Introduction

Urban is the center of human activities that have a significant impact on the surface of the Earth. Rapid urbanization not only causes significant changes in the utilization of land but also concentrates intensive energy consumption, both of which release large amounts of CO2 into the atmosphere. Some research showed that urban areas account for about 75% of carbon emissions worldwide (Li et al. 2022). Moreover, the United Nations Habitat predicts that the proportion of urban carbon emissions will increase to 76% by 2030 (Gan et al. 2022). Therefore, the mitigation of urban carbon emissions has played a vital role in the process of global carbon reduction (Gong et al. 2022a).

Numerous studies have been conducted on urban carbon emissions at different scales, such as Peng used logarithmic mean Divisia index (LMDI) method to analyze the energy consumption of CO2 emissions in Hunan province based on the Long-range Energy Alternatives Planning (LEAP) system model (Wang et al. 2017). Hu set up four scenarios to forecast and analyze the energy production and consumption of Shenzhen, a postindustrial city, from 2015 to 2030 (Hu et al. 2019).

However, similar studies regarded the city as a whole, but have no idea about the specific source and spatial distribution of carbon emissions. Some scholars used statistical data to compare the spatial heterogeneity between different administrations, and these studies generally concentrated on regions with low spatial resolution (Gan et al. 2022; Xia and Yang 2022). In recent years, the spatial study of urban carbon emissions has become an important research hotspot with the advancement of remote sensing technology and the abundance of multi-source data (AbdelRahman et al. 2021; Yang et al. 2022). The Defense Meteorological Satellite Program–Operational Linescan System (DMSP-OLS), a visible imaging linear scanning service system carried by the US Defense Meteorological Satellite, can obtain the weak light intensity on the ground (Wang et al. 2017; Zhang et al. 2022), which has been widely applied to carbon emission estimation. For example, Oda et al. used DMSP-OLS data to invert carbon footprint in city regions and used demographic data as an auxiliary inversion to non-city regions to generate a global carbon footprint grid map with 1-km spatial resolution from 2000 to 2019 (Oda and Maksyutov 2011). Based on National Polar-orbiting Operational Environmental Satellite System Preparatory Pro Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) nighttime light remote sensing data, Narit found that the factors affecting carbon emissions of land holdings include the share of primary industry output, land area per capita, and land use index (Narit et al. 2016).

However, nighttime light remote sensing data is inconsistent in different years, and the detection performance will gradually weaken, leading to problem like discontinuous image data (Liu et al. 2022). It has also been shown that nighttime lighting data are better suited to describe population activity, but not for reflecting energy consumption and CO2 emissions (Zhang et al. 2021). Furthermore, there were some studies generated fossil fuel carbon emissions at grid scale. Yang et al. (2022) simulated the spatial distribution of CO2 emissions based on fossil fuel carbon emission grid data and land use data. Gurney et al. (2020) used the carbon emission survey results of industry, commerce, household consumption, and other industries in American to form a carbon emission product at 1-km resolution. However, this method cannot be applied widely to detect the spatial distribution of carbon emissions due to difficulties in energy data availability and validity (Huynh et al. 2017).

Spatial heterogeneity refers to the unevenness and complexity of the spatial distribution of ecological processes and patterns, which can be generally understood as the sum of patchiness and gradient (Zhenyue et al. 2023). To distinguish the spatial heterogeneity of CO2 emissions, many studies take land use type as the basic unit for carbon emission calculation (Xiaowei and Jianxi 2019; Zeng et al. 2022). For instance, Muntean et al. (2014) used land use and human activity data to build a global greenhouse gas emission dataset with 0.1° resolution from 1970 to 2008 and tried to reduce CO2 emissions by optimizing the structure of land use (Li et al. 2022). However, in the existing studies on land use CO2 emissions, most of them only took into account the differences between built and non-built land and ignored the carbon intensity in detail between different functional areas, such as transport land, commercial land, and industrial land (Wang et al. 2019; Upadhyay et al. 2021). As the carbon emission intensity is significantly diverse with these different functional areas, it is necessary to further explore the difference of CO2 emissions among different functional areas.

