AI Soil Mapping Pilots Across Three Countries

AI-assisted soil mapping is moving from research to field pilots. Three programmes in East and West Africa are worth watching for methodology, accuracy, and — most importantly — adoption.

Shared Pilot Design

• Sentinel-2 imagery as the spatial base layer
• Ground sampling at calibration points to anchor model accuracy
• Plot-level fertilizer recommendations via SMS or USSD
• Extension-officer validation loops to refine outputs

Early Lesson

Model accuracy is approaching useful ranges on key nutrients, but adoption depends on extension trust. Without an officer willing to stand behind a recommendation, the SMS arrives and little changes in the field.