IntroductionCarbon emissions, which cause global warming, are one of the most important challenges facing society today. Cities are responsible for almost 70% of global carbon emissions associated with energy consumption1,2,3. Urban agglomerations, consisting of multiple large and megacities in close proximity, represent the most advanced type of urban spatial organization4,5,6. This high concentration of population and economic activity results in elevated carbon emissions7. Urban agglomerations can improve the efficiency of resource concentration and utilization, and are considered an important support for collaborative carbon reduction8,9. China’s urban agglomerations account for only 30% of the country’s total land area, but account for more than 75% of the population and 80% of GDP10. However, China exhibits notable regional disparities, with urban agglomerations reflecting varying levels of local development11. These agglomeration areas are not only spatial clusters, but also functionally integrated economic systems, in which trade flows greatly affect production and consumption dynamics. Neglecting agglomeration interval trade may lead to incorrect emission allocation and undermine the effectiveness of emission reduction policies targeted at specific regions12,13. Thus, carbon footprint accounting (i.e., consumption-based accounting), which tracks greenhouse gas emissions embodied in supply chains and attributes mitigation responsibilities to final consumers14—has emerged as a critical tool for understanding the indirect drivers of emissions and guiding regional low-carbon transitions. Understanding the current economic development, carbon footprint trends, and their correlations in these urban agglomerations is essential15. This knowledge helps delineate specific emission patterns of urban agglomerations, crucial for devising more effective mitigation strategies.Since the IPCC Fifth Assessment Report first systematically focused on the mitigation potential of cities16, there have been many studies about carbon emissions at city level17,18,19. Cities are becoming increasingly interconnected, and no longer exist in isolation20. Urban agglomerations, which bring together core cities, dense with industry and population, provide a critical but underexplored perspective21. For example, Li et al.22 and Wang, et al.6 highlight the potential for urban integration and spatial agglomeration to enhance productivity and environmental conservation, potentially reducing emissions. However, these studies do not fully consider the complexity of carbon emissions dynamics in urban networks23. The emerging phenomenon of urban agglomeration strengthens supply chains across cities and sectors, thereby unveiling opportunities to elevate resource efficiency and facilitate large-scale decarbonization24,25,26. Concurrently, this interconnectedness increases the potential to shift emission reduction obligations down the supply chain, thereby complicating accountability27. Previous research has explored the spatial relationships and spillover effects of carbon emissions reduction in urban agglomerations13,28, using modes like social network analysis29, geographically weighted regression30,31, and the spatial Durbin model32,33. A few studies have addressed the impact of such supply chain spillovers on carbon emissions in individual urban agglomerations34, such as Beijing-Tianjin-Hebei35, Yangtze River Delt36 and Pearl River Delta37, but lack of systematic analysis that examines internal and external carbon emissions related to evolution of more emerging urban agglomerations. This oversight highlights the need for a more holistic examination of the role of urban agglomerations in carbon emissions dynamics38, considering the multiple impacts of technology, economy, population, and consumption linkages on emission patterns and mitigation pathways.To bridge the gap, we assessed the evolution of carbon footprints of China’s urban agglomerations, traced the carbon flows hidden in the intercity supply chain of urban agglomerations, and identified key factors affecting carbon footprint changes. Our focus was on China’s 16 national-level urban agglomerations, which consist of 199 cities (Supplementary Table 1), accounting for approximately 72.5% of the national population and 88% of GDP. We focused on the pattern of carbon footprints in these urban agglomerations during the “new normal” phase from 2012 to 2017, when China officially bids farewell to the phase of GDP growth above 10% and changes in economic structure become more apparent39. Additionally, China’s overall carbon emissions have plateaued during this time40. As regional disparities and environmental pressures are common challenges faced by urban areas, the results of our study can help to identify opportunities and challenges for reducing emissions in urban areas that are conducive to tackling climate change and its impacts.ResultsStructural change in carbon footprint of the urban agglomerations from 2012 to 2017Carbon footprint of the 16 urban agglomerations was 5981.6 Mt, 5823.1 Mt and 6038.9 Mt in 2012, 2015 and 2017, respectively, accounting for 80.1%, 79.4% and 78.7% of the total emissions in China. The top three carbon footprint urban agglomerations in 2012 were Yangtze River Delta, Middle reaches of the Yangtze River, and Beijing-Tianjin-Hebei, which accounted for 40.2% of the total carbon footprint of all 16 urban agglomerations (Fig. 1). These three urban agglomerations were also the top three in terms of GDP in 2012 (Supplementary Fig. 1). However, in 2017, the Central Plains replaced Beijing-Tianjin-Hebei as the third-highest carbon footprint urban agglomeration due to the increasing proportion of heavy industry, although its GDP only ranked fifth. Within an urban agglomeration, the carbon footprint is not evenly distributed and is often concentrated in the core cities, whose carbon footprint exceeded one standard deviation from the mean of the urban agglomeration (as marked in Fig. 1). In urban agglomerations with relatively weak economies, the carbon footprint is more concentrated in the core cities. For example, five urban agglomerations have only one core city, and the economy of the Shandong Peninsula is more developed than the other four regions. Qingdao accounted for only 27.2% of the carbon footprint in the Shandong Peninsula, which is lower than the proportion of the carbon footprint of core cities in the other four urban agglomerations, especially Xi’an, Yinchuan, and Urumqi, which accounted for more than 50% of the carbon footprint in their respective urban agglomerations. The weaker the economy, the more dependent it is on a single pole in the urban agglomeration. In an urban agglomeration, the larger the proportion of the carbon footprint accounted for by the core cities, the more it indicates a high concentration of industrial and economic activities in those cities. In contrast, when the carbon footprint is more evenly distributed among multiple cities, it reflects a more diversified economic structure with less reliance on a few key cities. For example, the Yangtze River Delta has four carbon core cities, and Shanghai, with the highest carbon footprint, only accounted for 13.6% of the carbon footprint because cities in the Yangtze River Delta are developed evenly (Fig. 1S).Fig. 1: Seven sectors of carbon footprint in 16 urban agglomerations and non-urban agglomerations in China from 2012 to 2017.The urban agglomerations with carbon footprint peak are filled in green, the urban agglomerations with carbon footprint rising are filled in yellow, and the urban agglomerations with carbon footprint rebounded are filled in red. Grey area indicates other cities, and white area indicates no data.Full size imageOf the 16 urban agglomerations, five demonstrated a reduction in their carbon footprint over the period from 2012 to 2017, six exhibited an increase, and five reached a plateau (Fig. 1). The five urban agglomerations with carbon footprint decreased are Central Shanxi, Harbin-Changchun, Liaozhongnan, Beijing-Tianjin-Hebei, and Shandong Peninsula. These urban agglomerations are situated in North China and Northeast China, where industrialisation commenced at an early stage. The transformation and upgrading of traditional industries leaded their carbon mitigation. For example, the heavy manufacturing and construction sectors contributing 74.1% and 38.6%, respectively, to their carbon reduction from 2012 to 2017. Notably, the carbon footprint of heavy manufacturing in Shandong Peninsula decreased the most by 136.9 Mt, contributing 80.3% to the total carbon reduction, due to the removal of outdated capacity and reduction of excess capacity in the region. The share of heavy manufacturing in the region’s total carbon footprint declined from 44.1% in 2012 to 23.6% in 2017. In Central Shanxi, which is rich in coal resources, the carbon footprint of the mining sector decreased by 3.3% more than in other agglomerations. However, the power and service sectors increased in their carbon footprint by 8.5% and 14.6%, respectively.The carbon footprint of six urban agglomerations increased from 2012 to 2017. These are the Tianshan North Slope, Guanzhong Plain, Qianzhong, Yangtze River Delta, the West Side of the Straits and the Pearl River Delta. It should be noted that the Yangtze River Delta, the West Side of the Straits and the Pearl River Delta are situated on the southeastern coast of China. The economies of these urban agglomerations continue to expand at a rapid rate, with growth of 45.9%, 55.5% and 72.0% from 2012 to 2017. In such urban agglomerations, the service sector is the primary factor responsible for the observed growth in carbon footprint. The carbon footprint of the service sector in the Yangtze River Delta increased by 54.5%, in the Pearl River Delta by 63.8%, and in the West Side of Straits by 85.1%. In contrast, the Tianshan North Slope, Guanzhong Plain and Qianzhong are situated in the western region of China. To provide support for the development of underdeveloped areas in the western regions, these urban agglomerations have given priority to the implementation of development strategies. The economic growth mode of the Tianshan North Slope and Qianzhong is dependent on infrastructure, which consequently results in the construction sector becoming a contributor to the growth of the carbon footprint. The carbon footprint of the construction in the Tianshan North Slope and Qianzhong increased by 39.2% and 57.7%, respectively. However, in the Guanzhong Plain, heavy manufactory has become the dominant contributor to carbon footprint growth, with an increase of 72.1%.Carbon footprint of the remaining five urban agglomerations showed a plateau from 2012 to 2017. Among them, carbon footprint of Ningxia along the Yellow River, Central Plains, Hubaoeyu and the Middle reaches of the Yangtze River declined from 2012 to 2015 but rebounded from 2015 to 2017. These urban agglomerations are usually stimulated by the economy without considering environment protect after 2015, leading to a sudden increase in their emissions. Ningxia along the Yellow River is the urban agglomeration with the largest rebound, with a decrease of 4.2% from 2012 to 2015 but an increase of 70.2% from 2015 to 2017. Because its value is relatively small, and it is more vulnerable to the impact of policy and the economic environment. Power, construction, and service were mainly promoted economic recovery in these regions, especially contributing 86.7% of the rebound in Ningxia along the Yellow River. In addition, Chengyu was the only one urban agglomeration showing a carbon peak during 2012 to 2017, whose carbon footprint increased by 12.4% from 2012 to 2015, but decreased by 7.2% from 2015 to 2017. The decrease in carbon footprint of the construction sector by 43.5 Mt from 2015 to 2017 is the main factor contributing to Chengyu’s carbon peak. Meanwhile, the share of heavy manufacturing