Full-scope carbon dioxide emission dataset for Chinese cities in 2023

1 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata Full-scope carbon dioxide emission dataset for Chinese cities in 2023 Fanxin Meng 1 ✉ , Hanbo Hu1, Yutong Sun1, Li Zhang 2 ✉ , Jiaqi Hou1, Zhe Zhang3 ✉ , Lingyun Pang3, Bofeng Cai3 & Yuli Shan 4,5 ✉ Cities play a crucial role in implementing carbon reduction strategies and are essential administrative in this effort. A full-scope city CO2 emission inventory is important for designing effective emission control strategies and is thus crucial for achieving China’s carbon peaking and neutrality goals. However, recent research has focused on Scope 1 and Scope 2 emissions, with insufficient attention paid to Scope 3 emissions. In this study, we construct a methodological model for full-scope carbon emission accounting at the city level and establish a dataset that includes Scope 1, 2 and 3 emissions for Chinese cities in 2023. This dataset provides valuable data support for intercity comparisons and intracity management of city-wide carbon emissions. The total carbon emissions across Chinese cities in 2023 show significant quantitative and spatial differences. Notably, the emissions in the top 10 cities are almost 90 to 160 times higher than those in the bottom 10 cities. Cities with higher Scope 1 and 2 carbon emissions are predominantly located in the southeastern coastal areas, whereas cities with higher Scope 3 carbon emissions are concentrated in rapidly developing areas of central China. Background & Summary As the largest greenhouse gas (GHG) emitter globally, China plays an essential role in combating climate change. In alignment with international climate commitments, China has introduced carbon peaking and neutrality strategies, which aim to peak carbon emissions by 2030 and achieve carbon neutrality by 2060, respectively. Urban areas are associated with the consumption of more than 66% of energy and the production of more than 70% of carbon dioxide (CO2), which is a major contributor to emissions and global climate impacts and is pivotal in realizing these strategies1. Given that cities are responsible for a significant proportion of global carbon emissions, comprehensive carbon emission disclosure is critical to inform and support the formulation and execution of net-zero carbon climate action plans2. Comprehensive and consistent carbon emission reporting is crucial, as it provides cities with a reliable basis for formulating policies to reduce their carbon footprint. Wang et al.3 examined the challenges and opportunities for Chinese cities in implementing the country’s carbon peaking and carbon neutrality strategies, thereby proposing that accurate urban emission accounting is central to achieving carbon neutrality. Zhang et al.4 demonstrate that by the end of 2020 nearly one-third of Chinese cities have already reached carbon emission peaks; however, a systematic quantification of total urban emissions remains indispensable for elucidating the temporal evolution of aggregate city-level carbon trajectories. Standards such as the Global Protocol for Community-Scale Greenhouse Gas Emission Inventories5 and the US Community Protocol for Accounting and Reporting of Greenhouse Gas Emissions6 provide structured methodologies for urban GHG accounting. These protocols provide definitions of the key emission scope categories that must be considered: direct emissions from sources located within the city boundary (Scope 1), emissions from purchased electricity and heat, steam and/or cooling (Scope 2), and all other emissions occurring outside the city boundary (Scope 3)5. By offering clear and comprehensive guidelines, these standards enable cities to assess their emissions holistically and develop targeted strategies to reduce their climate impacts7. Moreover, recent studies have highlighted the importance of improving the data quality and enhancing transparency in urban emissions reporting. For example, Akpuokwe et al.8 and Baker et al.9 emphasized the need for more localized adaptation of global frameworks to meet specific regional 1School of Environment, Beijing Normal University, Beijing, 100875, China. 2School of the Environment and Safety Engineering, Jiangsu University, Zhenjiang, 212013, China. 3Center for Carbon Neutrality, Chinese Academy of Environmental Planning, Beijing, 100043, China. 4School of Geography, Earth, and Environmental Sciences, University of Birmingham, Birmingham, B15 2TT, UK. 5Birmingham Institute for Sustainability and Climate Action (BISCA), University of Birmingham, Birmingham, B15 2TT, United Kingdom. ✉e-mail: [email protected]; [email protected]; [email protected]; y[email protected] DATA DeSCrIpTOr OpeN 2 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ or national policy needs; Linton et al.10 noted the lack of integration between urban emission data and broader climate policies, and they advocated for more coordinated approaches to emission governance. Additionally, the United Nations Sustainable Development Goals (SDGs) report explicitly describes the requirement of accurate urban carbon emission accounting through two targeted objectives (i.e., SDG 11 on sustainable cities and communities and SDG 13 on climate action)11; while compelling nations to institutionalize climate-responsive governance frameworks for urban systems and to develop advanced city-scale carbon accounting methods is urgent, and many efforts have been made in recent years12, Huovila, et al.13 emphasized the need for standardized assessment methods to align and guide cities towards carbon neutrality, thereby reducing problems caused by the lack of consistency in urban carbon accounting methods and emission ranges. These studies collectively demonstrate the increasing recognition of the need for standardized yet locally adaptable frameworks, which reinforces the role of urban carbon accounting in achieving climate change mitigation and urban sustainable development. However, given the inherently open-system nature of urban environments, the delineation of transboundary carbon accounting boundaries remains a persistent challenge14, necessitating further scholarly research on methodological frameworks and data collection protocols for establishing city-scale emission inventories15. Although existing research highlights the importance of