Satellite-based precision agriculture
Monitor crop health and forecast yields at regional scale, from space.
The challenge
In West Africa, agriculture employs a major share of the population yet remains vulnerable: water stress, pests, climate variability. Decisions — irrigation, fertilisation, harvest timing — are often made without objective data on the real state of plots scattered across vast, hard-to-reach territories.
For crop insurers and development funders, the lack of reliable yield and loss measurement complicates pricing, claims and aid targeting. What is missing is a way to track crop health and productivity at large scale and low cost.
Our approach
ADST leverages free, recurring satellite imagery — notably Sentinel-2 — to monitor crops plot by plot. We compute vegetation indices (NDVI, EVI, water-stress indices) and track their evolution over time to detect anomalies early: stress, disease, growth delays.
By cross-referencing these image time series with weather and soil data, we train yield-forecasting models that estimate expected production well before harvest, at both plot and regional scale.
These indicators power concrete use cases: agronomic alerts for producers, index insurance automatically triggered in case of drought, impact monitoring for development programmes, and mapping of cultivated areas for public policy.
Architecture
- Sources: Sentinel-2 (optical), Sentinel-1 (radar), weather and soil data
- Processing: index computation (NDVI/EVI), cloud correction, per-plot time series
- Models: regression and LSTM/Transformer for yield forecasting, CNN for crop classification
- Delivery: web mapping, early alerts, API for index insurance
A crop insurer automates index-insurance triggering and cuts field-assessment costs while speeding up farmer payouts.