in Chengyu’s carbon footprint dropped from 24.2% in 2012 to 18.0% in 2017, causing it to fall from the second-largest to the third-largest emitting sector in the region.The mechanisms of carbon footprint changes in urban agglomerationThe patterns of carbon footprint changes for these three models are driven by different socio-economic factors. We selected five factors, including two production factors (carbon intensity and production structure) and three consumption factors (consumption structure, per capita consumption and population). The carbon footprint reduction achieved by five urban agglomerations before 2015 was mainly due to production side factors, and there were differences after 2015 (Fig. 2), which indicates that the marginal emission reduction benefits on the production side will decrease. This trend is closely related to the Chinese economy entering a “new normal” during the research period. As economic growth slowed, the lock-in effects of traditional high-carbon infrastructure made it difficult to decrease carbon emissions in line with the slowdown, resulting in a rebound in carbon intensity per unit of output39. Of the five urban agglomerations whose carbon footprint decreased, four were mainly driven by consumption factors from 2015 to 2017. Although the production structure played a dominant role in reducing emissions, the increase in carbon intensity offset part of this effect, leading to a net emission increase from the production side in Beijing-Tianjin-Hebei, Harbin-Changchun, Liaozhongnan, and Central Shanxi. Among the four urban agglomerations analyzed, Harbin-Changchun, Liaozhongnan, and Central Shanxi followed an economic decline model. Declines in per capita consumption and population contributed to carbon footprint reductions of 51.2 Mt in Harbin-Changchun, 22.5 Mt in Liaozhongnan, and 18.8 Mt in Central Shanxi, respectively. These reductions, driven primarily by constrained economic activity and weakened consumption demand, reflected the economic hardships faced by these regions. In contrast, the Beijing-Tianjin-Hebei achieved a 47.7 Mt carbon footprint reduction driven by shifts in consumption structure, despite increases of 36.6 Mt from per capita consumption and 3.3 Mt from population growth, reflecting a low-carbon consumption model. The Shandong Peninsula was the only urban agglomeration to achieve carbon footprint reductions through a technological advancement model. Between 2015 and 2017, production structure promoted 148.7 Mt reduction in carbon footprint. This production-driven reduction was largely facilitated by technological advancements and enhanced supply chain management.Fig. 2: The contribution of various driving factors to carbon footprint change in China’s urban agglomerations from 2012 to 2015 and from 2012 to 2017.Urban agglomerations are grouped into three types based on their carbon footprint trends: decline, plateau, and increase. Positive bars indicate emission-increasing effects, while negative bars indicate emission-reducing impacts for each driver.Full size imageFor six urban agglomerations with carbon footprint increase from 2012 to 2017, they have been in a period of rapid economic development, and the growth of consumption led to the increase of these carbon footprints. Among consumption factors, per capita consumption contribution most for these urban agglomerations. For example, per capita consumption promoted of contributed carbon footprint increased by 473.4 Mt in Yangtze River Delta from 2012 to 2017. We found that the consumption structure contributed to the reduction of the carbon footprint in West side of the Straits and Yangtze River Delta from 2015 to 2017, indicating that these two have the potential to form a low-carbon consumption model. However, the consumption structure of the Pearl River Delta and Tianshan North Slope reduced the carbon footprint from 2012 to 2015, but increased the carbon footprint in 2015-2017, indicating that the consumption structure has greater uncertainty, easily to change according to the current policy. Consumption factors were the dominated drivers of carbon footprint increases, despite production factors slightly promoting a decline in footprint for all urban agglomerations, which shows that most urban agglomerations have made progress in production technology and supply government. However, both production and consumption factors contributed to the carbon footprint of the Pearl River Delta, because high rapid urbanization and economic development41.The plateau phase observed in five urban agglomerations can be attributed to the competing dynamics between production-side technological advancements, which drive emission reductions, and consumption-driven demand growth, which exerts upward pressure on emissions. The trajectory of carbon footprint changes in urban agglomerations during this stage remains uncertain and is expected to be shaped by external factors, like policy interventions. The sharp rise in carbon footprints in the Central Plains, Ningxia along the Yellow River, and the middle reaches of the Yangtze River by 133.8 Mt, 50.7 Mt, and 172.9 Mt, respectively, was predominantly driven by increased per capita consumption from 2015 to 2017. The increase in investment contributed to 62.7%, 45.6%, and 38.9% of the rise in per capita consumption in these urban agglomerations, respectively (Supplementary Fig. 2). These urban agglomerations should address the risk of investment-driven emissions by prioritizing investments in the photovoltaic new energy industry and adopting low-carbon consumption models to mitigate emissions. In Hubaoeyu, carbon intensity was the primary driver of the carbon footprint increase from 2015 to 2017. As a key energy supply base in China, the region experienced a surge in coal supply in 2017, and the high carbon intensity of coal contributed to the sharp rise in emissions during this period42. Chengyu is the only urban agglomeration whose carbon footprint turned down