Inverse Modeling of CO2 Emissions from Point Sources using Large-Eddy Simulations with Observations from Uncrewed Aerial Vehicles

There is increasing interest in using uncrewed aerial vehicles (UAVs) to make measurements of atmospheric trace gases. However, integrating these observations with atmospheric models is challenging because of the often high spatial and temporal resolution of the UAV observations and the relatively coarse resolution of atmospheric models. This scale mismatch is particularly challenging for inverse modeling of surface emissions of environmentally important trace gases, such as carbon dioxide (CO₂), when using these data. Consequently, studies estimating CO₂ emissions from UAV observations typically use a mass balance or Gaussian plume inverse modeling approach. Here we quantify CO₂ emissions based on UAV observations from an offshore oil and gas facility by explicitly modeling atmospheric transport processes using a large-eddy simulation (LES) with the Weather Research and Forecasting (WRF) model. In situ CO₂ observations were made by the Airborne Ultralight Spectrometer for Environmental Application (AUSEA) sensor on a UAV and the data were incorporated into a Bayesian inversion approach with WRF-LES simulations at a spatial resolution of 10 m. The reported emissions from the facility at the time of the UAV measurements were 42–48 tCO₂/h. Starting from prior estimates that ranged from 30 ± 20 tCO₂/h to 60 ± 40 tCO₂/h, the inversion results suggested posterior emission estimates from 35.6 ± 5.8 tCO₂/h to 43.7 ± 7 tCO₂/h, respectively, which is a range that is consistent with the reported emissions. We also used the mass balance method and estimated emissions of 38.4 ± 14.4 tCO₂/h, which are in agreement with the Bayesian inversion results as well as the reported emission estimates. Our results demonstrate the potential utility of high-resolution modeling in the context of the Bayesian inversion analysis to estimate point source emissions using UAV observations.