<p>We have constructed an atmospheric inversion framework based on TM5 4DVAR to jointly assimilate measurements of methane and <em>δ</em><sup>13</sup>C of methane in order to estimate source-specific methane emissions. Here we present global emission estimates from this framework for the period 1999–2016. We assimilate a newly constructed, multi-agency database of CH<sub>4</sub> and <em>δ</em><sup>13</sup>CH<sub>4</sub> measurements. We find that traditional CH<sub>4</sub>-only atmospheric inversions are unlikely to estimate emissions consistent with atmospheric <em>δ</em><sup>13</sup>CH<sub>4</sub> data, and assimilating <em>δ</em><sup>13</sup>CH<sub>4</sub> data is necessary to deriving emissions consistent with both measurements. Our framework attributes <em>ca</em>. 85 % of the post-2007 growth in atmospheric methane to microbial sources, with about half of that coming from the Tropics between 23.5° N and 23.5° S. This contradicts the attribution of the recent growth in the methane budget of the Global Carbon Project (GCP). We find that the GCP attribution is only consistent with our top-down estimate in the absence of <em>δ</em><sup>13</sup>CH<sub>4</sub> data. We find that at global and continental scales, <em>δ</em><sup>13</sup>CH<sub>4</sub> data can separate microbial from fossil methane emissions much better than CH<sub>4</sub> data alone can, and at smaller scales this ability is limited by the current <em>δ</em><sup>13</sup>CH<sub>4</sub> measurement coverage. Finally, we find that the largest uncertainty in using <em>δ</em><sup>13</sup>CH<sub>4</sub> data to separate different methane source types comes from our knowledge of atmospheric chemistry, specifically the distribution of tropospheric chlorine and the isotopic discrimination of the methane sink.</p>