Air pollution mitigation in North China through flexible heating policies

Central heating in North China produces severe air pollution, although the need for heating may be reduced by rising temperatures associated with climate change. The regional trend of mean heating length (HL) for North China was −0.32 d per year during 1961–2019. Compared with the 2010–2015 mean values, the start and end dates for central heating in the North China Plain (NCP) during 2050–2055 will delayed by 9 d and advanced by 12 d, respectively, under the Shared Socioeconomic Pathway 5–8.5 (SSP5-85), and by 5 and 8 d under the carbon-neutral (CN) scenario, based on Coupled Model Intercomparison Project phase 6 model simulations. Here we propose a flexible heating policy (FHP), such that HL is determined strictly by temperature, and the associated air pollution benefit of shortening HL are examined by a global 3D chemical transport model GEOS-Chem. The study focused on the year 2019 with the current goal of elimination of severe PM2.5 pollution, and with the minimum HL estimated to provide up to a 24% reduction in severe PM2.5 pollution (daily mean PM2.5 > 150 μg m−3) over the NCP during periods of FHP implementation. For future CN policies, the NCP can achieve great air quality improvements by 2050, with more than 60% of days throughout the heating season with daily PM2.5 concentrations of <10 μg m−3, and 95% with <35 μg m−3. Although the SSP5-85 scenario may lead to reduced HLs, pollutant emissions are likely much higher than under CN scenarios, with pollution days of PM2.5> 100 μg m−3 still occurring frequently by 2050. Our results highlight that FHPs may effectively reduce severe PM2.5 pollution, and China’s carbon neutrality goals will play critical roles in mitigating air pollution and prolonged heating welfare during future heating season.


Introduction
North China is in a central-heating zone because of low temperatures in the boreal winter. The heating period refers the period each year when centralheating systems provide heat to local homes and is also associated with the bulk of residential and industrial energy consumption. Owing to rapid urbanization and population growth, China's energy consumption and emissions from residential central heating have increased dramatically over the last few decades (Zhang et al 2020a). As a result, the combustion of fossil fuels for winter heating has become a major air-quality and public-health concern (Xiao et al 2015). High PM 2.5 (particles with an aerodynamic diameter smaller or equal to 2.5 micrometers) exposure and premature mortality in China have been attributed in large part to residential coal combustion in rural areas (Tao et al 2018, Zhao et al 2018. Previous findings have indicated that winter heating is one of the major causes of China's severe particulate pollution, with heating contributing up to 13.7% of PM 2.5 emissions in Beijing (Yuan et al 2015). The control of residential sector emissions in the Beijing-Tianjin-Hebei region could significantly improve regional PM 2.5 air quality . The promotion of clean-burning fuels in the residential sector could lead to a 13% reduction in national PM 2.5 emissions and a 12% reduction in premature deaths nationwide (Zhang et al 2019). Since 2013, China has introduced the Clean Air Action Plan (2013)(2014)(2015)(2016)(2017) and Blue Sky Protection Campaign (2018)(2019)(2020) to mitigate air pollution, and national annual mean PM 2.5 concentrations decreased by 32% between 2013 and 2017 owing to regulatory control of power and industry sector emissions. The implementation of residential clean-heating programs also resulted in a rapid decrease in winter PM 2.5 levels in North China during 2018-2020 (Li et al 2021a), indicating that emissions from the residential sector are a major contributor to PM 2.5 pollution.
The winter heating region of China comprises 15 main provinces, typically referred to as being north of the Qinling-Huaihe Line, where low temperatures are most severe (figure 1). Heating demand is directly influenced by near-surface temperatures, which are sensitive to climate change (Chen et al 2021). Previous studies have explored historical and future variations in heating indices, including the Heating Degree Day, the length of heating period (heating length (HL)), the start date for central heating (heating start (HS)), and the end date for central heating (heating end (HE)). Based on daily mean temperatures over 437 stations in China, Shen and Liu (2016)  First, most previous studies regarding climate-change effects on heating indices placed more emphasis on high-carbon-emission scenarios (i.e. RCP4.5 and RCP8.5). As China is likely to take a low-carbonemission path after the proposal of the carbonneutral (CN) policy, it is essential to quantify variations in heating indices under low-carbon-emission scenarios. Second, no previous studies have investigated the associated air-pollution mitigation benefits.
In North China, the heating season is typically from October-November to March-April. Although local governments can decide when heating services start based on local weather conditions, such regulations are not strictly promoted in the central-heating zone in North China. Given that climate change can alter HL substantially, we hypothesize that the implementation of a flexible heating policy (FHP) where the heating period is determined strictly by temperature may achieve significant air pollution reductions.
Here we applied the GEOS-Chem model (Bey et al 2001) to assess such benefits from two points of view. First, for the current goal of eliminating severe PM 2.5 pollution before 2025 through the '14th Five-Year Plan' , we selected the year with minimum HL during 2013-2020 and evaluated the maximum extent of severe pollution abatement by FHPs. Second, for the future mid-term (2050), changes in HL were estimated using the Coupled Model Intercomparison Project 6 (CMIP6) model under CN and SSP5-85 (shared socioeconomic pathway 5-8.5) (Riahi et al 2017), scenarios. Associated air pollution mitigation benefits were investigated.

