The Spatiotemporal Evolution of Buildings’ Carbon Emissions in Siping, a Chinese Industrial City

Abstract

Industrial cities in transition face multiple pressures of socio-economic development and carbon emission reduction. Studying the spatiotemporal evolution of urban carbon emissions helps us understand the spatial adaptability of low-carbon cities. In this study, we took Siping, an industrial city in China, as an example; spatially mapped buildings’ carbon emissions by combining statistical data and points of interest; and used exploratory spatiotemporal analysis to dynamically evolve the spatial distribution and spatiotemporal-dependent paths of carbon emissions over the years. The results presented the spatial aggregation and heterogeneity of four types of buildings’ carbon emissions in Siping. In contrast, the spatial aggregation of block-scale carbon emissions related to residential buildings and commercial buildings was stronger, and the standard deviation ellipses showed a trend of expanding outward. However, with a large total volume of carbon emissions related to industrial buildings and a large standard deviation ellipse of the distribution, targeting industrial carbon emissions remains a priority for carbon reduction. With the expansion of urban land use, the population density and carbon emission intensity of the central area decreased. Therefore, Siping should slow down its rate of land expansion, improve land use efficiency, and achieve a new balance in the complex relationship between society, economy, and the environment.

1. Introduction

The Global Carbon Budget report indicated that global CO₂ emissions were projected to reach 41.6 billion tonnes in 2024, marking an increase from the 40.6 billion tonnes recorded in 2023 [1]. This underscores the pressing need to reduce emissions on a global scale to limit global warming to 1.5 °C, as outlined in the Paris Agreement 2015. Cities are responsible for 70% of global carbon emissions and are recognised as pivotal in achieving net-zero greenhouse gas emissions [2]. The concept of low-carbon cities has been proposed for nearly two decades, yet cities remain predominantly focused on low-carbon industrial development, among a range of feasible low-carbon measures, such as optimising industrial structure, enhancing energy use efficiency, and developing clean energy industrial technologies [3]. The development of low-carbon cities necessitates a comprehensive approach encompassing various urban spaces and socio-economic elements, such as the construction of low-carbon buildings and infrastructure, the promotion of low-carbon transportation options, and the enhancement in the carbon sink capacity of green spaces [4,5]. There is a growing body of research that highlights the nexus between urban space and carbon emissions. Cities with low fragmentation and high compactness in their external morphology are more conducive to reducing carbon dioxide emissions [6,7,8]. The efficiency of carbon emissions is specifically influenced by urban inner density, land use function, traffic structure, and other factors, and a compact development with relatively high population density is encouraged [9,10,11]. Consequently, the analysis of spatial and temporal changes in urban carbon emissions is crucial for evaluating the effectiveness of urban emission reduction and adaptation to climate change.

At the regional scale, the estimation of urban carbon emissions primarily depends on statistical data and certain spatial data that serve an intermediary function. In large, developed cities with a high availability of energy statistics, carbon emissions are directly calculated using a carbon emission inventory and the carbon emission coefficient [12]. In China, however, obtaining energy statistics for many urban areas is challenging due to the energy balance being calculated at the provincial level. Consequently, spatial data pertaining to the degree of urban development, such as night-time lighting data and land use data, are utilised to estimate urban carbon emissions [13]. In recent years, night-time lighting data have emerged as a significant spatial data source for studying carbon emissions [13]. The county-level carbon emission inventory table of China, compiled by Chen [14], was produced by downscaling provincial energy carbon emissions based on night-time lighting data. Xu [15] utilised this method to examine the relationship between carbon emissions and economic development in China’s Yangtze River Delta. It is acknowledged that night-time lighting data are an externalised representation of urban power energy use, and differences in urban industrial structures may cause large errors. The carbon emission coefficient of the land use type plays a pivotal role in assessing the carbon emissions of urban areas other than construction land; for example, the carbon emission coefficient of farmland is 0.0422 kg·m⁻²·a⁻¹ [16]. However, given the heterogeneity of natural resources and environmental conditions across different regions, the carbon emission coefficients of land use types may vary significantly. This necessitates the correction and utilisation of these coefficients to ensure the accuracy of carbon emission assessments at the urban scale. Consequently, while supporting regional development strategies, urban-scale carbon emission estimation continues to encounter challenges related to the scarcity of spatial data sources, imprecise calculation boundaries, and inadequate geographic space matching.

