China’s changing diets improve human health but increase environmental pressures

communications earth & environment Article
A Nature Portfolio journal
Article in Press https://doi.org/10.1038/s43247-026-04063-0
China’s changing diets improve human
health but increase environmental
pressures
Check for updates
Yufan Shi1,2, Fan Ning1,2, Miaomiao Liu 1,2,3 , Jianxun Yang 1,2,3,F a n gW e n 1,2,3, Zongwei Ma 1,2,3 &
Jun Bi 1,2,3
Understanding how dietary transitions shape environmental and health outcomes across socio-
economic and spatial contexts is essential for informing sustainable food system transformations.
Here we investigate evolving trade-offs between environmental and health outcomes during urban
and rural dietary transitions in China from 1990 to 2021, and examines how these trade-offs vary with
regional development levels. Here we show that China’s dietary transition has shifted from initial health
and environment synergies to trade-offs, with rising environmental pressures alongside health gains.
These trade-offs are more pronounced in regions with a human development index above 0.55. Urban
and rural dietary transitions show broadly similar patterns but differ in trade-off progression and
associated risks: urban diets are linked to faster health improvements with rising environmental
impacts, while rural diets show slower health gains alongside disproportionately high environmental
burdens. Scenario analyses indicate that integrating precision agriculture, food waste reduction, and
dietary optimization can substantially reduce both health and environmental burdens while narrowing
urban and rural inequalities.
The transformation of dietary patternsplays a pivotal role in advancing both
environmental sustainability and public health. Notably, unhealthy dietary
patterns exacerbate the dual burden on both environmental systems and
human health1,2. On the environmental side, global food systems impose
substantial pressures on the Earth system. They are responsible for
approximately 25–34% of total anthropogenic greenhouse gas (GHG)
emissions3, are a major driver of land degradation4, account for about 70% of
global freshwater withdrawals5 and contribute significantly to biodiversity
loss6. Meanwhile, from a health perspe ctive, poor dietary patterns are
associated with 10.8 –11.6 million premature deaths annually, primarily
from cardiovascular diseases, diabetes, and certain cancers7.P r e v a i l i n gs h i f t s
towards unhealthy diets—characterized by high consumption of sugary
beverages and red meat, along with low milk intake—continue to exacerbate
the global burden of diet-attributable diseases7,8.A c c o r d i n g l y ,t h ep r o m o -
tion of healthy and sustainable diets has emerged as a key priority in global
policy discourse, as evidenced by the EAT–Lancet Commission1,9 and the
United Nations Sustainable Development Goals (SDGs)10.
These agendas assign particular importance to rapidly developing
countries such as China, where advancing this transition is both urgent and
strategically vital11. Compared to high-income countries, developing nations
such as China remain in the early to middle stages of dietary system
transition12. Dietary patterns are undergoing rapid change 13,w h i l ei n f r a -
structure and institutional frameworks are still in the process of develop-
ment. This transitional phase is characterized by considerable uncertainty,
but also offers a crucial window of opportunity for targeted policy inter-
ventions and systemic enhancement. At the same time, China ’sv a s t
population means that its dietary choices and food system dynamics exert a
significant influence on global sustainability14. For example, China’sa g r i -
culture contributes 9–14% of global GHG emissions 15 and is the world ’s
second-largest user of agricultural freshwater, consuming over 300 billion
cubic meters annually 16. As such, guiding China ’s dietary transition is
therefore essential not only for domestic goals but also for advancing global
environment–health synergies.
Despite the growing consensus on the importance of promoting sus-
tainable dietary change, translating this vision into practice remains highly
challenging2,17. In China, sustained socioeconomic growth over recent
decades has driven profound dietar y transformations, shifting from
subsistence-oriented diets toward higher caloric intake and greater nutri-
tional diversity13. Previous studies have shown that such transitions often
entail environment –health trade-offs, underscoring the dif ficulty of
1State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing, China. 2Jiangsu Collaborative Innovation
Center of Atmospheric Environment and Equipment Technology (CICAEET), Nanjing University of Information Science & Technology, Nanjing, China. 3Basic
Science Center for Energy and Climate Change, Beijing, China. e-mail: liumm@nju.edu.cn
Communications Earth & Environment Article in Press
1234567890():,;
1234567890():,;

achieving synergies between the two domains. However, these studies have
primarily relied on cross-sectional analyses18, offering limited insights into
how these trade-offs evolve over time and alongside different stages of
socioeconomic development. Understanding such dynamics is key to
operationalizing and monitoring sustainable food system transformations19.
Moreover, significant disparities persist between urban and rural areas in
China regarding food availability, purchasing power, health awareness, and
policy support20–22. Such disparities may result in asynchronous or uneven
dietary transitions between urban and rural areas across regions and
populations, thereby complicatin g the design of effective policy
interventions2. Yet, current research and policy frameworks lack adequate
f o c u so nu r b a n–rural disparities. These ga ps constrain a comprehensive
understanding of sustainable dietary transition pathways and hinder the
development of context-specific, targeted strategies.
In this study, we address three questions on China’s dietary transition:
how diet-related environmental pressures and health burdens differ
between urban and rural populations; how environment and health trade-
off patterns evolve across developmen t stages; and which interventions
across production, supply chains, and consumption could support a more
sustainable food system. To guide the analysis, we construct a panel dataset
for 31 provinces over 1990–2021, covering the per capita consumption of
ten major food groups (grains, vegetables, fruits, pork, beef, lamb, poultry,
fish, eggs, and milk) among urban and rural residents. We translate food
consumption into consumption-based environmental footprints (GHG
emissions, cropland use, blue water u se, nitrogen and phosphorus appli-
cation) using life cycle assessment (LCA)23, and estimate disability-adjusted
life years (DALYs) from three key diet-related non-communicable diseases
(cardiovascular diseases, type 2 diabetes, and neoplasms) using the Global
Burden of Disease (GBD) comparative risk assessment24.B u i l d i n go nt h e s e
assessments, we characterize the evolving environment–health trade-offs of
dietary change and investigate how their trajectories vary with regional
development levels, as reflected by the Human Development Index (HDI)25.
Finally, we explore policy-relevant options across the food system by eval-
uating scenarios spanning production ef ficiency (precision agriculture),
supply-chain loss reduction (food-waste reduction), and demand-side shifts
(dietary adjustments), and assess thei r potential to deliver simultaneous
environmental and health gains. Findings from these analyses are presented
in three parts in the Results. The Discussion then synthesizes broader
implications, summarizes robustnesschecks, and acknowledges limitations,
while methodological details are provided in the Methods. An overview of
the study design and analytical workflow is shown in Supplementary Fig. 1.
Our analysis reveals that dietary transitions in China have evolved from
initial health and environmental syner gies toward increasing trade-offs,
with health gains accompanied by rising environmental pressures. These
trade-offs are more pronounced in pr ovinces with higher human devel-
opment levels and differ between ur ban and rural populations. Urban
populations experienced faster reductions in health burdens but greater
increases in environmental pressures, whereas rural populations achieved
slower health improvements while continuing to bear high environmental
burdens. Scenario analyses further indicate that integrated strategies com-
bining precision agriculture, food-waste reduction and dietary optimization
could simultaneously reduce healt h and environmental burdens while
narrowing urban–rural inequalities. Ourfindings advance understanding of
how health–environment trade-offs evolve during dietary transitions across
urban and rural China, highlighting t he need for development-sensitive,
regionally tailored strategies to promote a more sustainable and equitable
food-system transition.
Results
Diet-related environmental and health burdens in urban and
rural areas
From 1990 to 2021, dietary transitions in both urban and rural China were
accompanied by complex and divergent changes in environmental and diet-
related health burdens. Rather than exhibiting uniform increase or decrease
trends, both environmental and health burdens showed nonlinear
dynamics, with notable disparities between urban and rural areas in both
magnitude and trajectory.
To assess environmental pressure, we calculated five diet-related
environmental footprints, including GHG emissions, cropland use, blue
water use, and nitrogen and phosphorus application, and normalized them
by per capita planetary boundary th resholds for cross-dimensional
comparison26 (Methods). As shown in Fig. 1a, alongside shifts in dietary
patterns, four of the five categories of environmental pressure associated
with urban food consumption (cropl and use, blue water use, nitrogen
application, and phosphorus application) exhibited U-shaped trajectories,
with clear declines around 2000, followed by sharp increases through 2021.
Specifically, per capita dietary cropland use, blue water use, nitrogen
application, and phosphorus application in urban areas decreased from
822.45 m², 101.71 m³, 6.90 kg, and 1.12 kg in 1990 to 682.1 m², 82.08 m³,
5.81 kg, and 0.96 kg in 2000, before rising again to 956.08 m², 112.62 m³,
7.84 kg, and 1.29 kg, respectively, by 2021. In contrast, dietary GHG emis-
sions showed a steady rise over the entire period, from 279.40 kg CO₂-eq per
capita in 1990 to 416.84 kg CO ₂-eq per capita in 2021. By the end of the
study period, environmental pressures from urban diets had exceeded their
1990 levels across allfive dimensions. Rural environmental burdens showed
a different dynamic. Here, most envir onmental indicators also exhibited
broadly U-shaped trends, with modest declines around 2010 but overall
variation remaining within a narrower range than in urban areas. Moreover,
rural diets consistently carried higher burdens in cropland use, blue water
use, and nitrogen and phosphorus application, while urban diets had higher
GHG emissions. In 2021, per capita cropland use, blue water use, nitrogen,
and phosphorus application in ru ral diets were 1081.50 m², 126.41 m³,
8.65 kg, and 1.40 kg, respectively—all higher than their urban counterparts
and remaining close to their 1990 values. This persistence suggests that while
urban environmental burdens have increased markedly in recent years,
rural diets have long maintained structurally high environmental pressure.
To assess the robustness of the temporal patterns, we perturbed food-group
consumption values by ±10% and propagated uncertainty via Monte Carlo
simulation (100 replicates) (Methods). The results indicate that the main
temporal changes observed in both urban and rural diets are generally
robust under the uncertainty framework used. Supplementary Table 1
summarizes adjacent-period differences using the median, 95% uncertainty
interval, empiricalP value, and trend support, while Supplementary Fig. 2
visualizes the corresponding simulated trajectories.
The observed urban-rural disparities in magnitude and trajectory of
environmental pressures primarily reflect differences in dietary composi-
tion and quantities, as estimated under the footprint accounting approach
adopted here. While both urban and rural populations have moved toward
more diverse diets since 1990, rural diets remain dominated by staple grains
(more information on urban and ru ral dietary patterns can be found in
Supplementary Fig. 3), which account for substantial shares of land, water,
and nutrient footprints. As shown in Fig. 1b, the breakdown of rural
environmental burdens by food cate gory indicates that staple grains
accounted for 58.59% of cropland use, 62.87% of blue water consumption,
51.08% of nitrogen footprints, and 47.78% of phosphorus footprints in 2021.
In contrast, urban diets feature higher proportions of red meat, milk, and
eggs, which have considerably higherGHG intensities per unit weight and
correspond to relatively larger dietary GHG emissions under the footprint
accounting framework. In 1990, rural residents consumed 122.82 g more
staple grains per capita than urban residents; under the footprint accounting
approach adopted here, this differ ence accounted for 76.10%, 75.91%,
71.92%, and 69.62% of the urban –rural gaps in cropland, blue water,
nitrogen, and phosphorus footprints, respectively. By contrast, higher urban
consumption of animal-source food s, particularly red meat (10.28 g),
accounted for 58.02% of the observed urban –rural gap in dietary GHG
emissions. Similar patterns persistedin 2021: the higher rural intake of staple
grains (44.59 g) accounted for 77.42%, 73.41%, 73.74%, and 70.29% of the
observed urban–rural gaps in cropland, blue water, nitrogen, and phos-
phorus footprints, whereas the higher urban intake of red meat and milk,
with differences of 3.41 g and 9.41 g, respectively, accounted for 64.81% and
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

