Introduction

Climate warming, a result of anthropogenic greenhouse gas emissions (GHGs), has attracted significant global attention1,2,3. In recent years, sustained improvements in living standards, driven by economic development, have led to substantial changes in diets. This, in turn, has resulted in a dramatic increase in GHGs from the food system4,5,6,7. The food system has been recognized as a significant source of anthropogenic GHGs8,9. Without effective measures, the GHGs from the food system are expected to increase by an additional 80% by 205010. Establishing an environmentally sustainable and health-promoting food system is crucial for mitigating climate change and maintaining well-being, particularly to achieve the climate change mitigation target of limiting the increase to below 1.5 °C11,12,13.

Research on GHGs from the food system has primarily been conducted from the perspectives of production and consumption. On one hand, GHGs are closely associated with agricultural production. GHGs such as CO2, CH4, and N2O originate from agricultural processes, including the use of fertilizers and pesticides, energy consumption for transportation and irrigation, land use change, and livestock enteric fermentation, among others. The agricultural production sector has attracted significant attention due to its considerable potential for reducing GHGs, making effective mitigation measures a research priority for the academic community. Several economies have implemented initiatives such as organic farming, green agriculture, and ecological farming, aiming to mitigate climate change by reducing GHGs14,15,16. Specific measures, such as reducing fertilizer use (particularly nitrogen fertilizers), practicing no-till farming, and utilizing biochar, show promise for controlling GHGs within a certain range. On the other hand, research from the consumption side tends to focus on reducing GHGs through waste reduction and changing dietary habits. Many countries have implemented strategies such as dietary guidelines17,18,19, food environment labeling20,21, and imposing taxes22,23,24 to decrease GHGs from the food system.

GHGs from the food system exhibit spatial heterogeneity. The prevalence of the Western dietary model in North America and Northern European countries contributes to higher GHG emissions. For instance, food-related GHGs account for 15% of total emissions in the United States25,26, 35% in the United Kingdom27,28, and 31% in the European Union29. In contrast, the Mediterranean dietary model, characterized by a plant-based and diverse food intake, is prevalent in Southern Europe and is more environmentally friendly. Similarly, in Asia, there is a sixteen-fold difference in GHG emissions between South Asia and Central Asia, largely due to disparities in meat consumption30. Besides dietary habits, levels of economic and social development are often considered primary drivers contributing to differences in food system GHGs. For example, an increase in per capita GDP leads to a greater demand for protein-rich red meat, seafood, and dairy products, which in turn raises the consumption of animal-based foods with higher carbon footprints31,32. The process of urbanization elevates residents’ income levels and facilitates dietary diversification and dining out, thereby exacerbating GHG emissions33,34,35. Other factors such as education level, consumption spending, population growth, and technological advancement also play significant roles in causing regional heterogeneity36,37,38,39.

China is a major GHGs emitter, with the proportion from food systems accounting for 18% of the total emissions and showing a trend of continual growth39. In recent years, with improvements in living standards, dietary patterns have shifted from predominantly coarse grains and carbohydrates with minimal animal-based foods to diets characterized by increased consumption of refined rice, wheat, and animal products, alongside a reduction in carbohydrates40,41. Numerous studies have examined the impact of dietary changes on climate change using multiple methods, including input-output analysis42,43, life cycle assessment38,44, and footprint analysis45,46, consistently concluding that changes in Chinese residents’ diets have intensified GHGs, with rising income and consumption levels identified as the principal drivers. As a vast country, China exhibits significant regional inequities in geography, socioeconomy, and culture, making national average assessments of emissions inadequate to reflect micro-level differences in food-related GHGs among residents. Although a limited number of studies have estimated the regional disparities of per capita food greenhouse gas emissions (PFGHGs) using the Gini coefficient47,48, there has been a lack of quantification regarding the associations and differences between PFGHGs and socioeconomic factors. Furthermore, China’s distinct urban-rural dualism is also reflected in PFGHGs, which is crucial for the formulation of differentiated carbon reduction policies. Typically, urban residents, due to their higher consumption of animal-based foods, exhibit higher PFGHGs compared to their rural counterparts31,49. While some studies have assessed urban-rural disparities in PFGHGs at the national level50,51, they often overlook sub-national urban-rural variations and tend to discuss the drivers of PFGHGs in a macro context, lacking a systematic evaluation of the drivers and mechanisms behind the urban-rural disparities.

To address these gaps, this study explores the spatial differences and inequalities of PFGHGs from both provincial and urban-rural perspectives. Firstly, it employs life cycle methods and emission coefficients to estimate PFGHGs from 1990 to 2022. Secondly, it quantifies regional disparities in PFGHGs using the Gini coefficient and β-convergence models. Finally, it identifies the drivers contributing to urban-rural disparities through the Oaxaca-Blinder decomposition method. This research quantifies the degree and trend of inequality in PFGHGs across regions from an economic and social perspective, investigates the primary causes behind the changes in urban-rural disparities, and provides new insights for research on GHGs reductions in food systems, as well as theoretical foundations for relevant policy formulation.

