(64d) Data Assimilation Method for Quantification of Controlled Methane Release Using a Drone and Ground Sensors

This work assessed the particle filter data-assimilation technique to estimate the methane emission rate during the 3-day CH4 controlled-release experiments in September 2020 and 2021. Several controlled methane releases have taken place on a platform with 40m x 50m area named TADI, in France. These leaks were ranging from 0.01 to 5 g CH4 s-1 over 24 to 71 mins. A methane-detecting drone and five ground sensors recorded the methane concentration simultaneously. In this paper, a data assimilation method using a particle filter is developed to improve the accuracy of air contaminant dispersion predictions based on Gaussian-based models. In terms of data-driven modelling, diffusion coefficients and release rate are regarded as the system states, and a particle filter is then used to update these parameters during each step of the calculation. We evaluated different types of atmospheric data assimilation frameworks for the monitoring of CH4 emissions from industrial sites and facilities. In all cases, using fixed or mobile data and any of the specific assimilation configurations, the assimilations provide precise rate estimates for most releases. The average relative errors in the estimated CH4 release rates typically range between approximately 35% - 84% for 2020 campaign, and 29% - 72% for 2021 campaign. Using the drone measurements provide a better estimate of the emission rates than using the stationary measurements with both observation binning approaches, with average relative errors of approximately 51% and approximately 72%, respectively. The objective of the study is not limited only to assessing the performance of the ground measurements and drone technology on its accuracy in quantifying emissions using data assimilation. The interesting part is to compare the results of assimilation for individual technology with the hybrid observation, which combines these two approaches. The highest correlation (R2 = 0.97) and lowest error (29%) are observed in the data assimilation result with the hybrid measurement case when both drone and stationary approaches were evaluated simultaneously using a particle filter.