Context
An international NGO running food-security programs across a multi-country region needed earlier warning of emerging drought conditions than their existing quarterly reporting cycle could provide.
Approach
We built an automated pipeline computing SPI (1/3/6-month) from CHHIRPS precipitation data and Vegetation Condition Index from MODIS NDVI, refreshed monthly and surfaced through a lightweight web dashboard with region-level alert thresholds.
Outcome
The dashboard is now the NGO's primary early-warning tool, triggering field assessment visits in 3 regions ahead of visible crop stress.
Methodology
Automated monthly SPI computation (1/3/6-month) from CHIRPS precipitation via Python; MODIS NDVI-derived Vegetation Condition Index computation via Google Earth Engine; alert threshold calibration against 8 years of historical drought-declaration records; dashboard delivery via a lightweight Next.js/Mapbox GL frontend.
Results & Outcomes
- Automated monthly drought index computation replacing a manual quarterly process
- Triggered early field assessments in 3 regions ahead of visible crop stress
- Reduced time-to-alert from ~90 days to under 30 days
Challenges
Balancing alert sensitivity against false-positive fatigue required calibrating thresholds against 8 years of historical drought-declaration records rather than using generic index thresholds from the literature.
“Their SWAT calibration held up under donor technical review with zero revisions requested — that doesn't happen often.”
Thomas Berg, Water Resources Lead
An International NGO