Results

National development trend in PFGHGs

From 1990 to 2022, China’s PFGHGs increased from 252.34 to 423.52 kg CO2e, marking a cumulative growth of 54.40% and an average annual growth rate of 1.7% (Fig. 1). The PFGHGs associated with animal-based foods surged from 126.65 kg CO2e to 326.56 kg CO2e, whereas those related to plant-based foods decreased from 125.65 kg CO2e to kg 96.96 CO2e. The proportion of pork-sourced emissions in total PFGHGs rose from 38% to 48.3%, making it the category with the highest growth rate. In contrast, the PFGHGs from grains and vegetables experienced significant declines, falling from 64 kg CO2e and 54.5 kg CO2e to 35.8 kg CO2e and 43.4 kg CO2e, respectively, with their shares decreasing by 16.9% and 11.3%. Except for sugar, the proportions of other food categories increased to varying extents, with beef rising by 4.5%, mutton by 2.6%, poultry by 3.6%, and seafood by 3.7%, while changes in the proportions of other food types remained below 2%.

Fig. 1: Analysis of trends in PFGHGs from Chinese residents (1990–2022).
Fig. 1: Analysis of trends in PFGHGs from Chinese residents (1990–2022).
Full size image

The X axis represents the year, while the Y axis shows the cumulative PFGHGs of all food categories.

Regional differences in PFGHGs

The spatial pattern of PFGHGs displays a “low in the center and high on the periphery” distribution (Fig. 2a). In 1990, the provinces with the lowest PFGHGs were situated around Shanxi Province, encompassing Henan, Shaanxi, Gansu, and Shandong. Chongqing, Beijing, and Shanghai were ranked as the top three provinces with the highest PFGHGs, while other provinces with relatively high emissions were primarily concentrated in the southwestern, northwestern, and the middle and lower reaches of the Yangtze River. This distribution persisted in 2001 and 2012, with a rapid increase in PFGHGs in South China, where Guangdong Province emerged as the province with the highest PFGHGs by 2012. By 2022, this trend had become even more pronounced. Shanxi, Shaanxi, Gansu, Henan, and Shandong continued to be the provinces with the lowest PFGHGs, while Chongqing recorded the highest emissions, followed by Xizang and Zhejiang, with the southeastern, northwestern, and southwestern regions exhibiting even higher PFGHGs.

Fig. 2: Spatial distribution patterns of PFGHGs in 1990, 2001, 2012, 2022.
Fig. 2: Spatial distribution patterns of PFGHGs in 1990, 2001, 2012, 2022.
Full size image

The spatial distribution is presented in three vertical panel groups (a–c), with each group containing four thematic maps arranged vertically. a shows the spatial distribution of PFGHGs across all foods in 1990, 2001, 2012, and 2022. b shows the spatial distribution of PFGHGs across plant-based foods in 1990, 2001, 2012, and 2022. c shows the spatial distribution patterns of PFGHGs across animal-based foods in 1990, 2001, 2012, and 2022.

In 1990, the gap in plant-based PFGHGs between the provinces with the highest and lowest emissions was 98 kg CO2e, with the higher-emission provinces mainly located in the northeast and the middle and lower reaches of the Yangtze River (Fig. 2b). By 2001, nearly 90% of the provinces had seen a reduction in plant-sourced PFGHGs, with high-emission areas shifting towards the eastern and southern provinces, while the northwestern regions and Inner Mongolia exhibited relatively lower PFGHGs. By 2012, the vast majority of provinces continued to experience a decrease in plant-based PFGHGs, with high-emission areas becoming more concentrated in the middle and lower reaches of the Yangtze River. In 2022, half of the provinces witnessed a rebound in PFGHGs, primarily in the north, northeast, and eastern regions, with high-emission areas predominantly in the middle and lower reaches of the Yangtze River and the northeastern regions. Compared to 1990, all provinces showed a decline in PFGHGs, except for Xizang and Gansu. The provinces with the most significant decreases were Jilin, Inner Mongolia, Hubei, Guizhou, and Fujian, while the decline in other provinces was within 40 kg CO2e. The gap between the provinces with the highest and lowest plant-based PFGHGs narrowed to 50 kg CO2e.