At the micro-scale within urban areas, low-carbon or zero-carbon blocks have been shown to be effective in enhancing energy conservation and carbon reduction and reshaping the spatial structures and internal connections of urban blocks [17]. The micro-scale analysis of carbon emissions is primarily conducted considering two distinct aspects: (1) carbon sources, including emissions from buildings, transportation, and residential consumption, and (2) carbon sinks in blue and green spaces [18,19]. In 2021, it was estimated that the energy consumption and carbon emissions of the construction industry and associated sectors accounted for 36% of the global total energy consumption and carbon emissions [20]. Of these, residential emissions accounted for 16% of global emissions [20]. In China, the carbon emissions of urban residential buildings and public buildings account for approximately 40% and up to 41% of the building operation stage, respectively [21]. The development of the tertiary industry has been accompanied by an increase in carbon emissions from commercial buildings [22,23]. Consequently, as a pivotal space for urban production and habitation, buildings possess considerable potential for emission reduction [24,25,26,27,28], and measures can be taken to achieve this, such as altering design parameters in different temperature zones, optimising layout, and utilising natural resources [29,30].

The direct calculation of block-scale carbon emissions through a carbon emission inventory has been demonstrated to be a more accurate method than others [31]. However, relevant open databases are often lacking, such as those containing data on building types, construction materials, construction dates, and related energy use and geometry information on the building stock. Carbon emissions from buildings are mainly incurred during the construction and operation stages. In the context of the building operation stage, the direct energy carbon emissions of residents are the primary source, encompassing emissions from natural gas and liquefied petroleum gas utilised for culinary purposes, domestic hot water provision, and space heating. In addition, indirect energy carbon emissions have been identified as significant contributors, including electric energy consumed by various electrical appliances and heat energy from central heating [32]. Consequently, estimating carbon emissions at the micro-scale is primarily conducted by utilising a building energy consumption model or a questionnaire survey [33,34,35]. Wang [24] used the urban modelling interface (UMI) platform to conduct numerical simulations of block samples and typical layout patterns, employing the carbon emission calculation model to generate the carbon emission index of the entire life cycle of the building. Wang [18] adopted urban building energy modelling (UBEM) to simulate the carbon emissions of residential and commercial buildings. At present, by combining statistical data with spatiotemporal data, the top-down downscaling method is the main measure for spatially mapping micro-scale carbon emissions according to land or building function and carbon emission intensity. In a related study, Carpio [36] adopted a combination of statistical data and remote sensing images to examine the carbon emissions of commercial and residential buildings in the Monterrey metropolitan area of Mexico and the relationship between urban expansion and carbon emissions. Point of interest (POI) data have emerged as a significant spatial data source for urban research in recent years. In this vein, Huang [37] and Wang [38] employed POI data to map carbon emissions in Shanghai and Beijing, respectively, with high spatial and temporal resolution. It is evident that, owing to the influence of data quality and source, the spatial estimation of micro-scale carbon emissions remains contingent on the availability of more accurate spatial data.

Cities encounter varying pressures to reduce carbon emissions, particularly in the industry cities of China that have yet to achieve high levels of low-carbon production [3,39]. The deceleration of urbanisation has engendered challenges in the transformation of industrial cities, whilst cities undergoing shrinkage are confronted with a more intricate array of social and environmental issues [40,41,42]. Th authors herein conducted a case study of Siping, an industrial city in Northeast China, to illustrate these issues. Block-scale carbon emission maps were created, and an exploratory spatiotemporal analysis was conducted to study the evolution of carbon emissions. This analysis reveals the spatial adaptability and potential low-carbon pathways for Siping in the domains of energy conservation, emission reduction, and climate change response.

2. Materials and Methods

2.1. Study Area

There are nearly 100 old industrial cities in China, and the industrial cities in the northeast region have experienced massive population shrinkage [43]. In the choice between adapting and inhibiting shrinkage, the carbon emission pressure and development strategy of cities will be different. For instance, the urban construction land area of Jixi, Hegang, and Yichun has decreased significantly, while the construction land area of Anshan, Yingkou, Chaoyang, and Siping has increased significantly (

Table A1. Several old industrial cities in Northeast China.

