Session: 2026-05-22

Summary

Extended the riverine flooding analysis to cover both Adamawa and Benue states, producing a consistent set of outputs for each. The analysis is now parameterisable by state — switching between Adamawa (Wuroboki station) and Benue (Makurdi station) reuses the same notebooks and logic.

Analysis outputs

LGA selection map (`10_riverine_flooding_analysis.ipynb`)

Map of the state showing which LGAs are selected as riverine (i.e. within 500m of the Benue river). The Benue river is overlaid, along with the locations of both the GloFAS station and the Google GRRR gauge for the state. This map frames the spatial scope of the SFED analysis below.

Floodscan SFED time series (`10_riverine_flooding_analysis.ipynb`)

Daily mean SFED across all selected riverine pixels, 1998–2025. Each year is plotted as a single seasonal line (Jan–Dec). Years exceeding the 3-, 4-, and 5-year return period thresholds (derived from wet-season Jul–Nov annual maxima) are coloured and labelled. This gives a visual read of which years had significant flooding and how the current season compares historically.

Geospatial context maps (`12_geospatial_layers.ipynb`)

Three supporting maps for each state:

Google inundation history — areas with historically recorded flood inundation from Google Flood Hub, overlaid on the LGA map
NiHSA at-risk communities — locations of communities identified by NiHSA as flood-prone, coloured by depth zone (Low / Medium / High)
WorldPop population density — 2020 population estimates at 1km resolution, clipped to the state

Notebook index

Notebook	Purpose
`10_riverine_flooding_analysis.ipynb`	LGA map + Floodscan SFED time series and RP analysis
`11_floodscan_processing.ipynb`	Floodscan COG pipeline — processes 1998–2025 archive and writes pixel-level parquet to blob
`12_geospatial_layers.ipynb`	Google inundation, NiHSA risk communities, WorldPop maps

Next steps

Trigger definitions and accuracy — Makurdi and Wuroboki

Define candidate trigger thresholds at both stations using GloFAS and Google GRRR reanalysis
Produce accuracy plots (hit rate, false alarm rate, lead time) for each station and data source, following the same approach as 08_readiness_trigger_wurobokki.ipynb
Compare GloFAS vs Google performance at Makurdi and Wuroboki side-by-side
Evaluate a combined OR trigger and its overall return period