Global daily TROPOMI XCH₄ reconstruction and methane emission hotspot identification using machine learning

High-spatiotemporal-resolution column-averaged methane (XCH₄) data are crucial for mapping emission sources and supporting methane quantification and source attribution, but current satellite and ground-based observations are spatially and temporally discontinuous due to observational and atmospheric limitations, hindering the identification of local methane emission hotspots. To address this problem, we use TROPOMI XCH₄ retrievals as the target variable and ERA5 reanalysis, MCD18A1 surface downward shortwave radiation, and MCD18C2 photosynthetically active radiation as predictors in a machine-learning reconstruction framework to obtain a globally continuous daily XCH₄ dataset. The reconstructed data agree well with the original TROPOMI observations and are evaluated against independent TCCON measurements, land-cover-stratified statistics, and time-series analysis, supporting their accuracy and temporal consistency. The reconstructed XCH₄ fields are then applied to monitor anomalous methane emission sources in oil and gas production areas in Texas and in coal mining areas in Shanxi Province. The anomalous emission regions identified from the reconstructed data are generally consistent in location with emission sources detected by existing high-resolution satellite observations, suggesting their potential for identifying emission hotspots. This study provides reliable data for methane monitoring at global and regional scales and has important implications for greenhouse gas emissions accounting and climate change research.