Cities	Urban Population (10,000 Persons)			Construction Land Area (Square Kilometre)			The Proportion of Secondary Industry in GRP (Municipal District)
	2012	2022	Growth Rate	2012	2022	Growth Rate	2012	2022
Anshan	151.88	125.24	−17.54%	151.44	178.55	17.90%	52.20%	49.58%
Fushun	132.68	109.82	−17.23%	131.03	123.18	−5.99%	61.01%	55.03%
Yingkou	91.10	77.57	−14.85%	110.00	180.07	63.70%	56.80%	48.35%
Fuxin	77.90	71.42	−8.32%	70.37	76.50	8.71%	62.10%	35.73%
Chaoyang	56.00	48.37	−13.63%	39.80	61.64	54.87%	47.48%	31.03%
Siping	60.57	44.91	−25.85%	53.51	66.20	23.72%	61.39%	37.22%
Liaoyuan	50.00	37.80	−24.40%	46.33	46.70	0.80%	59.12%	30.29%
Jixi	71.58	61.00	−14.78%	78.85	58.22	−26.16%	60.30%	52.12%
Hegang	57.52	49.14	−14.57%	50.71	48.63	−4.10%	68.29%	55.62%
Yichun	75.69	41.44	−45.25%	156.96	91.92	−41.44%	39.78%	19.52%
Jiamusi	60.00	54.71	−8.82%	98.08	73.20	−25.37%	36.59%	24.16%
Mudanjiang	70.29	55.04	−21.70%	82.92	61.59	−25.72%	46.69%	24.33%

). At the same time, the proportion of secondary industry in Anshan and Yingkou is still high, while the proportion of secondary industry in the gross regional product of Chaoyang and Siping has declined significantly. Post-industrial cities such as Siping and Chaoyang have lagged behind in their spatial reconstruction, and there is still a big jump to becoming sustainable low-carbon cities. Taking Siping as an example, this study reveals the dynamic coupling between carbon emission and urban space through the spatiotemporal changes in carbon emission, and provides a reference for the spatial strategy of low-carbon cities.

Siping is a prefecture-level city in Northeast China’s Jilin Province, which includes Tiexi District, Tiedong District, Lishu County, Shuangliao City, and Yitong Manchu Autonomous County (Figure 1). In 2022, Siping’s urban population was 449,100, thus making it a Type Ⅰ small city according to the State Council’s standards for the size of cities. Winters are long and cold, and the heating period lasts for up to six months. Siping’s population loss is serious and could be caused by factors such as climate, employment, policies, and multiple other reasons [44]. According to China’s sixth (2010) and seventh (2020) censuses, the Siping area has lost nearly 470,000 people in one decade.

The industrial structure of Siping was dominated by agriculture at the beginning of the reform and opening-up policy. Under the guidance of the industrial policy to revitalise the old industrial base in Northeast China, the industrial-dominated economic pattern continued for ten years after 2008. In 2016, due to the impact of many adverse factors, such as overcapacity and insufficient demand, the Siping industry entered the most difficult period. At present, the tertiary industry accounts for nearly half of Siping’s gross regional product and plays an important role in increasing employment and mitigating population loss. In line with the development direction of the “low-carbon, circular, ecological, and green” approach, the internal structure of the industry has been continuously adjusted, and the development of the high-energy industry has gradually slowed down. In the future, the direction of industrial development will shift from the traditional machinery, energy, chemical, and food industries to the digital economy and pharmaceutical and health, warehousing and logistics, and cultural tourism and healthcare industries [45]. While the low-carbon industry constitutes the fundamental element of low-carbon cities, within the paradigm of shrinking cities, the low-carbon reconstruction of spatial form attains greater strategic significance.