20.44% of the observed urban –rural gap in dietary GHG emissions,
respectively. Additional details are presented in Supplementary Table 2. The
detailed contributions of major food groups to each environmental foot-
print for urban and rural areas can be found in Supplementary Figs. 5 –9.
Notably, phosphorus use in both urban and rural diets has already surpassed
the sustainable threshold of the Earth system (as shown by the dashed line in
Fig. 1a), and nitrogen use is rapidly approaching it, raising urgent concerns
for long-term ecosystem health. Thesefindings underscore the unsustain-
ability of current dietary patterns in terms of resource utilization.
In terms of health impacts, we examined three major non-
communicable diseases (cardiovascular diseases, type 2 diabetes, and neo-
plasms) linked to five key diet-related risk factors (Methods). Diet-related
disease burdens—measured by disability-adjusted life years (DALYs)—also
evolved in a nonlinear and spatia lly heterogeneous manner (Fig. 1c).
Overall, urban populations experien ced consistently lower diet-related
health risks than their rural counterparts. DALYs declined over time in both
urban and rural areas, but the magnitude and composition of these changes
varied substantially. From 1990 to 2021, urban diet-related DALYs per 1000
population decreased by approximately 38.01%, from 60.41 DALYs to 37.45
DALYs, while rural DALYs fell by only 28.24%, from 63.58 DALYs to 45.62
DALYs. The most substantial improvements were linked to increased fruit
consumption. During the study period, average fruit intake in urban diets
rose by 45.72 g /day, leading to reductions of 20.74, 2.58 and 0.99 DALYs per
1000 population for cardiovascular diseases, neoplasms, and type 2 diabetes,
respectively. Rural improvements from fruit intake were smaller, with
reductions of 16.22, 1.56, and 0.16 DALYs per 1000 population for the same
disease categories. Meanwhile, burdens associated with low vegetable intake
initially increased and then declined, but the net trend diverged between
urban and rural: urban diet-related D ALYs attributable to low vegetable
intake rose by1.34 DALYs per 1000 population, whi le rural DALYs
decreased by 0.47 per 1000 population from 1990 to 2021. Health risks
related to high red meat consumption increased steadily, with a more
pronounced relative rise in rural areas. DALYs attributable to a diet high in
red meat in rural populations more than doubled—from 0.55 to 1.17 per
Fig. 1 | Temporal trends and food group contributions to diet-related environ-
mental footprints and health burdens in urban and rural China from 1990
to 2021. a Urban-rural comparison of five diet-related environmental footprints
relative to planetary boundaries from 1990 to 2021; b Contributions of major food
groups to each environmental footprint in 2021; c Number of DALYs per 1000
population attributable to individual dietary risk factors in urban and rural areas
over time. Source data for a–c are provided as Supplementary Data 1.
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

1000 population—between 1990 and 2021, nearly matching urban levels,
which rose from 1.00 to 1.17. Burdens associated with low milk intake
declined modestly by approximately 0.17 per 1000 population and similarly
in both urban and rural areas, while health impacts from insufficient whole
grain intake remained minimal and largely unchanged. Using the Monte
Carlo perturbation framework (Methods), the overall decline in diet-related
DALYs remained robust in both urba n and rural populations. Supple-
mentary Table 3 summarizes uncertainty in period-to-period differences
(median and 95% UI) together with empiricalP value, and trend support,
while Supplementary Fig.9 visualizes the corresponding simulated trajec-
tories for the three disease categories.
Environmental and health risk trade-off patterns across
developmental stages
To better understand the trade-offs between diet-related environmental and
health burdens, we established a four-pattern classi fication framework
(Win–Win, Win –Lose, Lose –Win, Lose –Lose) to track the dynamic
changes and trade-off trajectories of all provincial units from 1990 to 2021
(see Methods). Specifically, Win–Win indicates simultaneous reductions in
both environmental and health burdens, whereas Win –Lose re flects
reduced environmental pressure but increased health burden, Lose–Win
reflects increased environmental pressure but improved health, and
Lose–Lose denotes worsening of both. Although urban and rural areas
eventually converged toward a similar Lose–Win dominant pattern by 2021,
they followed distinct transition pa thways and progressed at different
paces (Fig. 2).
In urban areas (Fig. 2a, c), the early stage (1990 –2000) was char-
acterized by a combination of Win –Win (71% of urban areas) and
Win–Lose (29%) patterns. This indicates that while diet-related environ-
mental burdens declined across all urban areas during this period, health
burdens decreased in most, but not all provinces. Notably, in southwestern
and northwestern provinces —including Chongqing, Sichuan, Yunnan,
Shaanxi, Qinghai, and Ningxia —urban diet-related health burdens
increased rather than declined, reflecting uneven progress in dietary health
improvement across regions. Between 2000 and 2010, a signi ficant shift
occurred: over 45% of urban areas previously exhibiting the Win –Win
pattern shifted to the Lose–Win pattern, and more than 66% of urban areas
characterized by the Win–Lose pattern also transitioned to the Lose–Win.
This period was marked by a continued decline in health risks across the vast
majority of urban areas, while environmental risks started to exhibit an
upward trend, indicating a growing trade-off between environmental and
health outcomes. Between 2010 and 2021, all urban regions that initially
followed the Win–Win pattern had transitioned to the Lose–Win state. The
Lose–Win pattern has expanded to encompass 94% of urban regions,
emerging as the predominant trade-of f mode. This sequential transition
reveals a clear path dependence: early re d u c t i o n si ne n v i r o n m e n t a lb u r d e n s
may have exhausted the most accessible opportunities for joint improve-
ments, making it harder to sustain co-benefits and leading to a shift toward
improving health outcomes even as environmental pressures began to
rise again.
In contrast to the more synchronized and earlier transitions observed
in urban areas, rural areas exhibited generally higher exposure to both
environmental and health risks as well as a more fragmented and delayed
pattern of change (Fig. 2b, d). During the 1990 –2000 period, rural areas
showed a dispersed distribution across all four trade-off states: Win–Win
(23%), Win–Lose (29%), Lose–Win (34%), Lose–Lose (12%). Between 2000
and 2010, approximately 68% of rural areas, encompassing most regions in
East, South, North, Northeast, and Northwest China, transitioned into the
Win–Win state. This shift suggests a temporary decoupling of diet-related
environmental and health burdens and reflects a catch-up phase following
the earlier improvements observed inurban regions. After 2010, rural areas
began to follow a transition trajectory similar to that of urban areas, with a
rapid convergence toward the Lose–Win pattern. Between 2010 and 2021,
about 90% of Win –Win and 80% of Win –Lose rural regions shifted to
Lose–Win, resulting in Lose–Win becoming the predominant state with a
prevalence of 87%. These trends underscore a key rural–urban divergence:
although both settings experienced a shift from co-benefits to trade-offs over
time, rural areas entered this transition later and remained at higher levels of
health and environmental burdens for longer. This temporal lag highlights
the need for regionally differentiated policy interventions that account for
the distinct transition trajectories andlocal constraints in rural versus urban
dietary-environment-health systems.
Building on the distinct urban–rural transition trajectories observed
above, we further explored whether regional development levels shape the
evolution of dietary environmental –health risk trade-offs. We used the
Human Development Index (HDI) as a composite indicator to assess the
development levels of urban and rural areas (see Methods). We applied
change-point detection to identify structural breakpoints in the relationship
between HDI and the four trade-off patterns (Win–Win, Win–Lose, Lose-
Win, and Lose-Lose) (see Methods). We identified four distinct HDI stages
separated by three breakpoints (0.36, 0.44, and 0.55), which we used to
delineate four HDI stages (Fig. 3). To formally assess whether trade-off
category compositions differ acrossthese HDI groups, we then constructed
a contingency table of HDI stages and trade-off categories and applied a chi-
square test of independence (Supplementary Table 4). Using a Monte Carlo
procedure with 10,000 replicates to obtain thep-value, the chi-square test of
independence yielded a test statistic of 73.10 with a simulated p < 0.001,
indicating that the composition of trade-off categories differs significantly
across HDI groups rather than being randomly distributed. The strength of
this association was further quantified using Cramér’sV ,w h i c hw a s0 . 3 9
based on 162 observations, suggesting a moderately strong association
between development stage and trade-off pattern composition.
We identified four distinct HDI stages associated with different trade-
off con figurations (Fig. 3). At the lowest development stage
(0 < HDI≤ 0.36), no dominant trade-off pattern emerged. Regions, were
relatively scattered across the four quadrants, re flecting heterogeneous
changes in both environmental and health risks. Speci fically, 32.3% of
regions fell into the Win–Win pattern, 25.8% into the Lose–Win pattern,
29.0% into the Win–Lose pattern, and 12.9% into the Lose–Lose pattern.
This diversity suggests that in early development phases, dietary systems had
not yet followed a consistent trajectory toward either improvement or
deterioration. Notably, most regions at this stage were rural, indicating that
urban areas had already moved beyond this lagging stage of development.
At the second development stage (0.36 < HDI ≤ 0.44), trade-off patterns
between environmental and health risks began to converge, with 57.5% of
regions achieving the Win–Win category. This convergence was observed in
more than half of the regions across both urban and rural areas, representing
a relatively short decoupling window during which development appeared
to enable synergistic dietary bene fits. At the third development stage
(0.44 < HDI≤ 0.55), trade-off patterns further shifted toward the two
categories characterized by declining health risks. Speci fically, 53.8% of
regions fell into the Lose–Win pattern and 35.9% into the Win–Win pat-
tern. This indicates that, at this stag e, most regions reduced health risks
through dietary improvements, while environmental risks began to rise,
especially in rural areas, where this trend was more pronounced. At the
highest development stage (0.55 < HDI ≤ 1), trade-off patterns became
highly skewed, with 82.7% of regions falling into the Lose–Win category,
reflecting a widespread pattern of he alth gains accompanied by rising
environmental burdens. Notably, almostall regions at this stage were urban,
indicating that rural a reas had not yet reach such advanced levels of
development. Together, these findings underscore a development-linked
transition in dietary trade-offs: from heterogeneity and uncertainty at low
HDI levels, to a brief window of synergistic improvement, followed by a
widening health –environment trade-off—especially in urban contexts.
These results call for urgent policy interventions that promote sustainable
diets capable of advancing health wit hout exacerbating environmental
pressures, particularly in the face of China ’s ongoing socioeconomic
transformation.
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

