Introduction

Carbon emissions generated by burning fossil fuels have seriously threatened global climate stability and sustainable human development (Schuur et al. 2015; Tollefson 2016; Cramer et al. 2018; Yang et al. 2021). Under the severe situation of global climate issues, carbon emission reduction has gradually become a consensus in the world (Wang et al. 2019a; Andiappan et al. 2019; Chen et al. 2022a). For a long time, developing countries have faced the dilemma of balancing economic growth and carbon emissions, posing a huge challenge to global carbon emission reduction (Xu and Yang 2019; Jin et al. 2020; Baloch et al. 2021; Wang et al. 2021a). As the largest developing country around the world, China has become the largest carbon emitter in 2007 (Dong et al. 2013; Wang et al. 2021a). Notably, China’s carbon emissions reached 10.67 Gt in 2020, accounting for about 30.65% of global carbon emissions (Ritchie et al. 2020). To address global climate issues, China has taken an active part in global climate governance, proposing the ambitious “double carbon” targets of carbon peaking by 2030 and carbon neutrality by 2060 (Normile 2020; Wang et al. 2021a).

As the concentration areas of socio-economic activities, urban areas cover less than 3% of the global land areas, consume more than 66% of the global energy, and generate more than 70% of the world’s carbon emissions (Grimm et al. 2008; Romero-Lankao et al. 2014). In this case, urban areas play a key role in seeking solutions and mitigation strategies for global climate change issues (Ribeiro et al. 2019; Wang et al. 2021b; Chen et al. 2022b). More importantly, urban land utilization is not only closely related to urban energy consumption and industrial activities (Carpio et al. 2021; Zhang et al. 2022), but also carries the carbon emission intensity at the urban spatial scale (Wang et al. 2021b). Compared with total carbon emissions or per capita carbon emissions, it is more practical to reduce carbon emission intensity to promote carbon emission reduction (Jotzo and Pezzey 2007). In order to explore the internal mechanism between urban land use system and carbon emission system, this paper proposes a conceptual framework of urban land use carbon emission intensity (ULUCEI).

The concept of ULUCEI is developed on the basis of “urban land use intensity.” The biggest difference between them is that ULUCEI reflects the carrying pressure of carbon emissions on urban land areas (Huang et al. 2013; Cui et al. 2019), focusing on the coordinated development of urban land use system and carbon emission system, while urban land use intensity emphasizes the coupling between economic development and urban land areas (Kuang et al. 2020a). Undoubtedly, the higher value of ULUCEI means that per unit of urban land carries more carbon emissions. Reducing ULUCEI within the limited urban space is helpful for promoting urban carbon emission reduction under the “double carbon” targets. Therefore, exploring the spatiotemporal patterns and evolution trend of China’s ULUCEI can contribute to reveal the internal relationship between urban land use and carbon emissions, which is of great practical significance for formulating reasonable urban carbon reduction strategies, exploring low-carbon urban development models, and achieving the “double carbon” targets.

The remainder of this paper is organized as follows. “Literature review” reviews the literature of relevant studies on ULUCEI. “Methods and data” presents the research methods and data in this paper. “Results and discussion” analyzes the spatiotemporal patterns and evolution trend of ULUCEI in China from 2000 to 2017. “Conclusions and policy implications” gives the conclusions and provides some policy implications.

Literature review

Relevant studies on ULUCEI have focused on the following three aspects, which are the impact of land use/cover change (LUCC) on carbon emissions, the association between land use and carbon balance, and the inner relationship between urban land use and carbon emissions. LUCC caused by human activities is a major source of carbon emissions and an important driver of global climate change (Mendelsohn and Sohngen 2019; Simmonds et al. 2021; Tang et al. 2021). Exploring the carbon emissions from LUCC is a key way to gain insight into the impact of human activities on regional sustainability. Therefore, the impact of LUCC on carbon emissions has gradually attracted the attention of scholars and policymakers. Studies on the impact mechanism of LUCC on regional carbon emissions show that carbon emissions continue to increase in the process of LUCC (Zhu et al. 2019; Zhang et al. 2022), and the increased carbon emissions are mainly concentrated in urban central areas (He and Zhang 2022; Zhang et al. 2022). There are also studies focused on the role of urban spatial form in reducing carbon emissions (Liu et al. 2020; Wang et al. 2020). Some scholars believe that the small-scale, compact, and dense urban land space patterns can produce less carbon emissions (Wang et al. 2019b, 2020), and should effectively control urban land sprawl to achieve low-carbon urban development (Xia et al. 2019).

