High-resolution inversion of urban methane emissions in the Chengdu–Chongqing economic circle using ground-based observations and a dynamic error Bayesian framework

Methane (CH 4 ) is a potent greenhouse gas and a major target for near-term climate mitigation. However, bottom-up inventories often overestimate emissions, particularly at the urban scale. Here, we present a high-resolution inversion of anthropogenic and wetland CH 4 emissions in the Chengdu–Chongqing Economic Circle (CCEC), a rapidly urbanizing region in southwestern China. We developed a coupled WRF-STILT modeling framework integrated with hourly ground-based CH 4 measurements and the EDGAR v8.0 emission inventory. To reduce uncertainty, we implemented a dynamic error-weighted Bayesian inversion algorithm that adapts to seasonal and spatial variability. Our approach yields posterior methane emissions at 10 km resolution and daily timescales for the year 2021. The EDGAR inventory overestimated anthropogenic and wetland CH 4 emissions by 37.28% and 25.81%, respectively. The dynamic error-based method significantly outperformed the fixed-error approach, improving the correlation (R) between simulated and observed concentrations from 0.41 to 0.62 and reducing RMSE from 127.51 ppb to 78.80 ppb. Sectoral analysis highlights agriculture and waste as dominant contributors, with distinct seasonal dynamics. High-emission zones correspond to major urban centers, while wetland emissions peaked during summer. This study demonstrates the value of combining atmospheric transport modeling with dynamic uncertainty quantification for methane monitoring in urban regions. The findings support more accurate emission inventories and inform targeted mitigation strategies under China's methane reduction commitments.