Nitrogen cycles in global croplands altered by elevated CO2

Croplands are the foundation of global food security and represent the largest nitrogen flows on Earth. Elevated atmospheric CO2 levels are a key driver of climate change with multiple impacts on food production and environmental sustainability. However, our understanding of how the cropland nitrogen cycle responds to elevated CO2 levels is not well developed. Here we demonstrate that elevated CO2 (eCO2) alone would induce a synergistic intensification of the nitrogen and carbon cycles, promoting nitrogen-use efficiency by 19% (95% confidence interval, 14–26%) and biological nitrogen fixation by 55% (95% confidence interval, 28–85%) in global croplands. This would lead to increased crop nitrogen harvest (+12 Tg yr−1), substantially lower fertilizer input requirements (−34 Tg yr−1) and an overall decline in reactive nitrogen loss (−46 Tg yr−1) under future eCO2 scenarios by 2050. The impact of eCO2 on the altered cropland nitrogen cycle would amount to US$668 bn of societal benefits by avoiding damages to human and ecosystem health. The largest benefits are expected to materialize in China, India, North America and Europe. It is paramount to incorporate the effect of rising CO2 on the nitrogen cycle into state-of-the-art Earth system models to provide robust scientific evidence for policymaking. Current understanding of how the cropland nitrogen cycle will respond to elevated atmospheric CO2 is limited. By modelling global nitrogen budgets under elevated CO2 and providing a monetized impact assessment, this study shows the synergistic effects of elevated CO2 alone on global croplands.

Croplands are the foundation of global food security and represent the largest nitrogen flows on Earth.Elevated atmospheric CO 2 levels are a key driver of climate change with multiple impacts on food production and environmental sustainability.However, our understanding of how the cropland nitrogen cycle responds to elevated CO 2 levels is not well developed.Here we demonstrate that elevated CO 2 (eCO 2 ) alone would induce a synergistic intensification of the nitrogen and carbon cycles, promoting nitrogen-use efficiency by 19% (95% confidence interval, 14-26%) and biological nitrogen fixation by 55% (95% confidence interval, 28-85%) in global croplands.This would lead to increased crop nitrogen harvest (+12 Tg yr −1 ), substantially lower fertilizer input requirements (−34 Tg yr −1 ) and an overall decline in reactive nitrogen loss (−46 Tg yr −1 ) under future eCO 2 scenarios by 2050.The impact of eCO 2 on the altered cropland nitrogen cycle would amount to US$668 bn of societal benefits by avoiding damages to human and ecosystem health.The largest benefits are expected to materialize in China, India, North America and Europe.It is paramount to incorporate the effect of rising CO 2 on the nitrogen cycle into state-of-the-art Earth system models to provide robust scientific evidence for policymaking.
Croplands are the major ecosystems supporting food security and human health, representing the largest nitrogen fluxes on Earth 1,2 .Climate change, associated with a continued rise in greenhouse gas emissions, could increase vulnerability of croplands and threaten global food security 3 .Atmospheric levels of CO 2 have increased by 47% since the Industrial Revolution, reaching a level unprecedented in at least two million years and set to exceed 600-1,000 ppm by the end of the twenty-first century 4 .As the primary greenhouse gas, CO 2 also acts as a gaseous fertilizer stimulating plant photosynthesis and productivity, and elevated CO 2 (eCO 2 , also known as CO 2 fertilization, CO 2 enrichment) accordingly enhances carbon sequestration in the terrestrial biosphere 5,6 .Meanwhile, nitrogen is a vital element to constitute protein in flora and fauna, and the capability of carbon sequestration in the biosphere is largely limited by nitrogen availability 7 .The nitrogen cycle will critically determine the potential carbon sinks or sources in croplands under elevated CO 2 levels, and hence it is vital to improve our understanding of the coupling between these two crucial biogeochemical cycles, nitrogen and carbon 8 .It is challenging to determine the response of the nitrogen cycle to climate change due to the complex nature and multiple variables involved in the nitrogen cascade processes.Compared to the extensive studies on the response of the carbon cycle (that is, net primary productivity, soil organic carbon) to climate change 9,10 , the delayed incorporation of nitrogen-cycle feedback into Earth system models has compromised the accuracy and Article https://doi.org/10.1038/s41893-023-01154-0countries worldwide 15 .Currently, the projection of global nitrogen flows and nitrogen budgets under future eCO 2 is lacking, particularly with a monetized impact assessment.Filling this knowledge gap is crucial to constrain Earth system models, which are widely applied for simulating and projecting potential policy interventions to inform policymaking for global sustainable development of agriculture 4,25 .
Manipulation experiments that simulate elevated CO 2 levels provide useful tools for studying the effects of eCO 2 as a single climate change driver on nitrogen cycle features.In this study, we assess the responses of key nitrogen and carbon cycle variables to eCO 2 in croplands based on a global dataset of eCO 2 experiments.Next, we project future cropland nitrogen budgets at a spatial resolution of 0.5° × 0.5° under multiple scenarios utilizing the Coupled Human and Natural Systems (CHANS) model 26 .Finally, we undertake an impact assessment to achieve a quantitative monetized valuation of eCO 2 impacts on the ecosystem and human health.