Land utilization not only changes the carbon sequestration in land ecosystems (Bordoloi et al. 2022), but also alters carbon emissions in the intensive areas of human activities (Tian et al. 2021), thereby significantly affecting the global carbon balance (Neupane et al. 2017). Research on the relationship between land use and carbon balance has been proliferating in recent years (Cui et al. 2019; Huang et al. 2020; Tong et al. 2020). Integrating land use dynamics and carbon balance into the urban management system, the impact of land use on the carbon balance indicated that the carbon balance rate showed a sharp downward trend in the Beijing-Tianjin-Hebei region from 2000 to 2015 (Wang et al. 2021c). In addition, the natural land use type exhibited competition relationships in the urban fringe areas, and exacerbated the local carbon imbalance (Xia and Chen 2020). Actually, scholars mainly evaluate the carbon balance changes caused by land utilization through carbon sources and carbon sinks (Huang et al. 2020; Tong et al. 2020). Studies of urban land use conversion rates affecting carbon balance find that the increase in construction land and decrease in vegetation coverage directly caused a significant decline in carbon storage (Zhu et al. 2019; Chuai et al. 2019; Feng et al. 2020). Especially in megacities and urban agglomerations, urban expansion leads to higher depletion of carbon storage (Chuai et al. 2015a; Liu et al. 2019).

For a long time, scholars have been highly concerned about the relationship between urban land use and carbon emissions (Fang et al. 2015; Wang et al. 2016; Zhang et al. 2018; Xia et al. 2020; Wu et al. 2022). Human activities can influence regional carbon emissions by changing land use patterns (Ou et al. 2013; Li et al. 2018). Related studies consider carbon emissions of urban construction land utilization as the main contributors of urban carbon emissions (Chuai et al. 2015b; Ali et al. 2018; Wang et al. 2019c; Zhang et al. 2022). Optimizing urban land use structure plays a key role in carbon reduction and is an important policy tool worthy of adoption by land managers and policymakers (Cen et al. 2015; Chuai et al. 2015b, 2016; Tang and Hu 2021). Due to the fact that the type of urban industrial activities directly affects the carbon emission effect of urban land, rational configuration of urban land resources is an important way to promote carbon emission reduction (Shu and Xiong 2019). Moreover, carbon emissions have gradually become an obvious obstacle to the realization of intensive urban land utilization (Wang et al. 2019c). Implementing the green model of urban land utilization with “low emission, low energy consumption, high growth, and high output” can continuously reduce carbon emissions (Campbell et al. 2012).

The common feature of the above-mentioned literature is that they devote to exploring the relationship between land use and carbon emission system. These studies can provide solid theoretical support for further in-depth analysis of ULUCEI. However, existing studies on ULUCEI need to further explore the following aspects. First, most of these studies mainly focus on urban land use and carbon emissions as two independent systems, and lack of considering them as a whole to reveal their development mechanisms. Second, previous studies are short of revealing the spatiotemporal patterns of China’s ULUCEI. Third, existing studies lack of exploring the evolution trend of ULUCEI. To fill the gap, the contribution of this study mainly concentrates on the following three parts. One is to construct the conceptual framework of ULUCEI and measure the ULUCEI of 334 cities in China from 2000 to 2017. The second one is to analyze the spatiotemporal patterns of China’s ULUCEI by using kernel density estimation (KDE) and exploratory spatial data analysis (ESDA). The third one is to explore the evolution trend by using spatial Markov chains. Through this study, we can obtain a systematical understanding of the current situation of China’s ULUCEI, which is conducive to exploring the practice model of China’s urban land low-carbon utilization. What’s more, China’s practical experiences about ULUCEI can provide valuable references for other developing countries that face the pressure to reduce urban carbon emissions.

