In order to explore the influences of factors on carbon intensity in depth, the LMDI model was used to decompose the carbon intensity for 620 county-level cities. Carbon intensity was decomposed into six factors: carbon coefficient (CC), energy structure (ES), energy consumption per capita (EP), population density (PD), urban sprawl (URS), and land demand (LD). Since changes in the carbon coefficients were negligible, their effects on carbon intensity were not considered.
3.2.1. Decomposition Analysis on the Provincial Scale
The decomposition results showed that the effects of factors on carbon intensity varied greatly in different spatial units. For macroscopic analysis, the effects of influencing factors on carbon intensity at the provincial level were averaged based on the decomposition results for the county-level cities. The results are displayed in
Table 2.
The effect of energy structure (Δ
CIES) on carbon intensity fluctuated over time in most provinces, and the change in energy structure had the least impact on carbon intensity compared to other factors, which was similar to the results of Jiang [
62]. The effects of energy consumption per capita on carbon intensity (Δ
CIEP) were mainly positive in the increase in carbon intensity, which was in accordance with the results of Han et al. [
63], and its degree of influence varied in different provinces. In some cities, such as Tianjin, the degree of influence of energy consumption per capita on carbon intensity tended to decrease, while in Shaanxi and Shanghai, its degree of influence exhibited the opposite trend, with an upward tendency. The effects of population density (Δ
CIPD) on carbon intensity presented obvious variations in provinces, and its effects fluctuated over time, mainly contributing to the decrease in carbon intensity; these results are consistent with those found in Song et al. [
64]. The land-demand effects (Δ
CILD) reduced carbon intensity for every province. The carbon intensities of Beijing, Tianjin, Chongqing, and Shanghai were greatly affected by LD; however, they tended to decrease over time, except in Chongqing.
The urban sprawl effects (Δ
CIURS) were positive in stimulating the increase in carbon intensity in all provinces, which was likely because urban sprawl would increase the per capita carbon emissions from energy consumption [
44]. The degree of impact of URS on carbon intensity and its temporal trend varied from province to province. In some provinces, such as Beijing, Shanghai, and Guangzhou, the degree of the positive effect of URS on the increase in carbon intensity gradually decreased with time, mainly due to limited urban expansion. Notably, in Shanghai, urban sprawl remained almost zero during 2010–2015. However, in provinces such as Tianjin, Hebei, Ningxia, and Hunan, there was no significant change in the impact of URS on carbon intensity, indicating that their built-up area kept expanding at a certain rate, and the impact on carbon intensity reached a stable level. Some provinces, such as Shanxi, Jiangsu, Guangxi, and Hainan, possessed an increasing influence on carbon intensity with time. Chongqing, Tianjin, Beijing, and Ningxia were the most affected by urban sprawl. The degrees of influence of urban sprawl were seemingly related to the level of economic development.
3.2.2. Decomposition Analysis for Different Types of Cities
• Classification according to the industrial type
There is a big gap in the industrial structure, energy structure, and energy efficiency between traditional and newly industrial regions. We selected some typical cities for the two industrial types displayed in
Table 3, and the decomposition results during 2010–2015 are shown.
The effect of EP in traditional industrial cities was slightly higher than that in newly industrial cities. With the aid of technical progress, traditional industrial cities have a lower energy consumption per capita, which contributes to the slower growth of carbon intensity. The urban sprawl effect was slightly lower than that in newly industrial cities; however, the difference was not obvious. The effects of ES, PD, and LD were quite different between the two industrial types. Due to structural adjustments and optimization, the carbon intensity of newly industrial cities was obviously reduced compared with traditional industrial cities. The PD effect reduced the carbon intensity more in newly industrial cities. High-tech industrial zones are usually located in the suburbs, which attracted people away from urban areas, and the population density decreased in built-up areas. In the meanwhile, the total energy consumption and carbon emissions decreased. Furthermore, thanks to the improvement of technological level, the carbon intensity declined in newly industrial cities. The LD effect was bigger in newly industrial cities than that in traditional cities. The development of high-tech zones had a positive effect on urban land-use efficiency [
65], usually with a high level of innovation and promotion of the economic development.
• Classification according to the administrative and economic level
The 620 county-level cities were divided into three categories: municipal districts of provincial capital cities (Type A cities), municipal districts of prefecture-level cities, except for Type A cities (Type B cities), and county-level cities, except for Type A and B cities (Type C cities). The results of decomposition for the different types of cities are shown in
Table 4.
The changes in ES were related to the increase in carbon intensity for the periods 2001–2005 and 2005–2010. In Type A and B cities, the degree of the effect of ES on carbon intensity showed a downward trend, while it showed an upward trend in Type C cities. During the period 2010–2015, the changes in ES were related to the decrease in carbon intensity for the three types of cities. Even so, the effect of ES on carbon intensity was small compared with those of other factors. In addition, ES had the most obvious effect on the carbon intensity of Type A cities. There was no significant difference in the effect of ES on carbon intensity between Type B and C cities.
Changes in EP had a positive and relatively large effect on the increase in carbon intensity. The degree of influence of EP on carbon intensity first increased and then decreased with time. The influence of EP on carbon intensity was greatest in Type A cities, followed by Type B cities, and then in Type C cities.
From 2001 to 2015, the influence of PD on carbon intensity fluctuated. The degree of the effect of PD on carbon intensity was relatively small—only slightly greater than that of ES. PD was more influential in Type A cities than in other types of cities.