Scenario analysis for reducing diet-related environmental and
health risks
Building on the observed urban –rural disparities and development-stage
regularities in environment–health trade-offs, we use scenario analysis as an
exploratory, policy-relevant tool to examine whether interventions at key
leverage points across the food system can reduce environmental pressures
and diet-related health burdens. These scenarios represent actionable
interventions across key stages of the food supply chain spanning produc-
tion efficiency, supply-chain loss reduction, and demand-side dietary shifts,
with the aim of informing holistic a nd regionally adaptive food policy
discussions27,28.S p e c ifically, we considered three individual scenarios: (1)
precision agriculture during food production29, designed based on a meta-
analysis of existing literature assessin gi t si m p a c to ne n v i r o n m e n t a lf o o t -
prints; (2) a 50% reduction in food waste at the food supply stage, aligned
with Sustainable Development Goal targets 30; and (3) optimized dietary
shifts designed to minimize environmental and health burdens, while
satisfying dietary acceptability and meeting intake recommendations from
the Chinese Dietary Guidelines (CDGs) (see Methods). The third scenario
was developed using a non-dominated sorting genetic algorithm (NSGA-II)
to achieve multi-objective optimization31. These scenarios were selected not
only for their policy individual relevance but also for their complementarity
when applied in combination, which may yield additional synergistic gains.
Individually, each scenario demonstrated measurable reductions in
environmental risks, with varying degrees of effectiveness across environ-
mental dimensions. The combined scenario generally produced the largest
and most balanced overall reductions (Fig. 4a). Precision agriculture is
projected to reduce blue water use by 8.4%, cropland use by 2.7%, GHG
emissions by 5.0%, nitrogen use by 13.4%, and phosphorus use by 8.2% in
urban areas. Similar reductions are observed in rural areas: 6.2% (blue water
use), 2.6% (cropland use), 5.4% (GHG emissions), 12.0% (nitrogen use), and
7.5% (phosphorus use). Among these, nitrogen use reduction is particularly
notable, underscoring the strong potential of precision agriculture to
Fig. 2 | Environmental and health risk trade-offs associated with dietary changes
in urban and rural from 1990 to 2021. Normalized environmental pressure levels
(0–1) and diet-related health risks (DALYs per 1000 population) across provinces in
urban (a) and rural ( b) areas from 1990 to 2021. Temporal transitions of
environmental–health risk trade-off states (Win –Win, Win–Lose, Lose–Win, and
Lose–Lose) across three time periods (1990–2000, 2000–2010, 2010–2021) for urban
(c) and rural (d) areas. Provincial abbreviations are used in thefigure; a complete list
of abbreviations for all 31 provincial units is provided in Supplementary Table 5.
Source data for a–d are provided as Supplementary Data 2.
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

optimize nutrient management and mitigate agricultural non-point source
pollution. In the food waste reduction scenario, all five types of environ-
mental footprints declined consistently in both urban and rural areas, with
reductions ranging from 5.3% to 9.3%.This suggests that this measure can
achieve balanced improvements across multiple environmental indicators.
Dietary shifts showed the most substa ntial reductions in environmental
footprints across most dimensions. In urban areas, this scenario reduces
blue water use, land use, nitrogen application, and phosphorus application
by 18.4%, 20.2%, 20.3%, and 15.3%, respectively. The effects were even more
pronounced in rural areas, with corresponding reductions of 27.9%, 29.2%,
27.2%, and 21.4%. However, its impact on GHG emissions remains rela-
tively limited. This outcome arises because healthier diets often include
more fruits, vegetables, and milk products, which have relatively high GHG
emission intensities per kilocalorie or per gram of protein compared to
staple grains. Overall, the combined scenario achieved the greatest reduc-
tions across most environmental dimensions. In particular, it produced the
largest improvements in blue water, cropland, nitrogen, and phosphorus
use. This pattern suggests that coordinated interventions across production
efficiency, supply-chain loss reduction, and demand-side dietary shifts can
generate complementary environmental benefits.
In contrast to the environmental outcomes, health improvements were
driven primarily by the dietary shift scenario, as the precision agriculture
and food waste reduction scenarios assumed unchanged food consumption.
A ss h o w ni nF i g .4b, c, dietary transition scenario led to substantial
adjustments in dietary structure. Compared to 2021 baselines, both urban
and rural areas exhibited major reductions in staple food intake (109.00 g/
day/person and 185.51 g/day/person, respectively), accompanied by
increases in the intake of fruits, vegetables, milk,fish, and eggs. Milk intake
increased the most, reaching 244.59 g/day/person and 269.34 g/day/person
in urban and rural areas, reflecting improved protein and calcium intake.
Vegetable and fruit consumption rose by 28.10 g/day/person and 31.76 g/
day/person in urban areas, and by 39.18 g/day/person and 54.95 g/day/
person in rural areas, respectively. Modest increases were also observed in
fish and egg consumption, with urban increases of 11.75 g/day/person and
9.03 g/day/person, and rural increases of 16.82 g/day/person and 10.20 g/
day/person. Meanwhile, the intake of animal-based foods such as red meat
and poultry declined, indicating a shif t toward healthier protein sources.
Detailed recommendations tailored to each urban and rural area are pro-
vided in the Supplementary Information (Supplementary Table 6). These
dietary adjustments indicate a clear shift toward healthier and more
balanced dietary patterns, particularly in rural areas where the changes were
more pronounced, thereby lowering the burden of major diet-related dis-
eases. For example, DALYs attributableto cardiovascular diseases declined
by 5.19 per 1000 persons in urban areas and by 11.07 per 1000 persons in
rural areas, primarily due to increas ed fruit consumption. Likewise, the
burden associated with neoplasms decreased by 2.22 and 3.43 DALYs per
1000 persons in urban and rural areas, respectively. Moderate improve-
ments were also observed for type 2 diabetes, particularly in rural popula-
tions (2.44 DALYs per 1000 persons), compared to urban populations (1.53
DALYs per 1000 persons), largely driven by reductions in red meat intake
(Fig.4d, e).
Importantly, the combined scenario achieved simultaneous reductions
in both environmental and health risks across urban and rural areas, with
larger relative benefits observed in rural regions (Fig.5a ) .O na v e r a g e ,h e a l t h
risks in rural areas declined by 31.10%, which was notably higher than the
20.25% reduction observed in urban areas. Similarly, the average reduction
in environmental risks reached 37.44% in rural areas, exceeding the 32.58%
reduction in urban areas. A similar pattern was observed across regions with
different development levels. Reg ions with lower HDI scores tended to
achieve greater reductions in environmental and health risks. For example,
in regions with HDI≤ 0.55, diet-related environmental risks are projected to
decline by 36.90% and health risks by 33.06%, whereas in regions with
HDI > 0.55, the reductions are 33.55% and 19.97%, respectively.
We further find that the combined implementation of the three sce-
narios may help reduce spatial dispariti e si nd i e t - r e l a t e dr i s k s .A ss h o w ni n
Fig. 5b, in 2021, rural areas with relatively low HDI values generally faced
higher levels of diet-related enviro nmental and health risks than urban
areas, and considerable heterogeneity in these risks was observed across
Fig. 3 | Shift in diet-related environmental–health
risk trade-off patterns across different Human
Development Index (HDI) stages in urban and
rural areas. Scatter plots show changes in health and
environmental risks for urban (triangles) and rural
(circles) areas across four HDI intervals. Each
quadrant represents a trade-off state: Win –Win
(WW, blue), Win –Lose (WL, yellow), Lose –Win
(LW, green), and Lose –Lose (LL, red). Percentages
indicate the proportion of population falling into
each quadrant within each HDI stage. Source data
for Fig. 3 are provided as Supplementary Data 3.
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

