An optimized methane retrieval approach based on Kalman filter for mapping methane point emissions from spaceborne imaging spectrometer

Abstract

Methane (CH4) emissions from point sources in the energy sector plays a crucial role in global CH4 budget. Spaceborne imaging spectrometer has shown superior ability for monitoring these emission events over large spatial coverages and extended timeframes. Presently, the Matched Filter (MF), which is a data-driven method, is widely employed for satellite-based retrieval of CH4 emission flux rates. However, traditional MF method faces challenges such as omission of CH4 plumes and underestimation of CH4 emission flux rate, which may lead to significant uncertainties in the CH4 inventories for the energy industry at global scales. In this study, we propose a new Kalman-Filtering Matched Filter (KMF) algorithm as the improvement of the MF method, incorporating a linear combination of MF results in different CH4 absorption channels to reduce the retrieval bias of point-source CH4 enhancement values. We validate this algorithm in two stages. Retrieval accuracy for the enhancement of column-averaged dry-air mole fraction of CH4 (XCH4) relative to the background (ΔXCH4) is tested using end-to-end simulation and emission-free scenario analysis. Additionally, data from hyperspectral satellites including Chinese Gaofen 5B, Ziyuan 1F, Italian PRISMA, and German EnMAP are used to retrieve CH4 emissions from a ground-based controlled-releasing experiment using our proposed KMF and traditional MF algorithm. We then compare the retrieval results with ground metered-measurements to evaluate the performance on estimating CH4 emission flux rate of our method. The results from end-to-end simulations show that the KMF method exhibits a 34.3% higher fitted-line slope and 25.5% lower root mean square error (RMSE) on ΔXCH4 retrieval compared with MF method. Further analysis in an emission-free scenario indicates that the retrieval precision with the KMF method can improve by up to 42.2% compared to the conventional MF method. Comparison from controlled-releasing experiment data reveals the capability of the KMF method to detect minor CH4 emissions that traditional MF methods fail to identify. Meanwhile, the KMF-based emission rate quantifications have an R2 of 0.99 and a small RMSE of 0.18 ton of CH4 per hour, which shows an approximately 62% reduction on RMSE. We further apply the KMF algorithm to Gaofen 5B and Gaofen 5 data in various regions, including the Delaware Basin (USA), Libya, Oman, and Shanxi (China) from 2021 to 2023, and focus on 16 plumes identified through case studies, highlighting the KMF algorithm's robust detection capabilities for CH4 point source emissions in the energy industry.