carbon emissions from urban supply chains, current carbon accounting studies focus primarily on Scope 1 and 2 emissions; Shan et al.16 developed an inventory of direct emissions (Scope 1) over a long time scale for 287 cities to support the formulation of carbon mitigation policies. Adopting Shanghai as a case study, Wei et al.17 quantified Scope 1 electricity-related carbon emissions and comprehensively analysed the underlying driving factors, Liu et al.18 examine methodologies for quantifying the carbon footprints of megacities, underscoring the imperative of conducting comprehensive greenhouse-gas emission inventories in densely populated, economically dynamic urban centers. Zhang et al.19 conducted a comprehensive assessment of Shanxi Province’s pathways toward carbon-peak and carbon-neutrality, rigorously quantifying the influences of regional characteristics—including resource endowments, energy-consumption patterns, and industrial structures—on provincial emissions. Shan et al.20 investigated Scope 1 carbon emissions attributable to both production processes and fossil fuel consumption in the Chinese lime manufacturing sector, providing a systematic assessment of emission sources in this energy-intensive industry. Additionally, our previous research aimed to provide a high-resolution carbon emission inventory for China, covering both Scope 1 and 2 emissions21. However, less attention has been given to Scope 3 emissions. Zhu et al.22 reported that Scope 3 emissions generally exceed Scope 1 or 2 carbon emissions, even surpassing the total of both types of emissions. Kucukvar et al.23 revealed that supply chain-related indirect emissions (represented by Scope 3) are responsible for nearly 56.5% of the total carbon emissions of various sectors in Turkey. Without a complete and accurate accounting of city-level carbon emissions, promoting an equitable distribution of emission reduction responsibilities among cities becomes challenging. This inefficiency may disproportionately burden upstream cities within the supply chain with the responsibility of reducing emissions. A comprehensive city-level accounting of Scope 3 emissions has been implemented for 79 members of the C40 Cities Climate Leadership Group24, and the results revealed that ignoring Scope 3 carbon emissions could result in underestimation of the global GHG emissions by 4%. Zhang et al.25 provide a comprehensive quantification of Wuyishan’s greenhouse-gas emission profile, revealing that 42% of the city’s total inventory originates beyond its administrative boundary. This findings underscore the imperative for cities to explicitly incorporate extra-territorial sources when compiling emission accounts and formulating mitigation strategies. Zheng et al.26 investigated carbon leakage induced by intercity supply chains across 309 Chinese cities and revealed that urban supply chain networks are associated with approximately 80% of the total carbon emissions. Moreover, Xia et al.27 reported that in 2017, the emission reduction achievements of nearly 186 Chinese cities were significantly influenced by the mitigation efforts of upstream cities within their supply chains. Hence, there is a lack of a unified feasibility framework for considering comprehensive Scope 1–3 emissions at the city level, posing challenges in establishing a unified urban carbon emissions dataset. To bridge these research gaps, we compiled city-level full-scope CO2 emission data for all 339 administrative cities in China. Our dataset includes Scope 1 emissions from industrial energy use, industrial processes, buildings, transportation, and agriculture; Scope 2 emissions from external power transfers; and Scope 3 emissions, which encompass emissions from supply chain activities associated with 10 key materials imported into cities. In this study, we employed the Monte Carlo simulation method to quantify the uncertainties in carbon emission data, with contrast validation against prior research findings to ensure the credibility of the dataset. Our full-scope emission dataset offers a novel, holistic perspective for evaluating the total carbon emissions in cities. By providing up-to-date, consistent, and transparent data across Scopes 1, 2, and 3, this dataset plays an important role in advancing the development of carbon data in China. It not only provides a more accurate assessment of the carbon footprint but also serves as a crucial tool for achieving an equitable distribution of carbon responsibility among cities, supporting informed policy decisions and effective climate action at the local level. Methods We adopted the IPCC-recommended baseline methodology28 (i.e., the emission factor approach) to account for carbon emissions in cities across all three scopes. The full-scope carbon emission accounting and verification model workflow comprised the following five principal phases: (1) Compilation of activity data, the primary data for Scope 1 and Scope 2 were obtained from the China High-Resolution Emission Database(CHRED)4,29 and China Carbon Watch (CCW)30, whereas the primary data for Scope 3 were acquired through the China Economic and Social Big Data Research Platform31. Following data compilation and cleaning, the requisite activity-level data were derived. Detailed sources of these data are provided in the “Activity Data” subsection of this chapter.Table 1 delineated, in detail, the primary data sources employed for the quantification of Scope 1, Scope 2, and Scope 3 emissions; 3 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ (2) For the acquisition of emission factors, please refer to SupplementaryTable A2for sector-specific emission factors and their sources categorized by scope; (3) Calculation of carbon emissions by Scope, the carbon emissions for each scope were calculated by substituting the corresponding activity-level data and emission factor data into the respective formulas presented below, thereby deriving the emissions for each scope. (4) Implementation of uncertainty analysis and validation protocols, wherein Monte Carlo simulations were employed to quantify data uncertainties, complemented by systematic contrast verification against established databases to ensure methodological reliability. (5) Data visualization, the spatial distribution of carbon emissions was visualized using ArcGIS (version 10.8) to generate carbon