An optimized methane retrieval approach based on morphological fusion for mapping methane point emissions from spaceborne imaging spectrometry

Methane (CH 4 ) emissions from point sources in the energy sector play a crucial role in the global CH 4 budget. Spaceborne imaging spectrometry has demonstrated superior capability for monitoring such events over large areas and extended periods. At present, the data-driven Matched Filter (MF) technique has been widely employed for satellite-based retrieval of CH 4 emission rates. However, the traditional single-channel MF approach often omits small plumes and underestimates fluxes, introducing significant uncertainties in the CH 4 inventories for the energy industry at the global scale. Here, we propose a morphological fusion matched-filter algorithm (Fused-MF) that spatially decouples plume detection from concentration quantification. A full-shortwave infrared MF (SWMF) is first used to produce a low-noise map of the column-averaged dry-air mole fraction of CH 4 (XCH 4 ) enhancement relative to the background (ΔXCH 4 ) to delineate the plume through morphological segmentation. Within the resulting mask, ΔXCH 4 values are then retrieved by a lognormal MF (LMF) corrected with a sensor-specific effective factor ( k SRF ), whereas SWMF result is retained outside the mask. This design preserves the operational efficiency of scene-wide MF screening while reducing the systematic underestimation of large ΔXCH 4 values. The proposed algorithm is validated via two-stage assessments. First, retrieval accuracy for ΔXCH 4 is assessed using end-to-end simulation. Second, hyperspectral observations from Chinese Gaofen 5B, Ziyuan 1F, Italian PRISMA, and German EnMAP are used to retrieve CH 4 emissions from a ground-based controlled-release experiment using our proposed Fused-MF and traditional MF algorithms. We then compare the retrieval results with ground-based measurements to evaluate our method's performance in estimating the CH 4 emission rate. The results from end-to-end simulations show that the Fused-MF method achieves the most accurate ΔXCH 4 quantification among the four methods evaluated, yielding a near-unity regression slope of 0.98, a BIAS of −8.48 ppb, an root-mean-square-error (RMSE) of 43.17 ppb, and a mean absolute error (MAE) of 34.02 ppb. Scenario-level statistics across 36 simulations confirm that its reduction in retrieval error is significant relative to all three single-channel methods. Further analysis using the field controlled-release experiment data reveals the capability of the Fused-MF method to detect minor CH 4 emissions that traditional MF methods fail to identify. Meanwhile, the Fused-MF-based emission rate quantifications show an R 2 of 0.99 and an RMSE of 0.19 t(CH 4 )/h, representing reductions of approximately 36.7% in RMSE, 56.0% in MAE, and 33.3% in standard deviation (STD) compared with the mostly used MF method applying spectral channel with strong CH 4 absorption signal (SAMF). We further apply the Fused-MF algorithm to Gaofen 5/5A/5B imagery acquired between 2019 and 2023 over the Delaware Basin (United States), Libya, Algeria, Oman, and Shanxi (China). Sixteen plumes are identified through case studies, confirming the Fused-MF algorithm's robust capability to detect and quantify CH 4 point-source emissions from the energy sector.