Methods and data

Research area

China consists of 34 provinces, with marked differences in economic development, industrial structure, and resource endowments among the provinces. Considering the availability of the data, this paper focuses on 334 cities in the Chinese mainland from 2000 to 2017, in which Hong Kong, Macau, Taiwan, and Tibet are not taken into account. The Seventh Five-Year Plan for the National Economic and Social Development of the People’s Republic of China (NESDC) in 1985 has divided the Chinese mainland into three large regions, namely, eastern, central, and western regions (Fig. 1). This regional classification has been used by many scholars to analyze spatial differences in Chinese economic phenomena (Guan et al. 2018; Liu et al. 2021). In this paper, this regional classification method is used to systematically reveal the spatiotemporal patterns and evolution trend of ULUCEI in China. As shown in Fig. 1, blue, green, and yellow colors identify the eastern, central, and western regions, respectively.

Fig. 1
figure 1

Research scope and regional classification

Full size image

Empirical methods

Research methods adopted in this study mainly include KDE, ESDA, and spatial Markov chains. First, KDE, a nonparametric method, is used to explore the temporal evolution features of China’s ULUCEI. Then, ESDA is proposed to reveal the spatiotemporal patterns of China’s ULUCEI. Moreover, the method of spatial Markov chains is used to analyze the evolution trend of China’s ULUCEI.

Kernel density estimation

Kernel density estimation (KDE) is a popular nonparametric density estimation method (Katkovnik and Shmulevich 2002). This approach uses the kernel density curve to capture the distribution of random variables, which can effectively avoid the subjectivity of function settings in parameter estimation, thereby improving the authenticity of the estimation results (Qin et al. 2011). This advantage has made KDE a typical method for revealing the temporal evolution features (Kuang et al. 2020b). Therefore, this paper describes the evolution characteristics of ULUCEI by observing the position, shape, and ductility of the density function. The principle of KDE is as follows. Assuming that the density function of the random variable X is f(x), the probability density at point x can be expressed by Eq. (1):

$$f(x)=\frac{1}{Nh}\sum_{i=1}^NK\left(\frac{X_i-\overline{x}}{h}\right)$$
(1)

In Eq. (1), K(·) is the kernel function; N is the number of observations. h is the bandwidth. Xi is the independent and identically distributed random variable, and \(\overline{x}\) is the mean value. The kernel function satisfies K(x) ≥ 0, K(x) = K(−x), \(\int \frac{+\infty }{+\infty }K(x) dx=1\), supK(x) <  + ∞, and \(\int \frac{+\infty }{-\infty }{K}^2(x) dx<+\infty\). Generally speaking, h is determined by its relationship with N, and they should satisfy Eq. (2):

$$K(x)=\frac{1}{\sqrt{2\pi }}\exp \left(-\frac{x^2}{2}\right)$$
(2)

Exploratory spatial data analysis

Exploratory spatial data analysis (ESDA) is usually used to describe and visualize the spatial distribution pattern and agglomeration characteristics of attribute values (Kim 2021). This paper uses ESDA to test the spatial patterns of China’s ULUCEI. ESDA includes the global spatial autocorrelation analysis and the local spatial autocorrelation analysis, which can be characterized by global Moran’s I and local Moran’s I, respectively. Moreover, global Moran’s I is used to reveal the overall spatial relationships, and local Moran’s I is used to evaluate the local spatial association pattern (Anselin 1995; Anselin et al. 2006). This paper uses ESDA to analyze the spatial patterns of China’s ULUCEI. The calculation of global Moran’s I and local Moran’s I can be calculated as follows:

$${I}_{\boldsymbol{global}}=\frac{\sum \limits_{i=1}^n\sum \limits_{j=1}^n{w}_{ij}\left({x}_i-\overline{x}\right)\left({x}_j-\overline{x}\right)}{\sum \limits_{i=1}^n\sum \limits_{j=1}^n{w}_{ij}{\left({x}_i-\overline{x}\right)}^2}$$
(3)
$${I}_{\boldsymbol{local}}={x}_i\sum_{j=1}^n{w}_{ij}{x}_j$$
(4)

where n represents the number of sample cities, \(\left({x}_i-\overline{x}\right)\) and \(\left({x}_j-\overline{x}\right)\) are the deviations of the attribute value and mean value on the city i and the city j, respectively. wij is the spatial weight matrix. When the city i and the city j are adjacent, wij = 1. Otherwise, it is 0.