regions. Following the implementation of the scenarios, both the overall
levels and regional disparities in diet-related environmental and health risks
were reduced across areas with varying HDI scores. Scenario-speci fic
comparisons further showed that this reduction in heterogeneity was not
uniform across interventions, but was driven primarily by the dietary shift
scenario. By contrast, implementing precision agriculture or food waste
reduction alone had relatively limited effects on the urban–rural risk gap
(Fig. 5c). For environmental risk, the urban–rural gap changed little under
precision agriculture (0.0502, compared with 0.0493 at baseline) and
declined only modestly under food waste reduction (0.0418), whereas it was
reduced sharply under dietary shifts (0.0022) and reached its lowest level
under the combined scenario (0.0016). An even clearer contrast was
observed for health risk: the urban–rural gap remained unchanged under
precision agriculture and food waste r eduction (both 502.7, identical to
baseline), but fell dramatically to 1.78 under both dietary shifts and the
combined scenario. Taken together, these results suggest that coordinating
interventions across multiple stages of the food system not only broadens
and strengthens their overall impact, but can also help narrow urban–rural
and regional gaps in diet-related risks, thereby supporting a transition
toward more balanced, efficient, and sustainable dietary systems.
Discussion
This study systematically assessed t he evolution of trade-offs between
environmental and health burdens associated with dietary transitions in
China from 1990 to 2021, with a particular focus on urban-rural disparities
and socioeconomic development levels. In addition, we conduct scenario
analysis as an exploratory, policy-relevant“what-if” assessment to illustrate
how interventions at key leverage points across production, supply chains,
and consumption might shift trade-offs. Ourfindings reveal that although
dietary improvements have led to notable reductions in health risks, par-
ticularly in urban areas, these gains are frequently accompanied by
increasing environmental burdens. Rural areas, in contrast, have experi-
enced slower improvements in diet-related health outcomes while shoul-
dering disproportionateenvironmental pressure, highlighting a persistent
dual burden. These urban–rural disparities underscore the need for dietary
transition strategies that are both geo graphically and so cioeconomically
tailored. In addition, our analysis ind icates that as socioeconomic devel-
opment improves, diet-related environmental degradation may intensify,
even as health outcomes continue to improve. This pattern deviates from the
ideal “win–win” trajectory in which health benefits are achieved alongside
reduced environmental impacts. Building on these findings, the scenario
analysis indicates that integrating precision agriculture, food waste reduc-
tion, and dietary optimization is associated with concurrent reductions in
environmental pressures and diet-related health risks in both urban and
rural populations, and may facilitate a shift toward a win –win trajectory,
while narrowing urban–rural disparities in diet-related risk exposure.
From a policy perspective, our scenario analysis shows that dietary
shifts are among the most powerful i ntervention pathways, delivering
simultaneous environmental and health co-benefits. However, the distinct
pathways of dietary transition observed across urban–rural settings suggest
that the main challenges, practical con straints, and policy priorities vary
across development contexts and local contexts. China’s future food system
Fig. 4 | Environmental and health effects of food intervention scenarios.
a Percentage reduction in five environmental footprints under three individual
scenarios and combined scenario; b, c comparison of adjusted and baseline dietary
patterns in urban and rural populations; d, e changes in DALYs in urban and rural
areas under dietary shift scenarios. Source data for a–e are provided as Supple-
mentary Data 4.
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

transition should therefore place dietary improvement at its center, sup-
ported by complementary measures such as precision agriculture and food
waste reduction, to advance more targ eted and integrated intervention
pathways.
First, as for dietary adjustment, policy priorities should be differ-
entiated across development stages.In areas with lower levels of develop-
ment, particularly rural areas, dietary choices are largely constrained by food
availability, poor market access, an d underdeveloped infrastructure32.I n
these contexts, the policy priorities in these areas should focus on improving
the supply, affordability, and accessibility of healthy foods, such as vege-
tables, fruits, and legumes, through targeted agricultural subsidies,
improvements to rural logistics and distribution infrastructure33.I na r e a s
undergoing rapid development, where diets are becoming more diversified
but health outcomes remain uneven, the main task is to steer dietary
transition in a healthier and more sustainable direction. Public nutrition
campaigns, school meal reforms, and supportive food environments should
be strengthened to promote the EAT-Lancet recommendations and the
Chinese Dietary Guidelines, and to guide dietary transitions toward heal-
thier and more sustainable consumption patterns34,35.I na r e a sw i t hh i g h e r
consumption levels and more resource-intensive diets, the main challenge
shifts toward the combined environmental and health pressures associated
with overconsumption, high intake o f animal-based foods, and elevated
levels of food waste. At this stage, dietary adjustment should focus more
directly on reducing excessive meat intake, promoting plant-forward diets,
lowering dependence on highly processed foods36.
Second, food waste reduction should be promoted as an important
complementary strategy to reduce unnecessary resource losses in produc-
tion, distribution and consumption stages. In China, future efforts to
manage and reduce food waste should place greater emphasis on the inte-
gration of technologies such as artificial intelligence, the Internet of Things
(IoT), and blockchain to improve ef ficiency along food production and
transport chains, enhance packaging and storage systems, and minimize
food spoilage and physical loss. For example, IoT-based sensors deployed in
storage facilities, refrigeration units, and transport vehicles can enable real-
time monitoring of temperature, humidity, and other environmental con-
ditions that affect food quality37.A r t ificial intelligence can further support
more efficient inventory management by using data analytics and machine
learning to improve tracking, forecasting, and stock optimization 38.I n
addition, blockchain can strengthen traceability across the supply chain by
recording information on product origin, production methods, transport
routes, and storage conditions from farm to fork 39. On the consumption
side, stronger education and media campaigns on food waste reduction,
together with interventions such as adjusting plate sizes in public catering
settings, could also help red uce household food waste 40.P i l o ti m p l e -
mentation of these measures may be especially valuable in more highly
developed regions, where higher consumption levels are often accompanied
by greater waste generation.
Third, precision agriculture can serve as an important production-side
strategy for reducing environmental burdens by improving fertilizer and
water-use efficiency, lowering nutrient losses, and reducing the environ-
mental intensity of food production. Specifically, precision agriculture can
contribute through three main channels: (1) improving field-level mon-
itoring and data collection using tools such as GPS, sensors, and drones41;( 2 )
enhancing decision-making for planting, irrigation, fertilization, and pest
management through data analytics and intelligent systems 42;a n d( 3 )
enabling more targeted input use and resource allocation to reduce waste
Fig. 5 | Combined scenario effects on environ-
mental and health risks in urban and rural areas.
a Combined effects of the three scenarios on pro-
portional reductions in environmental and health
risks in urban and rural areas; b distributions of
environmental and health risks in urban and rural
areas before and after scenario implementation;
c urban–rural differences in normalized risk under
different scenarios. Both environmental and health
risk indicators are normalized to the 0 –1 range and
are dimensionless. Source data for ( a–c) are pro-
vided as Supplementary Data 5.
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

while maintaining productivity43. However, the large-scale adoption of
precision agriculture may be constrained by equipment costs, technical
training requirements, digital infrastructure gaps, and the prevalence of
smallholder farming systems44, especially in less-developed areas. To avoid
widening regional disparities in environmental performance through
uneven technological uptake, and to r educe the risk that less-developed
areas become more dependent on food sourced from other regions, thereby
shifting environmental pressures elsewhere, practical efforts are needed to
address these constraints, including improving rural digital infrastructure,
strengthening technical training and agricultural extension services, low-
ering adoption costs through targeted subsidies or cooperative service
platforms, and promoting technology solutions better suited to smallholder
production systems.
Because the environmental pressures associated with food consump-
tion are not necessarily borne where consumption occurs, to further char-
acterize this spatial mismatch, we conducted a supplementary analysis using
a recent dataset of interprovincial food tradeflo w si nC h i n a45. Specifically,
we decomposed provincial urban and rural food consumption by source
region and derived a net environmental burden indicator together with the
locally borne share of consumption-related environmental footprints. A net
environmental burden indicator, de fined as the difference between the
footprint outsourced from local consumption to other provinces and the
footprint borne locally for consumption in other provinces, to identify
whether a province functioned as a netoutsourcing or net burden-bearing
region. The locally borne share of co nsumption-related environmental
footprints, which captures the exten t to which environmental pressures
associated with local food consumption are retained within the consuming
province.
This supplementary analysis indicates a clear spatial redistribution of
diet-related environmental burdens, and this pattern is broadly consistent
across blue water, cropland, greenho use gas, nitrogen, and phosphorus
footprints. Overall, Henan, Shandong, Hebei, the three northeastern pro-
vinces, and several other northern provinces tend to emerge as net burden-
bearing regions across multiple footprint types, indicating that these areas
not only bear the environmental pre ssures associated with local con-
sumption but also absorb substantial burdens driven by food demand
elsewhere. By contrast, Beijing, Shang hai, Zhejiang, Tianjin, and several
more developed provinces more oftena p p e a ra sn e to u t s o u r c i n gr e g i o n s ,
meaning that a relatively large share of the environmental burdens asso-
ciated with their food consumption is borne outside the consuming pro-
vince. The overall spatial pattern is broadly similar between urban and rural
areas, although the divergence between net burden-bearing and net out-
sourcing regions tends to be more pronounced in rural areas. In addition to
this cross-sectional pattern, temporalcomparison of the locally borne share
between 2000 and 2021 suggests that the coupling between consumption
and local environmental burden remained relatively stable in many pro-
vinces, but the magnitude of change was not uniform. Taken together, these
findings suggest that the environmen tal burdens associated with food
consumption are not necessarily borne where consumption occurs, but are
substantially redistributed through interprovincial trade. This pattern of
spatial outsourcing may also extend beyond provincial boundaries through
food imports, indicating that the e nvironmental consequences of food
consumption can be displaced across mu ltiple geographic scales. Conse-
quently, assessments of diet-related environmental costs based solely on the
consuming province may underestimate the pressures imposed on major
producing regions.
Importantly, these additional results do not materially alter the central
conclusions of this study. Although accounting for interprovincial trade
changes the spatial attribution of environmental burdens, the overall evo-
lution of the environment–health trade-offs remains broadly similar over
the overlapping period from 2000 to 2021. Speci fically, we estimated the
trade-adjusted locally borne environmental burdens for urban and rural
populations in each province and r ecalculated the trade-offs between
environmental and health burdens (Supplementary Fig. 16). Across both
2000–2010 and 2010 –2021, the aggregate transition pattern is largely
maintained, with the same broad trade-off structure still dominating in both
urban and rural areas. Overall, this s uggests that interprovincial trade
mainly reshapes where environmental pressures are borne rather than
overturning the main temporal pattern identified in our analysis. Further
methodological details and additional results are provided in Supplemen-
tary Text 1 and Supplementary Figs. 10–15.
Our results show good agreement with previous studies, reinforcing
the credibility of ourfindings. First, our health risk estimates are built upon
the GBD comparative risk assessment framework, and the ranking of major
diet-related risk factors generally aligns with GBD 2017 estimates 46.O n e
notably deviation involves low whole grain intake, which ranks relatively
lower in our analysis. This discrepancy is likely due to data limitations: our
proxy indicator (total grain consumption from theChina Statistical Year-
book) does not distinguish between whole and re fined grains. Secondly,
although our study includes a limited range of food categories, resulting in
lower absolute estimates of environmental footprints compared to similar
studies, the overall upward trend in d ietary environmental pressures is
consistent with prior research 18. Finally, the dietary transition scenarios
identified in this study closely align with internationally recognized sus-
tainable dietary models (e.g., the“Win–Win” diet) that emphasize increased
intake of plant-based foods and reduced consumption of red meats1,2.
Importantly, beyond confirming these widely observed patterns, our
study adds further value by uncovering pronounced urban–rural hetero-
geneities and the dynamic evolution of trade-offs along the Human
Development Index (HDI) gradient—dimensions that have received lim-
ited attention in earlier literature. This enables a more nuanced under-
standing of how dietary impacts shiftwith socioeconomic development and
supports more context-specific dietary system transformation strategies. In
sum, the intersection of urban–rural status and HDI level creates distinct yet
overlapping dietary transition pathways. Regionally adaptive and stage-
specific food system strategies are needed to seize health–environment co-
benefits early on, avoid unintended consequences in mid-transition, and
correct course when trade-offs emergein advanced stages of development.
This study provides forward-looking empirical insights into these chal-
lenges, emphasizing the need for vigilance against the risk of dietary systems
becoming locked into a trajectory of dual environmental and health dete-
rioration amid rapid socioeconomic change. By retrospectively analyzing
three decades of dietary transitions and their associated risks in urban and
rural China, our findings also offer valuable policy guidance and reference
for other developing countries undergoing similar nutrition transitions and
rapid urbanization.
Our main conclusions demonstrate strong robustness. To evaluate the
impact of input uncertainties on diet-related environmental burdens, this
study applied a Monte Carlo-based sensitivity simulation framework with
1,000 iterations. Environmental footprint intensities (including GHG
intensity, cropland use, blue water use, and nitrogen and phosphorus
application) were perturbed within a ± 30% range around their default
values (see Methods). The simulation results exhibited stable distributions
under all perturbation conditions . For example, in 2021, the simulated
medians were 0.625 for urban and 0.674 for rural areas, closely matching the
original estimates produced in this study—0.624 for urban and 0.674 for
rural, respectively (Supplementary Fig. 15). These values are standardized,
unitless measures representing the c omposite environmental footprint,
calculated as the average of the ratios offive environmental footprints to per-
capita planetary boundaries (see Methods). This indicates strong con-
sistency between the perturbed and non-perturbed results. Moreover, the
simulated distributions showed narrowinterquartile ranges across all years,
with IQR values typically between 0.03 and 0.05 (Supplementary Table 7),
indicating limited variability in the perturbation outcomes. The 95%
uncertainty intervals for environmental indicators across all years are
reported in Supplementary Table 8.
In addition, we conducted 1000 Mont e Carlo simulations incorpor-
ating uncertainties in dietary exposure distribution, relative risk estimates,
and baseline DALYs rates to obtain robust estimates and systematically
assess their impact on health risk assessment. The robustness of the DALYs
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