2.2. Data Sources

Due to the limited statistical and spatial data available in small- and medium-sized cities, the feasibility and accuracy of carbon emission research are severely limited. The data used in this study mainly include the following: (1) Statistical yearbook data were obtained from the China Urban Construction Statistical Yearbook and the China Urban Statistical Yearbook to estimate the total carbon emissions of various types of energy use in Siping’s urban area. These data pertained to the consumption of various types of energy, including electricity, heating, natural gas, and liquefied petroleum gas. The land use area of residential land, public administration and public service land, commercial service land, industrial land, etc., was obtained to calculate the weight of carbon emission allocation based on the carbon emission intensity of different buildings. (2) Geospatial data were used to map carbon emissions at the block scale in Siping and mainly included the vector point data of POIs from six years: 2012, 2014, 2016, 2018, 2020, and 2022. These data were sourced from Amap (https://www.amap.com/ (accessed on 23 November 2024), Baidu Map (https://map.baidu.com/ (accessed on 23 November 2024)), etc. Via OpenStreetMap (https://download.geofabrik.de/ (accessed on 15 November 2024)) to gain access to Siping’s road network to block division. In addition, LandScan population datasets (https://landscan.ornl.gov/ (accessed on 3 December 2024)) provided information on the distribution of the population over the years.

2.3. Methods

In this study, block-scale carbon emissions mainly refer to carbon emissions associated with building functions, including those of residential, public, commercial, and industrial buildings. In the absence of sufficient supporting information, it is difficult to identify building types on a large scale. Therefore, the spatial estimation of block-scale carbon emissions is based on the assumption that the block contains multiple building functions (Figure 2). Firstly, through cleaning, reclassification, and kernel density analysis, the kernel density of POIs related to different building functions in the study area was fully covered, thus facilitating the spatial distribution of carbon emissions. Secondly, the carbon emissions of energy sources and the components of carbon emissions related to building functions were determined according to statistical data. Thirdly, in combination with the density distribution of POIs, the spatial distribution of carbon emissions related to various building functions in the block was analysed, and the spatial and temporal characteristics were identified.

2.3.1. POI Data Processing

The attribute table of POIs contains points divided into 22 broad categories and 167 subclasses. Although some categories of POIs have certain spatial and public attributes, they are not directly related to the reporting of building functions and carbon emissions, such as those of public toilets and scenic spots. Therefore, POIs were excluded and reclassified according to the classification used herein. The main categories related to the residential functions of buildings included life services, transport services, health services, automobile services, automobile maintenance, sports and recreation services, etc., while those related to the commercial functions of buildings included catering services, shopping services, financial and insurance services, car sales, etc. Additionally, the main categories related to the public service functions of buildings included scientific, educational, and cultural services; medical and health services; government and social services; and transport services, and those related to the industrial functions of buildings were companies and enterprises. Under the combined categories, these were subdivided according to the subcategories, and irrelevant categories were removed. For example, functions related to housing referred mainly to clinics and pharmacies under the category of medical services, while those related to public services referred mainly to specialised and general hospitals.

Due to the uneven distribution of POI data and the possible dislocation of points, kernel density analysis was carried out for all types of POIs to achieve the full coverage of the study area (Equation (1)). The higher the kernel density value, the higher the heat of a certain type of building function.

$$D={\displaystyle \frac{1}{R^2}}{\displaystyle \sum _{i=1}^n\left[\frac{3}{\pi }{(1-{(\frac{d_i}{R})}^2)}^2\right]}$$

(1)

where D is the kernel density value, and i = 1, …, n are the input points. Only points within a radius distance of the (x,y) position are included. d_i is the distance between point i and the (x,y) position. R is the search radius, which is the bandwidth. The raster size of kernel density analysis for POIs related to each type of building function is set to 30 m, and the bandwidth is 500 m.

2.3.2. Carbon Emissions Estimation

• Total carbon emissions

Building-related energy use mainly included the indirect energy use of electricity and heating and the direct energy use of natural gas and liquefied petroleum gas. According to the statistical indicators of household natural gas (10,000 cubic metres) and liquefied petroleum gas usage (tons); total heating supply (10,000 gigajoules); central heating area (10,000 square metres); residential heating area (10,000 square metres); annual electricity consumption (10,000 KWH); industrial electricity consumption (10,000 KWH); and the household electricity consumption of urban and rural residents (10,000 KWH), carbon emissions were calculated from energy use. After 2017, the annual electricity consumption statistical scope included the whole city; before then, it contained the municipal district, and in 2021, administrative divisions were adjusted. Therefore, according to the historical data, it was roughly estimated that the social and industrial electricity consumption of the municipal district in 2018 and 2020 was about 1/3 of the city’s electricity consumption, and it was about 2/5 of the city’s electricity consumption in 2022. The carbon emissions of various energy sources were calculated as follows:

$$C{E}_i=A{D}_i\times E{F}_i$$

(2)

where CE_i refers to the carbon emissions of electricity, heating, household natural gas, and liquefied petroleum gas; AD_i is the energy usage; and EF_i is the carbon emission coefficient, and all types of energy references [46,47]. Central heating was considered to estimate carbon emissions with 80% thermal efficiency based on the amount of low heat generated by coal converted into standard coal.