Data and methods
2.1. Near-surface air temperature CN05.1 gridded daily mean near-surface air temperature data provided by the National Climate Center of China were used, with a spatial resolution of 0.25 • × 0.25 • for the years 1961-2020 (Wu and Gao 2013). CN05.1 datasets were derived by interpolation of data from 2416 observation stations in China and have been widely used in studies of climate extremes including extreme precipitation (He et al 2021) and heatwaves (Wang et al 2022). Future climate simulation data for near-surface air temperatures were derived from 35 CMIP6 models (table S1). All model outputs were interpolated into a common 2.5 • × 2.5 • grid using the bilinear interpolation method (Li et al 2021b). Projections under SSP5-85 and SSP1-26 were considered as representing high carbon emission and carbon neutrality, respectively.

HS and HE
Beijing's central-heating regulations indicate that if the temperature is below 5 • C for five consecutive days, heating can be started (http://service.law-star. com/cacnew/201011/440067099.htm). Accordingly, HS denotes the last date of the period when the daily mean temperature is <5 • C for five consecutive days, and HE is the last date when the mean temperature is >5 • C for five consecutive days for the first time during the following year. HL is thus determined by HS and HE.

GEOS-chem model
Air quality over the North China Plain (NCP) during the heating season was simulated using the GEOS-Chem model (version 11.1; http://acmg.seas.harvard. edu/geos/) driven by the US NASA Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2), with meteorological data from NASA's Global Modeling and Assimilation Office (GMAO) (Gelaro et al 2017). The nested-grid capacity of GEOS-Chem over Asia (11 • S-55 • N, 60 • -150 • E) was used with a horizontal resolution of 0.5 • latitude by 0.625 • longitude and 47 layers up to an altitude of 0.01 hPa. Boundary conditions for gases and aerosols were updated every 3 h from coupled global GEOS-Chem simulations performed at 2 • × 2.5 • horizontal resolution. PM 2.5 concentrations were calculated as the sums of simulated mass concentrations of sulfate, nitrate, ammonium, black carbon (BC), and organic carbon (OC) (Dang and Liao 2019).

Residential emissions
For anthropogenic emissions, the Multi-resolution Emission Inventory of China (MEIC, http:// meicmodel.org) was used (Li et al 2017a, Zheng et al 2018, and the MIX (a mosaic anthropogenic emission inventory over Asia) emission inventory (Li et al 2017b) was used for rest domains. For projected anthropogenic emissions in 2050, the Dynamic Projection for Emission in China (DPEC) was obtained (Tong et al 2020). The DPEC was developed to explore China's future anthropogenic emission pathways and had integrated the energy system model, emission inventory model, dynamic projection model, and parameterized scheme of Chinese policies. Compared to CMIP6 emissions, the DPEC are more accurate in predicting future variations in emissions of China due to a comprehensive consideration of detailed local environmental and climate policies. All three emissions (MEIC, MIX, and DPEC) inventories provide monthly anthropogenic emissions of major gaseous and particulate pollutants from industry, transport, power, agriculture, and residential sectors. Monthly emissions of total BC, OC, nitrogen oxides (NOx), sulfur dioxide (SO 2 ), carbon monoxide (CO), and primary PM 2.5 in the NCP from the residential sector are compared in figure S1 for different scenarios. During the NCP heating season, emissions from the residential sector increase sharply owing to coal combustion for central heating, whereas variations in emissions are minimal during the non-heating season (figure S1). For each simulation (section 2.5), residential emissions outside the central-heating period were considered as those averaged over the nonheating season.