UFZs describe the human activities within a certain area and contain socio-economic characteristics (Zhu et al. 2022). As the most basic unit of urban development, UFZ reflects the socio-economic characteristics of a specific region and helps to realize the optimization of energy consumption. Through it, the allocation of public resources and the type of energy consumption can be clearly observed. All functional areas are aggregated by geographic objects and extracted from the land use, and the same functional zone has basically similar energy consumption and carbon metabolism patterns (Wang et al. 2022a). Relatively speaking, the estimation of CO2 emissions at the UFZ scale can help urban planners to anticipate the carbon emissions associated with urban development. Thus, UFZs are a reasonable basic unit for the estimation of CO2 emissions (Jing et al. 2022; Zheng et al. 2022). The explosive growth of urban big data like social media data, cell phone signaling data, and point-of-interest (POI) data is characterized by large sample size and rich types, which makes up for the poor recognition ability of remote sensing technology on urban functional interaction and can help accurately detect the spatial distribution of UFZs (Zhou 2022), so as to promote the refined research on urban carbon emissions.

This study further constructed a framework to estimate the spatial pattern of CO2 emissions based on the scale of UFZs, using the land use type as the basic unit, in the case of Pudong New Area, Shanghai, China. Firstly, we extracted UFZs by combining remote sensing image with POI and road network data. Then, the carbon emissions of each UFZ were obtained through the combination of top-down decomposition and bottom-up spatial estimation method. Finally, the spatial patterns of CO2 emissions at the UFZ scale were analyzed to provide policy implications for the process of building a low carbon city. Overall, this study mapped the spatial distribution of CO2 emissions intensity at the UFZ scale, which is necessary for urban planning to adjust and reshape the industrial layout to build a more sustainable city.

Data and methods

Study area

Pudong New Area is located in the eastern part of Shanghai, with an area of 1210 km2, making it the largest municipal district in Shanghai (Fig. 1). Since the reform and opening up, Pudong New Area has experienced rapid urban development, a huge economic volume and extremely frequent human activities (Gong et al. 2022b). It is also an important transportation hub with a three-dimensional integrated transportation system including international ports, railway transport, air transport, and intercity high-speed transport, etc. (Qingyu and Oh 2021). The huge energy consumption has increased the corresponding carbon emissions year by year. As one of the most well-developed zones in China, Pudong New Area has played an indispensable role in the transition to an energy efficient and low emission mode. Therefore, this study chose Pudong New Area as a typical case area.

Fig. 1
figure 1

Overview map of the study area

Full size image

Data source and preprocessing

Data source

POI is the most important element of population agglomeration, which can better reflect the spatial distribution of material resources in the city, but its spatial distribution pattern is different (Yang and Li, 2022). POI opens up a new way of thinking for the study of urban spatial structure, and it has been widely used in the fields of analysis of the spatial pattern of UFZs (Xia et al. 2022). Multiple-source data were used to estimate UFZ scale based on energy consumption and CO2 emissions in Pudong. Specifically, the POI data was obtained from the Gaode Map Open Platform in 2019. And the road network data was obtained from the map sharing plan OpenStreetMap (OSM) website. Energy consumption and other data related to CO2 emission calculation were provided by China Energy Statistical Yearbook and Pudong New Area Statistics Bureau.

Data preprocessing

There were about 201 thousand POI data of all categories in Pudong New Area, and major types of use cover sightseeing spots, enterprises, traffic devices, education, medical care, community activities, business residence, etc. During the collection process, there were some problems, such as duplication of raw data and missing data, which cannot be used directly. Thus, therefore, missing data should be added, and duplicate data should be deleted and reclassified. In addition, since the actual OSM data obtained had topological errors, there were a large number of dead ends, redundant cross roads, and unsealed roads that were not usable. Therefore, it is necessary to preprocess the road data in combination with the high-resolution images and select the main highways, trunk roads, urban primary roads, secondary roads, etc.

