For sources of a known location, but unknown or unreliable in situ source magnitude characterization, a ‘top-down' approach of inferring source strength from downstream concentration measurements can be an effective tool, provided that an adequate representation of the atmospheric structure is available at the relatively fine scales (~ few km) needed to resolve the urban environment. Mesoscale models such as the Weather Research and Forecasting (WRF) model have been shown to be able to provide accurate depictions of the boundary layer structure and flow fields needed to minimize transport uncertainty at scales down to 1 km or less. Nevertheless, a number of other potential sources of uncertainty, or error, exist in the use of mesoscale models for GHG transport and inversion applications. ‘Aggregation' errors of source representation can arise from not representing the height of an elevated release properly, or representing a point source on an Eulerian grid. Furthermore, the turbulent diffusion parameterization used in most mesoscale models was not designed to be used at length scales less than the size of the largest boundary layer eddies, which can be over a kilometer in convective conditions.
For this study we use the 28 Sep 2013 case study from INFLUX to model daytime carbon dioxide transport from a power plant in Indianapolis using the tracer capability of the chemistry-adapted version of WRF (WRF-Chem) at sub-km horizontal resolution. The emission rates from the power plant, as a function of stack height, were provided from the Hestia inventory. As one of the methods of determining uncertainties and biases of the mesoscale model carbon dioxide concentrations, we also ran WRF in large eddy simulation (LES) mode. In LES the largest boundary layer eddies are explicitly resolved rather than completely parameterized. Since the largest eddies dominate the turbulent transport, this process simply appears as advective transport by flow fields resolved by the LES, and (assuming LES fidelity) can be compared with mesoscale model output.
We will show comparisons between different choices of horizontal grid spacing (1 km, 333 m, 111 m) and choices of physics (LES, mesoscale model), as well as tests to determine the impact of aggregation-type errors. We will show how the mesoscale model and LES are similar in the far-field when concentrations are suitably averaged, but relative biases exist in within a certain distance (several km) of the source. In particular, the mesoscale model had lower concentrations and a more rapidly ascending concentration envelope in the near field. We will use inherent differences between the closure assumptions of the LES and mesoscale models to explain the relative bias. Source height was also shown to impact concentrations within a few km of the source, but made little impact at greater distances. The results suggest that mesoscale models, in conjunction with a dense tower observational network, can produce reasonable inversions of GHG emissions in the urban environment, but care should be taken to prevent tower proximity to large point sources.