Statistical methods
Ground-based PM 2.5 measurements undertaken by the China Ministry of Ecology and Environment (http://106.37.208.233:20035/) at 39 observation cities in the NCP were used to assess model performance in simulating PM 2.5 concentrations (figure 1). Following the quality control criterion in a previous study (Li et al 2019), the original hourly data were processed to daily time series for each city. The performance of the WRF-Chem model was assessed using the correlation coefficient (R), normalized mean bias (NMB), and index of agreement (IOA). The IOA is a scale of 0-1 that indicates how close model results and observations are to being in perfect agreement. In general, the performance of the model is excellent if R and IOA are close to 1 and NMB is close to 0. Moreover, the Bland-Altman approach (Bland and Altman 1986) is used to assess the difference between simulation and observation within 95% limits of agreement of the mean difference (1.96 of the standard deviation). The Bland-Altman plot illustrates the difference (y-axis) as a function of the average (x-axis), allowing additional information to be presented about the error structure. Metric formulae are as follows: where M i and O i denote simulated and observed PM 2.5 concentrations, respectively; N indicates the total number of daily samples compared; andM andŌ represent averaged simulations and observations, respectively.  year, respectively, with greater reductions than for HS. Overall, all 15 northern provinces exhibited consistent variations with delay of HS and advance of HE, consistent with the findings of previous studies (Shen andLiu 2016, Chen et al 2021). The heating season generally started in October-November and ended in March-April. HS and HE are thus dominated respectively by autumnal and spring temperatures (Chen et al 2021). We investigated the long-term trends in surface temperature from 1960 to 2020 for spring, summer, autumn, and winter. Spring warming trends were generally more marked than those of autumn (figure S3), possibly explaining why the linear trend in HE is steeper than that in HS. The whole of North China displayed a significant reduction in HL from 1960 to 2019 (figure 2(c)), indicating the influence of global warming. The regional mean trend of HL in the NCP was −0.32 d per year from 1961 to 2019. We also examined the mean absolute deviation (MAD) in HL, which has received little attention in previous studies. Interannual variations in HL have a varied distribution across North China, with a highest MAD value of 25-30 d in southern Qinghai Province and the Hexi Corridor. For the NCP, the regional mean MAD was approximately 10 d.

Historical changes in HS, HE, and HL
Changes in HS and HE during 2011-2020 are compared with changes during 1961-1970 in figure   S4. Over the central-heating zone, regions with earlier HS are mainly those north of Heilongjiang Province and the Qinghai Plateau, where heating is required as early as the beginning of October ( figure S4(a)). The regions with early HS typically correspond to those with late HE, which occurs at around the end of April ( figure S4(b)). During 2011-2020, HS for the NCP was delayed to the last ten days of November, and HE was brought forward to before March 15, which is the current official HE. Compared with 1961-1970, the regional mean HL in the NCP decreased by 15.8 d during 2011-2020. These results indicate the flexibility of the duration of heating over the NCP.

Future projections of heating period
Future changes in HS and HE under a CN scenario are indicated in figure 3. Climate change may substantially alter HL by delaying HS and advancing HE in the future medium-term (2050-2055) relative to the historical period of 2010-2015. The most significant delay of HS may occur in the southern Qinghai province (figure 3(a)), with a maximum delay of 15-20 d. The spatial pattern of likely changes in HE in North China is similar to that of HS, except that the greatest changes occur in southern Shandong and Henan provinces. For the NCP, the latest HS is projected to be later than December 1 during 2050-2055, and the earliest HE before March 11. Spatial changes in HS and HE are similar under the SSP1-26 and SSP5-85 scenarios, but with larger magnitudes in the latter (figure S5). Averaged over the whole NCP, HS and HE changes are estimated to be 5 and 8 d under a CN scenario, and 9 and 12 d under an SSP5-85 scenario, respectively. These changes are considered in future simulation by the GEOS-Chem model (table 1).

