Air quality improvement masks global cooling from CO₂ reductions under China’s carbon neutrality policies for half a century

Abstract

Achieving carbon neutrality and improving air quality are pivotal sustainability strategies for the Global South countries. However, their global climate impacts over a realistic timescale remain unclear. Here we evaluate the climate impacts of China’s carbon neutrality and Beautiful China policies using a fully coupled Earth system model and updated future anthropogenic emission scenarios. We find that, for an unexpectedly long time through ~2070, China’s air pollutant reductions can cause a large global surface warming (0.12 ± 0.09 K for 2050–2070) that almost offsets the cooling from concurrent CO₂ emission reduction (0.16 ± 0.05 K for 2050–2070), compared to a business-as-usual scenario. This warming is mainly attributed to reduced SO₂ and organic matter emissions. Moreover, combined air pollutants and CO₂ declines create a striking hemispheric temperature change contrast, because of the stronger aerosol-induced heating in the Northern Hemisphere. Considering that most future air pollutant reductions represent synergistic effects of carbon neutrality policies, the associated inevitable warming effect over decades highlights the importance of exploring more aggressive policies including early carbon neutrality, methane reductions, and negative carbon emissions.

China’s carbon neutrality and clean-air policies cause global warming from reduced aerosols that nearly offsets cooling from reduced CO2 before 2060, with warming emerging afterwards, calling for early carbon neutrality and more mitigation strategies.

Introduction

Earth’s mean temperature increased to more than 1.5 °C above pre-industrial levels for the first time in 2024¹. To limit the continuous warming and avoid more disastrous consequences of climate change, achieving global carbon neutrality by the second half of the century is essential and urgent². Meanwhile, large parts of the world, especially Global South countries like China and India, face grand challenges in alleviating severe aerosol pollution³, which causes over three million premature deaths annually⁴. Achieving carbon neutrality and improving air quality are thus pivotal strategies for the sustainable development of Global South countries. For example, as one of the large annual emitters of CO₂ and short-lived air pollutants, China has pledged to achieve carbon neutrality by 2060 and fully realize a Beautiful China by 2050. The latter can be interpreted as China should at least meet the World Health Organization’s (WHO) 2005 guideline (10 μg m⁻³ for annual mean concentration of fine particulate matter, PM_2.5) and strive to achieve the 2021 guideline (5 μg m⁻³), the attainment of which requires additional clean air measures beyond carbon neutrality^5–8. Such regional carbon neutrality and clean air policies, by changing both CO₂ and aerosols, will have complex region- and time-dependent climate impacts. Specifically, compared with the well-mixed CO₂, the climate impact of aerosol reduction depends strongly on the region where the policies are implemented, because of the large spatial heterogeneity in aerosol compositions, chemical regimes, and topographical and climate characteristics^9–13. Moreover, the climate impacts of aerosol and CO₂ reductions occur on distinct time scales, with the former manifesting much sooner because of the shorter lifetime of aerosols^14–17. It is thus important to quantify the global climate impacts of each region’s carbon neutrality and clean-air policies on a realistic time scale to develop mitigation policies tailored for individual regions.

Previous studies on the climate impacts of decarbonization and clean-air policies have mostly considered the global policies as a whole, with the most well-known being the Coupled Model Intercomparison Project Phase 6 (CMIP6)^11,18,19. Moreover, the Detection and Attribution Model Intercomparison Project (DAMIP), as part of CMIP6, has worked to separate the contributions of individual forcings (e.g., changes in aerosols and greenhouse gases) to global climate change^20,21. However, only a couple of studies have examined the impact of regional decarbonization or clean-air policies on global climate change. A previous modeling assessment²² showed that the net effect of United States reductions in all emissions consistent with 2 °C warming would avoid 0.03–0.07 °C global warming in 2030 and 0.15–0.25 °C in 2100. Without considering the effect of concurrent air pollutant reductions, a previous study²³ showed that China’s CO₂ reduction under carbon neutrality would lower global mean temperature by 0.40–0.48 °C by the end of this century compared to scenarios representing less stringent climate policies. In contrast, another study²⁴ evaluated the climate impacts of China’s aerosol reductions without considering the influence of CO₂ changes. They showed that the aerosol reduction associated with achieving carbon neutrality by 2060 might increase global mean temperature by 0.23 °C under climate equilibrium, which is a hypothetical condition that would not be achieved unless all emissions remained constant for nearly 100 years. Nevertheless, regional CO₂ and air pollutant emissions will inevitably change simultaneously and progressively, especially under anticipated carbon neutrality and clean air policies. The climate impacts of concurrent regional CO₂ and air pollutant emission reductions, as well as their temporal evolution over a realistic policy-relevant timescale (e.g., from the present to 2060 and 2100), remain unclear.

