badIS-6/Tunisia-Cereal-Forecasting-ML

Machine Learning framework for forecasting cereal production in Tunisia using climate data, NDVI/LST indicators, and governorate-level production statistics. # Machine Learning-Based Cereal Production Forecasting in Tunisia Machine Learning framework for forecasting cereal production in Tunisia using **ERA5-Land climate data**, **MODIS satellite indicators**, and governorate-level production statistics. Developed as part of the **Smart SDG Tunisia** project. ## Overview The workflow combines climate and remote-sensing indicators to model cereal production at the Tunisian governorate level. ### Environmental predictors * Rainfall: October–November * Rainfall: January–March * Temperature: December * Soil moisture: October–November * NDVI: February–April * VCI: Vegetation Condition Index * VHI: Vegetation Health Index The study period is **2000–2018**. ## Workflow ```text ERA5-Land + MODIS ↓ Climate & vegetation indicators ↓ Governorate-level dataset ↓ Cereal production data ↓ Machine learning models ↓ Production prediction & forecasting ``` ## Models Four models are compared: * Linear Regression * Random Forest * XGBoost * Support Vector Regression (SVR) ## Data sources * **ERA5-Land Daily Aggregated** * **MODIS MOD13Q1 NDVI** * **MODIS MOD11A2 LST** * Tunisian governorate boundaries * Cereal production statistics ## Outputs The project generates: * ML-ready datasets * Model performance metrics * Test predictions * Feature importance * Historical production analysis * Five-year production projections **Smart SDG Tunisia Project**