City-level process-related CO2 emissions in China 2000–2021

1 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata City-level process-related CO2 emissions in China 2000–2021 Sijia Cai1,2, Jinghang Xu2, Yuru Guan 2, Miaomaio Liu1,3 ✉ , Chang tan 4, Jun Bi1,3 & Yuli Shan 2,5 ✉ As the world’s largest CO2 emitter, China needs accurate city-level CO2 emission accounts to formulate effective low-carbon policies. However, previous studies mainly accounted for emissions from fossil fuel combustion and overlooked process-related CO2 emissions from industrial production (e.g., mineral, chemical, metal products), which account for approximately 13% of China’s total emissions. In this study, we built the first time-series dataset of process-related CO2 emissions for 289 Chinese cities from 2000 to 2021. The dataset covers 11 industrial products and adheres to the methodology recommended by the Intergovernmental Panel on Climate Change (IPCC). We applied China-specific emission factors and compiled industrial output data from city statistical yearbooks and bulletins. Missing output data were imputed using missForest models. The estimated uncertainty of the process-related emissions in our dataset ranges from 3.87% to 3.91%. Our dataset provides a robust foundation for analyzing emission patterns at the city level and for designing targeted low-carbon policies. Background & Summary Climate change is one of the most urgent challenges, and countries are actively seeking solutions to reduce carbon dioxide (CO2) emissions1,2. To design effective solutions, accurate and detailed CO2 emission inventories are essential3,4. According to the Intergovernmental Panel on Climate Change (IPCC), CO2 emissions in a city may originate from two main sources: energy-related emissions from fossil fuel combustion and process-related emissions from chemical or physical transformations of raw materials without combustion, such as carbonate decomposition or metal reduction5. While energy-related CO2 emissions have been extensively studied and widely incorporated into policy-making, largely due to their large share and the availability of mature accounting methods, process-related emissions have received far less attention6–8. According to the IPCC, more than 30 industrial processes worldwide generate process-related CO2 emissions5. Overlooking process-related emissions in policy-making may leave major emission sources (e.g., cement and crude steel) unregulated. As the world’s largest emitter of CO2, China plays a pivotal role in global climate mitigation9. In response, China has proposed “dual-carbon” goals—to peak CO2 emissions before 2030 and achieve carbon neutrality before 2060. China’s accounting for over 50% of global output in key industrial products—such as cement, flat glass, crude steel, and aluminum10— highlighting its role as the global manufacturing hub. The country’s large-scale industrial activities and substantial demand for raw material processing, has led to high levels of process-related emissions11,12. According to China’s First Biennial Transparency Report on Climate Change, China’s process-related CO2 emissions reached 1,524 million tonnes (Mt) in 2021, making up 13% of the nation’s total emissions13. This figure is approximately 3.2 times as large as the total fossil CO2 emissions of Russia, the world’s fourth-largest emitter14. Given the substantial variation in resource endowment and development stage among Chinese cities, the contribution of process-related CO2 emissions also varies across cities15,16. If these emissions are not properly accounted for, city-level carbon inventories may underestimate total emissions. In recent years, scholars have increasingly recognized the importance of accurately accounting for process-related CO2 emissions. However, existing studies face three major limitations. First, most previous studies usually use global default emission factors recommended by the IPCC, such as those used by Cui et al.17 and Hu et al.18. However, these default emission factors can not reflect country-specific industrial processes, 1State Key Laboratory of Pollution control and Resource Reuse, School of the environment, nanjing University, Nanjing, 210023, China. 2School of Geography, earth and environmental Sciences, University of Birmingham, Birmingham, B15 2TT, UK. 3Basic Science Center for Energy and Climate Change, Beijing, 100081, China. 4Department of Earth System Science, Tsinghua University, Beijing, 100084, China. 