URS was the most stable factor, and there was no obvious change in the degree of influence of URS on carbon intensity with time. Notably, the effect of URS on carbon intensity in Type A cities was significantly greater than in Type B and C cities, indicating that URS played a greater role in increasing carbon emissions in provincial capitals.
The LD factor showed a negative effect on the increase in carbon intensity, with the degree of the effect first increasing and then decreasing with time. The degree of the influence of LD on carbon intensity in Type A cities was obviously greater than in Type B cities and was smallest in Type C cities.
In short, the factors’ effects exhibited clear temporal variation. The most influential factors were urban sprawl, land demand, and energy consumption per capita—urban sprawl is closely related to land-use change. In addition, the factors showed the greatest influence in the municipal districts of provincial capital cities, mainly due to their high carbon intensity.
• Classification according to the spatial location
There are large climatic differences between the northern and southern regions. Because the northern regions have a lower air temperature, they consume more energy for heating in the winter. In the summer, the southern regions have a higher air temperature and spend more energy on cooling.
As shown in
Table 5, the change in carbon intensity was greater in the northern cities than in the southern cities. There were two reasons for this result. One was that carbon intensity in northern cities was higher than that in southern cities, and thus provided plenty of room for the carbon intensity to decrease. The other was that the change in carbon intensity was also larger than that in the southern cities. Another characteristic was that the gap in the changes in carbon intensity between the northern and southern cities was much greater in Type A cities than in Type B cities. This finding further confirmed that changes in carbon intensity in Type A cities should be paid more attention.
3.2.3. Decomposition Analysis for Capital Cities
Since the factors had the greatest impact on carbon intensity in the municipal districts of provincial capital cities, which are energy-intensive and densely populated areas, it is necessary to further explore the regularity and features of the influencing factors for carbon intensity in these cities. The results of the decomposition for the municipal districts of provincial capital cities are displayed in
Table 6.
Changes in ES had the least impact on carbon intensity. During the study period, changes in ES first exhibited a positive effect on the increase in carbon intensity, and then showed a negative effect in some provincial capitals, such as Tianjin, Shijiazhuang, Taiyuan, Hohhot, Shanghai, Nanjing, Hefei, Wuhan, Chongqing, Xi’an, and Lanzhou. However, in Beijing, Shenyang, and Fuzhou, the temporal trend of the effect of ES was the opposite. The impacts of ES remained unchanged in Hangzhou, Nanchang, Changsha, Nanning, Chengdu, and Urumchi, among which only Urumchi showed a constantly restraining effect on the increase in carbon intensity.
With regard to the impact of EP, there were 12 cities with constantly positive effects on the increase in carbon intensity. The effect of EP in Shenyang during 2001–2005 was largest, and in Taiyuan and Xi’an during 2010–2015. In terms of the temporal variation, there was no significant change in the impact of EP on carbon intensity in Beijing. The impact of EP on carbon intensity in Tianjin, Nanchang, and Changsha tended to decrease, while in Shijiazhuang, Harbin, Fuzhou, and Jinan, it tended to increase. Most cities showed a fluctuating growth trend in the effect of EP on carbon intensity. For example, Chongqing showed an effect of increasing carbon intensity during 2001–2005 and 2010–2015, as well as an effect of reducing carbon intensity during 2005–2010. The degree of the effect of EP in Hefei first increased, and then decreased. Guangzhou showed a positive effect on the increase in carbon intensity during 2001–2005 and a negative effect during 2005–2015.
There were spatiotemporal differences in the influence of PD on carbon intensity. The effects of PD on reducing carbon intensity during 2001–2005 were relatively greater than the effects on increasing carbon intensity, such as in Chongqing, Shenyang, Harbin, Nanjing, and Zhengzhou. During 2005–2010, the effects of increasing carbon intensity were relatively greater, such as in Zhengzhou, which had the largest degree of effect, followed by Hohhot, Harbin, and Beijing. Changchun and Wuhan exhibited relatively large effects of reducing carbon intensity. During 2010–2015, the overwhelming majority of cities showed a decreasing trend in the degree of influence of PD, except for Hangzhou, which showed an increasing trend.
The effects of URS on increasing carbon intensity showed slight regional differences. The greatest promotive effect of URS on carbon intensity was found in Beijing during 2001–2005, while during 2005–2010, the biggest URS effect was identified in Changchun and Urumchi. There were no obvious regional differences in the positive effects of URS on the increase in carbon intensity during 2010–2015. In terms of temporal variations, Tianjin, Nanchang, Jinan, Changsha, Nanning, and Haikou showed stable effects of URS on carbon intensity. The effects of URS in Taiyuan, Hohhot, Xi’an, Lanzhou, and Xining had a steady upward trend. However, the effects of URS in most cities—Shijiazhuang, Nanjing, Hangzhou, Hefei, Zhengzhou, and Beijing—had a steady downward trend. Shenyang, Changchun, Harbin, and Chengdu had a fluctuating downward trend for the effect of URS.
LD had a negative effect on the increase in carbon intensity in all provincial capitals, and the degree of the effect mainly had an increasing trend, such as in Shijiazhuang, Taiyuan, Zhengzhou, and Xi’an. A few large cities showed a decreasing trend in the degree of the effect of LD, such as Beijing, Shanghai, and Guangzhou. The effects of LD in Nanjing, Hangzhou, Hefei, and Haikou were stable over time.