Changes in animal-based food PFGHGs were more significant (Fig. 2c). In 1990, the provinces with higher animal-based PFGHGs included Beijing, Chongqing, and several minority-populated provinces (Xinjiang, Xizang, and Inner Mongolia). By 2001, all provinces except Xizang showed an increasing trend in animal-based PFGHGs, with significant increases concentrated in the southwestern region (Yunnan: +52 kg CO2e; Sichuan: +44 kg CO2e; Guizhou: +59 kg CO2e). By 2012, Xinjiang experienced a slight decrease in animal-based PFGHGs, while all other provinces continued to show an upward trend, particularly in South China, where growth rates were the fastest (Guangdong: +102 kg CO2e; Guangxi_ +90 kg CO2e; Hainan: +80 kg CO2e). The period from 2012 to 2022 saw the most rapid increase in animal-based PFGHGs, with all provinces experiencing increases exceeding 60 kg CO2e, except for Beijing and Tianjin, which saw slight declines; half of the provinces recorded increases exceeding 100 kg CO2e (Xizang: +223 kg CO2e; Chongqing: +190 kg CO2e). The provinces with the highest PFGHGs changed from Qinghai, Beijing, Shanghai, Xizang, and Xinjiang to Inner Mongolia, Guangdong, Yunnan, Xinjiang, and Zhejiang. The overall spatial pattern exhibited a trend of being low in the center and high on the periphery, becoming more pronounced over time.

From 1990 to 2022, the Gini coefficient for PFGHGs in all food types fluctuated around 0.11, with higher values concentrated before 2000 (Fig. 3). After 2016, there was a gradual reduction in inequality between regions. Mutton, beef, and seafood were identified as the three food types most associated with inequality, while cereals, vegetables, fruits, and sugars exhibited relatively lower levels of inequality. The degree of inequality in animal-based food PFGHGs gradually decreased, with milk showing the most notable change, as its Gini coefficient decreased from 0.66 to 0.19. Poultry and eggs also experienced reductions, with Gini coefficients falling from 0.42 and 0.37 to 0.30 and 0.22, respectively. Meanwhile, the Gini coefficient for pork remained stable at 0.24.

Fig. 3: Temporal trends in Gini coefficients of PFGHGs for food categories.
Fig. 3: Temporal trends in Gini coefficients of PFGHGs for food categories.
Full size image

The Gini coefficient results reflect the spatial inequality of PFGHGs. The central radar chart in purple color represents the spatial inequality across all food categories, while the 12 surrounding radar charts with pink color display the spatial inequality of PFGHGs for 12 food types. The charts are arranged clockwise in chronological order, with the vertical axis indicating the values of the Gini coefficient.

Table 1 presents the absolute β-convergence results for all foods, plant foods and animal foods. Although the β coefficients for all foods and animal-based foods PFGHGs were negative, they failed to reach statistical significance. Consequently, neither all foods nor animal-based foods PFGHGs demonstrated convergence trends, indicating that the initial lower PFGHGs levels in certain provinces had no measurable impact on their growth rates. In contrast, plant-based foods PFGHGs exhibited statistically significant convergence effects. This suggests that provinces with initially lower plant-based PFGHGs levels are systematically “catching up” to those with higher initial levels.

Table 1 Beta-convergence results of PFGHGs
Full size table

The β coefficients for plant-based, animal-based, and total food PFGHGs were –0.264, –0.187, and –0.181 respectively, all statistically significant at the 1% level, demonstrating conditional convergence characteristics in these food-related emissions (Table 2). The convergence patterns exhibited variations across different influencing factors, with animal-based and total food PFGHGs showing particularly high similarity in their convergence behaviors.

Table 2 Beta conditional convergence results of PFGHGs
Full size table

The urbanization rate, consumption level, and income level all exerted significant impacts on animal-based and total food PFGHGs. Specifically, the consumption level showed negative effects at the 5% significance level, indicating that higher consumption standards actually helped suppress PFGHG emissions. In contrast, urbanization level and income level displayed positive correlations, contributing to the growth of PFGHGs.

Diverging from the patterns observed in animal-based and total food emissions, plant-based PFGHGs showed distinct characteristics: a decrease in the Engel coefficient promoted the reduction of plant-based PFGHGs, and interestingly, increasing urbanization level demonstrated a similar emission-reducing effect. Furthermore, consumption level and income level exhibited opposing influences on plant-based PFGHGs - while higher consumption levels led to emission mitigation, increased income levels conversely contributed to emission expansion.

Urban-rural inequality in PFGHGs

From 1990 to 2022, based on the changes in urban-rural PFGHG disparities, PFGHGs can be categorized into three phases: a “slow decrease - rapid increase - rapid decrease” process, with urban PFGHGs consistently remained higher than those of rural areas (Fig. 4). During the first phase, from 1990 to 2001, the urban-rural gap shrank at an annual rate of 7.33%, stabilizing at 32.47 kg CO2e. In the second phase, from 2001 to 2012, urban-rural PFGHGs expanded at an annual growth rate of 13.75%. By 2012, urban residents’ PFGHGs reached 358.34 kg CO2e, exceeding rural levels by 115.18 kg CO2e. The third phase (2012–2022) saw the urban-rural disparity rapidly shrink at an average annual rate of –20.76%, with rural PFGHGs (417.52 kg CO2e) approaching urban PFGHGs (426.04 kg CO2e) by 2022.