5Birmingham institute of Sustainability and Climate Action (BISCA), University of Birmingham, Birmingham, B15 2TT, UK. ✉e-mail: liumm@ nju.edu.cn; y[email protected] DATA DeSCrIPTOr OPeN 2 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata www.nature.com/scientificdata/ technologies, and raw material structures. China-specific emission factors are substantially lower than IPCC default values—by 79% for ferroalloys and 65% for plate glass5,19,20. As a result, applying global default factors to China may lead to overestimation of emissions from certain products. Second, most existing studies focus on the national scale, making it difficult to capture city-level differences in emission trends and patterns. Studies such as those by Birgit et al.21 and Tan et al.22 provide only national-level estimates for regions like the EU or China. Third, the coverage of industrial product types remains limited. Even at the national level, many studies focus only on a small number of representative products. For example, Xiao et al. estimated process-related emissions for a single industrial product in Russia23, while Liu included five products in China24. Systematic accounting across a broader range of industrial sectors is still lacking, making it difficult to fully reflect the sectoral structure of process-related emissions. Fourth, although some recent studies have attempted city-level estimations, they have primarily focused on the cement production while neglecting other industrial products. Examples include the work by Guo et al.25 and Shan et al.6. Such narrow scope limits our understanding of the full landscape of process-related emissions at the city level. A key challenge in constructing a city-level inventory of process-related CO2 emissions lies in the limited availability of complete industrial output data and country-specific emission factors. To address these issues, this study undertook comprehensive data compilation and preprocessing, an extensive literature review, and an uncertainty analysis. Specifically, we compiled output data for 11 products across 289 Chinese cities for the period 2000–2021. Missing values were imputed using four missForest models to ensure data completeness and consistency. The selected products—cement, plate glass, calcium carbide, ethylene, ammonia, soda ash, crude steel, ferroalloy, aluminum, lead, and zinc—account for over 77% of China’s process-related CO2 emissions, and the 289 cities represent more than 90% of the national gross domestic product (GDP)13. Several other industrial products (e.g., lime, pig iron, carbon black, fluorochemicals, and ceramics) were excluded due to extensive data gaps at the city level. In parallel, we conducted an extensive review of the literature to collect country-specific emission factors that better reflect China’s industrial processes. To evaluate the robustness of the emission estimates, we incorporated the coefficients of variation (CVs) for both activity data and emission factors into a Monte Carlo simulation to quantify the uncertainty in the CO2 emission estimates. Building on this effort, we constructed the first long-term, consistent dataset covering 289 Chinese cities. To ensure transparency, our dataset includes not only the estimated emissions but also the underlying industrial output data used in the calculations, comprising a total of 139,876 records. Our dataset, therefore, provides robust data support for city-level low-carbon policy design and implementation. Methods Scope and accounting boundary. This study estimates process-related CO2 emissions from 11 industrial products across 289 Chinese cities over the period 2000–2021. These products span three major industrial sectors: the mineral industry (cement and plate glass), the chemical industry (calcium carbide, ethylene, ammonia, and soda ash), and the metal industry (crude steel, ferroalloy, aluminum, lead, and zinc). To perform the estimation, we adopt the IPCC-recommended administrative territorial-based approach, which accounts for emissions generated within the geographical boundaries of each city. Figure1 presents the framework used to construct the annual city-level inventory of process-related CO2 emissions. Process-related CO2 emission accounting approach. The emission factor approach is currently the most widely used method for estimating CO2 emissions. It calculates emissions by multiplying the activity data by the corresponding emission factors, making it suitable for large-scale assessments and for contexts where Fig. 1 Methodological framework for annual city-level territorial CO2 emissions from industrial processes. 