Modeling and spatio-temporal analysis of city-level carbon emissions based on nighttime light satellite imagery

Climate warming due to carbon emissions has been recognized as a threat to food security, human health and natural ecosystem, and carbon emission reduction is a challenging job for each country in the world. In this study, an ensemble structure-based neural network model (NNEnsemble) is proposed to analyze the nonlinear relationship between the Defense Meteorological Satellite Program Operational Line-Scan System (DMSP-OLS) nighttime stable light (NSL) data, and province-scale statistical data on carbon emissions. Given the challenge of obtaining urban-scale carbon emission data, a weighted coefficient strategy by using the NSL data were employed to analyze the carbon emissions at the urban scale. The performance of the proposed method was found to be superior to that of comparable methods with respect to various evaluation indices. Under these circumstances, a hot spot and Standard deviational ellipse analysis of three northeast provinces in China was conducted from 1998 to 2013. The results can promote a better understanding of the spatio-temporal characteristics of carbon emissions in three northeastern Chinese provinces. Moreover, the developed application software based on NNEnsemble can serve as a basis for the development of carbon emission mitigation policies for other provinces.