Fig. 4: Quantitive gap in PFGHGs between urban and rural regions.
Fig. 4: Quantitive gap in PFGHGs between urban and rural regions.
Full size image

The red dots represent the average PFGHGs of urban residents in China; Blue dots represent the average PFGHGs of rural residents in China.

In urban areas, over 60% experienced a decline in PFGHGs during the first phase, with these provinces concentrated in North China, Central China, South China, and Northwestern regions (Fig. 5a). The provinces with the most significant increases and decreases were Xizang ( + 101.54 kg CO2e) and Ningxia (–93.56 kg CO2e), respectively. In the second phase, only four provinces (Beijing, Shanxi, Xizang, and Qinghai) saw a decrease in urban PFGHGs, while Chongqing ( + 104.51 kg CO2e) recorded the most notable increase. During the third phase, PFGHGs in Beijing continued to decline, with additional decreases in Tianjin, Guangxi, and Sichuan, while the remaining provinces experienced increases. The provinces with the most pronounced growth were Xizang ( + 239.11 kg CO2e) and Zhejiang ( + 166.86 kg CO2e).

Fig. 5: Spatial distribution of PFGHGs of urban and rural regions, 1990, 1999, 2011, and 2022.
Fig. 5: Spatial distribution of PFGHGs of urban and rural regions, 1990, 1999, 2011, and 2022.
Full size image

The spatial distribution is presented in three vertical panel groups (a–c), with each group containing four thematic maps arranged vertically. Panel a shows the spatial distribution of PFGHGs of urban regions, 1990, 1999, 2011, and 2022. Panel b shows the spatial distribution of PFGHGs of rural regions, 1990, 1999, 2011, and 2022. Panel c shows the spatial distribution of urban-rural disparities in PFGHGs across provinces in 1990, 2001, 2012, and 2022.

In rural areas, over 60% of PFGHGs increased in the first phase, with Henan Province exhibiting the most significant rise, while declines were concentrated in North China and Northeast regions (Fig. 5b). In the second phase, the number of provinces experiencing growth trends slightly decreased but still exceeded 50%, with marked declines in rural areas of East and Central China. During the third phase, all provinces experienced an increase in rural PFGHGs, particularly notable in the East China region (with Zhejiang at +279.26 kg CO2e; Jiangxi at +226.98 kg CO2e; and Anhui at +209.74 kg CO2e).

From the perspective of urban-rural disparities, in 1990, all provinces, except for Jilin, exhibited higher urban PFGHGs compared to rural areas, with the largest gaps found in the Northwestern, Southwestern, and North China regions (Fig. 5c). By 2001, the number of provinces where rural PFGHGs surpassed urban levels rose to four, with significant disparities concentrated in North China as well as Xizang and Xinjiang. In 2012, Qinghai Province was the only province where rural PFGHGs exceeded those of urban areas, and the urban-rural gap in the eastern region was markedly greater than in the western region. By 2022, the urban-rural gap further narrowed, with 15 provinces reporting higher rural PFGHGs than urban ones. The disparity exhibited a pronounced spatial distribution characterized by a higher central area surrounded by lower regions.

In the first phase (1990–2001), urban PFGHGs decreased by 24.42 kg CO2e. Pork contributed 16.35% to this decrease, being the primary reason for the overall decline in total PFGHGs (Fig. 6a). Additionally, the food categories that experienced declines included cereals, vegetables, sugar, beef, and lamb. In contrast, poultry and seafood made the largest contribution to the increase. During the second phase (2001–2012), all food categories showed an increasing trend, with urban PFGHGs rising by a total of 67.95 kg CO2e. PFGHGs from pork accounted for 40.75% of this increase, followed by beef, seafood, and poultry, while other food contributions were minimal, under 3%. The third phase (2012–2022) continued the growth trend observed in the second phase, with PFGHGs increasing by 54.69 kg CO2e, with pork remaining the largest contributing food type. Except for dairy, all animal-based foods exhibited an upward trend, particularly beef and seafood. Meanwhile, cereals experienced significant growth, while other plant-based food types saw slight declines.

Fig. 6: The structural changes of PFGHGs during three stages in urban and rural areas.
Fig. 6: The structural changes of PFGHGs during three stages in urban and rural areas.
Full size image

Panel a shows the structural changes of PFGHGs during three stages in urban areas, and b displays the structural changes of PFGHGs during three stages in rural areas. The gray bars in the figure represent the PFGHG values for either urban or rural residents nationwide during the target years, while the colored bars show the contribution percentage changes of different food categories to PFGHGs from the previous target year to the next.

