Current bottom up estimates of CO 2 emission fluxes are based on a mixture of direct and indirect flux estimates relying to varying degrees on regulatory or self-reported data. Hence, it is important to use additional, independent information to assess biases and lower the flux uncertainty. We explore the use of a self-organizing map (SOM) as a tool to use multi-species observations to partition fossil fuel CO 2 ( CO 2 ff ) emissions by economic source sector. We use the Indianapolis Flux experiment (INFLUX) multi-species observations to provide constraints on the types of relationships we can expect to see, and show from the observations and existing knowledge of likely sources for these species that relationships do exist but can be complex. An Observing System Simulation Experiment (OSSE) is then created to test, in a pseudodata framework, the abilities and limitations of using an SOM to accurately attribute atmospheric tracers to their source sector. These tests are conducted for a variety of emission scenarios, and make use of the corresponding high-resolution footprints for the pseudo-measurements. We show here that the attribution of sector-specific emissions to measured trace gases cannot be addressed by investigating the atmospheric trace gas measurements alone. We conclude that additional a priori information such as inventories of sector-specific trace gases are required to evaluate sector-level emissions using atmospheric methods, to overcome the challenge of the spatial overlap of nearly every predefined source sector. Our OSSE additionally allows us to demonstrate that increasing the (already high) data density cannot solve the co-localization problem.

To effectively address climate change, aggressive mitigation policies need to be implemented to reduce greenhouse gas emissions. Anthropogenic carbon emissions are mostly generated from urban environments, where human activities are spatially concentrated. Improvements in uncertainty determinations and precision of measurement techniques are critical to permit accurate and precise tracking of emissions changes relative to the reduction targets. As part of the INFLUX project, we quantified carbon dioxide (CO 2 ), carbon monoxide (CO) and methane (CH 4 ) emission rates for the city of Indianapolis by averaging results from nine aircraft-based mass balance experiments performed in November-December 2014. Our goal was to assess the achievable precision of the aircraft-based mass balance method through averaging, assuming constant CO 2 , CH 4 and CO emissions during a three-week field campaign in late fall. The averaging method leads to an emission rate of 14,600 mol/s for CO 2 , assumed to be largely fossil-derived for this period of the year, and 108 mol/s for CO. The relative standard error of the mean is 17% and 16%, for CO 2 and CO, respectively, at the 95% confidence level (CL), i.e. a more than 2-fold improvement from the previous estimate of ~40% for single-flight measurements for Indianapolis. For CH 4 , the averaged emission rate is 67 mol/s, while the standard error of the mean at 95% CL is large, i.e. ±60%. Given the results for CO 2 and CO for the same flight data, we conclude that this much larger scatter in the observed CH 4 emission rate is most likely due to variability of CH 4 emissions, suggesting that the assumption of constant daily emissions is not correct for CH 4 sources. This work shows that repeated measurements using aircraft-based mass balance methods can yield sufficient precision of the mean to inform emissions reduction efforts by detecting changes over time in urban emissions.