Spatial allocation of anthropogenic carbon dioxide emission statistics data fusing multi-source data based on Bayesian network

A gridded social-economic data is essential for geoscience analysis and multidisciplinary application. Spatial allocation of carbon dioxide statistics data is an important issue in the context of global climate change, which involves the carbon emissions accounting and decomposition of responsibility for carbon emission reductions. In this research a new spatial allocation method for non-point source anthropogenic carbon dioxide emissions (ACDE) fusing multi-source data using Bayesian Network (BN) was introduced. In addition to common-used DMSP (Defense Meteorological Satellite Program), PD (population density) and GDP (Gross Domestic Production) data, the land cover and vegetation data was imported into the model as prior knowledge to optimize the model fitting. The prior knowledge here was based on the understanding that ACDE was dominated by human activities and has strong correlations with land cover and vegetation conditions. A 1 km gridded ACDE map integrated emissions form point-source and non-point source was generated and validated. The model predicts ACDE with high accuracies and great improvement can be observed when fusing land cover and vegetation as prior knowledge. The model can achieve successful statistics data downscaling on national scale provided adequate sample data are available, offering a novel method for ACDE accounting in China.

Automated summary

The study introduces a new spatial allocation method using Bayesian Network to generate a high-precision ACDE map by fusing multiple data sources. Prior knowledge is based on the understanding of the relationship between ACDE, land cover, and vegetation.