Spatial Markov chains

The method of traditional Markov chains is a Markov process with discrete time and state, which mainly discretizes the continuous data into different types to reveal the probability distribution and evolution trend of each type (Agovino et al. 2019). Accordingly, a 1 × k dimensional matrix Ft = [F1, t, F2, t, ⋯, Fk, t] is established to store the state probabilities of ULUCEI of each city in year t, and the transition probabilities of ULUCEI in different years can be represented by the k × k dimensional matrix M (Table 1). According to the principle that the number of each type is similar, this paper divides China’s ULUCEI into four types by using quartile, and their values are divided into A, B, C, and D from small to large. The matrix M is the Markov transition probability matrix; the element Pij indicates the probability of being type i in year t and type j in year t + 1. The expression Pij is as follows:

$${P}_{ij}={z}_{ij}/{z}_i$$
(5)
Table 1 Traditional Markov transition probability matrix
Full size table

In Eq. (5), zij denotes the total number of cities that transferred from type i to type j from year t to year t + 1 during the study period. zi represents the total number of cities belonging to type i in the study period. If the ULUCEI of a city is type i in year t and remains type i in year t + 1, the transition type of a city is set as “smooth.” If the ULUCEI of a city increases, the type is set as “upward shift.” Otherwise, it is “downward shift.”

The method of traditional Markov chains mainly focuses on the development of the region itself, while ignoring the adjacent correlation and spatial characteristics of economic phenomena. On the basis of the traditional Markov chains, the method of the spatial Markov chains expresses the domain state of the region by introducing the concept of “spatial lag,” and dynamically reveals the evolution trend of economic phenomena (Le Gallo 2004). By comparing the elements of the traditional Markov transition probability matrix and the spatial Markov transition probability matrix, we can judge the importance of the regional background to the regional change transition. Therefore, this paper introduces the concept of “spatial lag” as a condition and divides it into k types, decomposes the traditional k × k dimensional transition probability matrix into the k × k × k dimensional transition probability matrix (Table 2), and examines the dynamic evolution of China’s ULUCEI. The element Pij|k represents the probability of being type i in year t and type j in the next year under the condition of the spatial lag type k. The spatial lag type considers the units that are geographically adjacent to the city, the spatial lag value Laga of city a is the weighted average of the observed value of the geographic units surrounding the city, and its expression is as follows:

$${Lag}_a={\sum}_{b=1}^n{Y}_b{W}_{ab}$$
(6)

In Eq. (6), Yb is the observed value of city b. n is the total number of cities. The spatial weight matrix Wab represents the spatial relationship between city a and city b. This paper uses the adjacency principle to define the spatial relationship.

Table 2 Spatial Markov transition probability matrix
Full size table

Indicators and data sources

Indicators used to measure ULUCEI

This paper attempts to propose a conceptual framework for ULUCEI with reference to “urban land use intensity,” which refers to the carbon emissions carried by per unit of urban land area (Huang et al. 2013; Cui et al. 2019). As the spatial carrier of terrestrial ecosystems, urban land use carries carbon emissions generated by urban production and life (Carpio et al. 2021; Zhang et al. 2022). In the process of rapid urbanization, the disorderly expansion and blind construction of urban land have not only caused the loss of carbon storage in urban land (Chuai et al. 2016), but also led to the increase of carbon emissions associated with urban land use (Zhou et al. 2021). Specifically, ULUCEI aims to explore the inherent relationship between urban land use and carbon emissions by measuring the carbon emission carrying the pressure of urban land use, so as to guide the low-carbon use of urban land. Exploring the spatiotemporal patterns and evolution trend of China’s ULUCEI is of great value for formulating carbon reduction policies and achieving the “double carbon” targets on schedule. The formula ULUCEIit = Carbonit/Landit was used to calculate the value of China’s ULUCEI from 2000 to 2017. Where Carbonit represents the total carbon emissions of city i year t, Landit represents the total urban area land of city i year t, ULUCEIit expresses the ULUCEI of city i year t, and the unit is million tons/square kilometer.