Methods

UFZ extraction and verification

Extraction of single UFZs

Figure 2 demonstrates the process of UFZs identification. All the POI data were divided into six types according to the land use type: industrial (ID) zone, transport (TS) zone, residential (RS) zone, cultural tourism (CT) zone, commercial service (CS) zone, and public service (PS) zone. The frequency density and proportion of different POI types can be directly used to reflect the type of urban function. The frequency density can be expressed as

$${F}_i=\frac{n_i}{S_i}\left(i=1,2,\dots, 6\right)$$
(1)

where i represents the type of POI and ni represents the quantity of the different POI types, Si represents the total number of the different POI types, and Firepresents the frequency density of the different POI types in the total number of POIs.

Fig. 2
figure 2

The flow chart of UFZs identification

Full size image

In order to compare different types of POI, Ci vector was used to represent the proportion of each type of POI in the unit on the basis of density. The calculation formula is as follows:

$${C}_i=\frac{F_i}{\sum_{i=1}^6{F}_i}\times 100\%\left(\;i=1,2,\dots, 6\right)$$
(2)

According to the calculation results of Ci, the POI and the research unit were superposed and analyzed to build the underlying data of urban spatial pattern, the hooking point data, and the street unit. And the POI-related data was calculated through the field calculator.

Extraction of mixed UFZs

In the actual practice, many UFZs usually hold mixed functions. Concretely, this study defined the unit as a single functional area when the proportion of the POI type was greater than or equal to 50%. On the contrary, when the proportion of the POI type was less than 50%, it was determined that there was not dominant UFZ in this area, which meant it was mixed UFZ. When the proportion of all the POI types was less than 20%, it was considered an integrated UFZ. Since the integrated UFZ occupied little proportion of the study area, it was not considered in this study.

Accuracy evaluation of UFZ extraction results

For accuracy evaluation, this study selected some units and verified the UFZ classification by sampling and scoring. Specifically, the full score was set up to 3, meaning complete consistency, while 0 meant complete inconsistency. If a single UFZ was identified as a mixed UFZ, the score was 2. And if a mixed UFZ was identified as a single UFZ included in the corresponding mixed UFZ, or as a mixed UFZ containing a function included in the corresponding mixed UFZ, the score was 2. The calculation formula for precision evaluation is as follows:

$$A=\frac{\sum_{i=1}^n{x}_i}{\sum_{i=1}^n{X}_i}\times 100\%$$
(3)

where A is the verification accuracy of UFZ, xi is the score of randomly selected sample, and Xi is the total score of randomly selected samples.

Calculation of CO2 emissions at the UFZ scale

Calculation of CO2 emissions in industrial zones

Based on the range of CO2 emissions from energy consumption, this study adopted the top-down accounting method to calculate CO2 emissions in industrial zones. The construction model is as follows:

$$\textrm{CP}=\sum_i{C}_i\times \sum_i\textrm{E}{\textrm{F}}_{\textrm{i}}$$
(4)

where CP is the total CO2 emissions generated by the functional area, i is the energy type, C is the amount of energy consumption, and EFi is the CO2 emission coefficient of energy i.

Calculation of CO2 emissions in residential and commercial service zones

As the carbon emission sources in residential zones and commercial service zones are mainly liquefied petroleum gas, electricity, gas, coal, etc. and the energy consumption data is difficult to obtain, this study utilized carbon emission intensity obtained from previous research to calculate the CO2 emissions in residential zones and commercial service zones. Its calculation formula is as follows:

$${\textrm{CO}}_{2m}=\sum_{i=1}^n\left({k}_i\times {E}_i\times {d}_i\right)$$
(5)

where CO2m is the total amount of various CO2 emissions of these zones, i is the energy consumption category, ki is the conversion factor for different energies, Ei is the amount of energy consumption of category i, and di is the CO2 emission coefficient (constant value).