2.

Spatial estimation of carbon emissions

The carbon emissions generated by the residential, commercial, public service, and industrial buildings of each block were determined by calculating the proportion of the kernel density value of the block using Equation (3). For example, the industrial carbon emissions of a block were computed by finding the ratio of the sum of the block’s kernel density values to the sum of the kernel density values in the study area, multiplied by the total industrial carbon emissions. Therefore, it was also necessary to determine the component of the carbon emissions of various energy sources related to various building functions, which was jointly determined using the area of construction land use and the carbon emission intensity of buildings (Equation (4)). For example, to calculate the carbon emissions of residential electricity, the product of the proportion of residential land area and the carbon emission intensity of residential buildings in the four types of emissions was used.

$$C{E}_{ij,m}={\displaystyle \frac{{\displaystyle \sum D_m}}{{\displaystyle \sum D}}}C{E}_{ij}$$

(3)

$$C{E}_{ij}={\displaystyle \frac{p_j{e}_j}{{\displaystyle \sum p_j{e}_j}}}C{E}_i$$

(4)

where CE_ij,m is the carbon emissions of energy type i associated with building type j in block m, ${\displaystyle \sum D_m}$ is the sum of the kernel density value of block m, and ${\displaystyle \sum D}$ is the sum of the kernel density of the study area. CE_ij refers to the carbon emissions of energy type i associated with the function of building type j. p_j represents the proportion of the area of residential, public, commercial, and industrial land use, and e_j is a weight determined using the carbon emission intensities of residential, public, commercial, and industrial buildings. Although the difference in the carbon emission intensity of similar buildings cannot be ignored, the top-down method is based on the assumption that the carbon emission intensities of buildings with similar functions are close in value. According to analyses in the literature [18,31], the average ratio of the carbon emission intensity of residential buildings/public buildings/commercial buildings/industrial buildings is about 1:2:3:2.

2.3.3. Spatiotemporal Characteristic Analysis

The directional distribution, i.e., the standard deviation ellipse, was used to analyse the aggregation trend in and the directionality of the spatial distribution of all types of carbon emissions. The range of the first-level standard deviation (default) was chosen to include the centroid of approximately 68% of the total input elements weighted by carbon emissions.

To explore the influence of the time factor on the spatial concentration and differentiation of carbon emissions, based on exploratory spatial analysis (ESDA), ESTDA’s LISA time path was adopted to perform an evolutionary analysis on the spatiotemporal differentiation of carbon emission intensity in local space [48,49]. Based on the local autocorrelation analysis of the carbon emission intensity of blocks, the spatial distance of LISA coordinates, which move continuously over time in the Moran scatter plot, was used to explain the changes in spatiotemporal interactions and the dynamic characteristics of spatiotemporal differences at the local level. The main indicators were relative length (Equation (5)) and curvature (Equation (6)).

$$U={\displaystyle \frac{N{\displaystyle \sum _{t=1}^{T-1}d(L_{i,t},L_{i,t+1})}}{{\displaystyle \sum _{i=1}^N{\displaystyle \sum _{t=1}^{T-1}d(L_{i,t},L_{i,t+1})}}}}$$

(5)

$$\beta ={\displaystyle \frac{{\displaystyle \sum _{t=1}^{T-1}d(L_{i,t},L_{i,t+1})}}{d(L_{i,1},L_{i,T})}}$$

(6)

where U is the relative length, and β is the curvature. T is the number of time periods, N is the number of blocks, $d(L_{i,t},L_{i,t+1})$ is the coordinate distance of block i from period t to period t + 1, and L is the coordinate position, which is composed of the standardised value of the carbon emission intensity and the spatial lag. $d(L_{i,1},L_{i,T})$ is the distance from the first phase to the last phase. The larger the U, the more dynamic the local spatial dependence relationship and the spatial structure of carbon emission intensity. Additionally, the larger the β, the more tortuous the LISA time path, that is, the greater the influence of the neighbourhood space on block i and the more fluctuating the carbon emission intensity change process and the spatial dependent evolution process of block i. The specific steps of this process were to conduct a univariate local autocorrelation analysis of carbon emission intensity in each period through GeoDA 1.22 software (https://gitee.com/geoda/geoda_mirror/releases/tag/v1.18 (accessed on 8 December 2024)), output a Moran scatter plot, obtain the standardised values and the spatial lag data, and then calculate U and β.