after 2015, because production structure improvement reduced emissions by 125.5 Mt. This trajectory exemplifies how structural transformation, coupled with the deployment of cleaner technologies, can contribute to decarbonization—for instance, by replacing high-emission sectors (e.g., coal-fired power) with low-emission alternatives (e.g., wind or solar energy).The trade-induced emissions between urban agglomerationsThe carbon footprint dynamics of urban agglomerations reflect the interplay between local actions and intercity trade between urban agglomerations, which together meet demand through supply chains. Figure 3 shows the changes in the sources of carbon footprint composition in urban agglomerations. From 2012 to 2017, local carbon emissions decreased by 351.9 Mt, but emissions from other urban agglomerations surged by 801.7 Mt, driven by intensified trade relationships, as trade prioritizes cheaper goods over low-carbon alternatives. For urban agglomerations with declined carbon footprint, a reduction of 293.8 Mt in local emissions is the largest contributor to emission reductions, followed by a reduction of 287.4 Mt in emissions from other cities. Urban agglomerations in the economic decline model, such as Harbin-Changchun, reduced emissions primarily through reduced local activity, accounting for 89.3% of the carbon footprint reduction, with trade patterns playing a secondary role. In contrast, urban agglomerations in the low-carbon consumption model (i.e., Beijing-Tianjin-Hebei), the contribution of local carbon emissions reduction accounted for only 54.0% of the carbon footprint reduction, relying more on trade to reduce emissions by switching to low-carbon imports, especially from 2015 to 2017, carbon inflow decreased by 24%. For agglomerations following the technological advancement model, local production improvements drove emissions reductions, but trade dynamics often moderated these gains. For instance, the Chengyu achieved local reductions (−53.0 Mt from 2012 to 2017) but faced a net increase (+17.2 Mt) due to trade. These dynamics underscore the necessity of integrating trade considerations into carbon management strategies, as supply chain linkages influence emissions outcomes.Fig. 3: Contributions to carbon footprint change by different sources from 2012 to 2017.Carbon footprint changes are broken down into three sources: local emissions, emissions outsourced from other urban agglomerations, and emissions outsourced from external regions. Negative values indicate that the urban agglomeration contributed to a decrease in emissions from the corresponding source, while positive values indicate that it led to an increase in emissions from that source.Full size imageTo evaluate the impact of trade on carbon emissions, we constructed a no-trade counterfactual scenario in which local demand is fully met by local production. Based on Eq. (18), we calculated trade-induced emission changes shown in Fig. 4. From 2012 to 2017, total trade-induced carbon emissions reduction across the 16 urban agglomerations declined from 1562.7 Mt to 1374.5 Mt, indicating a weakening mitigation effect. In 2017, 11 out of 16 urban agglomerations still achieved net emission reductions through trade. Liaozhongnan, Hubaoeyu, and Harbin-Changchun achieved the largest trade-induced emission reductions, with decreases of 479.8 Mt, 309.0 Mt, and 101.2 Mt in 2017, respectively. In contrast, Beijing-Tianjin-Hebei exhibited the largest trade-induced emission increase, reaching 98.5 Mt in 2017. Several urban agglomerations, such as Shandong Peninsula, Central Shanxi, Ningxia along the Yellow River, and Tianshan North Slope, shifted from net mitigation to net increases, reflecting diverging regional trajectories in trade-related carbon flows. These shifts highlight how evolving regional trade networks and industrial specialization patterns are reshaping the spatial distribution of carbon emissions. While urban agglomerations remained key actors in trade-enabled decarbonization, non-agglomerated regions increasingly contributed, with trade-induced emission reductions rising from 472.1 Mt in 2012 to 1470.3 Mt in 2017. This shift indicates that non-agglomerated cities have become more active contributors to trade-based emission mitigation, likely due to structural transformation and changing patterns of interregional trade.Fig. 4: Trade-induced emissions in intra and inter-urban agglomerations in 2012, 2015 and 2017.Negative values indicate that trade contributed to emission reductions, while positive values indicate that trade led to increased emissions.Full size imageTo further identify the drivers of trade-induced emission changes in urban agglomerations, we distinguish between intra-agglomeration and inter-agglomeration trade effects. Even when intra-agglomeration trade is balanced, uneven emission intensities and production effectiveness cause net changes in total emissions. Intra-agglomeration trade contributed to mitigation overall, with associated reductions declining from 548.7 Mt in 2012 to 233.3 Mt in 2017. In 2017, 10 out of 16 urban agglomerations achieved emission reductions through intra-agglomeration trade, reflecting improved internal efficiency. The Middle Reaches of the Yangtze River recorded the highest intra-agglomeration trade-induced reduction at 142.1 Mt, likely due to well-coordinated industrial structures and relatively low emission intensities in key production cities, enabling more efficient internal trade. However, unbalanced or inefficient industrial division within agglomerations may result in the opposite effect. In 6 urban agglomerations, intra-agglomeration trade led to increased emissions, suggesting the potential for carbon leakage within agglomerations. For example, Liaozhongnan experienced an intra-agglomeration trade-induced emission increase of 21.4 Mt in 2017, due to an uneven internal division of labour and the concentration of high-emission industries, which may have resulted in