Rural PFGHGs, when broken down by food type, significantly differ from their urban counterparts. During the initial phase (1990–2001), pork accounted for 23.36% of the increase in rural PFGHGs, whereas vegetables and cereals contributed –8.58% and –6.73%, respectively, with other food types showing negligible variations (Fig. 6b). The second phase (2001–2012) was characterized by a swift downturn in plant-based food PFGHGs, with cereals and vegetables experiencing sharp declines. Changes in animal-based foods and other plant-based food PFGHGs were not substantial. In the third phase (2012–2022), all food types displayed a swift upward trend in PFGHGs, with animal-based foods particularly notable. Pork contributed the most significant increase at 103.77%, taking the lead, followed by beef, lamb, and poultry. There was also a resurgence in the PFGHGs from vegetables, which contributed 7.32% to the overall growth.

Three stages were identified, with PFGHGs in urban areas consistently higher than those in rural areas. In the first stage, four factors—education level, consumption level, Engel’s coefficient, and income level—collectively accounted for 40% of the difference (Fig. 7a). The income level had the highest contribution rate (490.67%), significantly widening the urban-rural PFGHGs gap. In contrast, the Engel’s coefficient (-193.3%) and consumption level (–274.92%) contributed to narrowing the gap. These four factors together explained 55% of the difference in the second stage, with income (204.67%) remaining the primary factor contributing to the widening urban-rural gap. The education level (42%) had a relatively weak explanatory power for the gap, but its influence shifted from widening to narrowing the gap during this stage. The consumption level (38.82%) and Engel’s coefficient (67.4%) also demonstrated a trend of narrowing the gap. In the third stage, these four factors continued the influence trends observed in the second stage and even expanded their impact. Overall, the income level contributed to widening the urban-rural difference, with its contribution growing to 718.86%. However, the education level (–117.73%), consumption level (–65.06%), and Engel’s coefficient (–581.37%) were favorable for narrowing the gap.

Fig. 7: Decomposition of driving factors for urban-rural differences in PFGHGs.
Fig. 7: Decomposition of driving factors for urban-rural differences in PFGHGs.
Full size image

The three horizontal graphs form a group, representing three phases of PFGHGs (1990–2001, 2001–2012, and 2012–2022). Panel a shows the decomposition of driving factors for urban-rural differences in PFGHGs across all foods. Panelb shows the decomposition of driving factors for urban-rural differences in PFGHGs across plant-based foods. Panelc dispalys the decomposition of driving factors for urban-rural differences in PFGHGs across animal-based foods. The figure summarize four key disparity characteristics affecting urban-rural differences: educational attainment, income level, Engel’s coefficient, and consumption level. The gray bars indicate the unexplained portion in the Oaxaca-Blinder decomposition model.

The urban-rural disparity in PFGHGs for plant-based foods was negative, indicating higher levels in rural areas compared to urban areas. Across the three stages, the level of education consistently contributed to widening the urban-rural gap in PFGHGs for plant-based foods (Fig. 7b). In contrast, Engel’s coefficient consistently narrowed the urban-rural gap for PFGHGs of plant-based foods. The income level narrowed the difference in the second stage (with a contribution of 274.63%), but it widened the difference in the first and third stages (–325.67% and –545.16%, respectively). The consumption level had the opposite influence of the income level, reducing the urban-rural gap in the first and third stages but widening it in the second stage.

Consistent with the overall PFGHGs, the income level (278.76%, 213.02%, 619.63%) influenced the widening of the urban-rural gap in PFGHGs for animal-based foods across the three stages (Fig. 7c). The difference in PFGHGs for animal-based foods was consistently positive, while the consumption level contributed oppositely across the three stages, indicating efforts to narrow the urban-rural gap (–146%, –91.59%, –503.69%, respectively). Engel’s coefficient and the education level also contributed to narrowing the urban-rural gap, albeit to a lesser extent.

Discussion

There are significant disparities in food consumption and PFGHGs among residents across different provinces. Provinces surrounding Shanxi exhibit lower PFGHGs, primarily due to their dietary structure, which is predominantly plant-based (mainly wheat), resulting in lower animal-based PFGHGs compared to other provinces52,53. In contrast, provinces in the middle and lower reaches of the Yangtze River, as well as those in the Northeast region, have higher plant-based PFGHGs than other areas, which is strongly associated with local dietary habits. These two regions host major grain production bases in China, where residents are more accustomed to deriving satiety from plant-based foods54,55. Although animal-based foods have a higher emission coefficient, plant-based foods dominate in terms of consumption volume; from the perspective of PFGHGs, however, animal-based foods are predominant.