3 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata www.nature.com/scientificdata/ direct monitoring data are unavailable22,25. The IPCC provide two approaches for selecting emission factors: Tier 1, which uses global default values, and Tier 2, which applies country-specific emission factors that better reflect local production processes5,26. To enhance the accuracy and reliability of our estimates, this study adopts the Tier 2 approach for all 11 industrial products, incorporating China-specific emission factors. The general formula used for emission estimation is presented as follows: EADEF(1) iii =× where Ei represents CO2 emissions from product i. ADi represents the activity data (i.e., the output of product i), and EFi is the emission factor, indicating the amount of CO2 emitted per tonne of product i. For certain products such as soda ash, crude steel, lead, and zinc, we follow the methodology proposed by Yu et al.27, using a weighted average of emission factors based on the share of different production processes to improve estimation accuracy. The specific calculation methods for these products are described below: (1) Soda ash: Among the three soda ash production processes, only the natural soda process generates CO2 emissions. According to the China Soda Ash Industry Association (CSIA), this process accounts for 5% of China’s total soda ash production. Emissions are calculated as: EADEF5% (2) soda ashsodaash soda ash =∗∗ where ADsoda ash is the total output of soda ash, and EFsoda ash is the emission factor for soda ash. (2) Crude steel: Process-related CO2 emissions from crude steel production arise from carbon oxidation during the conversion of pig iron to crude steel, and limestone decomposition used as a flux in steelmaking. Crude steel is primarily produced via two technological routes: the blast furnace-basic oxygen furnace (BF-BOF) and the electric arc furnace (EAF). The first component is calculated by assessing the difference in carbon content between pig iron (4.1%) and crude steel (0.25%), and converting to CO2 using the molecular weight ratio (44/12): =+∗. −. ∗ − EADAD()(4 1% 025%) 44 12 (3) crudesteel crudesteel BF BOFcrude steel EAF(1)()() where ADcrudesteel BF BOF() − and ADcrudesteel EAF() are the outputs of crude steel from each respective process. The second component arises from limestone decomposition. The specific limestone consumption is 0.3 tonnes per tonne of crude steel for BF–BOF and 0.064 tonnes for EAF28. Emissions from this source are calculated as: =∗.+ ∗. ∗ − EADADEF(030064) (4) crudesteel crudesteel BF BOFcrude steel EAFlimestone(2)()() where EFlimestone is the emission factor for limestone decomposition. (3) Lead: Lead production in China consists of primary and secondary sources. Among primary sources, the imperial smelting furnace (ISF) accounts for 2.5% of output, while direct smelting (DS) contributes 97.5%29. Secondary lead is primarily recovered via battery processing (BP) from used lead-acid batteries. The total emissions from lead production are calculated as: EADEFADEFADEF25%975%(5) lead lead lead ISFleadleadDSleadsecondary lead BP() () () () =∗.∗ +∗.∗ +∗ where ADlead and ADlead secondary() denote total and secondary lead outputs, respectively. EFlead ISF() , EFlead DS() and EFlead IBP() are the emission factors for ISF, DS, and BP processes, respectively. (4) Zinc: In China, zinc is mainly produced through ISF, electro-thermic distillation (ETD), and pyrometallurgical processing (PMP). Only the ISF process generates CO2 emissions, accounting for 5% of national zinc output29. Thus, emissions are calculated as: =∗∗EAD5%EF(6) zinc zinc ISFzinc () where ADzinc is the total zinc output, and EFzinc ISF() is the emission factor for zinc production using the ISF process. Activity data and emission factors. In this study, activity data refer to the annual output of 11 industrial products from industrial enterprises above designated size across 289 cities. These data were collected from city-level statistical yearbooks and bulletins30. Each city’s yearbook was accessed via its official government website (e.g., http://tjj.beijing.gov.cn/; http://tjj.nanjing.gov.cn/). However, due to inconsistencies in data quality at the city level, approximately 2% of the output data are missing. To overcome this problem, we applied missForest, a nonparametric imputation method based on random forests, for data interpolation31. This method is well-suited for imputing both continuous and categorical variables, particularly in datasets with complex interactions and nonlinear relationships32,33. To improve imputation accuracy, we developed four separate missForest models based on the industry classification introduced earlier. Specifically, the mineral and chemical sectors each used a single model, while the metal sector was further divided into ferrous and non-ferrous smelting and rolling. This sectoral division is informed by well-established distinctions in production technologies, statistical reporting practices, and data 4 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata www.nature.com/scientificdata/ distributions between ferrous and non-ferrous metals, which may lead to divergent patterns of missingness and variation in output data. Each model incorporated sector-specific predictors, including total industrial output value, average number of