Responses of carbon and nitrogen dynamics to eCO 2
We present a global atlas of eCO 2 experiments in croplands comprising FACE (free-air CO 2 enrichment), OTC (open-top chamber) and GC (greenhouse and growth chamber) sites (Fig. 1a).In total, 1,003 response ratios were generated for various crop types, including wheat, rice, maize, soybean and others.Elevated atmospheric CO 2 levels promote crop yield by 21% (95% confidence interval (CI), 18-25%) relative to ambient CO 2 level (Fig. 1b), and this effect is consistent across different manipulation methods and regardless of whether values are derived from field and chamber studies (Extended Data Fig. 1c).The response sensitivity of yield to CO 2 fertilization varies with crop type and magnitude of manipulation (ΔCO 2 ) (Extended Data Figs.1a and 2a).Mean annual temperature and mean annual precipitation also moderate the response ratios for the specific type of crop (Extended Data Fig. 3).Soil respiration (Rs), mainly CO 2 emissions from plant roots and soil fauna, increased by 25% (95% CI, 19-31%), which is much higher than the increase in soil organic carbon (SOC) of 6% (95% CI, 2-9%) (Fig. 1b).The relatively small change in SOC may be attributed to the large soil carbon stock.Overall, we found an accelerating trend of carbon cycling reliability of future predictions.Moreover, excessive reactive nitrogen (N r , all forms of nitrogen other than dinitrogen (N 2 )) use in agriculture has led to adverse impacts on ecosystem and human health, including eutrophication, acidification, air pollution (PM 2.5 , that is, particulate matter 2.5 μm or less in diameter) and biodiversity loss 11 .Understanding the climate change impact on the environment of excess N r is an essential research question that needs quantification.Forecast changes in future precipitation regimes are expected to exacerbate N r runoff and intensify regional eutrophication 12 .Conversely, future eCO 2 is projected to dampen global N 2 O emissions-an N r gas known as a potent greenhouse and ozone-depleting gas-from croplands and pastures 13 , illustrating the potential environmental benefits of CO 2 fertilization.
The climate impact on croplands driven by warming and extreme weather events has been highlighted, yet CO 2 fertilization with its interaction effect is rarely considered in future projections 14 .Historical data and experiments reveal that CO 2 fertilization offsets some negative climate impacts on crop yield and production 6,15,16 .Rising CO 2 levels may increase the optimal temperature of photosynthesis and suppress evapotranspiration with lower stomatal conductance, probably interacting with warming and drought to induce cascade effects 6 .Moreover, emerging evidence of large-scale declines in nitrogen availability in non-agricultural ecosystems 17 and human dietary protein 18 underlines rising CO 2 as a main driver of global changes for the nitrogen cycle.eCO 2 not only influences the nitrogen pool and C:N stoichiometry 19,20 but also improves nitrogen-use efficiency (NUE) and biological nitrogen fixation (BNF) in the terrestrial nitrogen cycle 21,22 .Nevertheless, a holistic quantification of nitrogen cycle responses to future eCO 2 in global croplands is lacking.Previous studies typically focused on single or selected nitrogen variables, whereas it is imperative to take account of all the interconnected nitrogen variables in the nitrogen cycle from a systematic perspective to assess the potential feedback of the vital food system to eCO 2 .Responses of the nitrogen cycle to eCO 2 are probably heterogeneous and regionally variable, and the corresponding changes in food production and impacts on environmental and human health may affect regional development and exacerbate inequalities between countries 23,24 .The production and commodities of food and fertilizer must consider the linkages among  in global croplands under eCO 2 , probably a result of stimulated plant productivity (crop yield), gaseous carbon losses to the atmosphere (soil respiration) and carbon sequestration in soil (SOC).The higher carbon availability could provide substrates for microorganism activities closely associated with nitrogen cycling, such as nitrogen-fixing bacteria and denitrifiers 27 .The rates of BNF significantly increase by 55% (95% CI, 28-85%) (Fig. 1b), suggesting strengthened capability of symbiotic and free-living nitrogen fixation microbes to transform N 2 to inorganic nitrogen available for crops in croplands 28 .Nitrogen mineralization is promoted by 22% (95% CI, 6-44%), in correspondence with stimulated soil respiration.Soil nitrous oxide (N 2 O) emissions increase by 29% (95% CI, 5-65%), mainly as a consequence of enhanced denitrification (+24%; 95% CI, 4-57%) 29 .Meanwhile, eCO 2 can facilitate nitrogen uptake by plants, leading to 19% (95% CI, 14-26%) higher NUE in croplands.Higher NUE leads to lower N r loss, including reduced nitrogen leaching and runoff to water bodies (NO 3 − , −45%; 95% CI, −76% to −13%), and to decreased emissions of ammonia (NH 3 ) (−21%; 95% CI, −41% to −1%) and nitrogen oxides (NO x ) (−33%; 95% CI, −50% to −9%) to air.
The foliar C:N ratio increased by 19% (95% CI, 15-24%) (Fig. 1b).In contrast, nitrogen content decreased in grain (−7%; 95% CI, −9% to −5%), leaf (−15%; 95% CI, −18% to −10%) and stem (−10%; 95% CI, −18% to −2%).The decrease in plant nitrogen content could be attributable to the dilution of nitrogen in plant tissues due to increased carbon assimilation and lower investment in Rubisco for photosynthesis 30,31 .Long-term observations indicate eCO 2 has driven a general trend of reduced nitrogen availability in forest and grassland ecosystems since the early twentieth century 17 .Similar to the nitrogen deficiency in unmanaged ecosystems, declining nitrogen content in crop harvest of grain, leaf and stem in croplands is a result of eCO 2 .Although additional mineral fertilizer application can complement nitrogen inputs and create a nitrogen-rich environment in agricultural systems, the preference for nitrogen allocation to roots rather than to leaves for acquiring higher nitrogen uptake under eCO 2 in plants results in a lower leaf N r content 32 .Furthermore, the soil C:N ratio and soil inorganic nitrogen show non-significant responses to eCO 2 , which may be due to the large soil carbon and nitrogen pool with substantial anthropogenic nitrogen input to croplands.Reduced soil nitrogen availability due to continuous nitrogen depletion may occur in natural nitrogen-poor ecosystems without any external nitrogen input source but is unlikely in the intensively managed croplands.
Overall, elevated atmospheric CO 2 levels induce a synergistic intensification of both carbon and nitrogen cycles in global croplands.Increased carbon availability under CO 2 enrichment would stimulate the intensification of the nitrogen cycle, while enhanced nitrogen cycling could in turn alleviate nitrogen limitations for carbon