The Southwest region (Sichuan, Chongqing) and provinces with significant ethnic minority populations (Qinghai, Xizang, Xinjiang, Inner Mongolia) lead the nation in PFGHGs due to meat consumption56. In addition to dietary habits, the livestock industry in these minority-dense provinces is typically well-developed, which reduces the need for complex transportation and makes meat consumption more convenient57. When examining the dietary differences between northern and southern regions, it is evident that the southern areas favor animal-based foods, resulting in higher PFGHG emissions compared to the north.

Absolute convergence suggests that the convergence of PFGHGs from animal-based foods and all foods is relatively weak. Considering the Gini coefficient, while some animal-based food categories exhibit a significant downward trend in their Gini coefficients, disparities remain substantial. Notably, regional differences in staple foods, which account for the highest proportion of food consumption, and pork, which constitutes the highest proportion of PFGHGs, have not shown significant declines. The trend in the Gini coefficient for total PFGHGs closely mirrors that of pork, with only a brief decline observed after 2016.

The levels of urbanization and income are contributing to the widening of inequality, as evidenced by their negative CI indices, indicating that regions with lower socioeconomic levels have higher PFGHGs (Fig. 8a)58. This disparity primarily arises from the regions with the lowest socioeconomic levels compared to those with moderate socioeconomic levels. The average resident PFGHGs in the top ten provinces with the highest socioeconomic levels (in terms of urbanization and income) show minimal differences from the average PFGHGs in the bottom ten provinces with the lowest socioeconomic levels; however, both are significantly higher than the average PFGHGs in the 11 provinces with intermediate socioeconomic levels (Fig. 8b, c).

Fig. 8: The concentration index of PFGHGs between urban and rural areas in China, 1990–2022.
Fig. 8: The concentration index of PFGHGs between urban and rural areas in China, 1990–2022.
Full size image

Panel a shows temporal variation in the concentration index (CI) ranked by urbanization and income levels. The horizontal axis represents the year, while the vertical axis indicates the CI values. Panels b, c display the distribution of PFGHGs across provinces grouped into tertiles based on urbanization and income rankings. The color coding represents: orange—lowest group (10 provinces); green - intermediate group (11 provinces); blue - highest group (10 provinces). In the box plot, the dots represent the mean values of PFGHGs, and whiskers extending to 1.5 times the interquartile range.

From a national perspective, in 2022, the per capita food group heterogeneity gap (PFGHG) between rural and urban areas narrowed, indicating a reduction in urban-rural inequality regarding PFGHGs. Analysis of dietary structures shows that the higher PFGHGs from staple foods consumed by rural residents counterbalanced the PFGHGs from fruits and vegetables. This finding aligns with previous research, which suggests that wealthier individuals typically have higher total PFGHGs and greater animal-based PFGHGs59. Although rural residents have exhibited an upward trend in animal-based PFGHGs, this is mainly seen in pork consumption, which has a higher proportion of PFGHGs compared to urban areas; however, the proportions of other types of animal-based PFGHGs remain lower than those in urban settings. Additionally, there continues to be a disparity in dietary diversification between rural and urban populations.

Regarding regional distribution, by 2022, 15 provinces had higher rural PFGHGs than urban areas, whereas the remaining provinces exhibited higher PFGHGs in urban settings. Spatially, the distribution patterns of urban and rural areas are largely similar. In both rural and urban contexts, the PFGHGs in North China, as well as in provinces such as Shaanxi, Gansu, and Ningxia, are lower than those in other regions. Beijing and Tianjin are the only two provinces where urban PFGHGs in 2022 were lower than those recorded in 1990. Regarding the differences between northern and southern regions, southern areas generally contribute more to poultry PFGHGs, particularly in South China. This can be attributed to the region’s unique consumption habits and dietary culture. Southern consumers typically tend to favor poultry as the main source of meat in their diets60, while northern consumers tend to favor other meats, such as pork.

Rising income levels have contributed to the reduction of the urban-rural divide, while increasing consumption levels are widening the gap between urban and rural areas. Since the reform and opening up, the incomes of both urban and rural residents have increased more than tenfold. With this rise in income, rural residents have gained the opportunity to choose a greater variety of foods, such as pork, fruits, milk, and other products. Although the gap between urban and rural areas is narrowing, the higher incomes and better environmental conditions of urban residents enable them to select a broader range of food options, exerting more environmental pressure than in rural areas. The educational attainment of urban residents ranges from 9 to 13 years, while that of rural residents ranges from 5 to 10 years. Higher educational levels typically indicate greater employment opportunities and increased income61,62. Similarly, urban areas offer more employment opportunities, and residents with higher education generally reside in urban centers. This has led to educational attainment becoming a factor that exacerbates the disparity in food carbon emissions between urban and rural areas. The narrowing Engel coefficient is contributing to a reduction in urban-rural disparities in food-related carbon emissions. Engel coefficient serving as an intermediary variable between food expenditure and total expenditure, we observe that although consumption levels for both urban and rural residents continue to improve, the persistent decline in the Engel coefficient indicates that food consumption expenditure has not expanded in tandem with rising overall consumption levels (Fig. 9).

