Considering the observation and illumination angular configuration for an improved detection and quantification of methane emissions

Abstract. A growing constellation of methane-sensitive instruments from space is making it possible to monitor methane point-source emissions from different industries and activities. Advanced plume simulation methods are becoming key to better understanding the uncertainties in the plume detection and quantification processes. This paper describes the impact of the observation and illumination angular configuration on the retrieval of methane concentrations and the calibration of semi-empirical emission rate estimation models using simulated datasets. We review the mathematical expression of the retrieval and quantification methodologies to determine the uncertainty sources. We implement a method to simulate the apparent displacement of the plume when projected on the ground due to tilted illumination and observation (i.e., parallax effect). We apply this method to methane plumes generated from 3D spatial distributions of methane using the Weather and Research Forecasting Model in large-eddy simulation mode (WRF-LES). The results in the methane enhancement (∆X_CH4 ) maps show large spatial variations of the plumes with respect to a reference case with nadir observation and zenith illumination. In short, it suggests that the spatial distribution of the plumes is largely determined not only by the turbulence in the atmosphere but also by the acquisition’s illumination and observation geometry. This also results in calibration errors of the methane flux rates. Assuming nadir observation, these errors have a polar dependence with the solar angles that for our training dataset at 20×20 m² reaches values as high as the 30 % error for high sun zenith angles orthogonal to the direction of the plume. These errors are explained by the changing plume area with angular projection on the ground. Furthermore, it also has an impact on the probability of plume detection, and here an example is presented. Our results illustrate the importance of considering the acquisition and observation geometry when analysing plume maps with a spatial sampling of 20 m or better.