Calculation of CO2 emissions in public service and cultural tourism zones

The carbon emission accounting of public service zones and cultural tourism zones referred to the previous studies, which used the carbon emission intensity to estimate the CO2 emissions in the corresponding zones (Gao and Li 2021). The general calculation model is as follows:

$$C=\sum \left({L}_j{A}_j\right)$$
(6)

where C is the CO2 emissions of UFZ, j is the UFZ type, Aj is the area of UFZ, and Lj is the CO2 intensity of UFZ.

Calculation of CO2 emissions in transport zones

This study adopted the carbon emission coefficient of four transportation modes, namely, railway, highway, water transport, and aviation (Table 1) to calculate the carbon emissions:

$${CO}_{2\textrm{T}}={CO}_{2\textrm{T}1}+{CO}_{2\textrm{T}2}$$
(7)
$${CO}_{2\textrm{T}1}=\sum \left({g}_{1\textrm{i}}\times {h}_{\textrm{i}}+{g}_{2\textrm{i}}\right)\times {k}_i$$
(8)
$${CO}_{2\textrm{T}2}=\alpha\ {CO}_{2\textrm{T}1}$$
(9)
Table 1 Conversion between passenger and cargo and CO2 emission coefficient
Full size table

where CO2T is the total CO2 emissions of the transportation system, CO2T1 is the total CO2 emissions from intercity transportation, and CO2T2 is the total CO2 emissions of urban passenger transportation, i is different types of transportation modes, g1i is the passenger turnover, g2iis the freight turnover, ki is the passenger cargo conversion coefficient, and α is the carbon emission ratio coefficient of intercity transportation and urban transportation.

Spatial pattern analysis of UFZ based on carbon emissions

Utilizing the reclassified POI and OSM data and combining the carbon emission accounting data of each UFZ, the carbon emission maps of residential zones, commercial service zones, industrial zones, and transport zones were conducted as follows:

$${\textrm{CO}}_{2\textrm{i}}=\frac{D_i\times C}{\sum_{i=1}^n{D}_i}$$
(10)

where CO2i is the CO2 emission value of UFZ, Di is the carbon emission intensity of UFZ, n is the total number of pixels in the study area, and C is the total CO2 emission of all the UFZs.

According to the spatial distribution of POI in each UFZ and the relevant energy consumption, the CO2 emission intensity of residential zones, industrial zones, and commercial service zones was calculated as follows:

$$P=\frac{\sum_{i=1}^n{W}_i}{A}$$
(11)

where P is the carbon emission intensity, Wi is the emission weight of POI in corresponding UFZs (the weight of residential zones, commercial service zones, industrial zones, and tertiary industry is 1, respectively), A is the total area of UFZ, and n is the total amount of POI in certain UFZ.

As the transport zone is a linear vector, its calculation model of CO2 emission intensity is as follows:

$$D=\frac{\sum_{i=1}^m{L}_i\times {W}_i}{A}$$
(12)

where D is the CO2 emission intensity, Li is the length of the road in the UFZ within the search range, Wi is the volume of traffic on the road, A is the total area, and m is the total number of roads in the search range.

Exploratory spatial data analysis (ESDA) describes and visualizes the spatial characteristics of data by establishing statistical relationships through spatial geographic location correlation, so as to recognize the characteristics of data spatial distribution and to explore spatial correlation information. Spatial autocorrelation analysis is the main module of ESDA, which reveals the degree of interdependence between data from different geographical locations. According to the mutual spatial relationship, it can be divided into positive correlation and negative correlation. According to the scope of study, it can be classified into two categories: global spatial autocorrelation and local spatial autocorrelation.

