OCHA-DAP/ds-external-aa-frameworks

Review of external (WFP/FAO) Southern Africa drought AA triggers: recovered activation records + return-period check (password-protected site) # ds-external-aa-frameworks Independent review of external (WFP / FAO) anticipatory-action drought frameworks in Southern Africa — Mozambique, Zimbabwe, Malawi, Madagascar, Lesotho — ahead of the CERF El Niño 2026/27 trigger-harmonisation exercise. Recovers the trigger/activation records that back the agencies' stated return periods from public endpoints: - **PRISM operational trigger CSVs** (public S3; URLs from the open-source WFP-VAM/prism-app frontend config): per district × index × severity × window trigger thresholds and issued forecast probabilities, seasons 2023-24 onward (MOZ, ZWE, MWI). - **IRI fbfmaproom2 export API** (backend of the AA design maprooms): full hindcast of forecast probabilities vs design thresholds and "bad years", with skill counts (LSO 1983–, MDG 1992–). From these it computes realized/hindcast activation series and Weibull return periods, compares them against the RP labels in the July 2026 FAO-WFP overview file (internal, not committed), and publishes a password-protected review site. Related KB pages: `ds-knowledge-base/external-frameworks/{wfp,fao}/*-drought.md`. ## Layout - `scripts/fetch.py` — download PRISM CSVs + maproom exports (national + district level, incl. the Malawi maproom) into `data/` (not committed) - `scripts/analyze.py` — activation series + Weibull RPs per region → `data/derived.json` - `scripts/fetch_met.py` — ERA5/SEAS5 adm0 + target-adm1 zonal means from the team rasterstats DB (`public.era5`/`public.seas5`; needs stratus prod DB env vars) - `scripts/analyze_met.py` — independent met cross-check (were triggered years dry? is the 2026/27 forecast dry?) → `data/derived_met.json` - `scripts/plots.py` — indicator-vs-threshold time-series SVGs → `data/plots/` (inlined into the pages so they are encrypted with the HTML) - `scripts/build_site.py` — render the review site into `site_build/` ## Running ```sh uv run python scripts/fetch.py uv run python scripts/analyze.py uv run python scripts/fetch_met.py # ne …