Magguire/Africa-Rainfall-Predictability

# Africa Rainfall Predictability > **Finding the most predictable rainfall region in Africa using a causal, cross-validated multi-driver regression approach.** --- ## Overview This project identifies **where in Africa inter-annual wet-season rainfall is most predictable** from large-scale ocean-atmosphere climate drivers known *before* the season begins, producing skill that is actionable for seasonal forecasting. --- ### SST Predictors Six de-collinearised indices are used in joint regression: | Index | Region | Notes | |---|---|---| | **Niño3.4** | Central equatorial Pacific | Core ENSO indicator | | **DMI** | Indian Ocean Dipole (west − east) | Constructed as dipole to avoid collinearity | | **SIOD** | Subtropical Indian Ocean Dipole | SW Indian Ocean influence on southern Africa | | **TAG** | Tropical Atlantic Gradient (TNA − TSA) | Atlantic cross-equatorial SST contrast | | **Atl3** | Equatorial Atlantic | Atlantic Niño | | **Benguela** | SE Atlantic upwelling zone | Benguela Niño/Niña | --- ## Project Structure ``` Africa-Rainfall-Predictability/ ├── search_region.ipynb # Main analysis notebook ├── data/ │ ├── noaaoisst.mon.mean.nc # NOAA OI SST (monthly) │ └── cmapprecip.mon.mean.nc# CMAP precipitation (monthly) ├── assets/ # All figures and CSV output │ ├── map_wet_seasons.png │ ├── map_predictability.png │ ├── map_dominant_driver.png │ ├── top_pixel_skill.png │ ├── top_pixel_attribution.png │ ├── field_significance_null.png │ ├── sst_indices_timeseries.png │ ├── sst_indices_corr.png │ ├── wet_season_validation.png │ ├── summary_dashboard.png │ └── skill_results.csv ├── requirements.txt └── README.md ``` --- ## Getting Started ### Prerequisites - Python 3.8 or higher - The two NetCDF data files (see Data Sources below) ### Installation ```bash # 1. Clone the repository git clone github.com cd Africa-Rainfall-Predictability # 2. (Recom …