Global spatial autocorrelation analysis

Global Moran’s I is a key index used to assess the global applicability of spatial data, and it takes values between −1 and 1. When I < 0, it means that there is a negative spatial correlation in the distribution of this spatial element, and its negative correlation is stronger when the value of I is closer to −1. When I > 0, it means that there is a positive spatial correlation, and its positive correlation is stronger when the value of I is closer to 1. When I = 0, it means that these are spatially randomly distributed and do not have any spatial correlation. In the study, Global Moran’s I calculation formula for urban carbon emissions is as follows:

$$I=\frac{n\sum_{i=1}^n\sum_{j=1}^n{w}_{ij}\left({x}_i-\overline{x}\right)\left({x}_j-\overline{x}\right)}{\left(\sum_{i=1}^n\sum_{j=1}^n{w}_{ij}\right)\sum_{i=1}^n{\left({x}_i-\overline{x}\right)}^2}$$
(13)

where I denotes Global Moran’s I, n denotes the total number of regions, xi is the CO2 emissions of region i, xj is the CO2 emissions of region j, \(\overline{x}\) is the average CO2 emissions, and wij is the normalized spatial weight matrix (Eq. (14)), which characterizes the neighborhood relationship between spatial objects. There are two types of spatial weight matrix, Rook and Queen. Rook denotes the adjacency of common edges, and Queen denotes the adjacency of common vertices. In the study, Queen adjacency was used, that is, if region i and region j have a common edge, wij = 1; otherwise, wij = 0.

$${w}_{ij}=\left[\begin{array}{ccc}{w}_{11}& \cdots & {w}_{j1}\\ {}\vdots & \ddots & \vdots \\ {}{w}_{i1}& \cdots & {w}_{jn}\end{array}\right]$$
(14)

By standardizing the Z value and P value, the significance of Global Moran’s I can be evaluated. The null hypothesis of no spatial autocorrelation among spatial elements is first established, and the decision to reject or accept the null hypothesis is made by comparing the P value with the significance level: if −1.96 < Z < 1.96 and P > 0.05, the null hypothesis is accepted; if Z ≥ 1.96 or Z ≤ −1.96 with P ≤ 0.05, the null hypothesis is rejected. The Z value was calculated as

$$Z(I)=\frac{I-E(I)}{\sqrt{\textrm{VAR}(I)}}$$
(15)

and

$$E(I)=\frac{1}{n-1}$$
(16)

where E(I) is the theoretical expected value and \(\sqrt{\textrm{VAR}(I)}\) is the theoretical variance.

Moran’s scatter plot is a visualization result of global autocorrelation analysis, which can visualize the type of spatial autocorrelation each spatial element, respectively, and qualitatively reveal the local spatial stability of the study object. It represents different types of spatial agglomerations in four quadrants in the form of Cartesian coordinate system: high-high (H-H), low-low (L-L), high-low (H-L), and low-high (L-H).

Local spatial autocorrelation analysis

Global Moran’s I can identify the overall spatiality of the study object, and Moran’s scatter plot can indicate the type of agglomeration of each spatial element, but neither of them can show the specific spatial location of the agglomeration. In the study, the Local Indicators of Spatial Association (LISA) index was further used to analyze the local spatial pattern of the urban CO2 emissions. The principle of LISA index is to decompose Global Moran’s I into various parts of the entire spatial range, which can effectively reveal the aggregation of various spatial elements at specific locations. The formula is as follows:

$${I}_i=\frac{x_i-\overline{x}}{S}\sum_{j=1}^n{W}_{ij}\left({x}_j-\overline{x}\right)$$
(17)

where

$$S=\left(\sum_{j=1\ j\ne i}^n{x}_j^2/\left(n-1\right)\right)\ {\overline{x}}^2$$
(18)

while the rest of the variables have the same meaning as in Eq. (13).

Results and discussion

Results and validation of UFZ extraction

Figure 3 shows the distribution of processed POI data excavated from Gaode Map. Among them, commercial service, residential, and public service POIs were most densely distributed in the center of Pudong New Area, while industrial POIs were concentrated in their surroundings. And the transport POIs were mostly concentrated in coastal ports, airports, and high-speed railway stations, with the cultural tourism POIs distributed far away from the urban center.

