nourbadraoui2017-cloud/crop-yield-morocco

# Satellite-Based Crop Yield Prediction ## The Problem Over 60% of Morocco's rural population depends directly on agriculture, mostly small farmers working plots under 5 hectares, mostly rain-fed rather than irrigated. Morocco just came out of a seven-year drought, and agriculture consumes roughly 84% of the country's water — the entire system is highly exposed to rainfall variability. The core issue: nobody in the chain has reliable early information about how the harvest will turn out. - **Farmers** don't know until harvest whether the season is good or bad, so they can't plan — whether to sell early, hold, seek off-farm income, or adjust next season's planting. - **Insurers** can't price fair drought insurance for small farmers because assessing risk plot-by-plot (site visits, manual surveys) is too expensive relative to small policy sizes. Today, only about **3% of small farmers in Morocco** have any climate risk insurance. - **Cooperatives and government** only learn about a bad harvest once it's happening — too late to plan storage, imports, or aid distribution ahead of time. Existing satellite-based crop monitoring tools (e.g. USDA, FAO systems) are built for large-scale national/continental forecasting, not packaged affordably for a local insurer or cooperative in Morocco. The gap isn't the science — it's that nobody has adapted it for this specific, smaller-scale use case. ## The Solution Satellites (e.g. Sentinel-2) photograph farmland regularly and for free. Healthy, well-growing crops reflect light differently than stressed or sparse ones — greener, denser. By tracking this "greenness" (via a vegetation index like NDVI) throughout a growing season and comparing it to historical yield outcomes, you can predict this season's yield *before* harvest, instead of waiting until it's too late. ### How it works, step by step 1. **Continuous data collection** — pull new satellite imagery every few days for the target region, combined with rainfall/tempera …