Here we employ a comprehensive and fully coupled Earth System Model, the Community Earth System Model version 2 (CESM2), to evaluate the climate impacts of both simultaneous and separate CO₂ and air pollutant emission reductions in China, under the carbon neutrality and Beautiful China strategies on a realistic timescale in the 21^st century. In contrast to previous studies, we conduct long-term, fully coupled, transient simulations (representing time-varying radiative forcing in a realistic manner) with full interactions of all key components of the climate system, including dynamic oceans, biogeochemical systems, and chemistry-climate interactions. Unexpectedly, our findings reveal that, for half a century through ~2070, the global surface temperature increase from China’s air pollutant reductions relative to a business-as-usual (BAU) scenario largely offsets the cooling from China’s CO₂ reduction. This contrasts with previous studies that suggested that global or United States decarbonization policies would quickly contribute to temperature decreases^11,19,22. Intriguingly, the combined air pollutant and CO₂ reductions create contrasting temperature and precipitation changes in the two hemispheres. The inevitable and persistent offsetting effect underscores the importance of achieving carbon neutrality early, promoting methane reduction, and achieving negative carbon emissions in the long run.

Results

Changes in CO₂ and aerosols for 2050–2070 under control policies

Three emission scenarios for China, developed by a technology-rich provincial-level integrated assessment model⁵, are investigated in this study: (1) a BAU scenario, defined by the climate and air pollution control policies enacted as of 2020, assumes that carbon emissions peak around 2030 and then decline slightly; (2) a carbon neutrality and clean air (CNCA) scenario, which assumes the achievement of carbon neutrality by 2060, followed by negative carbon emissions, along with stricter clean air policies in line with the 2050 Beautiful China vision; and (3) a sensitivity scenario (SENS), which combines the CO₂ emissions from CNCA with the air pollutant emissions from BAU to help isolate the impacts of CO₂ and air pollutant reductions. The integrated assessment model fully accounts for China’s recent development path and policies and thus realistically reproduces recent and anticipated near-future emission trends, e.g., the constantly increasing CO₂ emissions until about 2030 and the swiftly decreasing emissions of major air pollutants in the previous decade (2010–2020); such trends are not visible in the shared socioeconomic pathways (SSPs) with overall comparable stringency (SSP2-4.5, SSP1-2.6, and SSP1-1.9, see Supplementary Fig. 1), which show that China’s CO₂ emissions have been decreasing since 2020 but the air pollutant emissions have been decreasing much slower than in reality. For the rest of the world, we use SSP2-4.5, which represents a future with moderate climate policies. For each emission scenario, we average over five ensemble members with small perturbations to the initial conditions derived from historical-era simulations to minimize the impact of internal variability in the climate system.

We first focus our analysis on 2050–2070, a twenty-year period centering around 2060 when China is expected to achieve carbon neutrality. Figure 1 shows the difference in CO₂ concentration and PM_2.5 burden between the CNCA and BAU scenarios (CNCA minus BAU, or CNCA − BAU) for 2050–2070. Global mean CO₂ concentrations are reduced by 23.5 ± 0.7 ppm (mean ± standard deviation), as a result of the year-by-year accumulation of CO₂ emission reduction driven by China’s carbon neutrality goal. The CO₂ concentration reductions are rather uniform globally, except for eastern China where the largest CO₂ emission reduction occurs, as it takes time for CO₂ to disperse evenly around the world. In contrast, air quality improvement in terms of PM_2.5 burden (excluding dust and sea salt) primarily occurs in and near China, with global and China mean reductions of 0.31 ± 0.10 mg m⁻² and 3.74 ± 0.11 mg m⁻², respectively. The PM_2.5 burden reduction represents a combined effect of the carbon neutrality target and additional end-of-pipe control policies aligned with the Beautiful China vision. Supplementary Fig. 1 highlights that the implementation of carbon neutrality is responsible for over 80% of the emission reductions in all air pollutants except for volatile organic compounds (VOCs) around 2060.

Fig. 1. Overall and spatial differences in CO₂ concentration and PM_2.5 burden under China’s carbon neutrality and clean air (CNCA) policies with respect to the business-as-usual (BAU) scenario for 2050–2070.

A, C Mean differences in global and China (A) PM_2.5 burden and (C) CO₂ concentration. B, D Spatial distribution of the differences in (B) PM_2.5 burden and (D) CO₂ concentration. The results are averaged over 2050–2070, centered around 2060 when China is expected to achieve carbon neutrality. The calculation of PM_2.5 burden differences excluded dust and sea salt components that almost exclusively originate from natural sources, since differences in these components are not directly tied to mitigation policies. The spatial distributions of the differences in different PM_2.5 chemical components are given in Supplementary Fig. 2. The raincloud plots are constructed with 25 samples obtained by pairwise differencing of five ensemble members under two scenarios. The thick horizontal line represent median values, the central box indicate the interquartile range (25–75%), the thin vertical lines extending from the box display the data range (excluding outliers), and the width of the violin shape shows the kernel density estimation distribution of the data. The scattered points to the right of the box represent the individual data samples. The numbers on top of the raincloud plots represent mean values. Stippled areas in the spatial maps indicate statistical significance at the 95% confidence level based on a two-tailed Student’s t test. Coastline boundaries in this and all subsequent figures and Supplementary Figs. were made with Natural Earth. Source data are provided with this paper.