emission maps, facilitating the presentation of this study’s dataset. System boundary for full-scope emissions. As shown in Fig.1, in this dataset, Scope 1 carbon emissions referred to the direct emissions resulting from the combustion of fossil fuels across key sectors, including industrial energy, buildings, agriculture, and transportation, as well as emissions from industrial processes and activities. Carbon emissions or absorption associated with deforestation or land-use changes were not included. Scope 2 emissions referred to the carbon emissions caused by importing electricity from outside prefecture-level cities. As demonstrated by Hillman et al.32, food, water, and cement (shelter) constituted essential material pillars for urban operations. These critical materials were primarily produced beyond urban administrative boundaries, thus necessitating cross-border transportation networks to sustain metropolitan functions. On this basis, we employed hybrid analysis33 to identify ten key materials for urban construction and operation processes as specific products for Scope 3 carbon emission accounting, including seven major agricultural products—rice, wheat, corn, pork, beef, mutton, and poultry—and two key construction materials—steel and cement—as well as water, which was classified as a resource product. The rationale for selecting these materials in this accounting framework was based on the following considerations: the agricultural production sector, encompassing both crop cultivation and livestock farming, represented the second-largest source of carbon emissions in China34; the steel and cement industries collectively accounted for approximately 15% of the global anthropogenic CO2 emissions35; as the world’s largest developing nation, Chinese cities were projected to experience sustained growth in its demand for these materials36,37; and water-induced carbon emissions were manifested primarily through energy consumption associated with water-related activities, which currently accounted for approximately 2% of the nation’s total electricity usage. This proportion was projected to increase significantly over the coming decades38. Scope 1 methodology. Scope 1 emissions were calculated via Eq. (1). ∑ =×+ CE QEFCE(1) ScopeFueluse ii Industrial process1, Province City Scope 1 Data Source Scope 2 Data Source Scope 3 Data Source Agricultural products Steel Cement Water Beijing Beijing CHRED 3.04,29 Beijing Statistical Yearbook 202461 Beijing Statistical Yearbook 2024 Statistical Bulletin on National Economic and Social Development in 202362 Beijing 2023 Water Resources Bulletin63 Tianjin Tianjin CHRED 3.0 Tianjin Statistical Yearbook 202464 Tianjin Statistical Yearbook 2024 Tianjin 2023 Water Resources Bulletin65 Hebei Shijiazhuang CHRED 3.0 Hebei Statistical Yearbook 202466 Hebei Statistical Yearbook 2024 Bulletin of Water Resources of Hebei Province in 202367 Hebei Tangshan CHRED 3.0 Hebei Statistical Yearbook 2024 Hebei Statistical Yearbook 2023 Bulletin of Water Resources of Hebei Province in 2023 Hebei Qinhuangdao CHRED 3.0 Hebei Statistical Yearbook 2024 Hebei Statistical Yearbook 2023 Bulletin of Water Resources of Hebei Province in 2023 Hebei Handan CHRED 3.0 Hebei Statistical Yearbook 2024 Hebei Statistical Yearbook 2023 Bulletin of Water Resources of Hebei Province in 2023 Hebei Xingtai CHRED 3.0 Hebei Statistical Yearbook 2024 Hebei Statistical Yearbook 2023 Bulletin of Water Resources of Hebei Province in 2023 Table 1. Table of Primary Data Sources (Selected Cities). Note: For the Primary data sources of the remaining cities, please refer to Supplementary TableA1. 4 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ where QFuel use,i denotes the different types of fuels used in industrial, building, transportation and agricultural sectors; EFi denotes the emission factor of various types of fuels; and CEIndustrial process denotes the carbon emission data for industrial processes in local production. Scope 2 methodology. The assumptions made in the Scope 2 methodology were as follows: 1) The electricity produced in a given city was prioritized for local supply. Scope 2 emissions were calculated via Eqs. (2) and (3). CE QEF(2) ScopeCityelectricity import Powergrid2 =× QQ Q(3) City electricityimportCitypower consumptionCitypower generation =− where QCity electricity import denotes the amount of electricity imported; EFPower grid denotes the emission factors of regional power grids; QCity power consumption denotes the amount of electricity consumed; and QCity power generation denotes the amount of electricity produced within the city itself. We assumed that locally produced electricity is prioritized for local consumption. If QCity electricity import is negative, this indicates that this city exports electricity on a net basis and does not import outside electricity. Scope 3 methodology. The assumptions made in the Scope 3 methodology were as follows: 1) The import volume of the city was recorded as the difference between the city’s demand and supply. 2) Products manufactured within the city were prioritized for local supply, whereas critical material shortages resulting from final demand activities were addressed through external sourcing to achieve a supply‒demand balance without considering the inventory. Scope 3 emissions were calculated via Eqs. (4) and (5). ∑ =× = CE QEF (4) Scope i n import ii3 1 , =−QQ Q(5) import idemandi supplyi,,, where Qimprot,i denotes the import volume of the i-th material, i denotes the type of material being accounted for, EFi is the carbon emission factor of the i-th product’s supply chain, Qdemand,i denotes the demand for the i-th product, and Qsupply,i denotes the supply volume of the i-th product. activity data Scope 1 activity data. Activity data for Scope 1 carbon emissions were derived mainly from enterprise-level data of the China High-Resolution Emission Database4,29 (CHRED 3.0) and prefecture-level city data from China Carbon Watch30. Specifically, the accounting of emissions in the electricity sector was based on accounting models developed from data such as enterprise GHG verification reports, city statistical yearbooks, and industrial electricity consumption. Fig. 1 Multivariate model of full-scope carbon emission accounting. 