Fig. 3
figure 3

Distribution of various POIs in the study area

Full size image

Figure 4 shows the proportion of single UFZs and mixed UFZs, respectively. Thus, for single UFZs, transport and industrial zones accounted for 63%; for mixed UFZs, those with transport or industrial zones accounted for the majority at 76%. The results of UFZ extraction showed that there were 1083 functional zones, of which 124 single UFZs accounted for 11.45% and 959 mixed UFZs accounted for 88.55% (Fig. 5). Concretely, mixed UFZs were mainly distributed in areas with developed economy, complete infrastructure, and convenient life.

Fig. 4
figure 4

The proportion of different UFZs (a represents the proportion of single UFZ and b represents the proportion of mixed UFZ)

Full size image
Fig. 5
figure 5

UFZ identification results of the study area

Full size image

Generally, residential zones were mainly located in larger towns away from commercial centers. Similarly, in this study, most of the residential areas in Pudong were located in the old urban areas around the center, with small plots and many patches. Commercial service zones were located around large business districts, that is, in the center of Pudong or around the Huangpu River, providing support for citizens’ life services. Industrial zones were located in the southeast and southwest of Pudong, with obvious industrial function advantages and intensive production factors. Public service zones were mainly located around administrative organs, like education, culture, and medical facilities. Transport zones were mainly located in Pudong Airport, railway station, subway, port, and other large transport hub areas (Zhao et al. 2022).

Compared with the existing research conducted in the other cities (Chen et al. 2022), the single UFZs of this study area accounted for 63% of the total. That is mainly because the UFZs in Pudong were concentrated in the central area, or around coastal ports and other peripheral areas, different UFZs had a high degree of overlap. In addition, in terms of research methods, we divided UFZs by constructing frequency density vectors based on POI and OSM data, while the above research divided UFZs through pixel threshold method based on POI and land use data. This was different from our situation, because Pudong was relatively developed in economy, commerce, and transportation; it was difficult to use the threshold method to divide various functional categories in the road network.

For precision evaluation, 300 UFZs were randomly selected and compared with the real map. The score was shown in Table 2. The overall accuracy was 81.1%, meeting the experimental requirements. This indicated that the POI data can supplement the remote sensing data and improve the spatial refinement.

Table 2 Verification score of functional zone identification
Full size table

Results of urban CO2 emission accounting

Figure 6 shows the carbon emissions of each UFZ, with the industrial, transport, and commercial service zones ranking in the top 3, while the residential and cultural tourism zones were far below them. This is consistent with the existing studies (Wang et al. 2022b), which computed the CO2 emissions of various functional zones in Sichuan province and found out that the percentage of CO2 emissions in industrial zones and transport zones was much higher than those in other UFZs.

Fig. 6
figure 6

CO2 emissions of the six UFZs

Full size image

According to the framework of IPCC, the carbon emission sources of industrial zones include emissions in the production process, waste treatment, and changes in land use (Mir et al. 2021; Xu et al. 2021). Therefore, due to the limited availability of data, the CO2 emissions of some UFZs may be underestimated, and the total carbon emissions considered in this study may be slightly lower than the actual values. Nevertheless, compared with the bottom-up carbon emission accounting, the top-down method used in this study covered all the energy consumption, which may in turn improve the estimate accuracy (Chen et al. 2023). Overall, the CO2 emissions computed in this study were obtained from statistical data due to the additional time required for field surveys, but the emissions results were relatively feasible and valid.

Spatial patterns of UFZs based on CO2 emissions

Spatial distribution

The spatial distribution of CO2 emissions at the UFZ scale was shown in Fig. 7. In terms of spatial distribution, the CO2 emissions of industrial zones ranked first, because China’s economy still mainly relied heavily on industry. Therefore, the adjustment of the structure of industrial UFZs was crucial for carbon emission reduction. Secondly, the emission source generated by the industrial zones was also the main source of air pollutants, which greatly affected the human living environment (Ahmed et al. 2022; Piotr and Zbigniew 2015). Therefore, in urban planning, the relevant environmental agencies must strictly monitor the CO2 emissions generated by the design of the spatial distribution of industrial land.