employees, main business income, main business cost, and city type. The first four variables were obtained from city-level statistical yearbooks30, and city types were derived using K-means clustering based on the employment and GDP structure. For emission factors, we conducted a comprehensive review of the existing literature and compiled values from multiple sources (see Table1). Due to the lack of empirical data at the city level and over time, these emission factors are assumed to be uniform and time-invariant throughout the study period and across all cities. Data Record The dataset is available as open access via Figshare (https://doi.org/10.6084/m9.figshare.29085002.v3)34 and contains 139,876 records on process-related CO2 emissions and industrial product output. Specifically: • 69,938 records document the output of 11 industrial products in 289 cities from 2000 to 2021 (File “China industrial product output inventory at the city level”); • 69,938 records document process-related CO2 emissions for the same 11 products across 289 cities from 2000 to 2021 (File “China process-related CO2 emission inventory at the city level”). The CO2 emission inventory follows a standardized format and is organized across multiple spreadsheet sheets, with each sheet corresponding to a specific year from 2000 to 2021. Within each sheet, the data are structured into 12 columns and 289 rows. The columns represent process-related emissions for each of the 11 industrial products, along with the total process-related emissions for each city. The rows represent the 289 cities. Each entry records the emissions associated with producing a specific product in a given city. technical Validation Statistical overview of the dataset. Between 2000 and 2021, process-related CO2 emissions from the 11 industrial products showed a notable increase. Total emissions rose from 294 Mt in 2000 to 1,110 Mt in 2021, with an average annual growth rate of 6.5% and an annual increase of 39 Mt. Among the three industries, the mineral industry contributed the largest absolute increase in emissions, the metal industry exhibited the highest relative growth rate, while the chemical industry experienced comparatively slower growth. The mineral industry was the largest contributor to process-related CO2 emissions. Emissions increased from 168 Mt in 2000 to 632 Mt in 2021, with an average annual increase of 22 Mt and a growth rate of 6.5%. Figure2a shows that between 2014 and 2021, emission hotspots in this industry gradually shifted from northern and northeastern China to southern and southwestern regions. This spatial shift is largely driven by rapid economic growth and rising construction demand in southern provinces after 201435,36. The fastest-growing cities were Meizhou in Guangdong province (0.4 Mt/yr), Sanming in Fujian (0.36 Mt/yr), and Chengde in Sichuan (0.32 Mt/yr). In contrast, Shijiazhuang in Hebei (−0.7 Mt/yr), Xuzhou in Jiangsu (−0.59 Mt/yr), and Changchun in Jilin (−0.39 Mt/yr) experienced substantial declines. Figure2d presents the emission composition of the mineral industry in 2021, showing that cement accounted for 99%, while the contribution from plate glass was relatively minor. Industrial product Production process Value Unit Source Cement / 0.29 tCO2/t Liu et al.19 Plate glass / 0.0737 tCO2/t Hu et al.20 Calcium carbide / 1.15 tCO2/t NDRC, 201146 Ethylene / 2.25 tCO2/t IPCC, 20065 Ammonia / 2.97 tCO2/t Yu et al.27 Soda ash Natural soda 0.14 tCO2/t IPCC, 20065 IPCC, 20065 IPCC, 20065 Solvay soda 0.00 tCO2/t Hou’s soda 0.00 tCO2/t Limestone use / 0.43 tCO2/t Limestone NDRC, 201146 Carbon content of pig iron / 4.10 % NDRC, 201146 Carbon content of steel / 0.25 % NDRC, 201146 Ferroalloy / 0.28 tCO2/t NDRC, 202047 Aluminum / 1.50 tCO2/t NDRC, 202047 Lead Imperial smelting furnace 0.66 tCO2/t Sjardin, 200348 Direct smelting 0.25 tCO2/t Sjardin, 200348 Battery processing 0.20 tCO2/t Sjardin, 200348 Zinc Imperial smelting furnace 3.12 tCO2/t Sjardin, 200348 Electro-thermic distillation 0.00 tCO2/t Sjardin, 200348 Pyrometallurgical process 0.00 tCO2/t Sjardin, 200348 Table 1. Emission factors for 11 industrial products. 