estimates was con firmed through these simula tions. Across various
uncertainty conditions, the simulated median DALYs for urban and rural
populations consistently aligned closely with the primary estimates across
all years, with relatively narrow interquartile ranges observed. For example,
in 2021, the simulated medians for urban and rural populations were 37.45
and 45.62 DALYs per 1,000 persons, respectively, closely matching the main
estimates of 38.23 and 46.26 DALYs per 1000 persons (Supplementary
Fig. 17, Supplementary Table 9). The 95% uncertainty intervals for all dis-
ease categories over the study period are provided in Supplementary
Table 10. These findings indicate that parameter variability has limited
impact on the results, supporting the stability and reliability of the temporal
trends and urban-rural differences identified herein.
Some limitations of this study shouldbe noted. First, due to limitations
in available data, the study does not cover all food categories consumed
across China’s diverse population. Long-term and disaggregated dietary
data across different genders, age groups, and income levels in both urban
and rural areas remain scarce. While this constrains a fully granular
assessment of dietary behaviors, th e food groups included in this study
represent the primary components of the Chinese diet and are sufficient to
capture the broad trends and trade-offs in health and environmental risks.
As more detailed consumption data become available, future research could
further examine within-population h eterogeneities and explore equity
implications in greater depth.
Second, our estimates of environmental impacts are based on stan-
dardized footprints coefficients from Springmann et al.47, a widely used and
policy-relevant dataset. However, these coefficients may not fully re flect
China-specific agricultural conditions or evolving production practices over
time. Chinese agriculture has underg one substantial reforms in fertilizer
management, including efforts to red uce chemical fertilizer inputs and
improve phosphorus use efficiency. As a result, global average coefficients
may overestimate the absolute magnitude of some environmental footprints
in China, particularly for nutrient-related indicators. This limitation is more
relevant to the interpretation of absolute footprint levels than to the relative
differences across dietary patterns. It is also important to emphasize that the
use of standardized global coefficients is a deliberate methodological choice
aimed at isolating the effects of dietary change itself and ensuring com-
parability with planetary-boundary-based targets, rather than capturing
country-specific production heterogeneity. Accordingly, the resulting esti-
mates are more informative for understanding relative patterns and trade-
offs associated with dietary shifts than for precisely capturing China’st i m e -
specific production-side environmental intensities. At the same time, we
acknowledge the value of developing more localized and time-varying life-
cycle assessment (LCA) data, and we encourage future studies to build on
this foundation as such datasets become more available.
Third, while the dietary scenarios developed in this study are informed
by multi-objective optimization to balance health and environmental out-
comes, they do not explicitly consider economic dimensions such as food
affordability, price elasticity, or co nsumer preferences. These factors are
likely to vary across regions and may b e particularly constraining in less-
developed rural areas. Consequently, the dietary scenarios identified here
should be viewed as theoretically opt imized scenarios, whose practical
feasibility at scale remains to be further evaluated. Future research could
integrate economic, behavioral, and institutional dimensions more expli-
citly to better evaluate the feasibility, transferability, and scalability of sus-
tainable dietary transitions.
Fourth, several sensitivity analyses were conducted to evaluate the
robustness of the main findings under alternative assumptions, including
the ±10% perturbation of food-group consumption values, the ±30% var-
iation in environmental footprint intensities, and alternative exposure dis-
tributions and coefficients of variation in the health risk assessment. These
analyses were intended to assess the sensitivity of the results to plausible
variations in key inputs and modeling choices, rather than to characterize
their true uncertainty distributions. Across these analyses, the mainfindings
remained generally stable, suggesting that the main conclusions are robust
to the assumptions examined in this study.
Finally, the diet-related environmental and health risks assessed in this
study primarily capture the risks driven by food consumption patterns and
the trade-offs between them, rather than the full range of production-related
environmental exposures and health burdens across regions. Although our
supplementary analysis usinginterprovincial food tradeflows provides an
initial indication of how consumption-related environmental burdens may
be redistributed across producing regions, it remains focused on the spatial
allocation of environmental burden and does not yet link these pressures to
actual exposure pathways and health outcomes in the affected populations.
This limitation is largely driven by t he lack of essential data and metho-
dological support. Robustly linking c onsumption-driven environmental
burdens to downstream health impacts would require geographically
explicit information on food production locations and systems, as well as
interprovincial food trade data that distinguish urban and rural demand. It
would also require pollutant-specific fate and transport models to capture
atmospheric and hydrological sp illovers, together with robust
exposure–response functions andfine-scale data on population distribution
and baseline disease burden 48. Without these essential foundations,
extending the analysis from environmental footprints to health losses would
likely introduce greater uncertainty and bias than additional interpretive
value. We therefore regard this as an important direction for future research
rather than a robust component of the present analysis.
Methods
Environmental analysis
Our analysis primarily focused on the environmental impacts associated
with ten major food groups: grains, ve getables, fruits, pork, beef, lamb,
poultry, fish, eggs, and milk. These food groups were selected because they
represent the main components of the Chinese diet and collectively account
for the majority of diet-related environmental footprints, including GHG
emissions, cropland use, blue water use, and the application of nitrogen and
phosphorus. We collected annual per capita dietary consumption data from
1990 to 2021 for urban and rural populations across all 31 provinces from
the China Statistical Yearbook49. To ensure a consistent assessment of the
environmental pressure associated with dietary patterns across urban and
rural areas in different regions, we a pplied global average environmental
footprints. This approach helps avoid discrepancies caused by variations in
technology levels and resource endowments and allows for comparison with
planetary boundaries, enhancing the global relevance and interpretability of
the results. Global average footprintsfor GHG intensity, cropland use, blue
water use, nitrogen use, and phosph orus use across 10 food groups are
obtained from the previous study47 and applied to assess the environmental
pressure (see Supplementary Table 11). Consistent with the referenced
study, the environmental footprints ofplant-based foods represent impacts
occurring at the farmland stage. The footprints of animal-based foods pri-
marily encompass feed-related impacts, including emissions from both feed
production and animal farming pr ocesses, with GHG emissions from
livestock additionally including a direct emission component. Footprints for
aquatic products ( fish and seafood) is mainly derived from the feed
demands and associated emissions of aquaculture, adjusted by the expected
share of wild-caught versus farmed production. The GHG intensity metric
specifically focuses on the major agricultural greenhouse gases, namely
methane (CH₄) and nitrous oxide (N₂O). Cropland use accounts for arable
land but excludes grassland use and the inputs of grass for ruminant
animals.
To quantify cumulative enviro nmental pressure, we compared five
environmental indicators against thei r respective planetary boundaries.
Each indicator was normalized to a 0–1 scale based on its proportion of the
planetary threshold, and thesefive normalized values were then averaged to
derive a composite measure of cumulative environmental pressure. The
planetary boundary framework de fines a “safe operating space ” for
humanity, within which Earth system processes remain stable50,51.A b o v e
planetary boundaries, it is suggested that ecosystem processes are at risk of
becoming destabilized51,52. To facilitate equitable comparisons of diet-
related environmental pressures across urban and rural regions, we adopted
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