Fig. 9: Conceptual framework of urban-rural PFGHGs driving mechanisms.
Fig. 9: Conceptual framework of urban-rural PFGHGs driving mechanisms.
Full size image

The figure consists of two main parts: one showing household characteristics affecting PFGHG changes and urban-rural disparities, and the other showing changes in residential consumption behavior. It illustrates how household characteristics drive consumption behavior changes, which in turn lead to changes in PFGHGs and its urban-rural disparities.

While this study did not specifically verify the impact of urbanization and per capita GDP on urban-rural PFGHG disparities, these factors remain critically significant. This significance manifests in two aspects: first, it reflects higher levels of urban development; second, it accompanies population migration. Regions with more advanced economic development typically provide greater employment opportunities63, as evidenced by rural residents migrating to cities for work and residents from poorer provinces seeking employment across provincial boundaries. This phenomenon creates a dual effect: On one hand, the income growth of migrant workers leads to changes in dietary patterns, thereby widening the gap in food carbon emissions between them and residents in their original communities64. On the other hand, more convenient shopping environments enhance food accessibility, promoting faster and more convenient dining and shopping experiences33,34,35.

Methods

Research design

Our methodology combines quantitative and qualitative approaches and comprises several steps as follows (Fig. 10):

Fig. 10: The theoretical analysis framework.
Fig. 10: The theoretical analysis framework.
Full size image

The figure primarily comprises two parts: the upper part presents an analysis of spatiotemporal patterns in PFGHGs, while the lower part examines disparities in PFGHGs between urban and rural residents in China. Each part includes both methodological approaches and explanatory results of driving factors.

Accounting for food GHGs

This study employs the LCA method to synthesize food GHG emission coefficients from existing research38,65,66. It then calculates the food GHGs by multiplying the emission coefficients of food categories by their respective consumption quantities67. The PFGHGs are calculated as indicated in Eq.( 1).

$${{PFGHGs}}_{j}={\sum }_{i=1}^{12}{w}_{i}{C}_{{ij}}$$
(1)

Where, \({\rm{i}}\) represents the food category, \({\rm{j}}\) denotes either urban or rural areas, \({{\rm{w}}}_{{\rm{i}}}\) represents the emission coefficient for the ith food category in urban/rural areas (Table 3), and \({C}_{{ij}}\) represents the per capita consumption quantity of the ith food category for urban/rural residents.

Table 3 GHGs coefficient of food types (kgCO2e/kg food consumption)
Full size table

The Gini coefficient

The measurement for assessing inequality in PFGHGs or environmental impacts is designed to employ approaches initially used to measure income inequality, such as the Gini coefficient method48,68. The Gini coefficient ranges from complete equality (0) to complete inequality (1), explaining the inequality status of GHGs between urban and rural areas among provinces. The specific formula is as follows:

$${Gini}=A\,/(A+B)=1-{\sum }_{i=1}^{n}({x}_{i}-{x}_{i-1})({y}_{i}+{y}_{i-1})$$
(2)

Where \(n\) represents the number of provinces, \(i\) ranks provinces based on PFGHGs, \({x}_{i}\) and \({y}_{i}\) denotes the cumulative percentage of ranking for province and PFGHGs, respectively.

β-Convergence analysis method

The convergence model originates from the endogenous growth theory proposed by Robert Merton Solow69, initially used to examine the convergence of economic growth. β-convergence describes a phenomenon where relatively underdeveloped regions grow at a faster rate, gradually catching up with more developed regions, leading to a state of convergence with equal growth rates. β-convergence can be subdivided into absolute β-convergence and conditional β-convergence70,71. Absolute β-convergence suggests that, all else being equal, the PFGHGs across different regions will converge to the same level over time72,73. The study of convergence originated from neoclassical growth theory. The neoclassical growth model explains the disparity in economic growth levels among countries, suggesting that economies eventually reach a steady state owing to diminishing marginal returns on capital74,75. We applied this analytical framework to examine inequality in food-related carbon emissions. In this study, convergence signifies that provinces with initially lower per capita food carbon emissions exhibited faster emission growth rates, while those with higher initial emissions demonstrated relatively slower growth. This dynamic enables provinces with lower initial emissions to progressively narrow the gap with or even surpass higher-emitting provinces. Under such convergence conditions, interprovincial disparities in emissions tend to diminish. Conditional β-convergence indicates that, after controlling for other factors, the PFGHGs across regions tend to demonstrate a stable state over time. PFGHGs are closely associated with socioeconomic factors such as educational attainment, urbanization levels, consumption expenditures, Engel’s coefficient, and income levels (Table 4). To rigorously evaluate these factors’ impacts on carbon emissions, we first conducted a comprehensive multicollinearity test. Variance inflation factor (VIF) analysis confirmed acceptable collinearity levels across all key predictors (Table 5). This study selects appropriate models (fixed effects model and random effects model) based on the results of the Hausmann test, using the β-convergence model to investigate the evolutionary characteristics of PFGHGs.

