Probabilistic Methane Source Attribution In Greater Accra: A Hierarchical Bayesian Dual-Isotope Approach

Abstract

Methane is a potent greenhouse gas requiring accurate source apportionment to inform targeted mitigation, particularly in rapidly urbanizing regions where emissions arise from multiple overlapping sectors. This study aimed to quantify the methane source mix in Greater Accra, Ghana, using a hierarchical Bayesian mixing model constrained by methane mole fractions and dual stable isotope signatures (δ 13 C-CH 4 2 and δ H-CH 4 ). Atmospheric methane was characterised using mobile cavity ring-down spectroscopy across 26 localities during March 2023, with observations apportioned among four aggregated source classes: Biogenic, Fossil Gas Leaks, Combustion/ Transport, and Industrial/Enriched. Results were reported as posterior distributions to transparently represent attribution uncertainty. The posterior indicated Fossil Gas Leaks as the dominant contributor, with a mean contribution of 34.2% (95% HDI: 10.5–56.6%) and a 62.9% probability of being the maximal source, followed by Biogenic sources at 26.7% (95% HDI: 11.7–43.5%). Combustion/Transport and Industrial/Enriched showed substantial posterior overlap, reflecting weak separability between these anthropogenic categories. Locality-level classifications revealed high-confidence biogenic attribution at documented landfill sites (Ada West, Adenta, Ga South), while mixed-source localities exhibited lower posterior confidence. Cross-validation with Keeling- derived attributions yielded 62% agreement, with strongest consistency for Biogenic and Fossil Gas Leaks cat egories. This study demonstrates that dual-isotope Bayesian source apportionment can provide actionable, uncertainty-aware guidance for methane mitigation prioritization in data-limited urban environments,