per capita planetary boundary valuesfor GHG emissions, cropland use, blue
water use, and nitrogen and phospho r u sa p p l i c a t i o n .T h i sa p p r o a c h
accounts for population size differences and more accurately re flects
regional environmental impacts of dietary patterns. The per capita planetary
boundary data used in our analysis are derived from O ’Neill et al.26,w h o
downscaled global planetary boundaries to per capita values and compared
them with national consumption-basedenvironmental footprints for over
150 countries, including China. This dataset forms the basis for our envir-
onmental assessment. As no specificp e rc a p i t av a l u ef o rc r o p l a n du s ew a s
provided, we estimated it based on existing literature. According to Henry
et al.53, the global cropland planetary boundary is 2010 million hectares
(Mha). Using a global population estimate of 7 billion in 2011, consistent
with O ’Neill et al. 26, we derived a per capita cropland boundary of
approximately 2870 m² per person (see Supplementary Table 12).
DALYs analysis
We employed the GBD comparative risk assessment approach to estimate
the proportion of disease-specific burden attributable to each dietary risk
factor, also known as the population attributable fraction (PAF)54–56.T h i s
estimation was conducted across urban and rural areas in each of the 31
provinces over different years. The number of disability-adjusted life years
(DALYs) attributable to each dietary risk was then calculated by multiplying
the PAF by the total number of disease-speci ficD A L Y s .T oc l a r i f yo u r
analytical procedure, we followed three main steps:
First, we identified key dietary risk factors. Following the criteria used
in the GBD study, we selected dietary factors with suf ficient evidence of
causal relationships with c hronic disease endpoints54–56.W ef o c u s e do n
three major diet-related non-communicable diseases (NCDs)—cardiovas-
cular diseases, type 2 diabetes, and neoplasms. These diseases are associated
withfive key dietary risk factors: low intake of fruits, low intake of vegetables,
low intake of whole grains, low intake of milk and high intake of red meat.
Second, we collected data and estimated the population-level exposure
to each dietary risk. We collected a nnual per capita dietary consumption
data from 1990 to 2021 for urban and rural populations across all 31 pro-
vinces from the China Statistical Yearbook49. These consumption data were
harmonized to account for household food waste57 and the edible portion of
each food58, enabling estimation of per capita daily intake. Detailed infor-
mation about household food waste and the edible portion of each food item
can be found in Supplementary Table 13. Then we assumed that the intake
of each dietary factor follows a normal distribution and estimated the
population-level intake distribution for each dietary factor based on the
mean intake59,60.
Third, we calculated the PAF for each diet–disease pair using relative
risks (RRs) and theoretical minimum risk exposure levels (TMRELs). The
RRs per serving for each dietary component and disease outcome were
sourced from the GBD study, which is based on comprehensive
dose–response meta-analyses of prospective observational studies54 (Sup-
plementary Table 14). TMRELs, also from the GBD study, represent the
optimal intake levels associated with the lowest mortality risk or minimal
attributable burden at the population level56 (Supplementary Table 15). By
integrating exposure levels, RRs, and TMRELs, we estimated the PAF for
each diet-disease pair using formula (1). Formula (2) is used to estimate the
overall PAF when considering multiple independent risk factors, under the
assumption that there is no interaction among them61. Finally, we applied
disease-specific PAFs in conjunction with DALY estimates obtained from
the Institute for Health Metrics and Evaluation (IHME) Global Burden of
Disease Results Tool62 to calculate the total number of attributable DALYs
across all relevant outcomes using formula (3). To evaluate the overall
environmental health burden, we aggregated DALYs from all disease
categories.
PAFct ¼
R u
l RR xðÞ Pct xðÞ dx /C0 RRðTMRELÞR u
l RRðxÞPctðxÞdx ð1Þ
PAFoct ¼ 1 /C0
Yn
i
1 /C0 PAFioct
/C0/C1
ð2Þ
Total attributableDALYsct ¼
Xw
o¼1 DALYoctPAFoct ð3Þ
where l is the lowest level of intake and u is highest level of intake; i
represents theith dietary risk factor;r represents the region;t represents the
year; o represents the specificd i s e a s e ;w represents all relevant outcomes.
Environment-health trade-off analysis
To evaluate the trade-offs between en vironmental and health impacts of
different dietary patterns, we conducted an integrated analysis of changes in
diet-related environmental footprints (including GHG emissions, land use,
blue water use, and nitrogen and phosphorus use) and health burdens
(measured in DALYs) across urban and rural populations. Based on the
direction of changes in environmental and health risks, we established four
representative trade-off patterns: Win–Win (WW): reduced environmental
pressure alongside improved health outcomes; Lose –Environment/
Win–Health (LW): increased environmental pressure but improved health;
Win–Environment/Lose–Health (WL): reduced environmental pressure
but increased health burden; Lose –Lose (LL): increases in both environ-
mental pressure and health burden.
The level of socio-economic development is indicated by the Human
Development Index (HDI), developed by the United Nations Development
Programe (UNDP)25. HDI is a widely used composite indicator that inte-
grates income, education, and health to reflect the socio-economic devel-
opment level of a population. In environmental health research, HDI has
been commonly employed to explore di sparities in environmental expo-
sures and health burdens across countries and regions at different devel-
opment stages63. Although HDI includes a health component, it primarily
reflects general life expectancy rather than the specific diet-related health
outcomes considered in our analysis. Therefore, its use as a proxy for
development level is unlikely to bias the assessment of environment–health
trade-offs. In this study, HDI is used toe x p l o r eh o wt h et r a d e - o f f sa s s o c i a t e d
with dietary transitions vary across different stages of regional development.
To reflect regional development levels, we constructed a normalized
HDI for each urban and rural population across China’s3 1p r o v i n c e sf r o m
1990 to 2021. The HDI is a composite index capturing three key dimensions:
health (life expectancy at birth), education (mean years of schooling), and
income (per capita disposable income). All data were obtained from the
China Statistical Yearbook. To better reflect the variation and dynamics in
our data sample, we used sample-specific minimum and maximum values
observed in our dataset for norma lization. For each indicator x,t h e
dimension index was calculated as:
Indexx;i ¼ xi /C0 xminxmax /C0 xmin ð4Þ
where x denotes one of the three HDI components: health, education, or
income;xmin and xmax represent the minimum and maximum values of the
indicator x across all years and regions within the study sample; and i
indexes the year in the study period from 1990 to 2021.
Finally, the HDI was calculated as the geometric mean of the three-
dimension indices:
HDIi ¼ð IndexHealth;i × IndexEducation;i × IndexIncome;iIncome; iÞ1=3 ð5Þ
This normalized HDI reflects the relative development levels across
urban and rural populations in China over time, enabling a context-specific
analysis of its association with diet-related environmental and health
outcomes.
To examine potential structural changes in the relationship between
HDI and diet-related environmental–health trade-offs, we applied change
point detection to identify breakpoints along the HDI axis where the
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

composition of trade-off categories shifts. Specifically, wefitted a segmented
model across a grid of initial values to locate breakpoints and obtained three
HDI thresholds (0.357, 0.441, and 0.546), which define four HDI stages. To
formally test whether trade-off cate gory distributions differ across HDI
groups, we constructed a contingencytable crossing HDI group with trade-
off category and conducted a chi-square test of independence. Because some
expected cell counts may be small, we obtained thep-value via Monte Carlo
simulation with 10,000 replicates. We also calculated Cramér’sVf r o mt h e
chi-square statistic and sample size ofthe contingency table as an effect size
measure of the strength of association, and report both the contingency table
and Cramér’s V in Supplementary Table 4. The results indicated statistically
significant structural variation across HDI levels.
Scenario analysis
Several strategies have been identi fied to mitigate the environmental
impacts of food production. In our analysis, these strategies are categorized
into three main types: (1) improvements in agricultural production tech-
nologies and management practices, particularly through the adoption of
precision agriculture; (2) reductions in food waste, primarily driven by
behavioral changes at the consumer level; and (3) dietary transitions toward
healthier and more sustainable eating patterns.
Precision agriculture is a data-driven management strategy that inte-
grates spatial, temporal, and individual-level information to support site-
specific decisions, aiming to enhance input ef ficiency, productivity, and
sustainability in agricultural systems64. The scenario of precision agriculture
includes projected efficiency gains in GHG emissions intensities, land use,
water use, and nitrogen and phosphorus application. We conducted a meta-
analysis to quantify the environmental bene fits of precision agriculture
technologies in crop and livestock production, compared to conventional
business-as-usual (BAU) practices. The analysis coveredfive key environ-
mental indicators, including GHG emissions, cropland use, blue water use,
and nitrogen and phosphorus application.
The meta-analysis followed four key steps: (1) problem formulation,
(2) literature search, (3) data extraction, and (4) statistical analysis. In the
problem formulation step, this meta-analysis aimed to address the following
research question:“From a life cycle perspective, which precision agriculture
technologies effectively mitigate environmental impacts, and how do these
compare to conventional business-as-usual (BAU) practices in terms of
environmental performance?” Next, in the literature search step, a sys-
tematic search was conducted following the Preferred Reporting Items for
Systematic Reviews and Meta-Analyses (PRISMA) method65.T h es e a r c h
process included recording the total number of retrieved records; removing
duplicates; screening titles and abstracts; conducting full-text screening; and
finally including eligible studies for quantitative synthesis. A semi-inherited
meta-analysis approach was applied by incorporating 15 life cycle assess-
ment (LCA) studies extracted from a previous meta-review covering lit-
erature up to 202364. The search terms included: (“life cycle assessment” OR
“LCA”)A N D( “precision agriculture” OR “variable rate technology” OR
“VRT”)A N D(“agriculture” OR “crop production”). An additional search
was conducted to include studies published in 2024 and 2025 using the same
keywords, yielding 40 articles. After duplicate removal and title/abstract
screening, three articles remained, with one ultimately meeting the inclusion
criteria, increasing the total number of crop production-related studies to
16. To expand coverage to livestock systems, another search was conducted
in Web of Science using: ( “precision agriculture” OR “variable rate tech-
nology” OR “VRT” OR “precision farming” OR “Precision Livestock
Farming” OR “PLF”)A N D(“livestock” OR “beef” OR “pork” OR “milk” OR
“dairy” OR “lamb”)A N D(“life cycle assessment” OR “LCA”). Restricted to
articles published through 2025, this search retrieved 26 articles. Following
duplicate removal and title/abstract screening, 12 articles remained, with 5
included after full-text review. In tot al, 21 peer-reviewed articles were
analyzed, covering both crop and livestock production systems. Detailed
information on selected studies, data extraction, and results of the statistical
analysis are provided in the Supplementary Information (Supplementary
Table 16).
Reducing food waste plays a crucial role in lowering resource con-
sumption and environmental burdens, representing a key pathway to
enhancing the sustainability of food systems66,67. In the food waste scenario,
we assumed a 50% reduction in food w aste at the consumption stage,
aligning with the 2030 targets set by the UN Sustainable Development
Goals30.
Dietary shifts constitute a vital approach to mitigating environmental
risks associated with food systems1,58,68. In the dietary change scenarios, we
developed optimization models to identify dietary strategies for urban and
rural areas across 31 provinces that si multaneously achieve synergistic
reductions in environmental and health risks, while respecting local food
preferences and traditions. To this end, we designed corresponding objec-
tive functions and constraints. Specifically, the optimization aimed to (1)
reduce environmental risks relative to 2021 and (2) reduce health risks. The
methods used to calculate diet-related environmental risks and health risks
are detailed in the Methods section under Environmental Analysis and
DALYs Analysis. To facilitate the implementation of dietary shifts, we also
considered two constraints: (1) ensu red dietary cultural acceptability by
minimizing deviation from current diets, (2) and the second required
alignment with recommended intake levels specified by the Chinese Dietary
Guidelines (CDGs). In addition, given that the impacts of different meats on
the environment and health vary greatly, we grouped meats into 4 categories
(pork, beef, mutton and poultry) and calculated the recommended amount
of each (see Supplementary Table17). We used the non-dominated sorting
genetic algorithm (NSGA-II) 69,70 to estimate the production possibility
frontier of optimal diets based on multiple objectives and constraints. This is
a widely used multi-objective optimization method for calculating healthy
dietary strategies.
Uncertainty estimate
To evaluate the influence of methodological choices and input data on the
results, this study conducted a series of uncertainty analyses. As part of these
analyses, we specifically evaluated the robustness of the observed temporal
trends under moderate uncertainty in food consumption data. For each
Monte Carlo replicate, food-group consumption values were randomly
perturbed within ±10% of their reported values. The resulting uncertainty
was propagated using Monte Carlo si mulation with 100 replicates. The
±10% range was used as a relative perturbation around the statistical
yearbook-based food-group consumption values for sensitivity testing,
rather than as an estimate of the tru e measurement uncertainty. Because
empirical uncertainty distributions for food-group consumption data across
different food groups, years, and po pulation groups were unavailable, a
moderate relative perturbation wa s introduced to evaluate whether the
principal temporal patterns remainedrobust to plausible variations in the
input data. Larger perturbations would be more representative of alternative
dietary scenarios than uncertainty in the reported consumption data. For
each replicate, we recalculated diet-related environmental and health risks
for urban and rural populations and estimated the differences between
adjacent periods. We then summarizedthe simulated distributions of these
period-to-period differences using the median, 95% uncertainty interval
(UI), empiricalP value, and trend support. The 95% UI was defined as the
2.5th and 97.5th percentiles of the simulated difference distribution, and the
empiricalP value was calculated as a Monte Carlo–based two-sided measure
of support for temporal change. Trend support was then classified as sig-
nificant increase, significant decrease, or not significant according to the sign
of the median difference and whether the empiricalP value was below 0.05.
Supplementary Tables 1 and 3 present the detailed results, and Supple-
mentary Figs. 2 and 9 show the corresponding simulated trajectories.
Environmental burdens were assessed by developing a Monte Carlo-
based sensitivity simulation framework to examine how uncertainties in key
environmental footprint intensities (defined as the environmental impact
per unit weight of food, including GHG emissions, cropland use, blue water
use, nitrogen use, and phosphorus u se) may affect the outcomes. These
intensities were assumed to vary within a ± 30% range around their default
values. Considering that such parameters in real-world contexts typically
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press