intra-agglomeration carbon leakage.The internal emission spatial structure of urban agglomerations can influence industrial specialization and production efficiency, thereby shaping regional emission outcomes43. Building on this perspective, our study further incorporates trade-embedded carbon flows to quantify how these structural differences translate into inter-city emission redistribution. Network analysis provides further insight into how the internal structure of urban agglomerations affects trade-induced emissions. Modularity, which reflects the extent to which a network is divided into relatively independent subgroups, indicates weaker integration when values are higher. In 2017, the average modularity across all urban agglomerations was 0.62. In contrast, strength measures the overall trade intensity within an agglomeration, reflecting the degree of economic connectivity among cities, with an average value of 0.48 in 2017. As shown in Supplementary Fig. 3, urban agglomerations with lower modularity (e.g., Middle Reaches of the Yangtze River: 0.56) and higher strength (0.54) tend to exhibit more cohesive trade networks that facilitate efficient resource sharing and reduce redundant high-emission production. Conversely, agglomerations such as Liaozhongnan exhibited higher modularity (0.65) and lower strength (0.39), indicating fragmented internal structures that may hinder low-carbon coordination and lead to carbon leakage within the agglomeration. These findings suggest that intra-agglomeration trade contributes more effectively to decarbonization when supported by well-integrated and strongly connected internal trade networks.Trade between urban agglomerations serves as a critical mechanism for interregional product exchange and resource coordination. From 2012 to 2017, trade-induced emission reductions resulting from inter-agglomeration trade increased from 1,014.0 Mt to 1,141.2 Mt, underscoring its growing role in optimizing resource allocation and supporting low-carbon development across regions. Several urban agglomerations, including Liaozhongnan (501.2 Mt), achieved substantial emission reductions through trade with other regions (Fig. 5). These areas typically feature high-carbon industrial structures, and inter-agglomeration trade enables them to meet local demand by importing relatively low-carbon products. Notably, 24% of Liaozhongnan’s external inflows came from more developed agglomerations, such as Beijing-Tianjin-Hebei, the Yangtze River Delta, and the Pearl River Delta, reducing the need for high-emission local production and thereby lowering their consumption-based carbon footprints. In contrast, some urban agglomerations experienced increased emissions as a result of inter-agglomeration trade. For example, Beijing-Tianjin-Hebei recorded a trade-induced emission increase of 117.6 Mt in 2017. This region imports a volume of carbon-intensive products to meet growing demand for infrastructure and consumption, resulting in the displacement of carbon emissions to external production regions. At the same time, it also acts as a foundational supply base, exporting large quantities of industrial products to other regions. A defining feature of these dual-role agglomerations is that exports exceed imports, making trade a key driver of local production and associated emissions. For instance, Tangshan, a major steel-producing city within the agglomeration, contributed 108 Mt of emissions through steel exports alone. While Beijing-Tianjin-Hebei has made efforts to reduce emissions through low-carbon consumption models, the continued outsourcing of demand to its industrial base has offset much of these gains.Fig. 5: Carbon flow between inter-urban agglomerations in 2012, 2015, and 2017.The colour of the source of carbon flow is consistent with the colour of the urban agglomeration. The side length of the urban agglomeration represents the size of the carbon flow flowing through it, and the depth of the colour represents the net inflow and net inflow.Full size imageDiscussionThis study analysed the changes in carbon footprints across 16 urban agglomerations in China from 2012 to 2017, revealing spatial and temporal variations in carbon emissions while examining the driving forces behind these changes, including production factors, consumption factors, and inter-city trade. The heterogeneity in carbon trends reflects structural imbalances between production and consumption systems, uneven policy implementation, and varying capacities for low-carbon transition. These findings highlight that decarbonization is not solely a technical or economic process, but a spatial and institutional one, deeply embedded in how cities are functionally connected through supply chains and trade flows.Among urban agglomerations with decreased carbon footprints, different mitigation modes are observed, each reflecting distinct policy and structural pathways. Some regions reduced emissions primarily through production-side reforms, such as the optimization of industrial structure and the decline in carbon intensity driven by technological upgrading and energy transitions (e.g., Shandong Peninsula). These regions would benefit from continued industrial modernization, expansion of renewable energy, and targeted interventions in hard-to-decarbonize sectors such as heavy manufacturing. In other cases, reductions were led by consumption-side measures, such as changes in household behaviour, green building adoption, and cleaner mobility systems (e.g., Beijing-Tianjin-Hebei). These cities should further institutionalize sustainable consumption practices by integrating low-carbon criteria into public procurement standards—such as requiring energy efficiency labels and life-cycle carbon footprint disclosure for government-purchased goods and services. Additionally, they can establish targeted incentives for residents and businesses, such as electricity bill discounts for households adopting rooftop solar or reward points for