Global mean changes in climate variables for 2050–2070 due to control policies

We first examine global mean difference in surface temperature (Fig. 2A), the most commonly used climate metric, between the CNCA and BAU scenarios (CNCA−BAU) for 2050–2070, as well as the differences induced by CO₂ (SENS−BAU) and air pollutant (CNCA−SENS) emission reductions separately. Also shown are the corresponding differences in top-of-atmosphere (TOA) net radiation flux (shortwave + longwave) that are closely related to temperature differences (Fig. 2B). The CO₂ emission reduction causes a decrease in the net TOA radiation flux and cools the surface by 0.16 ± 0.05 K relative to BAU, consistent with the mitigation of the greenhouse effect of CO₂. In contrast, the reductions in air pollutant emissions increase the TOA net radiation, warming the surface by 0.12 ± 0.09 K, as a combined effect of aerosol–radiation interaction (ARI) and aerosol-cloud interactions (ACI). ACI likely plays a leading role in causing the temperature changes (see more in Supplementary Note 1). The global warming from air pollutant reductions can largely offset the climate cooling benefit of CO₂ reduction even by around 2060, when net-zero CO₂ emissions are reached. The combined global effect is an insignificant cooling of −0.03 ± 0.09 K. Note that the magnitude of the above temperature differences is substantial, considering that emission controls are only implemented in a single country.

Fig. 2. Global mean differences in key climate variables under China’s carbon neutrality and clean air (CNCA) with respect to the business-as-usual (BAU) scenario for 2050–2070.

(A) and (B) show the overall differences in (A) surface temperature and (B) top-of-atmosphere (TOA) net radiation under these policies relative to the BAU scenario (CNCA−BAU), as well as the respective differences induced by emission reductions of CO₂ (SENS − BAU, where SENS denotes sensitivity scenario) and air pollutants (CNCA−SENS). Panel (C) shows the Effective Radiative Forcing (ERF) due to the combined and individual emission reductions of air pollutants, including sulfur dioxide (SO₂), primary organic matter (POM), black carbon (BC), and volatile organic compounds (VOCs). Note that results in (A) and (B) are derived from fully coupled simulations while those in (C) are derived from atmosphere-only simulations (see Methods). Interpretation of the raincloud plots is the same as that in Fig. 1. The numbers on top of the violin plots represent mean values. Source data are provided with this paper.

We further investigate how much individual air pollutant contributes to global climate, by quantifying their global mean effective radiative forcing (ERF). The details of additional sensitivity experiments can be found in Methods. Figure 2C illustrates that the global mean ERF associated with the emission reductions of SO₂, primary organic matter (POM), black carbon (BC), and VOCs are 0.045 ± 0.016 W m⁻², 0.046 ± 0.024 W m⁻², −0.018 ± 0.020 W m⁻², and 0.017 ± 0.023 W m⁻². These results clearly show that SO₂ and POM reductions are the dominant contributors to positive ERF, namely, reductions in sulfate- and POM-induced cooling drive most of the warming effect associated with air pollutant reductions. BC reductions contribute a negative ERF due to the removal of a warming agent, which partly offsets the sulfate- and POM-driven warming. VOC reductions contribute a smaller but non-negligible positive ERF, indicating that the decrease of secondary organic aerosol (SOA) also contributes to the warming effect.

Precipitation is another key climate metric that affects water availability, agriculture, and the overall health of ecosystems. Under CNCA policies, the CO₂ reduction decreases global mean precipitation by 1.84 ± 1.53 mm year⁻¹ for 2050–2070, while the air pollutant reductions increase global mean precipitation remarkably by 2.28 ± 2.13 mm year⁻¹ relative to BAU (Supplementary Fig. 3). The larger sensitivity to the air pollution reduction results in a small net precipitation increase for 2050–2070. The changes in precipitation are closely tied to the aforementioned warming and cooling of surface temperature, since global warming (cooling) leads to more (less) evaporation and subsequently more (less) precipitation²⁵.

Spatial changes in climate variables for 2050–2070 due to control policies

The spatial pattern of temperature difference has important implications for assessing its regional impacts and developing mitigation and adaptation policies. While the emission reductions occur exclusively in China, the temperature and radiation differences exhibit an obvious global pattern (Fig. 3A–B,D–E). A prominent feature is the polar amplification (especially the Arctic amplification), where the mean temperature differences induced by CO₂ and air pollutant reductions in the Arctic region (−0.30 K and 0.47 K) are 1.9–3.9 times the global average values (−0.16 K and 0.12 K). While polar amplification has been previously reported in climate assessments of global emission changes^18,19, our study indicates that such a pattern persists even when only China’s emission reductions are implemented. A likely explanation is that, following either CO₂ and air pollutant reductions, energy perturbations originating at low- and mid-latitudes can be propagated poleward and get amplified through a positive feedback loop involving incoming energy surplus, sea ice melting, surface albedo, and surface temperature over the Arctic^26–28. Such feedback is confirmed by the simulated opposite albedo differences to temperature differences (see Supplementary Fig. 4). The stronger Arctic amplification due to air pollutant reductions than that due to CO₂ reduction arises because the concentrated energy perturbations at northern mid-latitudes in the former case enhance poleward energy transport.

