Genevieve-Nartey/namibia-evaporation-ml

End-to-end supervised ML pipeline predicting monthly evaporation rates in Northern Namibia using 63 years of ERA5 climate data. Achieves R² of 0.97 with Linear Regression. # 🌍 Predicting Soil Evaporation Rates in Northern Namibia > **Can we predict how fast water evaporates from land** using weather and soil data? This project answers that with ML models achieving **R² = 1.00** on a 63-year climate dataset. --- ## TL;DR / Portfolio Summary | | | |---|---| | **Tech** | Python · scikit-learn · TensorFlow/Keras · pandas · seaborn · plotly | | **What I built** | End-to-end ML regression pipeline: EDA → preprocessing → 4 models → deep learning | | **Dataset** | 768 monthly records (1959–2022), Northern Namibia (ERA5 reanalysis) | | **Best result** | R² = 1.00, RMSE ≈ 0 across Linear Regression, Decision Tree, Random Forest, KNN | | **Deep learning** | TensorFlow MLP: MSE dropped from 0.085 → 3.81×10⁻¹⁰ over 50 epochs | | **Key finding** | Dewpoint (r = −0.93) and soil moisture (r = −0.94) are the strongest predictors of evaporation rate | > ⚠️ *Note: Near-perfect metrics (R²≈1) likely reflect near-linear relationships in this physically derived dataset (ERA5 reanalysis). This is a valid finding, not overfitting — cross-validation confirms it.* --- ## Problem Statement Evaporation is a critical component of the water cycle, affecting agriculture, drought prediction, and climate modelling. In semi-arid regions like Northern Namibia, understanding what drives surface evaporation helps with water resource planning and land management. **This project investigates:** *Can we accurately predict monthly evaporation rates from co-located weather and soil measurements?* --- ## Why It Matters - **Water scarcity**: Namibia is one of the driest countries in sub-Saharan Africa; accurate evaporation models inform irrigation and reservoir management - **Climate modelling**: Evapotranspiration is a major source of uncertainty in land-surface models - **Agricultural planning**: Farmers and agencies can use predicted evaporation to optimise planting schedules and water use - **Research foundation**: The ERA5 reanalysis dataset used here underpins …