5 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata www.nature.com/scientificdata/ The chemical industry showed slower growth, with total emissions increasing from 93 Mt in 2000 to 185 Mt in 2021. The average annual growth rate was 3.3%, with an annual increase of 4 Mt. As illustrated in Fig.2b, emission growth was concentrated in northern cities, including Ordos in Inner Mongolia (0.32 Mt/yr), Dalian in Liaoning (0.31 Mt/yr), and Yulin in Shaanxi (0.3 Mt/yr). Meanwhile, emissions declined in Shijiazhuang in Hebei (−0.26 Mt/yr), Liuzhou in Guangxi (−0.13 Mt/yr), and Hengshui in Hebei (−0.12 Mt/yr). Figure2e shows the 2021 emission composition: ammonia contributed 65%, ethylene 25%, while calcium carbide and soda ash accounted for smaller shares. The metal industry experienced the largest relative growth, with emissions rising from 32 Mt in 2000 to 293 Mt in 2021. The average annual growth rate reached 11.1%, with an annual increase of 12 Mt. As shown in Fig.2c, between 2014 and 2021, emission growth was concentrated in resource-intensive cities. The fastest-growing cities were Tangshan in Hebei (1.39 Mt/yr), Binzhou in Shandong (0.86 Mt/yr), and Baotou in Inner Mongolia (0.48 Mt/yr). In contrast, emissions declined in Shanghai (−0.16 Mt/yr), Beijing (−0.1 Mt/yr), and Tianjin (−0.09 Mt/yr). Figure2f presents the 2021 emission composition for the metal industry: crude steel accounted for 82%, followed by aluminum (15%), while lead, zinc, and other metals contributed relatively minor shares. Uncertainty analysis. Uncertainty analysis is a key tool for evaluating the quality of CO2 emission inventories37,38. According to the IPCC, uncertainties primarily originate from activity data and emission factors5. Two commonly used methods for uncertainty assessment in previous studies are error propagation and Monte Carlo simulation. Although error propagation is relatively easy to implement, it relies on the assumptions of linearity and normally distributed input variables. These assumptions make it unsuitable when input parameters have large uncertainties or skewed distributions27. In contrast, Monte Carlo simulation does not depend on such assumptions and can flexibly accommodate probability density functions of any shape and range, providing a more accurate representation of input uncertainties and their propagation in emission estimates39. Therefore, this study adopts the Monte Carlo method to assess uncertainties in the estimation of process-related CO2 emissions. In our analysis, we assigned a coefficient of variation (CV) of 5% to activity data obtained from official statistical yearbooks and bulltins27,40. For data imputed using missForest models, the CV was set at 15%5,41. Emission factor CVs follow the values recommended by the IPCC (see Table2)5. Emissions from ferroalloy production were excluded from the uncertainty analysis due to the complexity of its carbon sources (e.g., petroleum coke, graphite, biomass), which makes the quantification of its emission factor uncertainty particularly challenging. Probability distributions were derived based on a comprehensive literature review42,43. When the uncertainty range is less than ±60%, it is typically assumed to follow a normal distribution44,45. Accordingly, we assumed normal distributions for both activity data and emission factors. A total of 20,000 random samples Fig. 2 Spatial distribution and structure of process-related CO2 emissions. (a–c) Average annual growth of process-related CO2 emissions in the mineral, chemical, and metal industries from 2014 to 2021. Warmer colors (yellow to red) indicate emission increases, while cooler colors (light blue to blue) represent emission reductions. d-f Process-related CO2 emissions composition of the mineral, chemical, and metal industries in 2021. 6 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata www.nature.com/scientificdata/ were generated to estimate the uncertainty range of process-related CO2 emissions at the 95% confidence level. The results show that the total inventory uncertainty ranges from −3.87% to +3.91%. At the city level, Zhoushan (Zhejiang province) exhibited the highest uncertainty (−18.74% to +18.78%), while Ürümqi (Xinjiang Uygur Autonomous Region) had the lowest (−1.90% to +1.93%). To evaluate the potential impact of excluding ferroalloy emissions, we conducted a sensitivity analysis assuming a conservative CV of 50%, which showed that their inclusion would alter the total uncertainty range by less than 0.1 percentage point. Comparison with existing emission estimates. In addition to uncertainty analysis, we further evaluated the reliability of our activity data and emission estimates by comparing them with national statistics and results from previous studies. First, we compared the collected output data for 11 industrial products with national-level statistics (see Fig.3). Since this study covers 289 cities and not the entire country, the activity data for all 11 products are generally lower than national statistics. This discrepancy is primarily due to the exclusion of cities not included in our study. Among all products, the outputs of calcium carbide, aluminum, lead, and zinc in our dataset are more than 20% lower than the corresponding national totals. These products are likely produced in smaller or less-developed cities that are not part of our study. Additionally, our data are primarily sourced from statistical yearbooks and bulletins, which only include industrial enterprises above the designated