Fig. 3. Spatial differences in key climate variables under China’s carbon neutrality and clean air (CNCA) policies with respect to the business-as-usual (BAU) scenario for 2050–2070.

A–B, D–E Spatial distribution of the differences in (A–B) surface temperature and (D–E) top-of-atmosphere (TOA) net radiation due to emission reductions in (A, D) CO₂ (SENS−BAU, where SENS denotes sensitivity scenario) and (B, E) air pollutants (CNCA−SENS). C, F Meridional distribution of the differences in (C) surface temperature and (F) TOA net radiation under the CNCA policies relative to BAU (CNCA−BAU), as well as the respective differences induced by emission reductions of CO₂ (SENS−BAU) and air pollutants (CNCA−SENS). Stippled areas in the spatial maps indicate statistical significance at the 95% confidence level based on a two-tailed Student’s t-test. Error bars in the bar charts represent the 95% confidence intervals, also based on a two-tailed Student’s t-test. Source data are provided with this paper.

Another important feature is the hemispheric contrast in the temperature differences. Figure 3C shows the surface temperature differences averaged over different latitude bands in response to China’s CO₂ and air pollutant reductions relative to BAU. The warming induced by air pollutant reductions is much greater in the Northern Hemisphere (0.22 ± 0.12 K) than in the Southern Hemisphere (0.02 ± 0.09 K), mainly because the dedicated emission reductions in China and the short lifetime of air pollutants result in their localized warming effect in the Northern Hemisphere. In contrast, the cooling effect of CO₂ reduction has a relatively small hemispheric difference, since CO₂ concentrations are much more evenly distributed over the world. The overall CO₂ and air pollutant reductions result in contrasting temperature responses in the extratropical regions, with a net warming effect in the Northern Hemisphere and a cooling effect in the Southern Hemisphere (both statistically significant with p-values of 0.007 and <0.001, respectively). This hemispheric asymmetry is consistent with recent observations of declining fine-mode aerosols globally and a greater trend in absorbed solar radiation (ASR) increase in the Northern Hemisphere relative to the Southern Hemisphere^29,30.

In addition to the global pattern of climate impacts, regional hotspots of positive TOA radiation and surface temperature changes due to air pollutant reductions occur in China and the downwind North Pacific region (Fig. 3B, E), since the short lifespan of aerosols causes their concentration changes to be localized near the source (Fig. 1D, Supplementary Fig. 2). Below we explain the reasons behind such changes. Figure 4A indicates that China’s air pollutant reductions cause a large decrease in aerosol optical depth (AOD), contributing to a positive clear-sky TOA radiation change via ARI. Moreover, following the aerosol loading decrease, the cloud droplet number decreases while the cloud droplet radius increases (Fig. 4B, C), in line with the aerosol first indirect effect. The cloud water path also increases (Fig. 4D), probably through the aerosol second indirect effect, where the increased droplet radius enhances the collision and coalescence efficiency and fosters precipitation formation, leaving less cloud water in the atmosphere. The altered cloud radiative effect (defined as all-sky minus clear-sky radiative fluxes) by anthropogenic aerosol changes contributes to a positive TOA radiation change (Fig. 4F), which far exceeds the clear-sky radiation change (Fig. 4E). This suggests that ACI dominates over ARI in producing the positive TOA and temperature changes. It occurs both near China, where aerosol changes are greatest, and far downwind over the Pacific, where there are few other sources of particles.

Fig. 4. Spatial distribution of changes in aerosol and cloud properties as well as radiation fluxes due to air pollutant reductions for 2050–2070.

The six panels depict the changes in (A) aerosol optical depth, (B) cloud droplet number concentration (C) cloud droplet radius, (D) cloud water path, (E) clear-sky net shortwave flux at top-of-atmosphere (TOA), and (F) cloud radiative effects at TOA, due to air pollutant emission reductions. Stippled areas indicate statistical significance at the 95% confidence level based on a two-tailed Student’s t-test. Source data are provided with this paper.

For precipitation differences induced by CO₂ and air pollutant reductions, hotspots mainly occur over the tropical Pacific (Supplementary Fig. 5A, B). The air pollutant reductions induce a strong precipitation increase in the northern tropics and a decrease in the southern tropics—a more prominent north-south contrast compared to the precipitation changes induced by global air pollutant changes^31,32. Such a precipitation shift occurs because the stronger aerosol-induced heating in the Northern Hemisphere relative to the Southern Hemisphere causes a northward migration of the intertropical convergence zone (ITCZ). A diagnosis of the zonal mean position of the ITCZ following an existing method³¹ confirms that ITCZ is displaced northward by 0.35 degree (statistically significant with a p-value of <0.001). Compared to air pollutants, the CO₂ reduction induces a relatively weaker north-south contrast, with moderate precipitation decrease in the northern tropics and increase in the southern tropics, corresponding to a slight 0.14-degree southward displacement of ITCZ (not statistically significant). Averaged over hemispheres, the overall CO₂ and air pollutant reductions (Supplementary Fig. 5C) result in a net positive precipitation change in the Northern Hemisphere and a negative precipitation change in the Southern Hemisphere (both statistically significant with p-values < 0.001). A more in-depth discussion of the spatial precipitation changes in given in Supplementary Note 2.