Article
https://doi.org/10.1038/s41893-023-01154-0assimilation in global cropland systems.Elevated atmospheric CO 2 levels have recently been found to enhance nitrogen cycling through higher nitrogen return from litterfall, for example stimulating consistent tree growth in Tibetan Plateau forests based on observations over a ten-year period 33 .Our global synthesis reveals that the synergistic intensification occurs between the nitrogen and carbon cycles in croplands under eCO 2 at global scale and indicates that the thus enhanced nitrogen cycle would sustain CO 2 fertilization effects on crop yield.

Spatial changes in nitrogen budgets under eCO 2
We estimate the changes in global nitrogen budgets for cropland utilizing the CHANS model and integrate the responses of nitrogen parameters to elevated CO 2 levels to derive annual nitrogen flows.We designed a set of baseline scenarios (no climate change) and eCO 2 scenarios for a near future period (2030-2050) based on Shared Socioeconomic Pathways (SSPs) and Representative Concentration Pathways (RCPs) (Extended Data Fig. 4).Future atmospheric CO 2 levels in the eCO 2 scenarios are derived from CMIP6 models for three eCO 2 subscenarios (SSP1-RCP1.9 'sustainable society', SSP2-RCP4.5 'business-as-usual' and SSP4-RCP6.0'stratified society') relative to three baseline scenarios (SSP1, SSP2 and SSP4).At the global scale, eCO 2 will decrease total nitrogen input (−27 TgN yr −1 ), increase nitrogen harvest (+12 TgN yr −1 ), and reduce nitrogen surplus (N r loss and N 2 , −39 TgN yr −1 ) under the eCO 2 SSP2-4.5 ('business-as-usual') scenario relative to a no-climate-change scenario by 2050 (Figs. 2 and 3).The net increase of N r harvest in future eCO 2 scenarios is due to the positive effects of CO 2 fertilization on yield, which outweigh the negative effects on grain N r content.Under the eCO 2 SSP2-4.5 scenario, a notable increase in N r harvest will occur in East and South Asia, the Great Lakes region in North America, southeast Latin America and central Africa (Fig. 2d-f).These regions are also hotspots for crop production and population density, hence the increasing yield and associated N r harvest can bring immediate food security benefits, especially for low-income countries with considerable famine issues 34 .
Global aggregated NUE is modelled to increase from 47% to 57% by 2050 under the SSP2-4.5 scenario (Fig. 2j-l), albeit with regional heterogeneity.The increase in NUE is projected to exceed 20% in the United States, south and east Latin America, Europe, central and southern Africa, much higher than the minimum increase in NUE of less than 5% in Central America, the Caribbean, most of Asia and eastern Africa.Although the positive response of NUE to eCO 2 can be moderated by mean annual precipitation (Extended Data Fig. 2b), the spatial variation of changes in NUE is closely related to background NUE, marked by a higher relative increase in high-NUE regions, which may increase regional inequality 35 .Improved NUE predominantly drives substantial reductions of total N r input and losses.
Total N r inputs are projected to decline, dominated by lower fertilizer (−34 TgN yr −1 ) and manure input (−5 TgN yr −1 ), and reduced atmospheric N r deposition (−3 TgN yr −1 ); except that BNF will increase drastically by 15 TgN yr −1 under the eCO 2 SSP2-4.5 scenario in 2050 (Fig. 3).The largest reductions in N r inputs are projected for East and South Asia, specifically India and eastern China.Moderate reductions are also modelled for other highly intensified agricultural regions, including central and western Europe (Germany, Czech Republic, France, Italy), central and eastern United States, southern Canada, Argentina and coastal South Australia (Fig. 2a-c).The increased BNF results from stimulated microbial nitrogen-fixing quantity given the enhanced carbon availability in croplands (Extended Data Fig. 5).The reduced N r deposition is largely attributable to the lower NH 3 and NO x emissions from croplands under eCO 2 , because the N r deposition mainly derives from these N r emissions 36 .The producers would probably reduce the use of anthropogenic N r input for adaptation to the local soil nutrient condition with climate change 37,38 .The reduction in mineral fertilizer and manure application would also result in reduced costs for agricultural production in most regions.In contrast, some regions in Brazil, central Africa (Cameroon, Nigeria), and Southeast Asia (Philippines, Laos) would experience a slight increase in total N r inputs because increased BNF, manure application, and atmospheric  N r losses are projected to decrease globally by 46 TgN yr −1 under the eCO 2 SSP2-4.5 scenario in 2050, suggesting net positive effects on environmental sustainability and human health 39 .Our results reveal substantial reductions of nitrogen losses across various regions, particularly in East and South Asia, central and eastern North America, south and east Latin America, followed by west and central Europe, sub-Saharan Africa, Southeast Asia and coastal South Australia under the eCO 2 SSP2-4.5 scenario (Fig. 2g-i).The reduction in N r losses to the environment is projected to be achieved through decreased emissions of NH 3 (−12 TgN yr −1 ), N 2 O (−2 TgN yr −1 ), NO x (−0.9 TgN yr −1 ) and leaching and runoff of NO 3 − to water bodies (−32 TgN yr −1 ), while non-reactive N 2 emissions would increase by 7 TgN yr −1 due to enhanced denitrification processes in croplands (Fig. 3).The decrease of atmospheric NH 3 emissions is mainly concentrated in India and eastern China, the Great Lakes region in the United States and east Argentina (Extended Data Fig. 6).On the other hand, N 2 O emissions would increase in Central America, southern and eastern Africa, South and East Asia, while decreasing in southern Latin America and other regions.This regional variation can be attributed to interactions between the increased emission factors of N 2 O and reduced total N r input to croplands under eCO 2 .NO x emissions are projected to slightly decline on a global scale.The estimated reduction in NO 3 − -nitrogen leaching and runoff to ground and surface water bodies is substantial, particularly in the river basins of India, China, Southeast Asia, Canada and the United States.

