Independent estimates of fossil fuel CO 2 (CO 2 ff) emissions are key to ensuring that emission reductions and regulations are effective and provide needed transparency and trust. Point source emissions are a key target because a small number of power plants represent a large portion of total global emissions. Currently, emission rates are known only from self-reported data. Atmospheric observations have the potential to meet the need for independent evaluation, but useful results from this method have been elusive, due to challenges in distinguishing CO 2 ff emissions from the large and varying CO 2 background and in relating atmospheric observations to emission flux rates with high accuracy. Here we use time-integrated observations of the radiocarbon content of CO 2 ( 14 CO 2 ) to quantify the recently added CO 2 ff mole fraction at surface sites surrounding a point source. We demonstrate that both fast-growing plant material (grass) and CO 2 collected by absorption into sodium hydroxide solution provide excellent time-integrated records of atmospheric 14 CO 2 . These timeintegrated samples allow us to evaluate emissions over a period of days to weeks with only a modest number of measurements. Applying the same time integration in an atmospheric transport model eliminates the need to resolve highly variable short-term turbulence. Together these techniques allow us to independently evaluate point source CO 2 ff emission rates from atmospheric observations with uncertainties of better than 10%. This uncertainty represents an improvement by a factor of 2 over current bottom-up inventory estimates and previous atmospheric observation estimates and allows reliable independent evaluation of emissions.fossil fuel CO 2 | radiocarbon | greenhouse gas emissions | emission verification F ossil fuel carbon dioxide (CO 2 ff) emissions are the main driver of the increasing atmospheric CO 2 mole fraction (1). Of the ∼10 GtC/y of CO 2 ff now emitted globally, the largest 1,000 power plants emit 22% of the total (2). Thus, large power plants are an obvious target for regulating and reducing CO 2 ff emissions. They are already subject to regulation or emission trading schemes in some regions (3, 4), and the Paris Agreement requires transparency in emission reporting (5). The success of such regulations requires the ability to reliably monitor and verify emissions, which is currently achieved through "bottom-up" inventory data, in which CO 2 ff emission estimates are based on self-reported fuel use and carbon content statistics (6). Uncertainties in this method are on the order of 20% for individual power plants (7). Some studies suggest significant errors in emission reporting may already be occurring (8, 9) and under a regulatory environment, there may be incentive for deliberate misreporting. Independent, objective evaluation of emissions is needed to establish trust and ensure that self-reported bottom-up emission rates are unbiased (5, 10, 11). As power plants represent such a large proportion of total emissions, reducing t...