size. As a result, outputs from small enterprises are often excluded, leading to further underreporting. In contrast, the reported outputs of cement, plate glass, ethylene, ammonia, soda ash, crude steel, and ferroalloy in this study are only slightly lower than the national totals. This consistent underestimation suggests that the dataset is reliable—it does not overreport industrial output, which is crucial for ensuring the accuracy of CO2 emission estimates. To further assess the accuracy of our emission estimates, we compared them with those reported by Yu et al.27 and Hu et al.18, two of the most comprehensive national-scale studies on China’s process-related CO2 emissions. Industrial product CV (%) Cement 5.5 Glass 30 Calcium carbide 10 Ethylene 30 Ammonia 6 Soda ash 0 Limestone use 6 Crude Steel 5 Aluminum 10 Lead 50 Zinc 50 Table 2. CVs for emission factors. Fig. 3 Relative differences in activity data between this study and national statistics. The relative difference in activity data is calculated as (this study’s data/national data) − 1. A negative value indicates that the activity data collected in this study are lower than the national statistics. 7 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata www.nature.com/scientificdata/ Since no existing dataset provides city-level coverage of multiple industrial products, these national-level estimates serve as the most appropriate reference point for validating our results. As shown in Fig.4, our relatively lower emission estimates reflect a narrower spatial coverage by design, rather than an underestimation. Unlike national-level studies, our study improves spatial resolution by providing city-level CO2 emission estimates. Furthermore, the use of localized emission factors and more complete output data further improves the accuracy and completeness of the estimates. In the mineral industry (Fig.4a–b), our estimates for cement and plate glass are consistently lower than those of Yu et al. and Hu et al. during 2000–2020. For cement, this is mainly due to methodological variation—our study calculates emissions based on cement output, whereas both Yu et al. and Hu et al. use clinker-based estimates. On average, our cement emissions are 16% lower than those of Yu et al. and 22% lower than those of Hu et al. For plate glass, both our study and Yu et al. both apply a China-specific emission factor (0.07 t CO2/t), while Hu et al. Fig. 4 Process-related CO2 emissions from 11 industrial products compared with other studies. Please note that the ranges of the y-axis are different in each subplot. 8 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata www.nature.com/scientificdata/ adopts the default value (0.1 t CO2/t), resulting in estimates that are 39% higher than ours. The 13% lower estimate compared to Yu et al. is mainly attributable to cities not included in our sample. In the chemical industry (Fig.4c–f), our emission estimates for calcium carbide, ethylene, ammonia, and soda ash are consistently lower than those of Yu et al. and Hu et al. All three studies use the same emission factor for calcium carbide (1.15 t CO2/t), so the 33% lower estimate in our study is due to unstudied cities. Similarly, we use the same emission factor for ethylene (2.25 t CO2/t) as Hu et al., yet our results are approximately 10% lower. For ammonia, both our study and Yu et al. use a China-specific emission factor (2.97 t CO2/t), while Hu et al. adopts the lower default emission factor (2.77 t CO2/t). As a result, our estimates are 14% lower than Yu et al. and 8% lower than Hu et al. In the case of soda ash, although all three studies adopt the same emission factor (0.67 t CO2/t), Hu et al. does not account for the fact that only 5% of China’s soda ash is produced via the natural soda process, which is the only one that generates CO2 emissions. This omission leads to substantial overestimation— Hu et al.’s estimate is 96% higher than ours, while ours are 20% lower than Yu et al.’s. In the metal industry (Fig.4g–k), our emission estimates for ferroalloy, aluminum, lead, and zinc are consistently lower than those of Yu et al. and Hu et al., whereas our estimates for crude steel are slightly higher in certain years. The higher estimates for crude steel are due to data limitations: we lack city-level data on pig iron consumption and the technological breakdown of crude steel production via BF–BOF and EAF routes. Therefore, we use total crude steel output as a proxy and apply national-level process shares to estimate emissions. For ferroalloy, aluminum, and zinc (Fig.4h,i,k), we use the same emission factors as Yu et al. (0.28, 1.5, and 3.12 t CO2/t respectively). However, our results are 21%, 35%, and 29% lower, respectively, due to differences in spatial coverage. By contrast, Hu et al. adopts default emission factors (1.3, 1.6, and 1.72 t CO2/t), which do not align well with China’s actual industrial practices. Consequently, our estimates are 83%, 29%, and 94% lower than those of Hu et al., respectively. For lead (Fig.4j), we follow the same methodology as Yu et al., applying a weighted average based on different smelting processes. But due to the lack of city-level process shares, we rely on national-level proportions. Hu et al., on the other hand, uses a single default emission factor without distinguishing between smelting processes, resulting in 63% higher emissions compared to our study. Code availability All code used to construct the industrial product output dataset is published under https://github.com/sijcai/ Industrial-Product-Output. The code is written in R 4.3.0. Received: 21 May 2025; Accepted: 6 August 2025; Published: xx xx xxxx References 1. IPCC. 2014 IPCC Fifth Assessment Report. https://www.ipcc.ch/assessment-report/ar5/2014/ (2014). 2. UNFCCC. Climate ambition alliance: nations renew their push to upscale action by 2020 and achieve net zero CO2 emissions by 2050. https://unfccc.int/news/climate-ambition-alliance-nations-renew-their-push-to-upscale-action-by-2020-and-achieve-net-zero (2019). 3. Huo, D. et al. Near-real-time daily estimates of fossil fuel CO2 emissions from major high-emission cities in China. Sci Data 9 (2022). 4. Xu, R. C. et al. MEIC-global-CO2: A new global CO2 emission inventory with highly-resolved source category and sub-country information. Sci China Earth Sci 67, 450–465 (2024). 5. IPCC. 2006 IPCC Guidelines for National Greenhouse Gas Inventories. https://www.ipcc-nggip.iges.or.jp/public/2006gl/ (2006). 6. Shan, Y. L. et al. City-level climate change mitigation in China. Sci Adv 4, eaaq0390 (2018). 7. Chen, Q. L. et al. CO2 emission data for Chinese cities. Resour Conserv Recy 126, 198–208 (2017). 8. Yang, Y. T. et al. Mapping global carbon footprint in China. Nat Commun 11, 2237 (2020). 9. Janssens-Maenhout, G. et al. EDGAR v4.3.2 Global Atlas of the three major greenhouse gas emissions for the period 1970–2012. Earth Syst Sci Data 11, 959–1002 (2019). 10. National Bureau of Statistics. China Statistical Yearbook 2021. (China Statistics Press, 2021). 11. Chen, W. Y., Yin, X. & Ma, D. A bottom-up analysis of China’s iron and steel industrial energy consumption and CO2 emissions. Appl Energ 136, 1174–1183 (2014). 12. Liu, Y. K., Wang, Y. M. & Chen, Q. S. Using cemented paste backfill to tackle the phosphogypsum stockpile in China: A down-toearth technology with new vitalities in pollutant retention and CO2 abatement. Int J Min Met Mater 31, 1480–1499 (2024). 13. MEE. The First Biennial Transparency Report on Climate Change of the People’s Republic of China. https://www.mee.gov.cn/ywgz/ ydqhbh/qhbhlf/202501/t20250110_1100393.shtml (2024). 14. GCP. Supplemental data of Global Carbon Budget 2022 (Version 1.0). https://www.icos-cp.eu/science-and-impact/global-carbonbudget/2022 (2022). 15. Wang, H. K. et al. China’s CO2 peak before 2030 implied from characteristics and growth of cities. Nat Sustain 2, 748–754 (2019). 16. Jiang, Y. X. & Huang, Z. Impact of urban vitality on carbon emission-an analysis of 222 Chinese cities based on the spatial Durbin model. Hum Soc Sci Commun 11, 1–15 (2024). 17. Cui, D., Deng, Z. & Liu, Z. China’s non-fossil fuel CO2 emissions from industrial processes. Appl Energ 254, 113537 (2019). 18. Hu, L. T. et al. Carbon dioxide emissions from industrial processes and product use are a non-ignorable factor in China’ s mitigation. Communications Earth & Environment 5, 800 (2024). 19. Liu, Z. et al. Reduced carbon emission estimates from fossil fuel combustion and cement production in China. Nature 524, 335–338 (2015). 20. Hu, P. P., Li, Y. Z., Zhang, X. Z., Guo, Z. & Zhang, P. D. CO2 emission from container glass in China, and emission reduction strategy analysis. Carbon Manag 9, 303–310 (2018). 21. Bednar-Friedl, B., Schinko, T. & Steininger, K. W. The relevance of process emissions for carbon leakage: A comparison of unilateral climate policy options with and without border carbon adjustment. Energ Econ 34, 168–180 (2012). 22. Tan, C. et al. Different technology packages for aluminium smelters worldwide to deliver the 1.5 °C target. Nat Clim Change 15, 51–58 (2025). 23. Xiao, H. J., Zhao, W. C., Shan, Y. L. & Guan, D. B. CO2 emission accounts of Russia’s constituent entities 2005–2019. Sci Data 8 (2021). 24. Liu, Z. National carbon emissions from the industry process: Production of glass, soda ash, ammonia, calcium carbide and alumina. Appl Energ 166, 239–244 (2016). 25. Guo, Z. H. et al. Heterogeneous variations on historical and future trends of CO2 and multiple air pollutants from the cement production process in China: emission inventory, spatial-temporal characteristics, and scenario projections. Environ Sci Technol 56, 14306–14314 (2022). 