3. Results

3.1. General Trend of Carbon Emissions in Siping

Apart from 2016, in which an unusually low-carbon emission value was observed, Siping’s carbon emissions showed an overall downward trend (Figure 3). In terms of the four types of carbon emissions, those related to industrial buildings (ICE) were the highest, exceeding 50% in all years except 2016. Carbon emissions related to commercial buildings (CCE) were higher than those related to residential buildings (RCE) until 2018, after which there was a reversal of RCE being higher than CCE. Carbon emissions related to public buildings (PCE) were the lowest. In 2022, ICE was about 53%, RCE was about 31%, CCE was about 10%, and PCE was about 6%. ICE and CCE were on a downward trend for 10 years, with ICE decreasing from 23.9 million tons in 2012 to 7.24 million tons in 2022 and CCE decreasing from 3.683 million tons to 1.511 million tons in 2022. Both RCE and PCE showed an upward trend, with RCE ranging from a high of 3.95 million tons in 2022 to a low of 2.161 million tons in 2018 and PCE ranging from a high of 1.703 million tons in 2018 to a low of 0.295 million tons in 2016. In addition to ICE, RCE will also be the focus of Siping’s carbon reduction efforts.

3.2. Spatial and Temporal Characteristics of Carbon Emissions in City Blocks

The spatialisation of carbon emissions related to various buildings based on POIs’ kernel density presented the following characteristics: although the four types of carbon emissions fluctuated with time, they all showed significantly concentrated high carbon emissions in a few blocks (Figure 4). Blocks with high RCE reached a maximum value of 0.515 million tons in 2022; those with high PCE were at 0.156 million tons in 2018, CCE were the highest at 0.474 million tons in 2012, and ICE were the highest at 2.89 million tons in 2012. The spatial polarisation phenomenon was serious; less than 10% of the blocks with high carbon emissions accounted for more than 50% of total carbon emissions. For a few blocks with high carbon emissions, all four types of carbon emissions were at high values.

According to the standard deviation ellipse, the carbon emissions of blocks showed a trend of first decreasing and then increasing, with their centre of gravity shifting to the west. Specifically, the area of the standard deviation ellipse covered by RCE expanded, and the blocks of high carbon emissions migrated to the southeast of the city. Additionally, the standard deviation ellipse of PCE shrank, that of CCE gradually expanded toward the southeast, and that of ICE migrated towards the west. The area of the standard deviation ellipse of ICE was significantly larger than that of the other three types of carbon emissions, and there were blocks of high carbon emissions distributed around the periphery of the city, indicating that the concentration of ICE was low. Blocks with high RCE, PCE, and CCE were closer to the city centre and relatively more clustered. In addition, the azimuth angle of the standard deviation ellipse of ICE was larger than that of the other three types of carbon emissions, indicating that the axis of industrial development was significantly different from that of residential, public, and commercial development.

3.3. Spatial and Temporal Characteristics of Carbon Emission Intensity

In terms of carbon emissions per unit of land use, the carbon emission intensity showed a transitional trend from high intensity in the urban centre to low intensity in the periphery (Figure 5). Although the carbon intensity of the central region has decreased since 2016, this concentration feature was still significant. Moran’s I index also showed that the values of each period were highly clustered (Figure 6). In 2012, 2014, 2016, and 2018, the block with the highest carbon emission intensity reached 1.720 million tons/km², 1.863 million tons/km², 0.478 million tons/km², 1.178 million tons/km², respectively. In 2020 and 2022, the block with the highest carbon emission intensity was around 0.8 million tons/km².

According to the relative length of the LISA time path, the central region was significantly larger than the peripheral region, indicating that the local spatial structure of the central region was more dynamic and that the peripheral region was stable (Figure 7). In 40.98% of the blocks, the relative length of the LISA time path was greater than 1, with an average value of 0.976. The relative length of the LISA time path of a few blocks with high carbon emissions outside the central region was small, indicating that the change in the carbon emission intensity of these blocks was relatively stable.

