Braim016/nigeria-methane-spatiotemporal

Spatiotemporal analysis of atmospheric methane over Nigeria using six years of TROPOMI/Sentinel-5P and MERRA-2 data. SARIMA and gradient boosting forecasting, HDBSCAN and Getis-Ord Gi* hotspot detection, SHAP interpretation. # Spatiotemporal Analysis of Atmospheric Methane Over Nigeria (2018 to 2024) Six years of TROPOMI/Sentinel-5P column methane retrievals fused with NASA MERRA-2 reanalysis to characterise where and when methane concentrates over Nigeria, which meteorological and land surface drivers explain it, and how the national mean has moved over the study period. ## Key findings - Mean column methane over Nigeria rose from **1869.29 ppb in 2018 to 1935.17 ppb in 2024**, an increase of **3.52%**. - A **SARIMA(0,1,1)(0,1,1,12)** model on the monthly national mean achieved **R² of 0.9977**, **MAPE of 0.15%** and **RMSE of 4.23 ppb** on a six month holdout. - Of five benchmarked regressors, the tree based models led the field, with the best achieving **R² above 0.93** on a held out 20% split. SHAP attribution ranked spatial and land surface moisture features above aerosol and radiation terms. - **HDBSCAN** clustering combined with **Getis-Ord Gi\*** statistics (p 0.5, bias corrected retrieval used throughout ## Pipeline ``` 01_data_download Authenticated retrieval of TROPOMI and MERRA-2 granules 02_preprocessing NetCDF parsing, flattening, QA filtering, clipping to Nigeria, CSV export 03_spatial_analysis Regridding onto a 0.063° mesh, monthly composite choropleths 04_temporal_analysis Inter-annual trend, wet and dry season contrast, STL, SARIMA 05_hotspot_detection HDBSCAN clustering and Getis-Ord Gi* hotspot statistics by year 06_machine_learning MERRA-2 to TROPOMI interpolation, five model benchmark, SHAP attribution ``` Notebooks are numbered in execution order. `02` writes the two CSVs that every later notebook reads, so it must run first. ## Reproducing ```bash git clone github.com cd nigeria-methane-spatiotemporal # conda is recommended: geopandas and cartopy pull binary geospatial dependencies conda env create -f environment.yml conda activate methane cp src/config.example.py src/conf …