Methane Emissions in Melbourne, Australia. Part 2: Improving Prior Estimates Using Synthesis Inversion

Abstract Urban methane emissions remain highly uncertain due to the complexity of overlapping anthropogenic sources and limitations in current bottom‐up inventories. In this study, we apply a synthesis inversion framework to refine sector‐specific methane emissions for the Melbourne region in Australia, using in situ atmospheric observations and a Bayesian inversion approach. Our method incorporates constrained and unconstrained inversions, observational filtering, and ensemble‐based uncertainty quantification, including both ensemble spread and posterior covariance structure. The inversion indicates that prior methane emissions are substantially underestimated, particularly for emissions from gas infrastructure, which are inferred to be approximately five to seven times larger than prior estimates. Posterior emission estimates show substantial improvements in agreement with observations, as evidenced by reduced root mean square error (RMSE), increased correlation, and lower model bias across five receptor sites. Posterior methane emissions for the Melbourne region range from 222 to 253 kt/year across the six study months, representing a 67%–90% increase relative to the prior inventory of 133.5 kt/year. Sensitivity analyses confirm that these total emission estimates are robust to prior perturbations and inversion configuration choices. However, source attribution remains sensitive in cases of spatial overlap or low observational coverage. This study demonstrates the effectiveness of synthesis inversion in urban methane quantification and highlights its value for supporting targeted mitigation strategies and improving emission reporting in urban Australia.