lokiray4/Meteorological-Variable-Importance-Analysis-and-Spatial-Prediction-of-West-African-Monsoon-Rainfall

Repository for my 2025–2026 MSc Data Science dissertation at the University of Leeds: Meteorological Variable Importance Analysis and Spatial Prediction of West African Monsoon Rainfall Using ERA5 Reanalysis Data. # West African Monsoon Rainfall Analysis This repository contains the code for my 2025-2026 MSc Data Science dissertation at the University of Leeds: Meteorological Variable Importance Analysis and Spatial Prediction of West African Monsoon Rainfall Using ERA5 Reanalysis Data The project uses ERA5 reanalysis data to analyse which meteorological variables are most related to West African monsoon rainfall, and then compares several models for spatial rainfall prediction. ## Notebooks The original notebook was too large, so I split it into smaller notebooks: 1. 00_data_preprocessing_study_area.ipynb Loads the ERA5 data and defines the study area, time period, and variables. 2. 01_chapter4_mean_fields_and_correlations.ipynb Produces the mean spatial fields and correlation analysis used in Chapter 4. 3. 02_chapter4_ridge_variable_importance.ipynb Runs the Ridge-based variable importance experiments, including single-variable and leave-one-variable-out analysis. 4. 04_chapter5_model_comparison_spatial_prediction.ipynb Trains and compares Linear Regression, Ridge Regression, Simple CNN, Residual CNN, and CBAM-Residual CNN models. 5. 03_appendix_geographical_regional_analysis.ipynb Contains extra regional and geographical analysis for the appendix. ## Data The notebooks expect ERA5 .npy files stored locally. The paths in the notebooks currently point to my local folders, so they may need to be changed before running on another computer. Main data folders used in the notebooks: /Users/camus/Desktop/dissertation/variable/npy /Users/camus/Desktop/dissertation/variable/npy_21x37_mean ## Requirements The main Python libraries used are: - numpy - pandas - matplotlib - scikit-learn - scipy - torch - seaborn - cartopy - shapely ## Notes All notebook outputs were cleared to keep the repository small and easier to upload to GitHub.