Multiple scenario analysis and impact assessment
The ensemble averages and the variations of nitrogen budgets in croplands under historical and future scenarios over 2000-2050 were estimated using Monte Carlo simulations with the CHANS model (Fig. 4).The time series of all baseline scenarios consistently show increasing nitrogen flows in the near to mid-term due to continuous growth in food demand by 2050 2 .For instance, N r harvested will increase from 89 ± 5 TgN in 2020 to 118 ± 20 TgN in 2050 under the baseline SSP2 scenario.The baseline nitrogen budget of SSP2, representing a 'business-as-usual' scenario, is higher than that of SSP1, representative of the 'sustainable society' scenario, but lower than that of the SSP4 'stratified society' scenario.All future eCO 2 scenarios show consistent nitrogen cycling responses to elevated atmospheric CO 2 , with the more sustainability-focused scenarios resulting in lower budgets and their changes due to eCO 2 .We project a 24-29 TgN yr −1 (9-14%) reduction in total nitrogen input, an 8-14 TgN yr −1 (9-10%) increase in nitrogen harvest and a 12-66 TgN yr −1 (14-37%) reduction in nitrogen surplus due to eCO 2 by 2050.Compared to the eCO 2 SSP1-1.9scenario, the higher eCO 2 -induced changes of nitrogen budgets (that is, ΔN harvest, ΔN surplus) in the eCO 2 SSP4-6.0scenario would make a positive difference in alleviating the regional inequality via mitigating food crisis and environmental pollution.
The impact assessment of eCO 2 as a single climate change driver on the global croplands, in the absence of other concurrent climate change, resulted in societal benefits of US$668 bn under the eCO 2 SSP2-4.5 scenario in 2050, in terms of ecosystem benefit, human health benefit, yield increase, fertilizer saving and climate impact (Fig. 5).Globally, the ecosystem benefit accounts for the largest proportion of the total benefit (US$359 bn), followed by human health benefit (US$128 bn) and yield benefit (US$124 bn).China, India, North America and Europe can gain the highest benefit, with the sum accounting for 65% of total benefits.Sub-Saharan Africa is projected to receive US$20.6 bn from yield benefit. Yield benefit is also notable for Brazil, China, India and other Asian countries.The majority of eCO 2 impacts will lead to positive benefits, except that climate impact is expected to cost US$4.0bn, US$4.4 bn and US$0.3 bn in China, India and other OECD (Organization for Economic Cooperation and Development) countries, respectively.Therefore, eCO 2 as a single climate change factor could bring more benefits rather than damages to the ecosystem and humans.There is a caveat that other climate impacts are not accounted for in the monetized assessment, including changes in air temperature and precipitation, and extreme weather 4 .These climate impacts will damage the ecosystem and human health in the long term 12,40 , leading to extra damage costs that are not considered in this assessment 41 .The scope of this paper is limited to the direct costs and benefits of eCO 2 on the cropland nitrogen cycle and their impacts on food benefits, the environment and human health.

Future perspective
Although global ambitions to achieve NetZero+ are under way, these are unlikely to be achieved by 2050.Atmospheric CO 2 levels will probably continue to rise for the foreseeable future 4 .This indicates that the changes in nitrogen cycles in global croplands as a response to eCO 2 are equally likely to become a reality.Our analyses suggest that, despite the evident global existential risks of climate change, we may have the potential to supply more food to alleviate hunger and safeguard food security with less pollution under elevated CO 2 levels.Our findings underscore the opportunity to expedite global progress towards several sustainable development goals, such as zero hunger, no poverty, clean water and sanitation, good health and well-being.Therefore, we must recognize and respond to these changes in nitrogen cycles in global croplands under elevated CO 2 levels, and scientists, policymakers, farmers and other stakeholders must work together to adapt to these changes and design novel approaches for sustainable agriculture management practices 42 .Expected increases in BNF need to be confirmed and managed to avoid excessive N r inputs and losses in croplands.The integrated management of N r inputs between individual components becomes vitally important in the context of the overall reduction of N r input requirements under eCO 2 .The expected nitrogen cycle changes provide a unique opportunity to reduce N r inputs from mineral fertilizers, while increasing the reuse of manure and other organic N r forms, such as straw recycling 43 .New crop varieties that are better adapted to higher CO 2 levels could be developed to further increase NUE and reduce N r losses to the environment 44 .However, the decline of N r concentrations in grain may adversely affect the supply of protein in human diets 18 .Thus, it may be necessary to adjust future dietary recommendations to balance human nutritional requirements with grain protein supply 45 .However, the complexities associated with global warming and altered precipitation regimes, which will accompany elevated atmospheric CO 2 levels, and related impacts, will require integrated assessment approaches based on state-of-the-art complexity science methods to fully analyse and understand the response mechanisms of the nitrogen cycle to future climate change.This will be an essential step towards designing an effective and efficient climate policy 46 .Adaptation of farming systems to higher crop yields and improved NUE will need to go hand in hand with measures to manage extreme climate impacts to reduce the uncertainties in future global crop production 15 .Nevertheless, the final net impact of the complex interactions between the carbon and nitrogen cycles is not yet fully quantifiable.Whether the potential benefits to crop production and cropland NUE due to elevated CO 2 levels could offset some of the negative impacts of climate change requires more attention and in-depth analyses.Our results highlight the importance of fully quantifying trade-offs and co-benefits between climate change factors which researchers and policymakers alike must navigate in meeting climate change mitigation and sustainable development goals.A comprehensive and robust understanding of the response mechanisms of the nitrogen cycle to climate change will be a key requirement to constrain Earth system models and inform agricultural management and policy development to design future agricultural systems in the context of climate change 47 .Such robust, climate-resilient agricultural systems with reduced impacts on human and environmental health, while safeguarding food security, are vital for feeding a growing world population in a changing climate 48 .