5 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ For the industrial sector, the model was based on data from the CHRED database, carbon emission verification data, annual reports of environmental statistics, quarterly reports of enterprises in key industries, and online monitoring data from enterprises. Road transport data were obtained via an accounting model established on the basis of the daily urban road network speed and congestion index. Water transport and railway data were modelled on the basis of the CHRED database and monthly turnover data. Aviation data were modelled on the basis of daily information on the number of flights per airport and the engine parameters of various aircraft types. Data for the construction sector were modelled on the basis of Suomi-NPP VIIRS night light data, land use data and CHRED data. Data for the agriculture sector were modelled on the basis of the value added of agriculture, forestry and fisheries and the CHRED database. Scope 2 activity data. Activity data for Scope 2 emissions included power generation from fossil energy sources and nonfossil energy sources. Fossil power generation data were derived mainly from government documents. Nonfossil power generation data included four categories: wind power, solar power, hydropower, and nuclear power. Single-point source data of wind power, photovoltaic power and hydropower in China were obtained via remote sensing and GIS technologies. Combined with the actual provincial installed power generation capacity in 2023, statistics and GIS technology were used to obtain urban wind power, solar power and hydropower generation data. Nuclear power generation data were obtained by using unit-level generation information from statistics provided by the China Nuclear Energy Industry Association39. Scope 3 activity data. The supply volume of products for Scope 3 emission accounting can be relatively easily obtained from government public data or statistical yearbooks, whereas estimating the demand volume is more complex. Therefore, in this study, product demand data were obtained via the hybrid analysis method. As shown in Fig.2, on the basis of the different sources of the original data, three methods were designed in this study to account for activity level data. The input data sources used for carbon accounting in each city are documented in Table1 (Note: For cities whose 2024 statistical yearbook has not yet been released, the corresponding provincial 2024 statistical yearbook is used in conjunction with each city’s 2023 statistical bulletin to complete the calculations). Method 1. Method 1 involved obtaining the inflow quantities through surveys conducted with relevant departments. Specifically, data on agricultural product inflows were collected through surveys with the city’s Department of Agriculture or other departments responsible for agriculture; data on the industrial product inflows needed for construction, road paving, and industrial production were obtained from the city’s Urban and Rural Construction Bureau, Traffic Bureau, and Industry and Commerce Bureau; and water inflows were measured through surveys with the city’s Water Resources Bureau. Method 2. When only municipal-level statistical sources were available, such as those documented in city statistical yearbooks and government bulletins, Method 2 was employed. In Method 2, the production and demand quantities for each product type were calculated separately, with the product inflow quantity derived from the difference between the two. The specific calculation method was as follows: (1) Three types of food crops CE QQ EF[( 80%)] (6) Scopeigrainproduct i demand iproductioni i3, ,, ∑ =−×× ∈ where Qproduction,i denotes the output of food crops in statistical yearbooks by city; Qdemand,i denotes the food crop demand data by city; and EFi denotes the emission factors of food crops. Production data were directly obtained from annual statistical reports. According to previous studies, 38% of China’s grain production was used to meet the food demand of urban and rural residents, while 42% was used to satisfy the demand for animal feed. The actual grain supply was estimated at 80% of the total grain production40. The demand data were divided into two parts: food consumption data for urban and rural residents and grain demand data for livestock feed. The demand was calculated as follows: =+ −− QQ Q(7) demand idemandgrain forpeopledemandgrain foranimal, QPFC (8) demand grainfor people igrain productcity levelper capita graini ,, =× −∈− QPFC EF (9) demand grainfor animal igrain productanimalsi perunitanimalgrain itypeofgrains ,,, =× × −∈ where Qdemand-grain for people, i∈grain product denotes the food demand quantity for the i-th grain type; Pcity-level denotes the urban resident population; FCper capita grain, i denotes the per capita annual consumption of grain type i in the city; Qdemand-grain for animals, i∈grain product denotes the feed grain demand quantity for the i-th type of livestock; Panimals, i denotes the annual livestock population data; FCper unit animal grain, i denotes the amount of grain used per unit of animal product production; and EFtype of grains denotes the allocation ratio of grains used as feed. 6 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ (2) Four major types of meat CE QQ EF[( 90%) ](10) Scopeimeat product i demand iproductioni i3, ,, ∑ =−×× ∈ where Qproduction,i denotes the production quantity of four major meat products from the annual statistical yearbooks of each city; Qdemand,i denotes the demand data for each type of meat product; and EFi denotes the emission factor for each type of meat. Production data were directly obtained from city statistical yearbooks. On the basis of previous studies and considering the consumption of meat waste and offal, the actual meat supply was calculated as 90% of the total meat production40. The demand data were solely based on the meat consumption data of urban and rural residents. The demand was calculated as follows: = − QQ (11) demand idemandmeat, =× −∈ − QPFC (12) demand meat imeatproduct city levelper capita meat i ,, where Qdemand-meat,i∈meat product denotes the demand quantity for the i-th type of meat product; Pcity-level denotes the urban resident population; and FCper capita meat,i denotes the per capita annual consumption of the i-th type of meat at the city level. (3) Steel CE QQ EF[( )] (13) Scopesteelsdemandproductionsteels3 ∑ =−× − where Qproduction denotes the steel production quantity from the annual statistical yearbooks of each city, Qdemand denotes the steel demand data for each city, and EFsteels denotes the emission factor for crude steel. Crude steel production data were obtained from city statistical yearbooks, whereas steel demand data were calculated on the basis of three components, namely, the steel demand for building construction, the steel demand for transportation infrastructure construction, and the steel demand for consumer goods production. The steel demand for building construction was derived by estimating the steel consumption for new buildings by construction type in the city. The steel demand for transportation infrastructure was calculated by estimating the steel consumption for new railways and highways in the city. The steel demand for consumer goods production was determined by measuring the steel content in consumption goods purchased and transported into the city from external regions over the past year, such as transportation, Fig. 2 Decision tree for Scope 3 demand product calculation. 7 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ machinery, and household appliances. The calculation was as follows: =+ + −− − QQ QQ (14) demand demand steel building demand steel transportation demand steel goods,, , QASC (15) demand steel building jcity levelj unit area j,,, ∑ =× −− =× +× − QLSC LSC(16) demand steel transportation railwayunitlengthrailway highroad unit length highroad, QPSC (17) demand steel goodsggunitgood, ∑ =× − where Qdemand-steel, building, Qdemand-steel, transportation and Qdemand-steel,goods denote the steel demands for building construction, transportation infrastructure construction, and consumer goods production, respectively; j denotes the different building types; Acity-level,j denotes the annual new construction area for building type j; and SCunit area,j denotes the steel consumption per unit area of new construction for each building type. Moreover, Lrailway and Lhighroad denote the annual new construction mileages of railways and highways, respectively; SCunit length railway and SCunit length highroad denote the steel consumption values per unit length of newly constructed railways and highways, respectively; Irailway and Ihighroad denote the annual investments in new railways and highways, respectively; SCunit length railway and SCunit length highroad denote the steel consumption values per unit investment in new railways and highways, respectively; g denotes the major types of consumer goods, including transportation, machinery, household appliances, and other consumer goods; Pg denotes the quantity of each major consumer good; and SCunit good denotes the steel consumption per unit of each major consumer good. (4) Cement CE QQ EF[( )] (18) Scopecementdemandproductioncement3 ∑ =−× − where Qproduction denotes the cement production quantity from the annual statistical yearbooks of each city, Qdemand denotes the cement demand data for each city, and EFcement denotes the emission factor for cement. Cement production data were obtained from statistical yearbooks. Cement demand data were calculated on the basis of three components, namely, the cement demand for urban building construction, which was estimated by calculating the cement consumption for new buildings by building type in the city; the cement demand for urban transportation infrastructure construction, which was calculated by estimating the cement consumption for new railways and highways in the city; and the cement demand for urban industrial infrastructure construction, which was calculated by converting the industrial fixed asset investment in the city into the level of cement consumption. The calculation was as follows: =+ + −− − QQ QQ (19) demand demand cement building demand cement transportation demand cement industry,, , ∑ =× −− QACC ( 20) demand cement building jcity levelj unit area j,,, QLCC LCC ( 21) demand cement transportation railwayrailway highroad highroad, =× +× − QICC (22) demand cement industry industry investment unit value,, =× − where Qdemand-cement, building, Qdemand-cement, transportation and Qdemand-cement,industry denote the cement demands for building construction, transportation infrastructure construction, and industrial infrastructure construction, respectively; Acity-level,j denotes the annual new construction area for building type j, CCunit area,j denotes the cement consumption per unit area of new construction for each building type; Lrailway and Lhighroad denote the annual new construction mileages of railways and highways, respectively; CC railway and CC highroad denote the cement consumption levels per unit length of newly constructed railways and highways, respectively; Iindustry,investment denotes the industrial fixed asset investment at the original price; and CCunit value denotes the cement material strength per-unit value. (5) Water ∑ =−× − CE QQ EF[( )] (23) Scopewater demand production water3 where Qproduction denotes the water resource production quantity from the water resource bulletins of each city, Qdemand denotes the water demand data for each city, and EFwater denotes the water resource emission factor. Water production datawere obtained from the water resource bulletins published by each city. The water demand of each city was calculated as follows: =+++ QQ QQQ(24) demand agricultural waterindustrialwater domestic waterecologicalwater 8 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ where Qagricultural water denotes the annual agricultural water consumption of each city; Qindustrial water denotes the annual industrial water consumption of each city; Qdomestic water denotes the annual domestic water consumption of each city; and Qecological water denotes the annual ecological water consumption of each city. When data availability was limited to provincial-level statistical sources, such as those obtained from provincial statistical yearbooks and government bulletins, Method 3 was applied. In cases of inaccessible provincial data, nationally statistics data served as the alternative data source. In Method 3, a downscaling approach was adopted to calculate the product demand, thereby using provincial or national data. Method 3 was applied only to estimate the demand for industrial products. Method 3. (1) Steel The steel