low-carbon mobility choices like biking and public transit usage. A third group of agglomerations achieved reductions passively due to economic decline or population contraction, which raises concerns about the long-term sustainability of such outcomes. For these regions, policies should focus on managing economic recovery through green investment, diversification of industrial bases, and building resilience in employment and infrastructure by expanding green job training programmes in sectors such as renewable energy and building retrofitting, and by upgrading urban infrastructure, such as drainage systems and power grids.The carbon footprints of urban agglomerations that have entered a plateau phase illustrate the challenges of sustaining emissions control amidst ongoing economic and spatial restructuring. These regions have shown initial success in stabilizing emissions, but the durability of such trends depends on deeper transformations in energy systems, technology adoption, and industrial coordination. Policymakers should treat the plateau not as a sign of sufficiency but as a critical window to accelerate the integration of energy, transport, and industry planning—such as aligning electric vehicle infrastructure development with renewable power expansion—reduce fossil fuel dependence through coal-to-clean energy substitution in power and heat supply, and institutionalize long-term emissions control mechanisms via carbon budgeting and performance-based regulation. In contrast, urban agglomerations experiencing a continued rise in carbon footprint often face a combination of rapid industrial expansion, rising per capita consumption, and high carbon intensity of energy use. In these contexts, more aggressive sectoral interventions are needed—such as energy system electrification, carbon performance standards in infrastructure and construction, and promote green supply chain upgrading. These agglomerations also need to invest in climate-smart infrastructure and adopt region-specific carbon management plans to align local growth with national emission targets.The non-trade scenario demonstrates that inter-city trade between urban agglomerations tends to facilitate emission reductions, suggesting that agglomeration-driven improvements in production efficiency contribute to mitigation—consistent with prior findings44. Beyond confirming this general pattern, our study adds a quantitative, trade-embedded perspective by tracing embodied carbon flows under a counterfactual scenario, offering a novel approach to identify where and how agglomeration-related trade impacts emissions. Inter-city trade among urban agglomerations plays a dual role in carbon governance—it can facilitate emission reductions by reallocating production to lower-carbon regions, but it can also undermine mitigation efforts when trade reinforces carbon-intensive production patterns. Uneven environmental policies may inadvertently shift production capacity from regions with low carbon intensity to those with higher carbon intensity, potentially increasing total emissions. To address this, it is essential to implement green trade policies and promote sustainable supply chains. Mitigation measures must be context-specific, tailored to the unique challenges and opportunities of each urban agglomeration, and grounded in comprehensive long-term planning. Policies that promote regional collaboration can optimizing resource allocation and reducing emissions across the entire region. For example, integrating renewable energy systems into industrial planning, incentivizing green innovation, and fostering collaboration among cities can amplify the impact of local mitigation efforts. Additionally, imposing a carbon tax on high-carbon products can help reduce emissions and incentivize the adoption of cleaner alternatives45. By aligning trade and industrial policies with emission reduction goals, governments can ensure that regional progress supports broader climate objectives, driving meaningful reductions in carbon footprints across urban agglomerations.This study highlights the diverse pathways and challenges faced by China’s urban agglomerations in reducing their carbon footprints. While progress has been made, the findings underscore the need for targeted, region-specific strategies that integrate production, consumption and trade dynamics. Although inter-city trade decisions are typically driven by market forces rather than environmental concerns, understanding the carbon consequences of these flows remains critical. Our results help identify regions that are net exporters or importers of carbon-intensive goods, revealing potential carbon leakage and imbalances in responsibility. These insights can inform broader policy frameworks such as carbon accounting reform, regional coordination strategies, and infrastructure planning. For instance, by assessing whether intra-agglomeration trade contributes to emission reductions, policymakers can evaluate if regional industrial specialization enhances overall carbon efficiency. While governments do not directly control inter-city trade, they shape it through fiscal incentives, carbon pricing, and industrial planning. In China, existing inter-regional compensation schemes—such as those in the Yellow River Basin—already provide institutional platforms for balancing upstream–downstream environmental interests46. Although embodied carbon accounting is not yet widely institutionalized, these frameworks offer potential entry points for incorporating emission spillovers into regional governance. Such metrics could support differentiated emission targets, guide fairer burden-sharing among cities, and provide a data foundation for compensation mechanisms or carbon budgeting. As China advances its carbon neutrality agenda, integrating embodied emissions into national market rules and regional coordination plans can enhance policy coherence and align economic cooperation with environmental accountability. Looking