Temporal evolutions of temperature and precipitation responses

From 2015 through 2100, the global CO₂ concentration reduction relative to the BAU scenario due to the CNCA policies (CNCA−BAU) keeps increasing swiftly, while the PM_2.5 burden reduction in China increases rapidly through ~2050 and stabilizes thereafter (Fig. 5A). The difference can be explained by the fact that CO₂ concentration hinges on the cumulative emission reduction over many decades, while the concentration of PM_2.5, a short-lived forcing agent, primarily depends on air pollutant reductions in a given year. After ~2055, the CO₂ concentration reduction continues to grow due to the ongoing accumulation of CO₂ emission reduction and the use of negative carbon technologies; nevertheless, the PM_2.5 reduction stabilizes as the potential for further reducing air pollutants relative to BAU is largely exhausted (Supplementary Fig. 1). Note that the PM_2.5 reduction even slightly decreases after 2075 (Fig. 5A), as air pollutant emissions in BAU continue to decline slowly with time due to gradual decarbonization (1.0–2.5% per year), while the air pollutant emissions in the CNCA scenario remain almost constant due to the limited reduction potential (Supplementary Fig. 1). Accordingly, Fig. 5B shows that the global temperature increase driven by air pollutant reductions intensifies until around 2055–2060 and then levels off, while the temperature decrease from CO₂ reduction accelerates over time. The stabilization of air pollutant-induced temperature increase occurs later than the point at which PM_2.5 burden reductions roughly plateau (2045–2050), largely due to the ocean’s thermal inertia and its slow uptake of heat. As a result, before ~2070, the net global temperature difference under CNCA policies relative to BAU is not significantly different from zero, but after ~2070, the net temperature difference becomes significantly negative (p-value < 0.001) and reaches −0.21 K for 2070–2100.

Fig. 5. Temporal evolution of climate responses induced by China’s carbon neutrality and clean air (CNCA) policies.

A Differences in global mean CO₂ concentration and China mean PM_2.5 burden under the CNCA policies relative to business-as-usual (BAU) (CNCA−BAU). B Differences in global mean surface temperature caused by CO₂ reduction (blue, SENS−BAU, where SENS denotes sensitivity scenario), air pollutant reductions (red, CNCA−SENS), and their combined effect (black, CNCA−BAU). The lines represent the mean differences between two scenarios, and the shaded areas indicate one standard deviation of 25 samples obtained by pairwise differencing of five ensembles under the two scenarios. All values shown are 10-year running means. Source data are provided with this paper.

The time evolution of net global precipitation difference (Supplementary Fig. 6A) closely resembles the trends in surface temperature, where air pollutant reductions drive an intensifying and then stabilizing precipitation increase, and CO₂ reduction drives an accelerating precipitation decrease. Slightly different from temperature, the net difference in global precipitation under CNCA policies relative to BAU remains insignificant until around 2080 (cf. ~2070 for temperature), after which a significant negative difference emerges. This indicates that, compared to temperature, air pollutant reductions tend to have a greater efficiency in increasing precipitation, probably because precipitation can be more directly modified by aerosols via ACI. In China, where the precipitation difference is much larger than the global average, the precipitation increase due to air pollutant reductions significantly outweighs the precipitation decrease from CO₂ reduction throughout the simulation period, resulting in a persistent net precipitation enhancement (Supplementary Fig. 6B).

The above findings update our understanding on the global climate impacts of regional decarbonization and clean air policies. The key conclusion that warming from air pollutant reductions can largely offset the cooling from CO₂ reductions for about half a century through ~2070 stand in clear contrast to previous studies^11,19,22 on the climate impacts of global and United States emission reductions. Specifically, on the global scale, the CMIP6 simulations showed that the differences in radiative forcing or temperature between different SSPs, which broadly reflect the effect of global greenhouse gas (GHG) mitigation or clean air policies, were mostly dominated by GHG reduction, with air pollutant reductions playing a relatively minor role^11,19. A previous study also revealed that, in the United States, where air pollutant emissions are relatively low, decarbonization policies will quickly contribute to a decrease in temperature before 2030²². The stronger and more persistent warming effect of China’s air pollutant reductions than that revealed by CMIP6 global simulations and United States simulations is probably explained by (1) the higher intensity of air pollutant emissions in China as compared to developed countries and (2) the insufficient consideration of China’s aggressive clean air policies in SSPs. Note that this strong pollutant reduction-induced warming is unlikely to be attributed to the uncertainty of aerosol radiative forcing estimate, since the aerosol effective radiative forcing predicted by CESM2 is at a medium level among all CMIP6 models³³.

