Spectral clustering algorithm using density-sensitive distance measure with global and local consistencies

Provincial-level industrial CO2 emission drivers and emission reduction strategies in China: Combining two-layer LMDI method with spectral clustering

Understanding the CO₂ (carbon dioxide) emissions mechanisms in each province is important to reduce China's CO₂ emissions and achieve carbon reduction targets. This paper combines two-layer LMDI (logarithmic mean divisia index) method and spectral clustering to analyze industrial CO₂ emission drivers in 30 provinces of China and proposes carbon reduction policies. First, each province's industrial CO₂ emission factors are decomposed into energy structure, energy intensity, industrialization level, per capita GDP and total population; energy intensity and per capita GDP play a dominant role in suppressing and promoting industrial CO₂ emissions, respectively. The industrialization level has gradually developed from the eastern coastal areas to the north and the central and western regions, especially Inner Mongolia, Jilin, Henan and Sichuan. The impact mechanisms of the provinces on China's CO₂ emissions are very different from temporal and spatial angle. Second, this paper optimizes spectral clustering though geodesic distance and utilizes it to cluster 30 provinces in China. Finally, based on the characteristics of CO₂ emissions and clustering results, more targeted carbon emission reduction strategies are proposed. Therefore, the formulation of carbon emission reduction policies should not only consider the surface characteristics of CO₂ emissions, but also explore the root causes of CO₂ emissions.

Exploring the driving forces for emission reduction strategies in Henan by combining spectral clustering with two-layer LMDI decomposition

To comprehend city-level driving mechanisms of carbon emissions, this paper utilizes spectral cluster and two-layer LMDI (logarithmic mean divisia index) method to systematically assess the contribution values of correlative factors from each cluster to Henan's carbon emissions increments, and accordingly comes up with more strategies about how to reduce carbon emissions for each cluster on the basis of driving forces of carbon emissions. The results of clustering and the decomposition are as follows: (1) the 18 prefecture-level cities in Henan were divided into five categories by spectral clustering, and there are similar development patterns within each category, so they can learn from each other to improve their own defects of development; (2) this paper utilizes the two-layer LMDI method to divide the factors affecting each cluster of carbon emissions into four types, which includes energy structure, energy intensity, per capita GDP, and population, and calculates the contribution value of each factor. It was concluded that the contribution value of per capita GDP dominantly drove up carbon emissions, while energy intensity played a significant role in offsetting them. Therefore, it is important for Henan's low-carbon development to control the expansion of economic activity and improve energy efficiency in the future.