Impacts of booming economic growth and urbanization on carbon dioxide emissions in Chinese megalopolises over 1985–2010: an index decomposition analysis

Abstract

Given the booming economic growth and urbanization in China, cities have become crucial to sustaining this development and curbing national emissions. Understanding the key drivers underlying the rapid emissions growth is critical to providing local solutions for national climate targets. By using index decomposition analysis, we explore the factors contributing to the carbon dioxide (CO₂) emissions in Chinese megalopolises from 1985 to 2010. An additional decomposition analysis of the industry sector is performed because of its dominant contribution to the total emissions. The booming economy and expanding urban areas are the major drivers to the increasing CO₂ emissions in Chinese megalopolises over the examined period. The significant improvement in energy intensity is the primary factor for reducing CO₂ emissions, the declining trend of which, however, has been suspended or reversed since 2000. The decoupling effect of the adjustments in the economic structure only occurred in three megalopolises, namely, the Yangtze River Delta (YRD), the Beijing-Tianjin-Heibei Megalopolis (BTJ), and the Pearl River Delta (PRD). In comparison, the impacts of urban density and carbon intensity are relatively marginal. The further disaggregated decomposition analysis in the industry sector shows that energy intensity improvements were widely achieved in 36 sub-industries in the PRD. The results also indicate the concentrations of energy-intensive industries in the PRD, posing a major challenge to local governments for a low-carbon economy. As economic growth and urbanization continue, reductions in energy intensity and clean energy therefore warrant much more policy attentions due to their crucial roles in reducing carbon emissions and satisfying the energy demand.

Appendix A The validity of the downscale model

The majority of the validities of the downscale model have been fully discussed in our previous work (Meng et al. 2014). We also describe them briefly here for easy of reference.

The downscale model starts with the assumption that spatially, there is a linear relationship between the nighttime light intensity and energy consumption, and CO₂ emissions:

$$ \hat{F_{it}}={\alpha}^{\ast }{NTL}_{it}+{\beta}_p+\varepsilon {}_{it} $$

(A.1)

where F is the estimated CO₂ emissions from pixel i in year t, NTL is the digital number (light intensity) of nighttime light imagery, α is the coefficient of NTL, and β is the provincial fixed effect for capturing the differences in economic, geographical and cultural conditions between provinces, which may affect the intensity of nighttime lights, and ε is the error term. We then use the provincial aggregated NTL and the calculated CO₂ emissions, which are calculated by the provincial energy balance table, to verify the linear relationship between the nighttime light intensity and CO₂ emissions. The same method is also used for the energy consumption data.

Table 3 shows the estimated results for the panel regression of CO₂ emissions. In columns (1) to (4) of Table 3, both the total CO₂ emissions and the sectoral CO₂ emissions, are used as the dependent variable, and the sum of NTL is the independent variables. As expected, the coefficients of α are all positively significant. The R² of the four regressions are 0.58 (service sector) to 0.77 (transportation sector). That is, the linear model can significantly explain the majority of the CO₂ emissions. To further check the robustness of the linear relationship, four variables, namely, provincial GDP in 2000 constant prices (2000 $), the rate of population urbanization (%), the industry share of the total GDP (%), and the length of roads (km), are used as the explained variables in Equation (A1). The results are reported in columns (5)–(8) of Table 3. Despite adding four other explained variables, the sign and the significance of α are constant. Meanwhile, the explained ability of the linear model is only slightly improved. That is, the assumption of a linear relationship between the nighttime light intensity and CO₂ emissions is valid.

Table 4 reports the estimated results for the panel regression of energy consumption, and the similar conclusions are presented. Thus, to simplify, we only use the NTL as the proxy variable, to downscale the calculated energy and CO₂ emissions, as well as the sectoral data, into the pixel scale.

Table 3 The estimated results for the panel regression between provincial CO₂ emissions and nighttime light intensity

Table 4 The estimated results for the panel regression between the provincial energy consumption and nighttime light intensity