Discussion

The findings have important implications for the optimization of future climate and clean air policies in China. Even aggressive climate policies like carbon neutrality are not sufficient to cool the climate by ~2070 due to the compensating effect of air pollutant reductions, although a significant cooling effect is expected to unfold by 2100. We anticipate that the compensating effect would remain strong regardless of whether additional clean air policies were considered, since the synergistic effect of carbon neutrality, rather than the clean air policies, plays a dominant role (~80%) in reducing the emissions of most air pollutants. Moreover, the quest for Beautiful China represents an indispensable and regret-free policy choice given its drastic public health benefits^5,6. For these reasons, the clean-air policies should be firmly upheld without delay. To achieve a net cooling effect earlier and with a maximized amplitude, we suggest that carbon neutrality should be achieved as early as possible and that negative carbon technologies be vigorously promoted in the long run, augmented by policies to rapidly reduce methane emissions³⁴ (see more in Supplementary Note 3). Furthermore, future efforts could optimize the combinations of reduction measures across sectors by simultaneously considering climate benefits from reducing GHG and short-lived climate forcers, as well as public health benefits. This approach would ultimately identify the reduction pathway that delivers the greatest climate and health benefits while achieving carbon neutrality and clean air.

More importantly, our study has implications for sustainable development strategies in many other Global South countries that simultaneously grapple with high CO₂ emissions and severe air pollution, as exemplified by India. Ongoing and future climate and clean air policies in these countries or regions are expected to substantially alleviate PM_2.5 pollution³⁵. The anticipated warming effect could potentially be even larger than that of China, as the relatively lenient control measures currently in place in those countries indicate greater potential for future reductions in air pollutants. The widely used SSPs have not fully accounted for ambitious clean air goals, exemplified by the Beautiful China vision, and potential clean air policies in other Global South countries. In view of the challenges brought by air quality improvement, it is crucial to promote earlier and more stringent GHG reduction strategies worldwide—such as achieving net zero sooner and pursuing negative carbon emissions—to ensure that the temperature targets set by the Paris Agreement are met or the overshoot of those targets is limited in time as much as possible.

Methods

Description of the Earth System Model

In this study, we evaluate the climate impacts of carbon neutrality and clean air policies consistent with the Beautiful China vision, using the Community Earth System Model version 2 (CESM2). We conduct fully coupled simulations with all components activated, including the atmosphere, land, ocean, sea-ice, land-ice, river, and wave components. The atmosphere component CAM6 uses a nominal 1° (1.25° in longitude and 0.9° in latitude) horizontal resolution with 32 vertical levels and a model top at 2.26 hPa (about 40 km).

CAM6 incorporates a range of advanced capabilities and new improvements that are directly relevant to the climate impacts of emission reductions. Aerosols are represented using the Modal Aerosol Model version 4 (MAM4)³⁶ which predicts aerosol compositions, including sulfate, BC, POM, SOA, mineral dust, and sea salt in four lognormal modes (i.e., Aitken, accumulation, coarse, and primary carbon modes). Aerosols are assumed internally mixed within each mode but externally mixed between different modes. Sulfate aerosol is formed from its precursor SO₂ via gas-phase oxidation and aqueous-phase oxidation in bulk cloud water. The SOA formation is parameterized with a single-lumped semivolatile organic gas-phase species called SOAG, regarded as a proxy for the oxidation products of VOCs. The preprocessed surface emissions of SOAG are derived from the emission five primary VOCs, using mass yields of 5% for BIGALK (lumped butanes and larger alkanes), 5% for BIGENE (lumped butenes and larger alkenes), 15% for aromatics, 4% for isoprene, and 25% for monoterpenes, which equivalently simulates the gas-phase oxidation of VOC precursors³⁷. SOAG then undergoes condensation and evaporation into/from the Aitken and accumulation modes. BC and POM are emitted into the primary carbon mode and then transferred to the accumulation mode via aging. As with most Earth system models, nitrate and ammonium are not included in the current model, which needs to be improved in future studies. The Rapid Radiative Transfer Model for General circulation models (RRTMG) is used for radiative transfer calculations. ARI is computed online by feeding the optical properties (including scattering, absorption, and asymmetry parameters) estimated by MAM4 to RRTMG to assess shortwave and longwave radiative effects. An updated version of the Morrison-Gettelman cloud microphysics scheme (MG2)³⁸ is employed to treat the microphysical processes of stratiform clouds, including ACI. In MG2, aerosol activation follows the parameterization of Abdul-Razzak and Ghan³⁹, which converts aerosol number concentrations into cloud droplet number concentrations. This subsequently influences both the droplet size and cloud albedo (the “first indirect effect”) and the efficiency of cloud water converting into rainwater, thereby affecting cloud lifetime and precipitation processes (the “second indirect effect”). Moreover, the mixed-phase ice nucleation has been improved to depend on both aerosols and temperature^40–42. Another major advance is the inclusion of the unified turbulence scheme, Cloud Layers Unified By Binormals (CLUBB)^43,44, which provides a uniform treatment of clouds across cloud types.