exhibit central tendencies, random sampling was performed using dis-
tributions centered on the default values. Two perturbation strategies were
employed in the analysis. The first involved single-factor perturbation,
whereby only one environmental factor was varied in each simulation run
while the others remainedfixed, enabling the assessment of the individual
sensitivity of each factor. The second involved multi-factor joint perturba-
tion, in which all five factors were simultaneously and independently
sampled within the speci fied range during each iteration, allowing for
evaluation of the combined effects of parameter variability and the
robustness of the urban–rural comparative results. A total of 1000 simula-
tion iterations were conducted under these two strategies to characterize the
distribution of per capita envir onmental burdens and their urban –rural
differences under parameter uncertainty.
In this health risk assessment, due to the lack of individual-level dietary
intake data, population exposure was modeled based on per capita food
consumption reported in statistical yearbooks, assuming a normal dis-
tribution. To assess the sensitivity of this assumption, multiple parameter
sets were established, including varying coef ficients of variation (CV;
0.05–0.2) and three alternative distribution models: normal, skewed log-
normal, and gamma distributions. In each Monte Carlo simulation, the
exposure distribution type was ran domly selected, while corresponding
relative risk (RR) and disease rate values were sampled. RR estimates and
their 95% confidence intervals were used to replace fixed point estimates
through random sampling, thereby b etter capturing RR uncertainty.
Additionally, baseline disease burden indicators, specifically DALYs rates,
were incorporated using 95% uncertainty intervals provided by the GBD
study, from which values were uniformlysampled to simulate variability. By
jointly simulating these uncertainties, a total of 1000 Monte Carlo iterations
were performed to obtain robust DALYs estimates, systematically evaluat-
ing the in fluence of exposure levels, RR, and baseline disease burden
uncertainties on health risk assessment.
The 95% uncertainty intervals reported in the Results were calculated
from the 2.5th and 97.5th percentiles of the simulated distributions for all
environmental footprints and diet-related health outcomes. These intervals
reflect the combined effects of input parameter variability on both envir-
onmental and health metrics, providing a robust range around the point
estimates. They capture the uncertainty arising from the input parameters
included in the Monte Carlo simulations, but do not account for all potential
sources of uncertainty in the system.
Data availability
The data analyzed in this study were compiled from publicly available
sources and published datasets, as described in the Methods and Supple-
mentary Information. Third-party raw data are available from the original
sources listed therein. The processed datasets necessary to interpret and
reproduce the main analyses, together with the numerical source data
underlying the mainfigures, are publicly available in Zenodo athttps://doi.
org/10.5281/zenodo.1968636471. Additional summary tables are provided
in the Supplementary Information.
Received: 15 October 2025; Accepted: 3 September 2026;
References

1. Willett, W. et al. Food in the Anthropocene: the EAT –Lancet
   Commission on healthy diets from sustainable food systems. Lancet
   393, 447–492 (2019).

2. Ye, B. et al. Adoption of region-speci fic diets in China can help achieve
   gains in health and environmental sustainability.Nat. Food 5, 764–774
   (2024).

3. Crippa, M. et al. Food systems are responsible for a third of global
   anthropogenic GHG emissions. Nat. Food 2, 198–209 (2021).

4. Springmann, M. et al. The healthiness and sustainability of national
   and global food based dietary guidelines: modelling study. bmj 370,
   (2020).

5. West, P. C. et al. Leverage points for improving global food security
   and the environment. Science 345, 325–328 (2014).

6. Dudley, N. & Alexander, S. Agriculture and biodiversity: a review.
   Biodiversity 18,4 5–49 (2017).

7. Brauer, M. et al. Global burden and strength of evidence for 88 risk
   factors in 204 countries and 811 subnational locations, 1990–2021: a
   systematic analysis for the Global Burden of Disease Study 2021.
   Lancet 403, 2162–2203 (2024).

8. Stanaway, J. D. et al. Global, regional, and national comparative risk
   assessment of 84 behavioural, environmental and occupational, and
   metabolic risks or clusters of risks for 195 countries and territories,
   1990–2017: a systematic analysis for the Global Burden of Disease
   Study 2017. Lancet 392, 1923–1994 (2018).

9. Tulloch, A. I. et al. How the EAT –Lancet Commission on food in the
   Anthropocene influenced discourse and research on food systems: a
   systematic review covering the first 2 years post-publication. Lancet
   Glob. Health 11, e1125–e1136 (2023).

10. Arora, N. K. & Mishra, I. United Nations Sustainable Development
    Goals 2030 and environmental sustainability: race against time.
    Environ. Sustain. 2, 339–342 (2019).

11. Hirvonen, K., Bai, Y., Headey, D. & Masters, W. A. Affordability of the
    EAT–Lancet reference diet: a global analysis. Lancet Glob. health 8,
    e59–e66 (2020).

12. Popkin, B. M. Synthesis and implications: China ’s nutrition transition
    in the context of changes across other low-and middle-income
    countries. Obes. Rev. 15,6 0–67 (2014).

13. He, P., Baiocchi, G., Hubacek, K., Feng, K. & Yu, Y. The environmental
    impacts of rapidly changing diets and their nutritional quality in China.
    Nat. Sustain. 1, 122–127 (2018).

14. Cai, H. et al. The multiple bene fits of Chinese dietary transformation.
    Nat. Sustain. 8, 606–618 (2025).

15. Hu, Y., Su, M. & Jiao, L. Peak and fall of China ’s agricultural GHG
    emissions. J. Clean. Prod. 389, 136035 (2023).

16. Piao, S. et al. The impacts of climate change on water resources and
    agriculturein China. Nature 467,4 3–51 (2010).

17. Zhu, X. et al. Multidimensional deconstruction and workable solutions
    for addressing China’s food security issues: From the perspective of
    sustainable diets. Land Use Policy 148, 107401 (2025).

18. He, P., Baiocchi, G., Feng, K., Hubacek, K. & Yu, Y. Environmental
    impacts of dietary quality improvement in China. J. Environ. Manag.
    240, 518–526 (2019).

19. He, P. et al. Health –environment efficiency of diets shows nonlinear
    trends over 1990–2011. Nat. Food 5, 116–124 (2024).

20. Lian, Y., Gu, L., Yang, L., Wang, L. & Li, H. The reasonableness and
    spatial differences of the food consumption structure of urban and
    rural residents in China, 2015 –2021. Foods 12, 1997 (2023).

21. Fan, L., Wang, Z., Zhao, Y. & Ma, Y. Urban –Rural disparities in
    knowledge, use and perceived bene fits of nutrition labels in China:
    evidence from 10 Provinces. Nutrients 15, 1171 (2023).

22. Li, J., Shi, L., Liang, H., Ding, G. & Xu, L. Urban-rural disparities in
    health care utilization among Chinese adults from 1993 to 2011.BMC
    Health Serv. Res. 18, 102 (2018).

23. Finnveden, G. et al. Recent developments in Life Cycle Assessment.
    J. Environ. Manag. 91,1 –21 (2009).

24. Murray, C. J. et al. Global burden of 87 risk factors in 204 countries and
    territories, 1990–2019: a systematic analysis for the Global Burden of
    Disease Study 2019. Lancet 396, 1223–1249 (2020).

25. UNDP. Human Development Report 2023-24. UNDP (United Nations
    Development Programme) (2024).

26. O ’Neill, D. W., Fanning, A. L., Lamb, W. F. & Steinberger, J. K. A good
    life for all within planetary boundaries. Nat. Sustain. 1,8 8–95 (2018).