demand for building construction was estimated by calculating the steel consumption of new buildings by building type at the provincial or national level. This demand was then allocated on the basis of the proportion of the total new construction area in the target city relative to the overall new construction area. The steel demand for transportation infrastructure construction was calculated by determining the steel consumption for new railways and highways at the provincial or national level, which was then allocated on the basis of the proportion of the transportation infrastructure investment in the target city relative to the overall investment in transportation infrastructure. The steel demand for consumer goods production was estimated by calculating the steel consumption for consumer goods at the provincial or national level, which was then allocated on the basis of the proportion of the household consumption expenditure for corresponding consumer goods categories in the target city relative to the total household consumption expenditure in those categories. The calculation was as follows: () QASC A A (25) demand steel building jprovince nation levelj unit area jcity level province nation level ,(), ,() ∑ =×× −− − − QQ I I (26) demand steel transdemandsteel transprovincenationcity leveltrans province nation leveltrans ,,,(), (), =× −− − − ∑ =      ×        −− −∈ −∈ QQ I I (27) demand steel goodsgdemand steel goodsprovincenationg city levelg G province nation levelg G ,,,(), , (), where Aprovince(nation)-level,j denotes the annual new construction area for building type j at the provincial or national level; Acity-level denotes the annual new construction area in the city; SCunit area,j denotes the steel consumption per unit area of new construction for each building type; and Aprovince(nation)-level denotes the annual new construction area at the provincial or national level. Moreover, Icity-level,trans denotes the annual transportation infrastructure investment in the city; Qdemand-steel,trans,province(nation) denotes the amount of steel consumed in transportation infrastructure construction at the national or provincial level; Iprovince(nation)-level,trans denotes the annual transportation infrastructure investment at the provincial or national level; Qdemand-steel,goods,province(nation) denotes the apparent crude steel consumption in the industrial consumer goods manufacturing sector at the national or provincial level; Icity-level,g∈G denotes the annual household consumption expenditure for consumer goods; and Iprovince(nation)-level,g∈G denotes the annual household consumption expenditure for consumer goods. If the above provincial-level statistical data were unavailable or could not be fully obtained, the national annual steel consumption was used and allocated to the city on the basis of the urban resident population. The calculation was as follows: QSC P P (28) demand steel nation city level nation level =× − − − where SCnation denotes the total national steel demand; Pcity-level denotes the urban resident population of the target city; and Pnation-level denotes the national resident population. (2) Cement The national annual cement consumption was obtained and allocated to a given city on the basis of the urban resident population as follows: QSC P P (29) demand cement nation city level nation level =× − − − where SCnation denotes the total national steel demand; Pcity denotes the urban resident population of the target city; and Pnation denotes the national resident population. emission factors. All emission factors required for the calculations are provided in Supplementary TableA2. 9 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ Data validation and uncertainty analysis. Cross-checking was widely recognized as a critical method for ensuring data integrity and consistency across multiple research teams41. This approach, rooted in the principles of rigorous scientific validation, allowed for the identification and correction of biases that might have resulted from individual analytical processes or team-specific methodologies. By engaging multiple teams in independent yet standardized analyses of different data subsets, cross-group validation facilitated the verification of results and ensured the robustness and reproducibility of the findings across varying experimental conditions. Furthermore, it enhanced transparency and promoted collaboration, which was essential when diverse datasets are integrated or when large-scale studies are conducted that required interdisciplinary input. In the context of scientific research and the quality assurance of large‐scale datasets, the critical importance of cross-checking was affirmed by multiple international standards and best practices. For example, the GHG Protocol’s “Tools & Guidance” module42 advises that when deploying its suite of cross-sectoral, departmental, and city-level methodologies, inventory results produced by different tools or teams should have been subjected to horizontal comparison and cross-validation to uncover anomalies and contradictions, thereby ensuring that the final report was comprehensive, transparent, and consistent. Likewise, the Berkeley Initiative for Transparency in the Social Sciences (BITSS)43 cautions that larger team sizes can amplify biases and recommended that sub-teams independently carried out—and then mutually reviewed—each stage of data collection, cleaning, analysis, and archiving. This sub-team cross-review process guaranteed that disparate researchers following the same procedures achieved concordant outcomes, thereby safeguarding the overall coherence and transparency of the study44. The adoption of cross-group validation thus serves not only to enhance the scientific rigor of the analysis but also to promote consensus among researchers, thus ensuring that the conclusions were grounded in consistent and reproducible data. The procedure of this method is was follows: (1) Data division (the dataset was split into multiple subsets on the basis of specific criteria, with each group analysing their respective subset independently), (2) independent analysis (each team applied standardized analytical methods and models to their designated data subset), (3) cross-validation (the teams validated each other’s results to ensure consistency and accuracy across the