Innovating Taiwan's greenhouse gas estimation: A case study of atmospheric methane using GeoAI-Based ensemble mixed...

This study addresses a gap in atmospheric greenhouse gas research, focusing on methane (CH₄), a gas with a global warming potential 80 times greater than carbon dioxide (CO₂). Unlike prior studies that focus on emission sources and reduction strategies, this research emphasizes the spatiotemporal variations in atmospheric CH₄ concentrations, providing new perspectives on global climate mitigation efforts. A novel GeoAI-based ensemble mixed spatial prediction model was developed, integrating multiple machine learning algorithms and considering various factors to accurately estimate CH₄ concentrations across Taiwan. In the context of global net-zero emissions, this study offers a robust approach to assess spatial variations in CH₄ concentrations, providing valuable insights into the effectiveness of greenhouse gas reduction policies and climate strategies. Key factors influencing CH₄ levels include aquaculture, livestock, transportation land use, wind speed, national CH₄ emissions, net greenhouse gas emissions, population density, quarry sites, solar radiation, seasonal variations, residential areas, temples, CO₂ removal levels, and primary pollutants (e.g., NO₂, NO_x, PM₂.₅, PM₁₀, CO, CO₂, SO₂, and O₃). Seasonal analysis revealed higher CH₄ concentrations in spring and winter, and lower levels in summer and autumn. The model demonstrated high explanatory power with R² values of 0.99, 0.82, 0.98, and 0.67 across training, testing, cross-validation, and external validation datasets. This study presents a model that enhances the understanding of the dynamic factors influencing methane concentration variations. The methodology developed in this research can serve as a reference for other regions and timeframes, potentially offering key insights for the formulation of effective global climate mitigation strategies.