Data sources

Taking 334 cities in the Chinese mainland as the research objects, this paper collected the annual statistical data of each city from 2000 to 2017. In fact, there is no official carbon emission data at the city level. Excitingly, the China Emission Accounts and Datasets (CEADs) released a set of county-level carbon emissions datasets covering the period 1997 to 2017 (https://www.ceads.net/data/county/). These datasets are estimated by unifying the scale of DMSP/OLS and NPP/VIIRS satellite imagery using the particle swarm optimization-back propagation (PSO-BP) algorithm (Chen et al. 2020). In addition, these datasets are considered comprehensive and effective by previous studies (Shan et al. 2021; Liu et al. 2022). Therefore, this paper obtains China’s city-level carbon emission data by aggregating county-level carbon emission data in CEADs. The land area data of each city is obtained from the China Urban Construction Statistical Yearbook (CUCSY) and China City Statistical Yearbook (CCSY).

Results and discussion

Temporal characteristics of ULUCEI

There was an increasing trend of ULUCEI in China and three large regions during the sample period. According to Fig. 2, China’s ULUCEI increased from 0.102 in 2000 to 0.283 in 2017. This temporal feature is closely related to China’s urbanization and industrialization trends. In the twenty-first century, China’s urbanization and industrialization have entered a stage of rapid development, and the extensive and sprawling urban land development model has occupied a large amount of ecological space, resulting in a rapid increase in carbon emissions (Zhou et al. 2021; Li et al. 2021). From the regional perspective, ULUCEI in China’s three large regions showed an upward trend from 2000 to 2017. Specifically, the average value of ULUCEI in the eastern region rose from 0.192 in 2000 to 0.525 in 2017, the central region increased from 0.089 in 2000 to 0.231 in 2017, and the western region increased from 0.041 in 2000 to 0.132 in 2017. Entering the twenty-first century, China’s urban population agglomeration, industrial scale expansion, and infrastructure construction promote urban space expansion, which also brings a lot of energy consumption and carbon emissions (Liu and Zhang 2022). Obviously, China’s ULUCEI shows a gradually decreasing regional difference from east to west, and the intensity level and growth rate in the eastern region are markedly higher than those in the central and western regions. Eastern cities with developed economies, agglomerated populations, and complete infrastructure have huge advantages in attracting foreign direct investment and large-scale industrial development, which not only stimulate urban spatial expansion but also exacerbate urban carbon emissions. Compared with the east, the cities in the central and western regions are relatively disadvantaged in terms of population size, economic level, and industrial structure, and the carbon emissions of the cities are relatively low.

Fig. 2
figure 2

Average value of ULUCEI in China and three large regions from 2000 to 2017

Full size image

From the perspective of changing trends, after 2011, China’s ULUCEI has changed from rapid improvement to stable development. Actually, this changing trend is closely related to urban carbon reduction policies enacted by the Chinese government. These policies mainly include the three batches of low-carbon city pilots released by the National Development and Reform Commission of the People’s Republic of China (NDRC) in 2010, 2012, and 2017 (Zou et al. 2022), the carbon emissions trading pilot in 2011 (Zhang et al. 2021), and the new-type urbanization in 2014 (Sheng and Guo 2016). With the successive implementation of these policies, the transformation of urban industrial structure, optimization of land resource allocation, and enhancement of green technology innovation reduce urban carbon emissions and lead to a gradual stabilization of ULUCEI.