The larger the LISA time path curvature value, the more the carbon footprint is influenced by the neighbourhood space (spillover/polarisation), and its spatial dependence has greater volatility [48,49]. The blocks with a large LISA time path curvature were concentrated in the central area and the southern and eastern peripheries, with an average value of 4.553 and 23 blocks greater than 10.

4. Discussion

4.1. Influencing Factors of Spatial and Temporal Trends in Carbon Emissions

In regional-scale studies, urban shrinkage, industrial structure, and urban expansion are the main factors affecting cities in Northeast China [50]. Chen [51] and Xiao [52] argued that urban expansion represents a decarbonisation trend, with new urbanisation reducing carbon emissions from buildings. Cheng and Hu [53] argued that urban expansion simultaneously increases carbon dioxide emissions from industry and carbon emissions from transport and buildings. For Siping, the decarbonisation process of urban expansion and the increase in building carbon emissions existed simultaneously. From 2012 to 2022, the expansion of construction land in Siping was mainly concentrated in the east, north, and southwest, adhering to the development of its industrial city (Figure 8). In 2022, the area of industrial land was 2.05 times that of 2012; the area of commercial land was significantly reduced to 40% of that of 2012; the area of public land increased by 27.19%; and residential land grew by only 3.39%. The results showed that the carbon emissions of Siping were still dominated by industrial carbon emissions, and the National Hongzui Economic and Technological Development Zone existed in the northwest of the city (Figure 9a) and the Provincial Siping Economic and Technological Development Zone in the east. Therefore, the coverage area and the azimuth of the standard deviation ellipse of ICE were larger than those of other types. After 2016, ICE in Siping was relatively stable and decreased significantly compared to before 2016, indicating that the adjustment of industrial structure and technological optimisation played a significant role in improving carbon emission efficiency. Before 2016, there were two hot spots of kernel density in Tiexi District and Tiedong District, but the value was higher in the commercial centre of Tiexi District. After 2018, the commercial hot spot in Tiedong was strengthened because Wanda Plaza (Figure 9b), the commercial centre of Tiedong District, was opened in 2016, so the standard deviation ellipse of CCE became larger. The expansion of residential areas was mainly concentrated in the southwest and east of the city. As the POIs related to residential buildings were still transitioning from high density in the central area to low density in the periphery, the standard deviation ellipse of RCE increased slightly.

In 2022, the urban population was 449,100, which means that nearly 30,000 people were lost compared to 2002. From the view of the population shrinkage (population change rate) of the grid, the population density of the urban centre decreased significantly, with a minimum decrease of −34.62% (Figure 10). Although the total population remained basically stable, the population moved to the outskirts of the city, especially to the southeast. The decrease in population concentration in the central area was part of the reason for the decrease in carbon emission intensity in the central area and could also explain why the time path of the central area was more dynamically influenced by the neighbouring areas. A reduction in population density can promote the construction of low-carbon blocks to a certain extent, but it comes with the risk of reducing social vitality and hindering the efficient use of infrastructure [54]. At present, a large commercial building is idle in the city centre. The reduction in emissions in the shrinking population in Oberhausen in Germany had been proved to be caused by the economic downturn rather than an active energy policy, and continued unemployment, low income, and population loss reduced the city’s coping capacity [55]. Xiao [52] believed that the urbanisation rate, population density, and total population were positively correlated with carbon emissions from urban buildings. Meanwhile, Hong [56] believed that for small cities with a population of less than 1 million, increasing urban density could promote carbon emission reduction. Although the total carbon emissions of Siping decreased, the increase in RCE was caused by urban expansion and the decrease in population density in the central areas. The decline in energy intensity and population in the residential sector plays an important role in the reduction in CO₂ emissions, but changes in population lifestyle such as household size and living area and increases in carbon factors can offset these effects [57]. Many countries in the world, such as Japan and the United States, emphasise the development of compact urban forms [58,59,60]. Therefore, Siping should pay attention to developing a reasonable threshold for inhibiting the effects of population density on CO₂ emissions [56,61]. By developing a compact urban form and improving land use efficiency, Siping will be able to adapt to urban shrinkage.