different groups), (4) comparison and adjustment (differences in the results were identified, and necessary adjustments to methods or data processing were made), and (5) final consensus (after reconciling inconsistencies, a unified report was generated, in which the findings from all groups were consolidated). In this study, 45 researchers from 31 institutions in the China City Greenhouse45 (CCG) working group participated in the basic data collection and carbon emission calculation processes. Prior to initiating data collection and cross-group validation, we convened a one-week intensive workshop on the carbon emission accounting methodology guidelines of this study for all 45 researchers. This training ensured a uniform understanding of the procedures, underlying formulas, parameter selections, and reporting formats. Following the workshop, each researcher independently calculated the scope-specific carbon emissions using the same guidance manual, and subsequent cross-checking further confirmed the methodological robustness and the absence of any systematic computational errors. Assessing uncertainty was indispensable for robust greenhouse-gas inventory development, as it both highlighted the reliability of the current dataset and guided subsequent methodological enhancements. Incomplete or inconsistent activity records and the intrinsic variability of emission factors were the principal contributors to overall uncertainty in emissions accounting. The IPCC46 recommends two principal approaches for quantifying these uncertainties: classical error‐propagation techniques and Monte Carlo simulation. While error propagation offered a straightforward estimate by combining variances analytically, Monte Carlo methods could accommodate complex, non‐Gaussian distributions and correlations among input parameters, yielding a more comprehensive uncertainty profile. Previous work had applied Monte Carlo simulation extensively to characterize uncertainties in greenhouse‐ gas and other pollutant inventories at the national scale47–49. Drawing on those applications, along with the IPCC’s uncertainty‐analysis guidance, national GHG inventory reports50, plus peer‐reviewed evaluations of urban carbon data quality and expert judgment, we derived uncertainty bounds for our city‐level activity datasets. Moreover, a number of studies had already implemented similar stochastic approaches to quantify the uncertainties associated with Scope 1 and Scope 2 CO2 emissions in individual cities. For the analysis presented in this paper, we employed the Monte Carlo simulation method recommended by the IPCC to estimate the uncertainty in the GHG accounting results for Chinese cities. This process involved three steps: (1) defined the probability distribution as a normal distribution for the estimation parameters such as activity data and emission factors across the different sectors and cities, along with corresponding uncertainty analysis statistics; (2) calculated the corresponding emissions for each category according to the method of calculating the carbon emissions resulting from producing the products in this study; and (3) repeated the simulation 10,000 times. Probability distributions for the different categories or the total emissions were subsequently obtained, as were corresponding uncertainty analysis statistics. For detailed methodological approaches and uncertainty quantification pertaining to Scope 1 and 2 carbon emission data, readers were referred to our prior study21. In quantifying the uncertainty associated with Scope 3 carbon emissions, we adhered to the IPCC’s46 tiered guidance by assigning distinct uncertainty bounds according to the provenance of activity data. Specifically, Method 1 paralleled a Tier 3 approach, relied on high‐fidelity, field‐surveyed data collected directly by municipal authorities; the IPCC recommended an uncertainty band of 2–5%51, and we have adopted a value of 4%. Method 2 corresponded to Tier 2, drew upon figures published in municipal statistical yearbooks and other publicly available datasets; consistent with the IPCC’s suggested 5–10% range51, we applied an uncertainty of 8%. Method 3 was equivalent to Tier 1, wherein city‐level estimates were derived by proportional down‐scaling of broader aggregates; given the IPCC’s 30–100% guidance for Tier 151, we imposed a 40% uncertainty when allocating provincial data to cities and 50% when allocating 16 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ 46. Intergovernmental Panel on Climate Change (IPCC). IPCC Guidelines for national greenhouse gas inventories. Institute for Global Environmental Strategies (IGES) (2006). 47. Sha n , Y. et al. Methodology and applications of city level CO2 emission accounts in China. Journal of Cleaner Production 161, 1215–1225, https://doi.org/10.1016/j.jclepro.2017.06.075 (2017). 48. Jamatutu, S. A. et al. Quantifying future carbon emissions uncertainties under stochastic modeling and Monte Carlo simulation: Insights for environmental policy consideration for the Belt and Road Initiative Region. Journal of Environmental Management 370, 122463, https://doi.org/10.1016/j.jenvman.2024.122463 (2024). 49. Eory, V., Topp, C. F., Butler, A. & Moran, D. Addressing uncertainty in efficient mitigation of agricultural greenhouse gas emissions. Journal of Agricultural Economics 69(3), 627–645, https://doi.org/10.1111/1477-9552.12269 (2018). 50. The Fourth National Communication on Climate Change of the People’s Republic of China. 51. Pulles, T., Meijer, J. & van Aardenne, J. Estimating Uncertainties in GHG Emissions from Fuel Combustion. Intergovernmental Panel on Climate Change (IPCC) (2006). 52. Meng, F. et al. China City-Level Full-scope Carbon Emissions Dataset 2023. Figshare. https://doi.org/10.6084/m9.figshare.28645499 (2025). 53. Xu, J., Guan, Y., Oldfield, J., Guan, D. & Shan, Y. China carbon emission accounts 2020-2021. Applied Energy 360, 122837, https:// doi.org/10.1016/j.apenergy.2024.122837 (2024). 54. Wang, H., Zhang, R., Liu, M. & Bi, J. The carbon emissions of Chinese cities. Atmospheric Chemistry and Physics 12(14), 6197–6206, https://doi.org/10.5194/acp-12-6197-2012 (2012). 55. Energy Institute(EI). Statistical Review of World Energy 2025 (74th ed.). In collaboration with Kearney and KPMG. https://www. energyinst.org/statistical-review (2025). 