To further reveal the temporal evolution features of China’s ULUCEI, this paper adopted the method of KDE to plot kernel density curves of ULUCEI in China and three large regions by using EViews 10.0 software (Fig. 3). From the perspective of the peak height of the curve, China and three large regions showed a gradually decreased trend from 2000 to 2017. From the perspective of the peak position of the curve, China and three large regions shifted to the right from 2000 to 2017. In terms of the number of curve peaks, from 2000 to 2017, China and the central region were dominated by a single peak, and the eastern and western regions changed from “multi-peak” to “single peak.” In terms of the shape of curve peaks, China and the western regions were dominated by “spiky peak,” and the eastern and central regions evolved from “spiky peak” to “broad peak” from 2000 to 2017. In general, the results of KDE indicate that China’s ULUCEI shows an obvious upward and polarized trend. The frequent socio-economic activities such as industrial restructuring, population growth, and urbanization construction aggravate the transition from carbon sink land to carbon source land (Zhang et al. 2022). Urban carbon emissions have increased significantly, which has led to the reduction in the polarization of ULUCEI.

Fig. 3
figure 3

Kernel density curve of ULUCEI in China and three large regions from 2000 to 2017. Note: this figure is calculated and drawn by using EViews 10.0

Full size image

Spatial pattern of ULUCEI

To visually present the spatial evolution disparities of ULUCEI in China from 2000 to 2017, this paper selected 2000, 2005, 2010, and 2017 as the years of investigation. According to the calculation results of ULUCEI, grading similar values into the same group facilitates comparison. Therefore, this study used the equal interval classification method of ArcGIS, with 0.10 as the equal interval, and China’s ULUCEI is divided into four level groups, namely low intensity (0 <ULUCEI≤ 0.10), medium-low intensity (0.10 <ULUCEI≤ 0.20), medium-high intensity (0.20<ULUCEI≤0.30), and high intensity (ULUCEI>0.30). As shown in Fig. 4, China’s ULUCEI shows an obvious increasing trend, with the number of high-intensity cities increasing nearly sixfold from 21 in 2000 to 121 in 2017. During the study period, there are significant regional differences in the spatial distribution of China’s ULUCEI. Low-intensity and medium-low-intensity cities are mainly distributed in the western region. The medium-high-intensity and high-intensity cities are mostly distributed in the eastern region, which are concentrated in the Beijing-Tianjin-Hebei, Yangtze River Delta, Pearl River Delta, and other urban agglomerations. Owing to the large populations, active economies, and developed industries, urban agglomerations have large and concentrated carbon emissions (Cai et al. 2018; Wang et al. 2022), which inevitably leads to a higher level of ULUCEI in these regions.

Fig. 4
figure 4

ULUCEI in China during 2000–2017. Note: this figure is calculated and drawn by using ArcGIS 10.2

Full size image

To reveal the global spatial characteristics of ULUCEI in China, this study conducted spatial autocorrelation tests on ULUCEI in China from 2000 to 2017, and obtained the value of Moran’s I and Z statistic (Fig. 5). The results showed that the values of Moran’s I were positive and passed the 1% significance test from 2000 to 2017, indicating that China’s ULUCEI showed a positive spatial autocorrelation. The values of Moran’s I have experienced a dynamic process from “rapid rise” to “high steady growth,” rising from 0.3942 in 2000 to 0.5295 in 2017, indicating that the spatial agglomeration of ULUCEI showed a rapid increase followed by a gradual slowdown. It should be noted that China’s urbanization has shown a spatially and geographically concentrated development, gradually forming a regional development pattern dominated by urban agglomerations (Ouyang et al. 2021), promoting the rapid enhancement of the spatial agglomeration of China’s ULUCEI. However, with the implementation of policies such as low-carbon city pilots, new-type urbanization construction, and carbon emissions trading pilots, the coordinated development of large, medium, and small cities in China has been advocated (Chen et al. 2018). The ULUCEI of each city has increased significantly, the gap between cities has generally shown a slowing trend, and the spatial agglomeration has gradually stabilized.