9 Scientific Data | (2025) 12:1435 | https://doi.org/10.1038/s41597-025-05782-3 www.nature.com/scientificdata www.nature.com/scientificdata/ 26. IPCC. 2019 Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories. https://www.ipcc-nggip.iges.or.jp/ public/2019rf/index.html (2019). 27. Yu, X. & Tan, C. China’s process-related greenhouse gas emission dataset 1990-2020. Sci Data 10, 55 (2023). 28. Ren, L., Zhou, S., Peng, T. D. & Ou, X. M. A review of CO2 emissions reduction technologies and low-carbon development in the iron and steel industry focusing on China. Renew Sust Energ Rev 143 (2021). 29. ILZSG. Brochure World Primary and Secondary Lead Plants 2022. https://www.ilzsg.org/static/home.aspx?from=1 (2022). 30. Municipal Bureaus of Statistics of China. City-level Statistical Yearbooks. (Local Statistical Bureaus, 2000–2021). 31. Troyanskaya, O. et al. Missing value estimation methods for DNA microarrays. Bioinformatics 17, 520–525 (2001). 32. Gatti, L. V. et al. Amazonia as a carbon source linked to deforestation and climate change. Nature 595, 388–393 (2021). 33. Tang, F. & Ishwaran, H. Random forest missing data algorithms. Stat Anal Data Min 10, 363–377 (2017). 34. Cai, S. et al. City-level process-related CO2 emissions in China 2000–2021. Figshare https://doi.org/10.6084/m9.figshare.29085002.v3 (2025). 35. Shan, Y. L. et al. Peak cement-related CO2 emissions and the changes in drivers in China. J Ind Ecol 23, 959–971 (2019). 36. Liao, S. M., Wang, D., Xia, C. Y. & Tang, J. China’s provincial process CO2 emissions from cement production during 1993–2019. Sci Data 9, 165 (2022). 37. Romano, D., Bernetti, A. & De Lauretis, R. Different methodologies to quantify uncertainties of air emissions. Environ Int 30, 1099–1107 (2004). 38. Han, P. F. et al. Evaluating China’s fossil-fuel CO2 emissions from a comprehensive dataset of nine inventories. Atmos Chem Phys 20, 11371–11385 (2020). 39. Fang, K., Li, C. L., Tang, Y. Q., He, J. J. & Song, J. N. China’s pathways to peak carbon emissions: New insights from various industrial sectors. Appl Energ 306 (2022). 40. Liu, J. G. & Yang, H. China Fights Against Statistical Corruption. Science 325, 675–676 (2009). 41. Zhang, X. M., Zhuang, G. Y. & J., L. Uncertainty analysis of urban greenhouse gas inventories. Journal of Environmental Economics 3, 8–18 (2018). 42. Winiwarter, W. & Muik, B. Statistical dependence in input data of national greenhouse gas inventories: effects on the overall inventory uncertainty. Climatic Change 103, 19–36 (2010). 43. Ramírez, A., de Keizer, C., Van der Sluijs, J. P., Olivier, J. & Brandes, L. Monte Carlo analysis of uncertainties in the Netherlands greenhouse gas emission inventory for 1990–2004. Atmos Environ 42, 8263–8272 (2008). 44. Shan, Y. L. et al. Data Descriptor: China CO2 emission accounts 1997–2015. Sci Data 5, 1–14 (2018). 45. Li, X. B., Liu, Y., Zhang, J., Zhou, M. F. & Meng, B. China’s product-level CO2 emissions dataset aligned with national input-output tables from 1997 to 2020. Sci Data 12, 30 (2025). 46. NDRC. Guidelines for Provincial Greenhouse Gas Inventories. http://www.cbcsd.org.cn/sjk/nengyuan/standard/home/20140113/ download/shengjiwenshiqiti.pdf (2011). 47. NDRC. First batch of 10 industry-specifc corporate greenhouse gas emission accounting methods and reporting guidelines (for trial implementation). http://www.ncsc.org.cn/SY/tpfqjy/202003/t20200319_769744.shtml (2020). 48. Sjardin, M. CO2 emission factors for non-energy use in the non-ferrous metal, ferroalloys and inorganics industry. Copernicus Institute Utrecht 63 (2003). Acknowledgements This work was supported by the National Natural Science Foundation of China (72222012, 72174084, 72234003 and 72488101), the Natural Science Foundation of Jiangsu Province (BK20220125), Horizon Europe Project EUCHINA-BRIDGE (101137971), which is funded by UKRI grant (10132630) at the University of Birmingham, and National Key Research and Development Program of China (2022YFE0208500). Author contributions S.C., M.L. and Y.S. designed the research. S.C. led the data collection, model development, and result analysis. J.X. and Y.G. assisted with the literature review. C.T. provided methodological and technical support. S.C. drafted the initial manuscript, and M.L., Y.S. and J.B. revised the paper. All authors reviewed and approved the final version of the manuscript. Competing interests The authors declare no competing interests. additional information Correspondence and requests for materials should be addressed to M.L. 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 4.0 International License, which permits use, sharing, adaptation, 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 changes were made. 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/4.0/. © The Author(s) 2025