Development of a spatiotemporal resolution enhancement method for satellite-observed XCH4 products based on spatiotemporal information and LightGBM-Bayesian Integration

Abstract

High spatiotemporal resolution XCH₄ observational data are crucial for the comprehensive prevention and control of CH₄ pollution. Satellite remote sensing has emerged as a key approach for XCH₄ monitoring; however, its effectiveness is constrained by satellite observation tracks, atmospheric disturbances, and sensor limitations. Consequently, data gaps persist in certain regions. Machine learning models have demonstrated remarkable success in generating high spatiotemporal resolution data for gases such as CO₂, yet research on their application to XCH₄ remains limited. Moreover, most existing studies fail to fully capture the complex spatiotemporal characteristics of XCH₄ due to insufficient feature selection. Therefore, this study proposes a novel spatiotemporal resolution enhancement method for satellite-derived XCH₄ data, integrating spatiotemporal information with a LightGBM-Bayesian framework. This approach establishes the latent relationships between satellite-derived XCH₄ measurements, auxiliary data, and precise spatiotemporal information. Using this method, we generated XCH₄ distribution maps for the Beijing-Tianjin-Hebei region from 2020 to 2022. Experimental results indicate that: (1) The LightGBM-Bayesian model outperforms traditional models such as LightGBM, XGBoost, and RF, demonstrating superior accuracy; (2) Model predictions exhibit strong consistency with TCCON station observations, validating its high precision; (3) Incorporating precise spatiotemporal information as input features significantly enhances the model's predictive performance; and (4) The spatiotemporal distribution of XCH₄ in the Beijing-Tianjin-Hebei region from 2020 to 2022 reveals a seasonal trend of higher concentrations in summer and autumn and lower concentrations in spring and winter, along with a year-on-year increase. Spatial patterns indicate elevated levels in the southwest and lower levels in the northeast.