56. International Energy Agency (IEA). CO2 Emissions in 2023. IEA, Paris. https://www.iea.org/reports/co2-emissions-in-2023 (2024). 57. Mickey, F. How has U.S. energy use changed since 1776? U.S. Energy Information Administration. https://www.eia.gov (2025). 58. Crippa, M., Guizzardi, D., Pagani, F., Pisoni, E. GHG Emissions at sub-national level. European Commission, Joint Research Centre. https://data.jrc.ec.europa.eu/dataset/d67eeda8-c03e-4421-95d0-0adc460b9658 (2023). 59. Li, W. et al. Assessment of greenhouse gasses and air pollutant emissions embodied in cross-province electricity trade in China. Resources, Conservation and Recycling 171, 105623, https://doi.org/10.1016/j.resconrec.2021.105623 (2021). 60. Tong, K. et al. Greenhouse gas emissions from key infrastructure sectors in larger and smaller Chinese cities: method development and benchmarking. Carbon Management 7(1–2), 27–39, https://doi.org/10.1080/17583004.2016.1165354 (2016). 61. Beijing Municipal Bureau of Statistics, National Bureau of Statistics Beijing Survey Team. Beijing Statistical Yearbook 2024. Beijing: China Statistics Press (2025). 62. National Bureau of Statistics of the People’s Republic of China. Statistical Bulletin on National Economic and Social Development in 2023 National Economic and Social Development. Beijing: China Statistics Press. (2024). 63. Beijing Municipal Water Affairs Bureau. Beijing Water Resources Bulletin 2023. Beijing: Beijing Municipal Water Affairs Bureau (2024). 64. Tianjin Municipal Bureau of Statistics, National Bureau of Statistics Tianjin Survey Team. Tianjin Statistical Yearbook 2024. Beijing: China Statistics Press (2025). 65. Tianjin Municipal Water Affairs Bureau. Tianjin Water Resources Bulletin 2023. Tianjin: Tianjin Municipal Water Affairs Bureau. (2024). 66. Hebei Municipal Bureau of Statistics, National Bureau of Statistics Hebei Survey Team. Hebei Statistical Yearbook 2024. Beijing: China Statistics Press (2025). 67. Hebei Municipal Water Affairs Bureau. Hebei Water Resources Bulletin 2023. Hebei: Hebei Municipal Water Affairs Bureau (2024). acknowledgements This research was funded by National Natural Science Foundation of China (72522012, 72174028, 72504111)and the Horizon Europe Project EU-CHINA-BRIDGE (101137971), which are supported by UKRI grant (10132630) at the University of Birmingham. This study was performed by volunteers from the China City Greenhouse Working Group. The authors would like to thank the following people for their contributions to the fullscope CO2 emissions database and this article: This study utilized a collaborative working group model for data collection. The authors gratefully acknowledge the 45 members of the China City Greenhouse Working Group for their significant contributions to the collection, extraction, and initial compilation of the underlying activity data used in the full-scope CO2 emissions database. These individuals served as data contributors and are not co-authors of this manuscript. Scope 1 and Scope 2 activity data were primarily sourced from publicly accessible national databases. All Scope 3 activity data were meticulously gathered from officially published citylevel statistical yearbooks and statistical bulletins; detailed sources for each city are provided in Supplementary TableA1. The authors would like to thank the following people for their data collection efforts: Chen Lv, Chinese Academy of Environmental Planning. Jing Guo, Chinese Academy of Environmental Planning. Lingyun Pang, Chinese Academy of Environmental Planning. Wanyue Shan, Beijing University of Chemical Technology. Xu Zhang, School of Environment, Beijing Normal University. Jiaqi Hou, School of Environment, Beijing Normal University. Jintao Sun, School of New Energy and Environment, Jilin University. Yimiao Wang, Beijing Normal University. Qiwen Ma, Beijing Normal University. Fumin Gu, Suzhou Xieren Environmental Protection Technology Service Co., Ltd. Yanqing Wang, Dalian Institute of Chemical Physics, Chinese Academy of Sciences. Yiwen Sun, Public Testing Evaluation and Identification Technology Center, Jiangsu Academy of Agricultural Sciences. Shijie Li, Guangdong University of Technology. Jianhui Cong, School of Economics and Management, Shanxi University. Dandan Zhang, Department of the Built Environment, Aalto University, Finland. Buyrayem, Xinjiang Agricultural University. Shengnan Zhao, Chifeng University. Yongliang Yang, Zhejiang Sci-Tech University. Zhiyuan Ning, Shanghai Hebang Certification Co., Ltd. (NSF International). Biao Liu, School of Land Science and Technology, China University of Geosciences (Beijing). Zhihong Li, Hong Kong Huayi Design Consulting (Shenzhen) Co., Ltd. Yiwei Xiong, Foshan Institute of Quality and Standardization. Mengli Wu, Tianjin Youmei Environmental Protection Technology Co., Ltd. Zhouye Zhao, Tongji University. Nan Li, Hefei University of Technology. Yue Hu, Dalian Institute of Chemical Physics, Chinese Academy of Sciences. Yao Tang, Wuhan University. Lu Yang, Wuhan Textile University. Mengbing Du, Wuhan University. YunLong Wu, Capital University of Economics and Business. YuShen Liu, Hongjie Ji, Shanxi University. Wei He, Wuhan University. Jian Huang, Wuhan University. Peng Li, Hangzhou Bingxin Environmental Protection Packaging Co., Ltd. Mingji Lao, Fangyuan Mark Certification Group Co., Ltd. 17 Scientific Data | (2025) 12:1672 | https://doi.org/10.1038/s41597-025-05949-y www.nature.com/scientificdata www.nature.com/scientificdata/ author contributions Fanxin Meng and Bofeng Cai led the project and designed the research; Hanbo Hu, Yutong Sun, Jiaqi Hou, Zhe Zhang and Lingyun Pang collected the raw data, assembled the data and participated in the dataset construction; Fanxin Meng, Hanbo Hu and Li Zhang wrote the manuscript; Yutong Sun, Li Zhang, Zhe Zhang and Yuli Shan revised the manuscript. Competing interests The authors declare no competing interests. additional information Supplementary information The online version contains supplementary material available at https://doi.org/ 10.1038/s41597-025-05949-y. Correspondence and requests for materials should be addressed to F.M., L.Z., Z.Z. or Y.S. Reprints and permissions information is available at www.nature.com/reprints. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution-NonCommercialNoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/. © The Author(s) 2025