Fig. 5
figure 5

Global Moran’s I of ULUCEI in China from 2000 to 2017

Full size image

To further present the local spatial characteristics of China’s ULUCEI, this paper used the LISA (Local Indicators of Spatial Association) cluster maps to divide China’s 334 cities into four types (Fig. 6), namely “high-high” agglomeration, “high-low” polarization, “low-high” collapse, and “low-low” homogeneity. Specifically, “high-high” agglomeration indicates that cities with relatively high ULUCEI are surrounded by cities with also high ULUCEI, showing a significant positive correlation. Cities of this type show an obvious expansion trend and concentrate in urban agglomerations such as Beijing-Tianjin-Hebei, Yangtze River Delta, Pearl River Delta, and Central Plains. “High-low” polarization represents cities with high ULUCEI surrounded by cities with low ULUCEI, showing a significant negative correlation. Cities of this type are mainly scattered in the central and western regions. “Low-high” collapse denotes cities with low ULUCEI surrounded by cities with high ULUCEI, showing a significant negative correlation and remaining relatively stable in spatial distribution and quantity. “Low-low” homogeneity indicates that cities with relatively low ULUCEI are surrounded by cities with also low ULUCEI, showing a significant positive correlation. Cities of this type are mainly distributed in Xinjiang, Inner Mongolia, Yunnan, and other provinces in the western region, presenting a pattern of concentrated contiguous distribution.

Fig. 6
figure 6

LISA map of ULUCEI in China during 2000–2017. Note: this figure is calculated and drawn by using GeoDa and ArcGIS 10.2

Full size image

Evolution trend of ULUCEI

This study applies the traditional Markov chains to explore the evolution trend of China’s ULUCEI. To be consistent with the above-mentioned content, the quartile method is used to divide China’s ULUCEI into four levels, namely low rank (A), medium-low rank (B), medium-high rank (C), and high rank (D). Subsequently, the transition probability matrix of the traditional Markov chains is obtained (Table 3). From the traditional Markov transition probability matrix, it can be seen that the transition probability of each rank has strong stability. Without the effect of the spatial lag, the probabilities of maintaining steady state for the rank of A, B, C, and D are 93.2%, 90.2%, 90.7%, and 99.3%, respectively. The values on the diagonal are significantly higher than other values in the peer group, indicating that it is relatively difficult to achieve the spatiotemporal transition of China’s ULUCEI. Moreover, the probabilities of transferring upwards range from 6.8 to 9.1%, while the probabilities of shifting downwards range from 0.6 to 1.6%. The probabilities of transferring upwards are significantly higher than that of shifting downwards, which indicates that there is an increasing evolution trend of China’s ULUCEI. Although there are possibilities of transferring to the adjacent states, the probability of cross-rank transition is smaller.

Table 3 Traditional Markov transition probability matrix of China’s ULUCEI from 2000 to 2017
Full size table

The “space lag” is added to the Markov transition probability matrix, and four types of space lags are set from low to high, namely I, II, III, and IV (Table 4). The results show that the spatial lag affects the transition between cities of different ranks. Under the influence of the spatial lag type I, the probabilities of maintaining steady state of ranks A, B, C, and D are 96%, 94.4%, 93.8%, and 100%, respectively. The probabilities of shifting upwards range from 3.9 to 4.9%, while the probabilities of transferring downwards range from 0.7 to 2.3%. When the spatial lag type is II, the probabilities of maintaining steady state of ranks A, B, C, and D are 86.7%, 90.7%, 92.6%, and 98.9%, respectively. The probabilities of shifting upwards range from 4.1 to 13.3%, while the probabilities of transferring downwards range from 0.7 to 3.3%. When the spatial lag type is III, the probabilities of maintaining steady state of ranks A, B, C, and D are 87.3%, 89%, 92.5%, and 99.2%, respectively. The probabilities of shifting upwards range from 6.9 to 12.7%, while the probabilities of transferring downwards range from 0.5 to 0.8%. When the spatial lag type is IV, the probabilities of maintaining steady state of ranks A, B, C, and D are 100%, 68.5%, 83.6%, and 99.3%, respectively. The probabilities of shifting upwards range from 14.1 to 31.5%, while the probabilities of transferring downwards range from 0.7 to 2.3%.

