Estimating greenhouse gas sources and sinks using atmospheric observations: From surface to space

Abstract

Carbon dioxide (CO2) and methane are the two strongest anthropogenic greenhouse gases (GHGs) and account for more than 85% of the GHG forcing since pre-industrial times. As such, their current and future emissions can have a profound impact on the future state of our climate. Quantifying their emissions is critically important for both projecting future climate and assessing the impact of environmental policy. This thesis develops novel methods for solving large-scale inverse problems, such as those encountered when estimating GHG emissions with atmospheric observations, quantifies the magnitude and spatio-temporal distributions of GHG sources and sinks, and develops statistical models that characterize the importance of precision and network density for quantifying urban GHG emissions. I quantify aggregation and smoothing errors as a function of state vector dimension. I then compare three methods for reducing the dimension of the state vector from its native resolution: (1) merging adjacent elements, (2) clustering with principal component analysis, and (3) using radial basis functions (RBFs) with Gaussian kernels. The RBF method is shown to retain resolution of major local features while smoothing weak and broad features. This novel RBF method is used to estimate North American methane emissions with satellite observations. Posterior emissions are found to be in good agreement with independent observations. US methane emissions are found to be underestimated in current inventories and this underestimate may be due to an increasing trend in US methane emissions. On the global scale, I find that the most likely explanation for the renewed growth of methane since 2007 involves a decrease in OH that is partially offset by a decrease in methane emissions. However, I also demonstrate that the problem of attributing methane trends from the current surface observation network is under-determined and does not allow for unambiguous attribution of decadal trends. Finally, this thesis addresses urban CO2 emissions by: (1) quantifying the trade-off between precision and network density and (2) characterizing the ability of various measurement networks to quantify urban CO2 emissions. I develop statistical models that estimate the efficacy of the combined model-measurement system in reducing uncertainty in CO2 emissions.