ahead, future research should assess not only the long-term impacts of mitigation policies and emerging low-carbon technologies, but also the robustness of inter-city carbon flow patterns under varying IO assumptions. Given the known structural uncertainty in input-output tables47, it is important to interpret marginal variations in results—such as those in Figs. 1 and 3—with caution. Many elements in IO matrices contribute little to multiplier effects, suggesting that small differences across cities or years may reflect data sparsity or model assumptions rather than meaningful trends. Future work should conduct sensitivity analyses to validate the robustness of these patterns. Despite these uncertainties, our findings offer practical entry points for policy: the identification of net carbon exporters/importers can inform inter-regional compensation, while insights on emission outsourcing can support regionally differentiated carbon pricing and quota allocation. Fostering collaborative governance and aligning carbon goals with development priorities remain essential. By grounding theoretical analysis in evolving policy mechanisms and existing institutional precedents, this study helps translate complex inter-city carbon dynamics into actionable strategies for advancing carbon neutrality.MethodsEnvironmentally extended multi-regional input-output analysis methodFootprint is accounted not only for the activity of a region area, but also for activity that occurs outside the area which can be attributed to activities within the region48. The input-output model, developed by Leontief49, is one of the most widely used methods of analysing footprint, which reveals the trade flow between different sectors through the supply chain. Multi-regional input-output model (MRIO) is a well-acknowledged tool employed to account for carbon footprint, which illustrates the inter-region and inter-industry relationships along supply chains and interregional trades50. Here, we used the MRIO of 2012, 2015 and 2017 to estimate the carbon footprint of China’s urban agglomerations.With the MRIO table, the equation can be expressed as:$${{\bf{X}}}={\left({{\bf{I}}}-{{\bf{A}}}\right)}^{-{{\bf{1}}}}{{\bf{F}}}$$(1)$${{\bf{X}}}=\left[\begin{array}{c}{x}^{1}\\ \begin{array}{c}{x}^{2}\\ \vdots \end{array}\\ {x}^{n}\end{array}\right],{{\bf{A}}}=\left[\begin{array}{ccc}\begin{array}{c}\begin{array}{cc}{a}^{11} & {a}^{21}\end{array}\\ \begin{array}{cc}{a}^{21} & {a}^{21}\end{array}\end{array} & \begin{array}{c}\cdots \\ \cdots \end{array} & \begin{array}{c}{a}^{n1}\\ {a}^{n2}\end{array}\\ \begin{array}{cc}\!\! \vdots & \,\,\,\,\,\,\,\,\vdots \end{array} & \ddots & \vdots \\ \begin{array}{cc}{a}^{n1} & {a}^{21}\end{array} & \cdots & {a}^{{nn}}\end{array}\right],F=\left[\begin{array}{ccc}\begin{array}{c}\begin{array}{cc}{f}^{11} & {f}^{21}\end{array}\\ \begin{array}{cc}{f}^{21} & {f}^{21}\end{array}\end{array} & \begin{array}{c}\cdots \\ \cdots \end{array} & \begin{array}{c}{f}^{n1}\\ {f}^{n2}\end{array}\\ \begin{array}{cc}\!\! \vdots & \,\,\,\,\,\,\,\,\vdots \end{array} & \ddots & \vdots \\ \begin{array}{cc}{f}^{n1} & {f}^{21}\end{array} & \cdots & {f}^{{nn}}\end{array}\right]$$(2)where \({{\rm{X}}}=\left({x}_{i}^{s}\right)\) is the vector of total output and \({x}_{i}^{s}\) is the total output of sector \(i\) in region \(s\), \({{\bf{I}}}\) is the identity matrix, and \({(I-A)}^{-1}\) is the Leontief inverse matrix. \({{\bf{A}}}=[{a}_{{ij}}^{{rs}}]\), and \({a}_{{ij}}^{{rs}}={z}_{{ij}}^{{rs}}/{x}_{j}^{s}\) is the technical coefficient matrix of each sector from region \(r\) to region \(s\); \({z}_{{ij}}^{{rs}}\) is the intersectoral flows from sector \(i\) in region \(r\) to sector \(j\) in region \(s\). \({{\bf{F}}}=\left({f}_{i}^{{rs}}\right)\) is the final demand matrix, and \({f}_{i}^{{rs}}\) is the final demand from sector \(i\) in region \(r\) to region \(s\). The carbon footprint is calculated using an environmental extended multi-regional input-output analysis, a widely adopted approach for consumption-based carbon accounting. This method attributes carbon emissions along regional supply chains to the final consumers who drive production. On the basis of the carbon intensity per output \({{\bf{E}}}\) (that is, CO2 emissions per unit of economic output), the carbon footprint is calculated:$${{\bf{C}}}={{\bf{E}}}{\left({{\bf{I}}}-{{\bf{A}}}\right)}^{-{{\bf{1}}}}{{\bf{F}}}$$(3)where \({{\bf{C}}}\) is a vector of total CO2 emissions in goods and services used for final demand; \({{\bf{E}}}=\left[{e}_{i}^{r}\right]\) is the matrix of carbon intensity per output; \({e}_{i}^{r}=\frac{{{CE}}_{i}^{r}}{{x}_{i}^{r}}\) represents carbon intensity per output of sector in \(i\) in region \(r\); \({{CE}}_{i}^{r}\) is CO2 emission inventory from sector \(i\) in region \(r\). This framework allows us to trace carbon emissions along interregional and intersectoral production chains and attribute them to the regions where final consumption occurs, in line with the carbon footprint accounting principle.CO2 emission inventory constructionThe fossil fuel-related CO2 emission formula is as follows:$${{CE}}_{{ij}}={\sum }_{i}{\sum }_{j}{{AD}}_{{ij}}\times {{NCV}}_{i}\times {{CC}}_{i}\times {O}_{{ij}}$$(4)where \({{CE}}_{{ij}}\) is the CO2 emissions caused by the sector j using the fossil fuel i; \({{AD}}_{{ij}}\) is the consumption of fossil fuel i by sector j; \({{NCV}}_{i}\) is the net calorific value of fossil fuel i; \({{CC}}_{i}\) is carbon content of fossil fuel i; \({O}_{{ij}}\) is oxygenation efficiency of fossil fuel i combusted in sector j.The fossil process-related emissions refer to the emissions escaping from chemical reactions in the industrial processes:$${{CE}}_{t}={{AD}}_{t}\times {{EF}}_{t}$$(5)where \({{CE}}_{t}\) is the carbon dioxide emissions induced in the industrial processes t; \({{{\rm{AD}}}}_{t}\) is the production amount of processes t; \({{EF}}_{t}\) is emission factor of processes t.Structural decomposition analysisStructural decomposition analysis (SDA) is a widely used approach to estimate the drivers of changes in carbon emissions and energy consumption51,52. Carbon footprint in Eq. (3) can be disassembled into carbon emissions per output (\(E\)), production structure (\(L={({{\rm{I}}}-{{\rm{A}}})}^{-1}\)) and final demand (\(F\)). The final demand in Eq. (5), can be further decomposed into consumption structure (\(S\)), consumption per capita (\(Q\)) and population (\(P\)):$$F=\left[\begin{array}{ccc}\begin{array}{c}\begin{array}{cc}{s}^{11} & {s}^{21}\end{array}\\ \begin{array}{cc}{s}^{21} & {s}^{21}\end{array}\end{array} & \begin{array}{c}\cdots \\ \cdots \end{array} & \begin{array}{c}{s}^{n1}\\ {s}^{n2}\end{array}\\ \begin{array}{cc}\!