4.2. Accuracy of Spatial Estimation

For small- and medium-sized cities, the spatial estimation and management of micro-scale carbon emissions are urgently required for the development of low-carbon cities. Based on a top-down approach, the authors combined energy statistics and POI kernel density analysis to estimate and spatiotemporally analyse block-scale carbon emissions in Siping. Huang [37] believed that the carbon emission estimation based on POIs showed more spatial heterogeneity than that based on night-time lighting. At the same time, the spatiotemporal evolution of carbon emissions in Siping reflected by this method is consistent with urban development. However, some limitations were found in this study. Firstly, statistical yearbook data were used to obtain the value of the total carbon emissions from energy use. Due to changes in the statistical calibre of the statistical yearbook data, it is difficult to ensure the spatial consistency of the data, which affects the long-term time series analysis. At the same time, most small- and medium-sized cities often do not provide their detailed energy data. Although the method used in this study is theoretically feasible, the types of energy sources that can be selected for analysis are limited. Secondly, if the components of carbon emissions related to different buildings are estimated based on the proportion of urban land use, there may be differences from the actual carbon emissions, which will lead to differences in the spatial analysis results. For example, in the case of RCE and CCE, a higher proportion of CCE may result in a strong agglomeration of block-scale carbon emissions; conversely, if the proportion of RCE is higher, block-scale carbon emissions may be relatively dispersed. In the absence of data, this is a suboptimal choice that can relatively show the spatiotemporal differences in carbon emissions. Thirdly, in the process of spatialising carbon emissions, the order of magnitude of the POI varied greatly over the years, and its accuracy could not be fully guaranteed, and there may be plenty of missing or untimely removals. At the same time, in the process of classifying POIs related to different building functions, due to the mixed use of buildings, it could only be assumed that the POI was related to a certain building function, while in fact, it may be an accessory to other building functions too (Figure 9c). Finally, the classification of industrial enterprises included a proportion of POIs in the tertiary sector, which were concentrated in urban centres and may have an impact on the spatial distribution of ICE.

4.3. Further Studies

The authors distinguished between the carbon emissions of four types of functional buildings and not the specific differences in similar buildings, which requires more localised parameters such as carbon emission coefficients and carbon emission intensity. Due to the lack of the official carbon inventories, the accuracy of POI-based estimation results is not tested. If we can obtain energy use data such as those on electricity, heating, and natural gas used to operate buildings, as well as physical inventory data such as those on building materials and construction time, this will further improve the accuracy of the estimate. Future research can be carried out with the following aspect in mind: bottom-up methods can be combined, such as building energy use surveys and multi-source remote sensing imagery, to obtain information on building geometry, subdivide building functions in the street area, and improve the accuracy of the estimation of buildings’ carbon emissions. For example, the carbon emissions of medical buildings (Figure 9d) in the public sector are significantly higher than those of other types of public buildings [32]. We can also study the carbon emissions generated by transportation, the carbon sink capacity at the block scale, and the optimisation of urban spatial form according to urban functional allocation. In addition, street view images can be combined with population density to further identify the shrinkage differences, and provide more specific suggestions for the city to adapt to the spatial compact form of shrinkage [62].

5. Conclusions

By combining statistical data with spatial data, the authors mapped the block-scale carbon emissions of Siping considering the four aspects of RCE, PCE, CCE, and ICE and studied the spatiotemporal variation in carbon emissions and the potential of building low-carbon blocks. The results showed the spatial aggregation and heterogeneity of different carbon emissions in Siping. Additionally, the spatial aggregation of RCE and CCE was strong, and the standard deviation ellipse showed a trend of outward expansion. With the expansion of the urban area, the population density of the central area and the carbon emission intensity decreased. Reducing carbon intensity helps achieve low-carbon blocks, but this comes with the risk of reduced urban vitality, decreased land use efficiency, and increased carbon emissions from transportation. Therefore, Siping should limit its urban land expansion and increase the population density in its central area. The year 2016 was the turning point for Siping, as this was the year that ICE decreased significantly. Even after the adjustment of the industrial structure, ICE remained the focus of the carbon emission reduction. The authors used a top-down method to estimate the carbon emissions of buildings, but the next comprehensive study on the carbon emissions of transportation and the carbon sequestration capacity of vegetation should be combined with a detailed energy use survey to make it more helpful for evaluating the construction potential of low-carbon cities.