# Methodology and evidence notes

## Scope

This story compares climate cause, exposure, and response across Pacific Island countries and territories. The charts use the JSON files in `data/`; `verify_evidence.py` checks the displayed headline numbers directly against those files.

The two emissions measures are deliberately kept separate:

- Our World in Data CO₂ per person is used for the global comparison and the 0.046% Pacific share.
- Pacific Data Hub greenhouse-gas emissions per person, expressed in CO₂-equivalent, is used in the Act I cross-country test.

They are related measures, but they are not interchangeable.

## Reproducible calculations

- **Pacific global CO₂ share:** sum for the 17 of 21 targeted island countries and territories available in OWID, divided by the world total. Displayed as 0.05%; unrounded value 0.046%.
- **Sea-level change:** change in the regional mean anomaly from 1993 to 2023, approximately 12.38 cm.
- **Act I test:** Pearson correlation across 17 countries between 2014–2023 mean GHG emissions per person and mean sea-level anomaly. Result: r = −0.296; two-sided permutation p = 0.282.
- **Low-elevation exposure:** national population share in the 0–10 m low-elevation coastal zone.
- **Disaster event shares:** directly affected people divided by the nearest available Pacific Data Hub national population year. Event counts can overlap categories and percentages should not be interpreted as unique people across multiple events.
- **Drinking water:** median of 15 available country series; WHO/UNICEF JMP modelled estimates.
- **Solar growth:** 602.4 GWh in 2023 divided by 1.4 GWh in 2000 = 430.3×.
- **2023 fossil share:** fossil generation divided by all five displayed generation groups = 72.6%.

## Interpretation limits

The Act I test was used as a challenge to the narrative before publication. It does not prove that emissions and climate impacts have “nothing to do” with each other, nor does it estimate causality. It asks a narrower question: whether differences in a Pacific country's own emissions account for the observed cross-country pattern of sea-level anomaly in this sample. They do not show a statistically detectable association here. That result narrowed the public claim toward the mismatch between local contribution and exposure.

The exploratory heatmap is shown only as an audit artifact: it documents that the team looked for alternative explanations before narrowing the story. The earlier “124 of 136” claim is still excluded from the public claim set because the bundle did not contain the pairwise result table, relationship threshold, or heatmap-generating code needed to reproduce that count.

Anomalies inherit the reference baselines encoded by the Pacific Data Hub series. Coverage differs by indicator, so every chart states its sample or source scope where it matters.

## Sources

- Pacific Data Hub (PDH.stat), Pacific Community: `DF_CLIMATE_CHANGE`, `DF_POP_LECZ`, `DF_SDG_13`, `DF_SDG_06`, `DF_POWER_GEN`, and `DF_NMDI_POP`. Accessed 21 June 2026.
- Our World in Data, “CO₂ and Greenhouse Gas Emissions,” based on the Global Carbon Project. Accessed 21 June 2026. CC BY 4.0.
- Hero image: NASA Goddard / EO-1, Tikehau Atoll, French Polynesia.

## Reproduce the evidence gate

From this folder, run:

```powershell
python verify_evidence.py
python -m http.server 8000
```

The first command exits with an error if a headline claim drifts from the live JSON and rewrites `data/evidence_manifest.json` after a successful check.

## AI disclosure

AI assisted early visual exploration and implementation review. The team remains responsible for the final data selection, calculations, narrative, and published output.