This project models how drought conditions, measured using precipitation‑derived indices, affect agricultural GDP per capita growth in Chad, Nigeria and Cameroon. We use Generalised Additive Models (GAMs) to capture non‑linear and lagged responses, while accounting for structural breaks in each economy.

Agriculture accounts for over 30% of Chad’s GDP and ~17% of Nigeria’s and Cameroon’s, making these economies vulnerable to climate variability. Since the 1960s, Lake Chad has shrunk by over 90%, intensifying dependence on annual rainfall. Many studies pool African countries together; our project fills the gap by analysing country‑specific impacts.

• Precipitation data: monthly rainfall from GPCC or other gridded climate datasets (stored locally in data/raw/).
• Agricultural GDP per capita: annual data from the World Bank.
• Data are processed and matched by time and country; raw files are excluded from the repo (see data/processed/README.md for schema and download instructions).

• Drought indices: compute Standardised Precipitation Index (SPI‑12) and Seasonal SPI using xarray, scipy and pymannkendall.
• Generalised Additive Models (GAMs): fit smooth terms for drought metrics, lags, time and country effects using pygam / statsmodels.
• Model specification allows for non‑linear thresholds, lagged recovery dynamics and structural breaks in macroeconomic regimes.
• Notebook DATA_EDA.ipynb explores data and computes drought metrics; 2Analysis-2.ipynb fits GAMs and evaluates results.

• Chad: strong, non‑linear drought effects – agricultural GDP losses accelerate once rainfall deficits cross a threshold.
• Nigeria: limited direct sensitivity to drought, but significant rebound dynamics in lagged drought severity.
• Cameroon: drought metrics not significant, suggesting macroeconomic shocks dominate agricultural outcomes.
• Full tables and figures are included in reports/lake_chad_drought_gdp_modelling.pdf.