Machine learning framework for estimating global Surface Water Fraction (SWF) from passive microwave brightness temperatures (WindSat/CIMR). Covers end-to-end data preprocessing, EDA, and a full regression pipeline including model selection, feature engineering, HPO, SHAP explainability, and error analysis.
data-science machine-learning geospatial regression xarray remote-sensing xgboost feature-engineering satellite-data earth-observation masters-thesis shap passive-microwave optuna cimr surface-water-fraction windsat lpdr
-
Updated
May 10, 2026 - Jupyter Notebook