abinittio/SoilIntelligence

Predicts quality of soil for any coordinate in Africa and recommends fertiliser to maximise yield # SoilIntelligence Precision agriculture decision support system. Analyses soil nutrient levels, predicts crop yield with uncertainty quantification, and prescribes exact fertiliser quantities — all backed by 244,000+ global soil samples and real-time market data. ## What it does 1. **Soil analysis** — Assesses N, P, K, Ca, Mg, pH, organic carbon and gives an overall quality score 2. **Yield prediction** — Monte Carlo forecasting using Liebig's Law (limiting factor) with 90% confidence intervals 3. **Fertiliser recommendations** — Exact product quantities (DAP, Urea, MOP) with cost and ROI 4. **Regional benchmarking** — Compares your soil against 49k real African + 195k global reference samples 5. **Risk assessment** — Probability statements ("82% chance of positive ROI") with LOW/MODERATE/HIGH tiers ## How it works ``` Soil test results or GPS coordinates ↓ Nutrient analysis + quality scoring ↓ Monte Carlo yield prediction (1000 simulations, 90% CI) ↓ Fertiliser prescription + ROI calculation ↓ PDF report for farmer ``` ## Data sources | Source | Coverage | Size | |--------|----------|------| | iSDA Africa | 49,225 field samples with measured nutrients | 10 MB | | WoSIS Global | 195,000+ modelled samples worldwide | 14 MB | | SoilGrids API | 250m resolution soil properties (ISRIC) | Live | | Open-Meteo | Weather forecasts and soil moisture | Live | | FAO GIEWS | Crop price data | Live | | World Bank | Fertiliser commodity prices | Live | ## Running locally ```bash pip install -r requirements.txt python web.py ``` Open 127.0.0.1. ## Tech stack - **Backend**: Flask, pandas, numpy, scipy - **Frontend**: HTML/CSS/JS, Leaflet.js maps - **ML**: Random forest benchmarking, Liebig's Law yield model, Monte Carlo uncertainty - **Reports**: fpdf for PDF generation ## License MIT