Database and global synthesis
A global database of elevated CO 2 simulation experiments was established through data extraction from site-based eCO 2 manipulation studies and data compilation from other data sources (Supplementary Table 1).The selection criteria for qualified studies mainly comprise of: (1) manipulation experiments with an eCO 2 group and a control group; (2) variables related to the nitrogen cycle or the carbon cycle monitored on a regular basis in both an eCO 2 group and a control group, and where it was possible to extract the values of these variables from the study; (3) studies published in peer-reviewed journals included in authoritative databases such as Web of Science, Google Scholar and Scopus.The systematic literature search contained but was not limited to the following key terms: {(elevated CO 2 /rising CO 2 /CO 2 fertilization) OR (FACE/OTC/GC)} AND {(nitrogen fixation/BNF/nitrogen use efficiency/ NUE/denitrification/NH 3 /ammonia/N 2 O/nitrous oxide/nitrogen leaching/nitrogen runoff/nitrogen mineralization/nitrification/nitrogen cycle) OR (yield/SOC/soil organic carbon/soil respiration/Rs/nitrogen content/C:N ratio/carbon cycle)}.We collected paper information, site information, study information and variable information.The terminology used in the study can be found in the Supplementary Information.
Data of variables were extracted from the text, tables and figures in the published papers.WebPlotDigitizer 4.4 was used to extract data from figures (https://apps.automeris.io/wpd/).Meanwhile, data from other sources were compiled into our database to supplement the missing information in some publications, that is, climate data, soil texture and climate zones.Climate data of study sites (that is, mean annual temperature, mean annual precipitation, maximum temperature and minimum temperature) were obtained from the WorldClim (https://worldclim.org/data/index.html#).Soil texture was from the Global Land Data Assimilation System (GLDAS) by NASA (https://ldas.gsfc.nasa.gov/gldas/soils).Assignment of climate zone was based on the Köppen-Geiger climate classification.https://doi.org/10.1038/s41893-023-01154-0 Meta-analysis was conducted to assess the response ratio (RR) of nitrogen and carbon cycling variables under eCO 2 .The response ratio of individual observations as a natural logarithm (ln RR) was calculated as 49 : where x eCO 2 and x aCO 2 are the means of parameters at elevated CO 2 level and ambient CO 2 level, respectively.The weight of individual observations was calculated based on the experimental replications as 50 : where n eCO 2 and n aCO 2 are the numbers of replications at elevated CO 2 level and ambient CO 2 level, respectively.The RRs and weights of individual observations were calculated in Microsoft Excel.The mean RRs and 95% CIs were generated following a randomization resampling procedure by bootstrapping (4,999 iterations) in the software MetaWin 2.0 51 .The results were reported as percentage changes: RR% = (e RR − 1) × 100% (3) An RR was considered significant (P < 0.05) if the 95% CI did not overlap with zero.
Meta-regressions were used to test whether the potential moderators can affect the response pattern across locations.The statistical analysis of meta-regression was done with the metafor package 52 in R v.4.1.3.

Global cropland nitrogen budget
The present and future global cropland nitrogen budgets at 0.5° × 0.5° resolution are generated based on the Global Nutrient Model (GNM) as part of the Integrated Model to Assess the Global Environment (IMAGE) 38,53 and the CHANS 54 models (Extended Data Fig. 7).IMAGE is an ecological-environmental model simulating the environmental consequences of human activities by integrating society, biosphere and climate systems in one framework (https://www.pbl.nl/en/image/about-image).CHANS is a process-based model that specializes in simulating nitrogen flow within the human-natural coupled system, including 14 subsystems (that is, cropland, grassland, forest, atmosphere, surface water and groundwater) 55 .A detailed introduction of these models can be found in the Supplementary Information.Both IMAGE and CHANS models can provide global cropland nitrogen budgets based on similar mass-balance principles and multiple-source data.
Here we incorporate the IMAGE model output into the CHANS model to deliver plausible nitrogen budgets with finer resolution and minimized uncertainties (Extended Data Fig. 8).Gridded data of the global cropland budget (0.5° × 0.5°) were exported from the IMAGE model, and then input to the CHANS model for validation and optimization with historical data at the country-level embedded in the CHANS model.Future crop harvests from 2030 to 2050 at 10 yr intervals are based on the future prediction data from the Food and Agriculture Organization (FAO), which projects future crop yield and harvest area based on food demand depending on population, gross domestic production and urbanization rates.
The compilation of cropland nitrogen budget is to identify nitrogen input (N input ), nitrogen harvest (N harvest ), nitrogen surplus (N surplus ) and NUE, on the foundation of the nitrogen mass-balance principle.The calculation formulas are as follows: N surplus,i = N gas,i + N water,i where N input contains five components, that is, synthetic fertilizer (N fer,i ), BNF (N BNF,i ), manure (N man,i ), deposition (N dep,i ) and other inputs (N other,i , that is, seed, lighting, sewage sludge); N harvest refers to harvested crops consisting of grain and straw; N surplus contains two components, that is, gaseous nitrogen loss (NH 3 , N 2 O, NO x , N 2 ) (N gas,i ) and nitrogen loss to water (leaching to groundwater and runoff to surface water, NO 3 − ) (N water,i ).
The input factor (F input,i ) and emission factor (F emit,i ) are defined as: where N input component,i could be any component of N input , that is, N fer,i , N BNF,i , N man,i , N dep,i and N other,i , and N emit component,i could be any component of N surplus,i , that is, N gas,i and N water,i .
The reactive nitrogen (N r,i ) flows include NH 3 flows ) and nitrogen loss to water (N water,i ):

