GaiSamuel/MIT_Emerging_Talent_ELO2-Uganda-Districts-Rainfall-Prediction-Using-Machine-Learning

This project predicts rainfall for Uganda's districts across the four regions from 2025_10 to 2026_12 to support farmers and the communities plan better. # **Uganda Districts' Rainfall Prediction using Machine Learning** ## 1. **The Problem Identification** ### The Problem Uganda’s economy and food security are fundamentally tied to rain-fed agriculture, making the country highly sensitive to climate variability. The recent Uganda National Meteorological Authority (UNMA) outlook predicts "above-normal" (enhanced) rainfall for the SOND 2025 season across most of the country. While increased rainfall is generally beneficial, this excess introduces critical risks that farmers are currently unprepared for, including water logging, nutrient leaching, and significant post-harvest losses due to poor drying conditions. In specific high-risk zones like the Elgon and Kigezi sub-regions, this enhanced rainfall threatens crop destruction through landslides and flash floods. The core issue is that without precise, district-level predictions, farmers lack the granular data needed to effectively time planting or prepare drainage infrastructure to mitigate these specific local excesses. ### **Research Question** > **Can we use historical ERA5 climate data and Climate Hazards data (CHIRPS) to** **generate granular, district-level rainfall forecasts that allow farmers to** **anticipate and mitigate the risks of "above-normal" rainfall events?** ### Domain Knowledge *Based on the UNMA SOND 2025 Seasonal Rainfall Outlook%202025%20Seasonal%20Rainfall%20Outlook%20(1)_compressed.pdf):* - **Key Risk 1: Agronomic Instability:** The forecasted "above-normal" rains pose a high risk of soil nutrient leaching and water logging, which can rot root crops like beans and cassava if drainage is not managed. - **Key Risk 2: Post-Harvest Loss:** Continuous heavy rains during harvest periods create poor drying conditions, leading to mold and storage losses for cereal crops. - **System View (Geographic Vulnerability):** The impact is not uniform; mountainous regions (Elgon, Kigezi, Rwenzori) face physical destruction of farmland via landslides, …