GIK-IceChain

Zero-Cost Cloud-Native Pipeline for Retrospective Flood Risk Decision Support in East Africa

GIK-IceChain unlocks over one petabyte of ECMWF IFS ensemble forecast data, currently inaccessible to modern data science tools, and transforms it into an operational flood risk decision support system for East Africa. Building on existing GIK Parquet manifests produced by ICPAC, the pipeline converts GRIB2 references into a fully interoperable IceChunk Zarr v3 store with zero data duplication, enabling single-line access to ~1,200 days of forecasts. The system then computes retrospective exceedance probabilities using adaptive GEV thresholds, delivers an interactive calendar-map visualisation, and integrates the ICPAC CRMA Bayesian network to produce daily admin-1 flood risk scores across 11 East African countries — validated against EM-DAT historical flood records.