Scenario and CHANS model simulation
Aiming to estimate the changes of nitrogen budget under future elevated CO 2 levels, we designed two scenarios: the counterfactual baseline scenario (no climate change) and the eCO 2 scenario, each containing three subscenarios with different SSPs (Extended Data Fig. 3).The baseline scenario hypothesizes that no climate change will occur in the future and the atmospheric CO 2 levels will remain fixed at 2020 levels.The eCO 2 scenario considers only elevated atmospheric CO 2 as a single factor of climate change, without considering associated warming and changing precipitation.Future atmospheric CO levels were applied from RCPs, including RCP1.9, RCP4.5 and RCP6.0.As climate change factor was set in the baseline and eCO 2 scenarios, SSPs provide storylines and narratives about socioeconomic aspects for future projections.Our study is based on SSP1, SSP2 and SSP4, corresponding to 'sustainable society' (low greenhouse gas emissions), 'business as usual' (intermediate greenhouse gas emissions) and 'stratified society' (high greenhouse gas emissions).The baseline scenario has three subscenarios, SSP1, SSP2 and SSP4, each considering different socioeconomic pathways that influence population and GDPs, leading to changes in food demand and supply, fertilizer use and NUEs.Accordingly, the eCO 2 scenarios have subscenarios of SSP1-RCP1.9,SSP2-4.5 and SSP4-RCP6.0,considering both socioeconomic pathways and eCO as a single indicator of climate change for modelling.We conducted CHANS model simulation for nitrogen budgeting based on the above scenarios.Future trends from 2030 to 2050 are projected (Extended Data Fig. 8).In the baseline scenarios, mainly the socioeconomic indexes of population, GDP and urbanization are considered to project future food production; but no climate change effects are considered in the modelling given the fixed CO 2 level since 2020.In the eCO 2 scenarios, in addition to the socioeconomic indexes, we simulate rising CO 2 levels by integrating the eCO 2 impacts on the nitrogen and carbon cycles derived from eCO 2 experiments in the meta-analysis into the CHANS model for optimizing parameterization.
The effects of eCO 2 on crop yield and grain content for various crop items were incorporated into the crop production dataset of future prediction in 2030-2050 from FAO.The crop production by country is summed up as follows: where N base harvest and N eCO 2 harvest indicate nitrogen harvests in the country under the baseline scenario and the eCO 2 scenario, respectively; Yield crop,i is the yield of the specific crop item and crop i indicates the specific crop item; GrainN crop,i is the nitrogen content in the grain; Area crop,i refers to the harvest area of the crop item; RR% yield and RR% GrainN denote the response ratios of yield and grain nitrogen content to eCO 2 , respectively.The responses of yield and grain nitrogen content are moderated with the regional ΔCO 2 , MAT and MAP, and constrained by maximum yield potential, and the upper and lower limit of 95% CIs from the meta-analysis.
The effects of eCO 2 on nitrogen cycling parameters (NUE i , N BNF,i , N fer,i , N NH 3 ,i , N water,i , etc.) were scaled up to modify the NUE, input factor (F input,i ) or emission factor (F emit,i ).The NUE under eCO 2 , NUE eCO 2 i , is calculated as: where RR% NUE denotes the response ratios in percentage change for NUE.
The input and emission factors coupled with eCO 2 effects, F eCO 2 input,i and F eCO 2 emit,i , are calculated as: where RR% input component and RR% emit component denote the response ratios in percentage change for the nitrogen input and emission factors, respectively.CHANS model simulation is used to predict future cropland nitrogen budget with the gridded data of the global cropland nitrogen budget, which takes into account regional patterns and geographical heterogeneity.Responses of NUE modulated by the local MAP within the CIs in the meta-analysis are allocated as RR% NUE,i to the gridded data (Extended Data Fig. 2).Responses of BNF in different climate zones are allocated to the gridded data (Extended Data Fig. 1).Responses of deposition depend on the summed NH 3 and NO x emissions, with the responses of NH 3 and N 2 O moderated by ΔCO 2 and then incorporated into the model.Because NO x emissions from croplands are minimum and the metadata on NO x in croplands are lacking, we use response ratios of NO x in the terrestrial ecosystem as a substitution.The responses of NO 3 − are allocated as RR% water,i .Anthropogenic nitrogen input is considered a flexible component of input because the use of fertilizer will adapt and evolve over the mid-to long term.