Table 4 Spatial Markov transition probability matrix of China’s ULUCEI from 2000 to 2017
Full size table

Analyzing the results can be found that the probabilities of maintaining steady state are higher than shifting upwards or transferring downwards, which implied that China’s ULUCEI presents strong stability and “club convergence” trend. The probabilities of shifting upwards are much greater than that of transferring downwards, indicating that China’s ULUCEI shows an increasing trend. Moreover, spatial factors significantly affect the dynamic transition of China’s ULUCEI, and its effect on the shifting upwards gradually enhances with increasing lag type.

Conclusions and policy implications

Conclusions

This paper constructed a conceptual framework of ULUCEI and measured the ULUCEI of 334 cities in China from 2000 to 2017. On this basis, KDE and ESDA are used to explore the spatiotemporal patterns of China’s ULUCEI. Moreover, the method of spatial Markov chains is used to analyze the evolution trend of China’s ULUCEI. The main conclusions are as follows:

  1. 1.

    There was an increasing trend of ULUCEI in China and three large regions during the sample period. China’s ULUCEI increased from 0.102 in 2000 to 0.283 in 2017. From the regional perspective, the ULUCEI in the eastern region is markedly higher than that in the central and western regions. Specifically, the ULUCEI in the eastern region rose from 0.192 in 2000 to 0.525 in 2017, the central region increased from 0.089 in 2000 to 0.231 in 2017, and the western region increased from 0.041 in 2000 to 0.132 in 2017. After 2012, China’s ULUCEI has changed from rapid improvement to stable development. Moreover, the results of nuclear density estimation indicate that China’s ULUCEI shows an obvious upward and polarized trend.

  2. 2.

    The spatial distribution of China’s ULUCEI has obvious regional differences during the sample period. The results of the global space autocorrelation indicate that China’s ULUCEI shows a positive spatial autocorrelation, and the spatial agglomeration shows a rapid increase followed by a gradual slowdown. There are obvious disparities in the distribution rules of cities with different spatial agglomeration forms. Specifically, “High-high” agglomeration cities show an obvious expansion trend and concentrate in urban agglomerations. “High-low” polarization cities are mainly scattered in the central and western regions. The spatial distribution and number of “low-high” collapse cities remain relatively stable. “Low-low” homogeneity cities present a pattern of concentrated contiguous distribution.

  3. 3.

    Without the effect of the spatial lag, the probabilities of remaining stable are higher than 90%, which indicates that it is relatively difficult to achieve the spatiotemporal transition of China’s ULUCEI. Under the influence of the spatial lag, the probabilities of maintaining steady state are still higher than shifting upwards or transferring downwards, which implied that China’s ULUCEI presents strong stability and “club convergence” trend. Moreover, spatial factors significantly affect the dynamic transition of China’s ULUCEI, and its effect on the shifting upwards gradually enhances with increasing lag type.

Policy implications

According to the main research results obtained in this paper, the following several policy implications for reducing ULUCEI and promoting urban carbon emission reduction in China were proposed.

First, formulating the differentiated urban land low-carbon use models and carbon emission reduction strategies based on the spatiotemporal patterns and evolution trend of China’s ULUCEI. The eastern region should exert its technological innovation capabilities and industrial development advantages, improve the economic output of urban land, minimize carbon emissions, and explore new models of low-carbon utilization of urban land. The central region should explore the integration mechanism of industrial development and urban land utilization, improve the regional integrated development model, and promote the coupled development between economic growth and carbon emission reduction. The western region should rely on the national development strategy of western development, promote the close connection between infrastructure construction, industrial development, and low-carbon economy, and improve the potential of urban land low-carbon utilization.

Secondly, transforming the urban land use model to promote urban carbon emission reduction. On the one hand, changing the urban land use mode from “high energy consumption, high emission, and low output” to “low energy consumption, low emission, and high output.” On the other hand, exploiting the potential of urban land use, limiting the disorderly expansion of urban space, and reducing carbon emissions caused by urban land expansion.

Finally, optimizing urban spatial layout to promote urban carbon emission reduction. Combining industrial structure upgrading with urban land use planning, strictly limiting energy-intensive, high-emission, and high-pollution industrial land indicators to reduce carbon emissions. Improving the land supply for high-tech industries, reducing urban carbon emissions, and promoting low-carbon use of urban land.