Scenarios and experimental design

We employ emission scenarios for China developed in our previous study⁵ using the GCAM-ABaCAS modeling system, including a BAU scenario (named “NDC-2015” in Sun et al.⁵) and a CNCA scenario (named “NDC-2021-AQE” in Sun et al.⁵). Supplementary Fig. 1 shows that, in 2060, the emissions of SO₂, NO_x, PM_2.5, BC, and organic carbon (OC) under CNCA are 93–97% lower than those under BAU, and VOCs emissions are 78% lower, resulting from a combination of the synergistic effect of carbon neutrality and additional clean air policies. As a result, the population-weighted annual mean PM_2.5 concentration in China under CNCA is estimated to reach 7.3 and 5.3 μg m⁻³ by 2050 and 2060, respectively, consistent with the expected improvement under the Beautiful China vision to attain the 2005 WHO guideline (10 μg m⁻³) and approach the 2021 guideline (5 μg m⁻³). Another independent emission projection⁶ showed that carbon neutrality and clean-air policies could reduce 2060 emissions of SO₂, NO_x and PM_2.5 by 85–90% and VOCs emissions by 63% compared to BAU, broadly consistent with those used in this study. The synergistic effect of carbon neutrality (represented by the difference between the BAU and an intermediate scenario where the carbon neutrality target is implemented but no additional clean air policies are considered, see Supplementary Fig. 1) contributes over 80% of the emission reductions in SO₂, PM_2.5, BC, and OC, and 60% of the emission reductions in VOCs.

We conduct fully coupled CESM2 simulations for 2015–2100 for the BAU and CNCA scenarios, as well as a sensitivity scenario (SENS) to separate the individual impacts of CO₂ and air pollutant reductions. In the SENS scenario, the CO₂ emissions and air pollutant emissions are set to the levels of the CNCA scenario and the BAU scenario, respectively. The total impact of CO₂ and air pollutant reductions under carbon neutrality and clean air policies aligned with the Beautiful China vision is obtained by differencing the CNCA and BAU scenarios (CNCA−BAU). The impact of CO₂ reduction is obtained by differencing the SENS and BAU scenarios (SENS−BAU) and that of air pollutant reductions is achieved by differencing the CNCA and SENS scenarios (CNCA−SENS).

For all scenarios (BAU, CNCA, and SENS), the fully coupled CESM2 simulations are transient runs initialized on January 1, 2015. The initial conditions are derived from CESM2 historical-era simulations that used standard SSP2-4.5 historical forcings. For each scenario, five ensemble members are performed with a small initial perturbation to atmospheric temperature, to minimize the impact of internal variability in the climate system. In all simulations, the anthropogenic emissions of CO₂ and air pollutants outside China, as well as global biomass burning emissions follow the default SSP2-4.5 scenario, which represents the medium level within the range of future forcing pathways. The base-year emissions in SSP2-4.5 are obtained from the Community Emissions Data System (CEDS)^44–46. The SSP2-4.5 scenario represents a medium pathway within the range of future forcing trajectories, which we consider to best reflect the most likely future emission trajectory outside China. Most analyses in this study focus on the time around 2060 (represented by the average of 2050–2070 to reduce the influence of internal variability), while the temporal evolution trends until 2100 are also examined.

In addition to the fully coupled simulations, we conduct a separate set of atmosphere-only sensitivity experiments to quantify the contribution of individual air pollutant emission changes to the ERF. These experiments are driven by prescribed present-day monthly sea surface temperatures and sea ice concentrations, allowing the differences between experiments to represent the ERF. The horizontal winds in these experiments are nudged towards reanalysis data in 2010, which helps to identify the ERF within a relatively short simulation time^24,47. The simulations are integrated for 10 years with the average of the last 9 years used for ERF estimates. In these experiments, the emissions of SO₂ (together with primary sulfate), POM, BC, and VOCs are sequentially adjusted from their BAU levels to the CNCA levels, with each experiment differing from the previous one by the change of a single pollutant. Therefore, by taking the difference between each pair of adjacent experiments, we can quantify the contribution of each pollutant to ERF. The CO₂ concentrations in all these experiments are prescribed based on the fully coupled simulation results under CNCA and are fed into the model via lower boundary conditions. Other configurations are the same as the fully coupled experiments.

Note that this study has not yet considered the changes in non-CO₂ greenhouse gases (such as methane and nitrous oxide) or ozone due to climate or clean air policies (see more in Supplementary Note 3). Future studies could further benefit from including these forcing agents for a more comprehensive modeling analysis, although the Chinese government has not set specific targets for non-CO₂ greenhouse gases, and the climate impact of ozone is considered to be smaller compared to CO₂ and aerosols⁴⁸.