Impact assessment
The potential impact of eCO 2 (I eCO 2 ) as a single climate change factor in global croplands consists of ecosystem impact (I eco ), human health impact (I human ), yield change (I yield ), fertilizer saving (I fer ) and climate impact (I climate ) related by the following equation 26 : The comprehensive monetary impact analysis of eCO 2 is conducted at the national scale and then scaled up to regional and global cropland by categorizing country groups.The methods have been developed and tested in our previous work on the cost-benefit assessment of nitrogen pollution and mitigation 26,56,57 .
The ecosystem impact is defined as the changed damage cost of N r effects on the ecosystem service.The ecosystem impact for country/ region i can be calculated as: where ΔN r,i are the changes of N r including NH 3 flows, N 2 O flows, NO x flows and nitrogen loss to water for country or area i; d eco,EU stands for the estimated ecosystem damage cost of N r losses in the European Union based on the European Nitrogen Assessment 58 ; WTP i and WTP EU denote the values of the willingness to pay for ecosystem service in country/area i and the European Union, respectively; PPP i and PPP EU denote the purchasing power parity of country/area i and the European Union, respectively.Here we apply the ecosystem damage cost of N r losses in the European Union to other countries after corrections for willingness to pay and purchasing power parity, aiming to attain comparable ecosystem benefit across the globe 59 .Several cost-benefit studies concerning the effects of N r on the ecosystem have been conducted in Europe and the United States, but there is a paucity of available data in other areas or countries 59,60 .The human health impact is defined as the changed health damage due to varied N r losses under elevated CO 2 levels.The monetary estimate of human health is as follows: where ΔN r,i are the changes in N r for country or area i; d human,i stands for the human health damage cost of N r losses for country/area i, which is calculated based on the metric of nitrogen-share to PM 2.5 pollution 56 , that is, the contribution of N r compounds to the total PM 2.5 concentration determined by modelling with and without N r losses.The monetary evaluation of yield change can be calculated as the changed crop revenues from crop harvest using the following equation: where ΔN eCO 2 harvest,i are the changes in nitrogen harvest in the eCO 2 scenario relative to the baseline scenario for country/area i; and p yield,i is the crop price in the specific country/area i, in US$ per kgN.We assume the crop price is equal to the gross production value of crops (constant 2014-2016 thousand US$) divided by total crop production quantity in each country.Data are derived from FAOSTAT (https://www.fao.org/faostat/en/#data) and missing values for some countries are substituted with the global mean value.
The fertilizer saving refers to the saved investment of nitrogen fertilizer to croplands due to reductions in synthetic fertilizer input under eCO 2 scenarios as: I fer,i = ΔN fer,i × P fer (20)   where ΔN fer,i are the changes in nitrogen fertilizer input in the eCO The climate impact can be positive or negative for different countries and regions.The potent greenhouse gas N 2 O contributes to global warming, implying a negative climate impact, whereas NO x and NH 3 are vital precursors of aerosols, which would reflect long-wave solar radiation and have a strong cooling impact on the climate system 57 .Therefore the cost-benefit analysis of climate impact is conducted as follows: where m climate,i are the unit climate damages or benefits of N r losses for country/area i in US$ per kgN.

Uncertainty analysis
An uncertainty analysis of the cropland nitrogen budget was performed by running Monte Carlo simulations with the CHANS model for 1,000 iterations.Monte Carlo simulation is a computational simulation method that utilizes random sampling and probability to analyse complex systems or processes.In the CHANS model of the nitrogen budget, the uncertainty sources and uncertainty ranges of input parameters are identified according to the data distribution and characteristics.The coefficients of variation were used to represent the relative uncertainty ranges of cropland nitrogen budget data and climate change impact under eCO 2 (Supplementary Table 2).After 1,000 iterations of CHANS model simulations, the mean and the variations of the nitrogen budgets can be calculated from projection ensembles.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Fig. 1 |
Fig. 1 | Effects of elevated CO 2 levels on nitrogen and carbon cycles in global croplands.a, Maps displaying the distribution of experimental sites simulating elevated atmospheric CO 2 levels, by different manipulation methods and by various crop types.b, Response ratios of main variables to elevated CO 2 levels based on the meta-analysis.Scatter plots in colour represent response ratios of observations included in the meta-database, and diamonds with error bars indicate mean values of response ratios with a 95% CI.The value of the

Fig. 3 |
Fig. 3 | Nitrogen flows in global croplands under eCO 2 scenario (SSP2-4.5)by 2050.a, Nitrogen input and output constitute the major nitrogen flows, represented by blue and yellow arrows, respectively.Values of nitrogen flows in dark grey denote flows in the baseline scenario with no climate change; the Articlehttps://doi.org/10.1038/s41893-023-01154-0deposition would outweigh the decrease in mineral fertilizer application in these regions.

Fig. 4 |
Fig. 4 | Time series of nitrogen budgets in global croplands over the period 2000-2050 under multiple scenarios.a-i, Solid lines represent mean values of total nitrogen input (a), nitrogen harvest (b), nitrogen surplus (N r loss and N 2 )

Fig. 5 |
Fig. 5 | Impact assessment of elevated atmospheric CO 2 levels as a single climate change factor under the eCO 2 SSP2-4.5 scenario relative to the baseline scenario (no climate change) in 2050.The positive values indicate benefit and negative values indicate damage cost.FSU, Former Soviet Union; MENA, Middle East and North Africa; OECD, Organization for Economic Cooperation and Development; SSA, sub-Saharan Africa.

