climate.copernicus.eu/api
key: <YOUR-API-KEY>

The Tier 1 / 2 sources (meteostat + HadCET) and the Tier 4 NOAA 20CRv3 OPeNDAP feed work without this key and together cover the overwhelming majority of the dataset; the CDS key only matters when the cascade falls through to ERA5 (1981+ events outside Britain and outside meteostat-station radius).

Verifying the installation¶

python -c "import thermostrife; print(thermostrife.__version__)"
pytest                 # runs the full test suite

Your first run¶

The pipeline is a sequence of small scripts in scripts/ that you run from the repo root.

# Sanity-check the cascade against the curated CSV
python scripts/validate_cascade.py

# Inference on the headline violent panel
python scripts/run_inference.py --panel violent

# Inference on the peaceful-control panel
python scripts/run_inference.py --panel peaceful

# Side-by-side comparison + two-panel raincloud
python scripts/compare_panels.py

# Sensitivity stratifications (era / hemisphere / duration / event type)
python scripts/run_stratifications.py

Each script writes a Markdown report under reports/ and a figure triple (SVG + PNG + CSV) under reports/figs/. The cascade caches per-(station, year-month) results under data/cache/, so re-runs after the first network-touching pass are fast.