Model evaluations

We compare simulated mean surface temperature for 2015–2020 with MERRA2 reanalysis at a 0.25° × 0.25° resolution. The CESM2 model shows strong agreement with MERRA2 reanalysis data for the global patterns of both temperature and precipitation, with the model effectively capturing major temperature gradients and precipitation features (Supplementary Fig. 7A–D). The temperature comparison demonstrates particularly high consistency with an R² value of 0.9914, while precipitation shows moderate agreement with an R² of 0.7492 (Supplementary Fig. 7E, F). CESM2 also performs well in reproducing the overall evolution of global historical temperature in the 20^th century. Danabasoglu et al.⁴⁹ show that, compared with observational datasets such as HADCRU4.5, GISTEMP and NOAAGlobalTemp, CESM2 successfully captures key multi-decadal evolution of global mean surface temperature change, including the early warming from 1920–1940, the mid-century warming hiatus from 1940–1970, and the accelerated warming after the 1970s (see Fig. 7 of Danabasoglu et al.⁴⁹).

The performance of CESM2 in simulating aerosol concentrations has been evaluated in previous studies^49–51. We use the same emissions as these studies, i.e., the CEDS (CMIP6) emission inventory, except that we replace the default Chinese emissions with our localized emission inventory derived from the GCAM-ABaCAS system and calibrate it against the ABaCAS-EI. For this reason, we further evaluate simulated concentrations of key PM_2.5 components, including sulfate, BC, and organic aerosol, against observations at more than 40 sites in eastern China—the main polluted region of China. These monitoring sites are distributed in major polluted regions including Beijing–Tianjin–Hebei, Shandong, Henan⁵², and Shanghai⁵³, covering both urban and mountainous areas. At these sites, we analyze online measurements of OC/EC and ionic species using OC/EC analyzers and ion chromatography systems during 2018 and 2019. The observational data sources and detailed measurement methodology are described in previous studies^52–54. Supplementary Fig. 8 shows that simulated concentrations of sulfate, BC, and organic aerosol are generally within a factor of 3, 2, and 2 of the observed values across various sites. The normalized mean biases (NMBs) for the three components are −46%, 12%, and −30%, respectively, indicating reasonable model-measurement agreement. The underestimation of sulfate suggests that the warming effects due to air pollutant reductions could be stronger than our current simulation results. Accordingly, the global surface temperature increase from China’s air pollutant reductions may offset the cooling from China’s CO₂ reduction for an even longer time.

Building on this, we further examine CESM2’s ability to simulate the separate responses to GHG and aerosols—two key forcing factors. Although their individual contributions cannot be directly verified through observations, we can indirectly assess the model by analyzing their roles in different historical phases of global temperature change. From the CESM2 single forcing large ensemble experiments, which isolate historical climate forcings (such as GHG and aerosols), the model reveals the independent effect of each forcing (see Fig. 2 of Simpson et al.⁵⁵). The results show that under GHG-only forcing, the model simulates a continuous warming trend, consistent with the physical understanding of GHG as the main driver of warming. Under anthropogenic aerosol-only forcing, the model produces a clear cooling signal, with the strongest cooling in the mid-20^th century, gradually weakening afterward due to the decrease in aerosol loading. When both GHG warming and aerosol cooling are combined, CESM2 successfully reproduces the phased characteristics of historical temperature change, i.e., mid-20^th-century warming hiatus and accelerated warming in the late 20^th century. This indicates that aerosols contributed significantly to the mid-century warming slowdown, while the continued strengthening of GHG impact and the weakening of aerosol impact drove later temperature rise. These results also align with previous attribution studies⁵⁶. This suggests that CESM2 can not only reproduce the overall historical temperature curve but also capture the separate effects of GHG and aerosols on temperature change from a physical mechanism perspective. Additionally, the effective radiative forcing of GHG and aerosols simulated by CESM2 falls within the range of the CMIP6 multi-model ensemble⁵⁷, further supporting the reasonableness of its forcing representations. Evaluation of precipitation simulations by CESM2 is given in Supplementary Note 4.

Author contributions

B.Z., Y.W., and S.W. conceived the study. B.Z., X.W., Y.W., Y.S., D.G., Y.G., J.Z., Y.Z., G.L., and S.W. developed the methods used in this study. B.Z., X.W., Y.W., Y.S., D.G., Q.G., Y.G., J.Z., Y.Z., D.S., S.J.D., G.L., Y.h.W., B.C., H.H., K.H., J.H., and S.W. carried out the research. B.Z. and X.W. wrote the original draft of the manuscript. B.Z., X.W., Y.W., Y.S., D.G., Q.G., Y.G., J.Z., Y.Z., D.S., S.J.D., G.L., Y.h.W., B.C., Z.J., H.H., K.H., J.H., and S.W. revised the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

Source Data corresponding to all figures in this paper have been provided in the Source Data file. In addition, the processed datasets, including emission inventories for the Business-As-Usual (BAU) and China Carbon Neutrality and Clean Air (CNCA) scenarios, as well as key model outputs covering the 2050-2069 period with 5 ensemble members, have been deposited in Figshare and are accessible via 10.6084/m9.figshare.30511721. Source data are provided with this paper.

Code availability

The codes of the CESM2 model used in this study are available at https://www.cesm.ucar.edu/models/cesm2.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68586-y.