Estimating carbon dioxide concentrations in urban areas from satellite imagery using Bayesian network

Carbon dioxide (CO₂) is the most important anthropogenic greenhouse gas causing global warming. An increasing number of studies have focused on urban areas recently because cities are major anthropogenic sources of CO₂ and also the main habitats of most human beings. However, the complicated nature of urban landscapes and the inhomogeneous distributions of CO₂ sources and sinks lead to methodological difficulties in CO₂ observation. This paper introduces a new approach to estimate CO₂ concentration from satellite imagery using a Bayesian network. An estimation model based on Bayesian network was built to characterize the quantitative relationships between remote-sensing data and CO₂ concentrations. Comparative analysis of the proposed model and multiple regression models was then carried out. The feasibility of estimating carbon dioxide concentrations in urban areas from satellite imagery was analyzed, and the advantages of modeling land-surface parameters using the Bayesian network were addressed.