Extended Data Fig. 1 |Extended Data Fig. 2 |
Article https://doi.org/10.1038/s41893-023-01154-0Effects of elevated CO 2 levels on crop yield and grain N content in global croplands.Effects of elevated CO 2 levels on (a) crop yield and (b) grain N content by crop groups; (c) crop yield by manipulation methods, including FACE (Free-air CO2 Enrichment Experiment), OTC (Open-top Chamber), and GC (Growth Chamber); (d) BNF by climate zones.The error bars indicate the 95% confidence interval of the mean, which is significant if the 95% confidence interval does not overlap zero.Numbers in the parentheses denote the number of observations in the meta-analysis.Article https://doi.org/10.1038/s41893-023-01154-0Meta-regressions between response ratios (RR%) of variables and environmental factors.(a) RR of crop yield versus ΔCO 2 (elevated CO 2 level relative to ambient CO 2 ); (b) RR of NUE versus MAP (mean annual precipitation); (c) RR of NH 3 versus ΔCO 2 ; (d) RR of N 2 O versus ΔCO 2 .The bubbles represent the response ratios of individual observations, with bubble sizes indicating the weights of the response ratios.The P values were obtained from a two-sided F-test for the fitting lines with no adjustments for multiple comparisons.Unit ppm denotes parts per million.Extended Data Fig. 3 | Meta-regressions between logarithm-transformed response ratios (LnRR) of variables for specific crop type and environmental factors.(a) LnRR of yield versus ΔCO 2 for barley; (b) LnRR of yield versus MAT (mean annual temperature) for maize; (c) LnRR of yield versus MAP (mean annual precipitation) for rice.The bubbles represent the response ratios of individual observations, with bubble sizes indicating the weights of the response ratios.The P values were obtained from a two-sided F-test for the fitting lines with no adjustments for multiple comparisons.https://doi.org/10.1038/s41893-023-01154-0Extended Data Fig. 5 | N input of global cropland and their changes under elevated CO 2 SSP2-4.5 scenario relative to baseline scenario in 2050.Biological N fixation (BNF) in baseline scenario (a), eCO 2 scenario (b), and ΔBNF (c); Fertilizer in baseline scenario (d), eCO 2 scenario (e), and ΔFertilizer (f); Manure in baseline scenario (g), eCO 2 scenario (h), and ΔManure (i); Deposition in baseline scenario (j), eCO 2 scenario (k), and ΔDeposition (l).Values in the legend reflect the average annual N budget from croplands within a grid cell (0.5 by 0.5 degree).The base map is applied without endorsement from GADM data (https://gadm.org/).

Extended Data Fig. 6 |
Article https://doi.org/10.1038/s41893-023-01154-0Reactive N loss of global cropland and their changes under elevated CO 2 SSP2-4.5 scenario relative to baseline scenario in 2050.NH 3 in baseline scenario (a), eCO 2 scenario (b), and ΔNH 3 (c); N 2 O in baseline scenario (d), eCO 2 scenario (e), and ΔN 2 O (f); NO x in baseline scenario (g), eCO 2scenario (h), and ΔNO x (i); N leaching and runoff in baseline scenario (j), eCO 2 scenario (k), and ΔN leaching and runoff (l).Values in the legend reflect the average annual N budget from croplands within a grid cell (0.5 by 0.5 degree).The base map is applied without endorsement from GADM data (https://gadm.org/).Extended Data Fig.7| CHANS (Coupled Human and Natural System) model with detailed representation of cropland sub-system.(a) Overview of the CHANS model.The CHANS model contains 14 subsystems that are divided into four functional groups: processors (indicated by a green outline), consumers (red outline), removers (seagreen outline), and life-supporters (blue outline).The focus of this study is on the cropland sub-system.(b) Detailed representation of cropland sub-system.The model utilizes input data from multiple sources including social-economic data (indicated by orange) and N cycling data (indicated by blue).The N flows associated with N input, harvest, and surplus serve as CHANS model variables in the N-balance operation, which generates outputs of cropland N budgets for different geographical regions.Additionally, the CHANS model enables impact assessment to estimate the monetized impacts on health and the environment.

Extended Data Fig. 8 |
Articlehttps://doi.org/10.1038/s41893-023-01154-0Methodology framework.Meta-analysis is conducted to assess the impacts of elevated CO 2 (eCO 2 ) on N and C cycles.Crop N budgets are generated based on the IMAGE and CHANS models using multi-source data.In scenario analysis, we simulate rising CO 2 levels by integrating the eCO 2 impacts on N and C cycles in meta-analysis into the CHANS model for optimizing parameterization and projecting future cropland N budgets.
Response ratios of main variables to elevated CO 2 levels based on the meta-analysis.Scatter plots in colour represent response ratios of observations included in the meta-database, and diamonds with error bars indicate mean values of response ratios with a 95% CI.The value of the response ratio is significant if the 95% CI does not overlap zero.The numbers in parentheses next to the variables represent the numbers of observations.Rs, soil respiration; Grain [N], grain nitrogen content; Leaf [N], leaf nitrogen content; Stem [N], stem nitrogen content; SIN, soil inorganic nitrogen; Min., nitrogen mineralization; Denitri., denitrification.The base map is applied without endorsement from GADM data (https://gadm.org/). b, of uncertainty.The fertilizer price for each country is estimated by dividing the value of fertilizers traded by the quantity.Data are derived from the UN Comtrade Database (https://comtrade.un.org/) and missing values for some countries are substituted with the global mean value.
2 scenario relative to the baseline scenario for country/area i; and P fer is the nitrogen fertilizer price, in US$ per kgN.Fertilizer prices fluctuate substantially with respect to time and region and are subject to a degree