misalisu-ai/nigeria-maize-weather-yield-data

# Nigeria Maize Weather-Yield Dataset Reproducible state-year dataset linking Nigerian maize yield observations with seasonal agro-meteorological features derived from NASA POWER daily data. ## Dataset v1.0 - States: 37 - Years: 2020–2024 - Observations: 185 - Target: `Yield_MT_Ha` - Weather source: NASA POWER - QC status: PASS WITH DOCUMENTED WARNINGS - Frozen SHA-256: `09ab62faed51d7f391595068587d33595f49058ae1687621925cf12408c2f97c` ## Repository purpose This repository contains the data-engineering and provenance layer: coordinates, season windows, extraction, feature engineering, QC history, QC outputs, provenance metadata, and the frozen Dataset v1.0. Modeling lives in the companion repository: `nigeria-maize-weather-yield-model`. ## Final dataset `data/processed/nigeria_maize_weather_yield_2020_2024_v1.0.csv` Do not modify Dataset v1.0. Derived datasets belong in the modeling repository. ## Known warnings - `Mean_Tmax_C` and `Mean_RH_pct` are highly correlated. - Ekiti and Ondo retain an identical weather profile after raw-source audit; this is documented as a spatial-resolution characteristic. - 18 observations have `|Rainfall_Anomaly_Z_2000_2019| > 5` and are retained for robustness/sensitivity analysis. - The pre-repair raw archive was not preserved; yield preservation was independently verified by QC v2.5.7. ## QC history `v2.1 → v2.2 → v2.3 → v2.4 → v2.5 → v2.5.1 → v2.5.2 → v2.5.3 → v2.5.4 → v2.5.5 → v2.5.6 → v2.5.7 → v2.6.1 → v2.6` See `docs/qc_history.md`.