27. Read, Q. D., Hondula, K. L. & Muth, M. K. Biodiversity effects of food
    system sustainability actions from farm to fork. Proc. Natl. Acad. Sci.
    USA 119, e2113884119 (2022).
    https://doi.org/10.1038/s43247-026-04063-0 Article
    Communications Earth & Environment Article in Press

28. Xing, Y. & Wang, X. Precise application of water and fertilizer to crops:
    challenges and opportunities. Front. Plant Sci. 15, 1444560 (2024).

29. Gebbers, R. & Adamchuk, V. I. Precision agriculture and food security.
    Science 327, 828–831 (2010).

30. United Nations. Transforming our world: the 2030 agenda for
    sustainable development. Report No. A/RES/70/1, (New York, 2015).

31. Deb, K., Pratap, A., Agarwal, S. & Meyarivan, T. A fast and elitist
    multiobjective genetic algorithm: NSGA-II. IEEE Trans. Evolut.
    Comput. 6, 182–197 (2002).

32. Zanello, G., Shankar, B. & Poole, N. Buy or make? Agricultural
    production diversity, markets and dietary diversity in Afghanistan.
    Food Policy 87, 101731 (2019).

33. Ruel, M. T. & Alderman, H. Nutrition-sensitive interventions and
    programmes: how can they help to accelerate progress in improving
    maternal and child nutrition? Lancet 382, 536–551 (2013).

34. Michie, S., Van Stralen, M. M. & West, R. The behaviour change wheel:
    a new method for characterising and designing behaviour change
    interventions. Implement. Sci. 6, 42 (2011).

35. Swinburn, B. A. et al. The global syndemic of obesity, undernutrition,
    and climate change: the Lancet Commission report. Lancet 393,
    791–846 (2019).

36. Roberto, C. A. et al. Patchy progress on obesity prevention: emerging
    examples, entrenched barriers, and new thinking. Lancet 385,
    2400–2409 (2015).

37. Sonwani, E., Bansal, U., Alroobaea, R., Baqasah, A. M. & Hedabou, M.
    An artificial intelligence approach toward food spoilage detection and
    analysis. Front. Public Health 9, 816226 (2022).

38. Jawad, Z. N. & Balázs, V. Machine learning-driven optimization of
    enterprise resource planning (ERP) systems: a comprehensive review.
    Beni-Suef Univ. J. Basic Appl. Sci. 13, 4 (2024).

39. Centobelli, P., Cerchione, R., Del Vecchio, P., Oropallo, E. & Secundo,
    G. Blockchain technology for bridging trust, traceability and
    transparency in circular supply chain. Inf. Manag. 59, 103508 (2022).

40. Reynolds, C. et al. Consumption-stage food waste reduction
    interventions–What works and how to design better interventions.
    Food Policy 83,7 –27 (2019).

41. Urugo, M. M. et al. A comprehensive review of current approaches on
    food waste reduction strategies. Compr. Rev. Food Sci. Food Saf. 23,
    e70011 (2024).

42. Javaid, M., Haleem, A., Khan, I. H. & Suman, R. Understanding the
    potential applications of Arti ficial Intelligence in Agriculture Sector.
    Adv. Agrochem. 2,1 5–30 (2023).

43. Shaikh, T. A., Rasool, T. & Lone, F. R. Towards leveraging the role of
    machine learning and artificial intelligence in precision agriculture and
    smart farming. Comput. Electron. Agric. 198, 107119 (2022).

44. Pierpaoli, E., Carli, G., Pignatti, E. & Canavari, M. Drivers of precision
    agriculture technologies adoption: a literature review. Procedia
    Technol. 8,6 1–69 (2013).

45. Luo, L. et al. A dataset of interprovincial food trade flows in China. Sci.
    Data 12, 943 (2025).

46. Zhou, M. et al. Mortality, morbidity, and risk factors in China and its
    provinces, 1990–2017: a systematic analysis for the Global Burden of
    Disease Study 2017. Lancet 394, 1145–1158 (2019).

47. Springmann, M. et al. Options for keeping the food system within
    environmental limits. Nature 562, 519–525 (2018).

48. Wang, H. et al. Trade-driven relocation of air pollution and health
    impacts in China. Nat. Commun. 8, 738 (2017).

49. National Bureau of Statistics of China. China statistical yearbook
    (China Statistics Press, various years).

50. Campbell, B. M. et al. Agriculture production as a major driver of the
    Earth system exceeding planetary boundaries. Ecol. Soc. 22 (2017).

51. Steffen, W. et al. Planetary boundaries: Guiding human development
    on a changing planet. Science 347, 1259855 (2015).

52. Rockström, J. et al. A safe operating space for humanity. Nature 461,
    472–475 (2009).

53. Henry, R. et al. Food supply and bioenergy production within the
    global cropland planetary boundary. PloS ONE 13, e0194695 (2018).

54. Afshin, A. et al. Health effects of dietary risks in 195 countries,
    1990–2017: a systematic analysis for the Global Burden of Disease
    Study 2017. Lancet 393, 1958–1972 (2019).

55. Gakidou, E. et al. Global, regional, and national comparative risk
    assessment of 84 behavioural, environmental and occupational, and
    metabolic risks or clusters of risks, 1990–2016: a systematic analysis
    for the Global Burden of Disease Study 2016. Lancet 390, 1345–1422
    (2017).

56. Forouzanfar, M. H. et al. Global, regional, and national comparative
    risk assessment of 79 behavioural, environmental and occupational,
    and metabolic risks or clusters of risks, 1990 –2015: a systematic
    analysis for the Global Burden of Disease Study 2015. Lancet 388,
    1659–1724 (2016).

57. Xue, L. et al. China ’s food loss and waste embodies increasing
    environmental impacts. Nat. Food 2, 519–528 (2021).

58. Liu, X. et al. Dietary shifts can reduce premature deaths related to
    particulate matter pollution in China. Nat. Food 2, 997–1004 (2021).

59. Micha, R. et al. Global, regional and national consumption of major
    food groups in 1990 and 2010: a systematic analysis including 266
    country-specific nutrition surveys worldwide. BMJ Open 5, e008705
    (2015).

60. Feigin, V. L., Kinfu, Y. & Collaborators, G. R. F. Global, regional, and
    national comparative risk assessment of 84 behavioural,
    environmental and occupational, and metabolic risks or clusters of
    risks for 195 countries and territories, 1990 –2017: a systematic
    analysis for the Global Burden of Disease Study 2017. Lancet 392,
    1923–1994 (2018).

61. Clark, M. A., Springmann, M., Hill, J. & Tilman, D. Multiple health and
    environmental impacts of foods. Proc. Natl. Acad. Sci. 116,
    23357–23362 (2019).

62. Global Burden of Disease Collaborative Network. Global Burden of
    Disease Study (GBD) 2021 Results. (Institute for Health Metrics and
    Evaluation (IHME), Seattle, United States, 2024).

63. Costa, L., Rybski, D. & Kropp, J. P. A human development framework
    for CO2 reductions. PloS ONE 6, e29262 (2011).

64. Bahmutsky, S., Grassauer, F., Arulnathan, V. & Pelletier, N. A review of
    life cycle impacts and costs of precision agriculture for cultivation of
    field crops. Sustain. Prod. Consum. 52, 347–362 (2024).

65. Moher, D., Liberati, A., Tetzlaff, J. & Altman, D. G. Preferred reporting
    items for systematic reviews and meta-analyses: the PRISMA
    statement. Bmj 339, (2009).

66. De Boni, A., Ottomano Palmisano, G., De Angelis, M. & Minervini, F.
    Challenges for a sustainable food supply chain: A review on food
    losses and waste. Sustainability 14, 16764 (2022).

67. Porter, S. D., Reay, D. S., Higgins, P. & Bomberg, E. A half-century of
    production-phase greenhouse gas emissions from food loss & waste
    in the global food supply chain. Sci. Total Environ. 571, 721–729
    (2016).

68. Li, Y. et al. Reducing climate change impacts from the global food
    system through diet shifts. Nat. Clim. Change 14, 943–953 (2024).

69. Yin, J. et al. Diet shift: considering environment, health and food
    culture. Sci. Total Environ. 719, 137484 (2020).

70. Pan, T. et al. Optimization of multi-objective multi-functional land-use
    zoning using a vector-based genetic algorithm. Cities 137, 104256
    (2023).

71. Shi, Y. et al. China ’s changing diets improve human health but
    increase environmental pressures [Dataset]. Preprint at https://doi.
    org/10.5281/zenodo.19686364 (2026).
    Author contributions
    Yufan Shi: Conceptualization, Data Collection, Formal analysis, Writing—
    original draft, Writing—Review & Editing. Fan Ning: Data Collection, Data
    analysis. Miaomiao Liu: Resources, Conceptualization, Methodology,
    https://doi.org/10.1038/s43247-026-04063-0 Article
    Communications Earth & Environment Article in Press

Formal Analysis, Writing—Review & Editing. Jianxun Yang: Validation,
Supervision. Wen Fang: Validation, Supervision. Zongwei Ma: Validation,
Supervision. Jun Bi: Validation, Supervision, Conceptualization.
Funding
The study was financially supported by The National Natural Science
Foundation of China (Grant No. 72234003, 72174084, 72488101).
Competing interests
The authors declare no competing interests.
Additional information
Supplementary informationThe online version contains
supplementary material available at
https://doi.org/10.1038/s43247-026-04063-0.
Correspondenceand requests for materials should be addressed to
Miaomiao Liu.
Peer review information Communications Earth & Environmentthanks
Cornelia F. A. van Wesenbeeck and Aoyu Hou for their contribution to the
peer review of this work. Primary Handling Editors: Martina Grecequet. A
peer review file is available
Reprints and permissions informationis available at
http://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-NonCommercial-NoDerivatives 4.0 International License,
which permits any non-commercial use, sharing, distribution and
reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative
Commons licence, and indicate if you modi fied the licensed material. You
do not have permission under this licence to share adapted material
derived from this article or parts of it. The images or other third party
material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material
is not included in the article's Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted use,
you will need to obtain permission directly from the copyright holder. To
view a copy of this licence, visit http://creativecommons.org/licenses/by-
nc-nd/4.0/.
© The Author(s) 2026
This Article in Press is shared early to give you faster access to new research.
It is citable and carries a permanent DOI. Thefinal edited version will replace it
automatically.
https://doi.org/10.1038/s43247-026-04063-0 Article
Communications Earth & Environment Article in Press