Fig. 7
figure 7

The spatial distribution of CO2 emissions (a) and the local spatial autocorrelation analysis (b)

Full size image

The transport zones were the second largest source area of urban CO2 emissions (Zhou et al. 2023), as the regional traffic facilities have undertaken the transportation task of serving the normal economic development of Pudong, especially the existence of numerous ports, which makes Pudong a world shipping node. Therefore, it should be properly adjusted the distribution of transport zones in the urban space, such as increasing the strength of water transportation to replace part of air transportation, in order to achieve the goal of reducing CO2 emissions from transportation (Wang et al. 2023).

Spatial autocorrelation

The global spatial autocorrelation analysis showed that Moran’s I value was 0.8116, which revealed that the positive correlation of urban carbon emissions in Pudong New Area was stronger in terms of spatial distribution. And the CO2 emissions of high and low carbon intensity regions were more significantly different from other regions. Furtherly, local spatial autocorrelation results illustrated that industrial zones and transport zones were the main locations of high emission concentrations in Fig. 7. In the LISA cluster map, high-high, high-low, and low-high clusters were majorly distributed in inland ports and areas around the city center. This was because inland ports were important transportation hubs in Pudong, while industrial enterprises were distributed in areas around the city center, which was the same as the location of these clusters in the LISA map. Besides, low-low clusters were majorly distributed in coastal areas and areas far from urban centers (Xia et al. 2023). This was mainly due to the fact that these areas have poorer economies and fewer large shopping malls (Zhu et al. 2023). And because of the large area, there was more agricultural and forestry land and parkland.

Compared with the extraction results of UFZs, it can be found that the areas with high carbon emissions were mainly located in the center of Pudong New Area and inland ports, while the areas with low carbon emission intensity were mainly located in coastal communities, such as Chuansha New Town and other areas with good ecological environment quality. Existing research failed to accurately show the difference in regional carbon emissions (Ji and Lin 2022; Zhou et al. 2023); this study took the impact of UFZ into account, so as to simulate regional carbon emissions distribution at a finer scale.

Conclusion

It is crucial to make accurate spatial distribution simulations of urban CO2 emissions for low carbon city construction. The study used POI and road network data to extract UFZs in Pudong New Area, China, and designed a framework based on the scale of UFZs to explore the spatial distribution pattern of CO2 emissions. The results showed that the spatial difference of the mixed utilization degree of the UFZs was obvious within the study area, which was specifically reflected in the downtown area, especially in the areas with developed transportation. Accordingly, there were significant spatial differences in urban CO2 emissions at the UFZ scale (Li et al. 2021; Xia et al. 2022). Specifically, the high CO2 emission areas were located in the typical secondary industry and transport functional areas around the inland ports and urban center, which was consistent with the actual situation (Tao et al. 2023; Yang et al. 2023). Besides, in terms of spatial autocorrelation analysis, the CO2 emissions in Pudong New Area presented a significant high spatial agglomeration, and the high CO2 emission clusters were majorly distributed in the zones with concentrated industries and transportation. This paper tried to decompose the CO2 emissions into corresponding UFZs and established the UFZ carbon emission correlation framework, fully considering the spatial heterogeneity within the area, which can help improve the policy implementation for low carbon city construction (Wang et al. 2018).

At present, the POI data used in this study only included the location, name, type, etc., while the corresponding housing area of each UFZ needed to be supplemented in the subsequent research. Furthermore, high-resolution remote sensing images can be utilized to obtain building profiles. And with the maturity of big data and artificial intelligence technology, future research can try to extend UFZ-based carbon emission spatialization to larger scope.