\! \vdots & \,\,\,\,\,\,\,\,\vdots \end{array} & \ddots & \vdots \\ \begin{array}{cc}{s}^{n1} & {s}^{21}\end{array} & \cdots & {s}^{{nn}}\end{array}\right]\times \left[\begin{array}{ccc}{q}_{1} & \begin{array}{cc}{q}_{2} & \cdots \end{array} & {q}_{n}\end{array}\right]\times \left[\begin{array}{ccc}{p}_{1} & \begin{array}{cc}{p}_{2} & \cdots \end{array} & {p}_{n}\end{array}\right]$$(6)where the total consumption of each city is obtained by summing the corresponding column of the final demand matrix. Dividing this total consumption by the city’s population yields the per capita consumption. The consumption structure is calculated by dividing each element in the city’s final demand vector by the total consumption of that city.Thus, the changes in carbon footprint can be decomposed as:$$\Delta C= \ {C}_{1}-{C}_{0}=\left(\Delta E\right){LSQP}+E\left(\Delta L\right){SQP}\\ +{EL}\left(\Delta S\right){QP}+{ELS}\left(\Delta Q\right)P+{ELSQ}\left(\Delta P\right)$$(7)where 0 refers to the initial stage, and 1 refers to the last stage. \(\Delta\) denotes the change in a factor. Each of the six factors in Eq. (7) represent the contributions to carbon emissions change induced by one force while other factors are kept constant. The six factors in our SDA model have \(6!=720\) first-order decompositions, but this approach is too time-consuming for the modelling. Instead, we use the average of two polar decompositions53. The influence of each factor can be respectively illustrated as:$$\Delta {C}_{E}=\frac{1}{2}\left(\Delta E{L}_{1}{S}_{1}{Q}_{1}{P}_{1}+\Delta E{L}_{0}{S}_{0}{Q}_{0}{P}_{0}\right)$$(8)$$\Delta {C}_{L}=\frac{1}{2}\left({E}_{0}\Delta L{S}_{1}{Q}_{1}{P}_{1}+{E}_{1}\Delta L{S}_{0}{Q}_{0}{P}_{0}\right)$$(9)$$\Delta {C}_{S}=\frac{1}{2}\left({E}_{0}{L}_{0}\Delta S{Q}_{1}{P}_{1}+{E}_{1}{L}_{1}\Delta S{Q}_{0}{P}_{0}\right)$$(10)$$\Delta {C}_{Q}=\frac{1}{2}\left({E}_{0}{L}_{0}{S}_{0}\Delta Q{P}_{1}+{E}_{1}{L}_{1}{S}_{1}\Delta Q{P}_{0}\right)$$(11)$$\Delta {C}_{P}=\frac{1}{2}\left({E}_{0}{L}_{0}{S}_{0}{Q}_{0}\Delta P+{E}_{1}{L}_{1}{S}_{1}{Q}_{1}\Delta P\right)$$(12)where \(\Delta {C}_{E}\) represents the impact of changes in carbon intensity per output, capturing the carbon efficiency of production technologies and energy use; \(\Delta {C}_{L}\) represents the impact of changes in the intersectoral relationships in the production system, and changes in L indicate shifts in the composition of intermediate inputs across sectors—for instance, a movement from carbon-intensive upstream industries to cleaner downstream or service sectors; \(\Delta {C}_{S}\) represents the impact of changes in the distribution of final demand across different sectors, and a shift in S reflects changes in consumption preferences or patterns—for example, decreasing demand for construction materials and increasing demand for low-carbon services; \(\Delta {C}_{Q}\) represents the impact of changes in the total level of final consumption per person, reflecting overall consumption intensity or affluence; \(\Delta {C}_{P}\) represents the impact of changes in population.No-trade counterfactual scenarioTo explore the impact of urbanization on carbon mitigation, we built a no-traded scenario based on input-output database54. In this scenario, each city satisfies its own final demand entirely through local production, while all intercity trade flows are assumed to be absent. The aim is to simulate a system in which cities are economically self-sufficient and spatial interactions are eliminated, thereby isolating the emission impacts of trade integration.City r’s export vector \({e}^{r}\) and its gross import vector \({m}^{r}\) are decomposed as follows:$${e}^{r}={\sum }_{s\ne r}{e}^{{rs}}={\sum }_{s\ne r}{z}^{{rs}}+{y}^{{rs}}$$(13)$${m}^{r}={\sum }_{s\ne r}{e}^{{sr}}={\sum }_{s\ne r}{z}^{{sr}}+{y}^{{sr}}$$(14)where \({z}^{{rs}}\) and \({y}^{{rs}}\) represent intermediate and final demand exports from region r to region s, respectively. Equations (15) and (16) define the emissions embodied in city’s exports and imports:$${{EEGX}}^{r}={c}^{r}{L}^{r}{\sum }_{s\ne r}{e}^{{rs}}={c}^{r}{L}^{r}{e}^{r}$$(15)$${{EEGM}}^{r}={c}^{s}{L}^{s}{\sum }_{s\ne r}{m}^{{sr}}={c}^{s}{L}^{s}{m}^{r}$$(16)where \({c}^{r}\) is the vector of direct carbon emission intensities for city r, and \({L}^{r}\) is the full Leontief inverse matrix for city r within the multi-regional system. These two equations represent actual emissions generated by local exports (onshoring) and emissions displaced via imports (offshoring), respectively.Under the no-trade scenario, all imported goods \({m}^{r}\) are assumed to be produced locally, using the intra-city Leontief structure \({L}^{{rr}}=(I-{A}^{{rr}})\). The emissions associated with this substitution are:$${{EEGM}}^{r}={c}^{r}{L}^{{rr}}{m}^{r}$$(17)We then define net onshoring emissions (NONSH) as the difference between local emissions from exports and emissions avoided by imports:$${{NONSH}}^{r}={{EEGX}}^{r}-{{EEGM}}^{r}={c}^{r}{L}^{{rr}}({e}^{r}-{m}^{r})$$(18)This indicator quantifies the net effect of trade on domestic emissions. If \({{EEGX}}^{r}\) > \({{EEGM}}^{r}\), trade leads to a net increase in local emissions, as more emissions are generated to supply external markets (onshoring). Conversely, if \({{EEGX}}^{r}\)