Table 4 Indicators of influencing factors for PFGHGs
Full size table
Table 5 Multicollinearity test results of variables
Full size table

Absolute β-convergence Eq. (3) is expressed as:

$$\tfrac{{\mathrm{ln}\mathrm{PFGHGs}}_{{\rm{i}},\,{\rm{t}}+{\rm{T}}}-{\mathrm{ln}\mathrm{PFGHGs}}_{{\rm{i}},{\rm{t}}}}{{\rm{T}}}={\rm{\alpha }}+{\rm{\beta }}{\mathrm{ln}\mathrm{PFGHGs}}_{{\rm{i}},{\rm{t}}}+{{\rm{\sigma }}}_{{\rm{i}},{\rm{t}}}$$
(3)

Conditional β-convergence Eq.(4) is expressed as:

$$\begin{array}{l}{\mathrm{lnPFGHGs}}_{{\rm{i}},{\rm{t}}+{\rm{T}}}-{\mathrm{lnPFGHGs}}_{{\rm{i}},{\rm{t}}}={\rm{\alpha }}+{\rm{\beta }}{\mathrm{lnPFGHGs}}_{{\rm{i}},{\rm{t}}}+\mathrm{\gamma lnedu}\\ +\mathrm{\varepsilon urban}+\mathrm{\theta cons}+\mathrm{\mu engeer}+\mathrm{\rho income}+\hskip -1px{{\rm{\sigma }}}_{{\rm{i}},{\rm{t}}}\end{array}$$
(4)

Where α denotes the constant term, and represents the initial year’s PFGHGs; \({edu}\),, \({urban}\), \({cons},{engeer}\), and \({income}\) correspond respectively to educational attainment, urbanization rates, consumption spending, Engel’s coefficient, and income levels. The term σi,t represents the error term. If β < 0 and is statistically significant, it indicates the presence of β-convergence in PFGHGs; otherwise, convergence does not exist.

Blinder-oaxaca decomposition method

The Oaxaca-Blinder decomposition method is a tool used to decompose changes in distributional statistics, and it has been applied across various fields58,76. This study employs the Oaxaca-Blinder decomposition method, incorporating educational attainment, consumption spending, Engel’s coefficient, and income levels as driving factors influencing PFGHGs. It explores the contributions of these socioeconomic drivers to the disparities in PFGHGs between urban and rural areas.

$$\bar{{{\rm{PFGHGs}}}_{u}}-\bar{{{\rm{PFGHGs}}}_{r}}=\bar{{X}_{u}}{\beta }_{u}-\bar{{X}_{{\rm{r}}}}{\beta }_{{\rm{r}}}=(\bar{{X}_{u}}-\bar{{X}_{{\rm{r}}}}){\beta }_{r}+{\int }_{-}({\beta }_{u}-{\beta }_{r})$$
(5)

Where \(u\) and \(r\) denote the urban and rural areas respectively, \(\overline{{\rm{PFGHGs}}}\) repersents the average PFGHGs, \(\bar{X}\) and \({\rm{\beta }}\) denote the mean and the corresponding regression coefficients of the covariates, respectively.

Data sources

This study focuses on China’s 31 provincial-level administrative regions from 1990 to 2022. Due to data availability, the research scope does not include Hong Kong, Macao, and Taiwan. The per capita consumption of various food types required for the calculation of GHGs was sourced from the provincial/regional statistical yearbooks, including The China Statistical Yearbook, The China Rural Household Survey Yearbook, The China Rural Statistical Yearbook, and the China Household Survey Yearbook, for the years 1991–2023. Urban/rural population were sourced from China Population and Employment Statistical Yearbook and China Population Statistics Yearbook. Socioeconomic indicator data were derived from the China Statistical Yearbook. The logical framework of the research is shown in Fig. 10. The urban and rural food consumption in this study were sourced from the China Statistical Yearbook and provincial statistical yearbooks. However, under China’s current statistical system, these datasets do not include the sub-provincial-level data, and city-level statistical yearbooks do not distinguish between urban and rural food consumption in their statistics and do not make this information public. Consequently, an assessment of urban-rural disparities in food consumption at the sub-provincial level is unavailable.