Air pollution mitigation with a FHP
Here we first evaluate model performances in the reproduction of PM 2.5 spatial patterns and temporal evolutions in the Base_13, Base_17, and Base_19 experiments. Spatial distributions of simulated and observed average near-surface PM 2.5 concentrations for November-March in 2013, 2017, and 2019 in the NCP are shown in figure 4(a). In 2013, the simulated average near-surface PM 2.5 concentrations exceeded 100 µg m −3 over large areas of the NCP. The underestimation over the western NCP is likely due to the coarse spatial resolution of GEOS-Chem (0.5 • × 0.625 • ), which cannot properly resolve the complex topography of the Taihang mountains (Long et al 2016). Since 2013, anthropogenic emissions have decreased substantially through clean-air actions (Zhang et al 2019). In 2017 and 2019, the simulated mean PM 2.5 concentrations were much lower than in 2013. Overall, the simulated PM 2.5 distributions were generally consistent with observations, albeit with slight model biases. The temporal evolution of simulated and observed near-surface PM 2.5 concentrations is described in figure 4  Differences in mean PM 2.5 concentrations and number of severe PM 2.5 pollution days (SPPDs) between the Base_19 and FHP_19 experiments during heating adjustment periods (15-22 November, 2019, and6-15 March, 2020) are compared in figure 5. Southern Hebei province had the greatest improvement in air quality, with daily mean PM 2.5 concentrations decreasing by as much as 9 µg m −3 . The adoption of FHPs effectively reduced the risk of SPPDs. When the entire NCP was taken into account, the number of SPPDs decreased from 79 to 60 d, a reduction of almost 24%. The '14th Five-Year Plan' for ecological and environmental protection now includes the elimination of severe pollution as a target for several cities in North China by the year 2025. Our findings imply that encouragement of FHPs may aid accomplishment of this objective. Differences in PM 2.5 concentrations between the FHP_CN and FHP_SSP5-85 scenarios during the future heating season (November 15 to March 15, 2050) are shown in figure 6. In the FHP_CN experiment, there was a significant improvement in air quality, with a mean PM 2.5 concentration of <20 µg m −3 during the heating season in the NCP. Mean PM 2.5 concentrations under the SSP5-85 scenario were much higher, even though the HL was lower. The highest PM 2.5 concentrations, exceeding 35 µg m −3 , were in the Beijing area. In the FHP_CN experiment, areas with high PM 2.5 concentrations also experienced a noticeable improvement in air quality. For example, the mean PM 2.5 concentration during the heating season was reduced by up to 30 µg m −3 in the northern NCP. Averaged over the whole NCP, mean PM 2.5 concentrations during the heating season ranged from 8.1 to 10.0 µg m −3 and from 18.4 to 22.6 µg m −3 under the CN and SSP5-85 scenarios ( figure 6(b)), respectively. The PM 2.5 concentration under the CN scenario was about half that under the SSP5-85 scenario due to variations in anthropogenic emissions. The emissions for major pollutants, including BC, OC, NOx, SO 2 , CO, and primary PM 2.5 , decreased more dramatically under a CN scenario in 2050 relative to a SSP5-85 scenario (figure S7). No SPPDs were detected during the heating season of 2050. However, high probabilities of pollution days with daily PM 2.5 exceeding 100 µg m −3 remain for the SSP5-85 scenario. Under the CN scenario, high air quality would be realized on >60% of days throughout the heating season with daily PM 2.5 concentrations of <10 µg m −3 , and on 95% of days with concentrations of <35 µg m −3 (figure 6(c)). The CN scenario may thus have a substantial positive impact on human health through reductions in PM 2.5 pollution, although more effort would be required to reduce PM 2.5 concentrations to the new World Health Organization recommended value of 5 µg m −3 (World Health Organization 2021).

Discussion and implications
There are inevitably various uncertainties and limitations in our investigation. First, future near-surface air temperature projections under the SSP1-26 scenario do not strictly match CN goals. We used future climate data from SSP1-26 (involving 35 models) because insufficient data were available under SSP1-19 (using only 15 models) in CMIP6. The CO 2 emission pathway under China's ambitious carbon neutrality goals was found to lie between those of SSP1-26 and SSP1-19 when compared with global climate predictions (Cheng et al 2021). Therefore, the surface temperature under SSP1-26 is expected to be slightly higher than that under the CN scenario, with future changes in HL possibly being overestimated. Second, there also existed biases from input emission. The relatively poor performance in 2017 may be attributed to the uncertainty from emission inventories. During the winter of 2017, the Chinese government implemented stringent emission controls to achieve the goal of Air Pollution Prevention and Control Action Plan, which started in 2013 and the first phase of plan ended in 2017. The residential emissions for the adjusted non-heating days during heating periods were configured roughly as averages over the non-heating season (April-October), which may lead to uncertainties. Third, unique changes in HS and HE were taken into account for the entire NCP in model simulations, resulting in uncertainties for subregions. The spatial distributions of future changes for both HS and HE are not uniform over the NCP, with southern NCP areas exhibiting larger changes than the northern NCP (figure S6). Changes in HL were thus under-and over-estimated for the southern and northern NCP, respectively.
This study investigated PM 2.5 pollution reduction through implementation of FHPs in the contexts of current and future climate-change scenarios, with two suggestions arising for air-quality improvement during the heating season in North China. First, the promotion of FHP may aid achievement of the goal of eliminating severe PM 2.5 pollution by 2025. Our analysis indicates that an FHP may lead to a 24% reduction in severe PM 2.5 pollution over the NCP. In addition to air-quality improvement, the FHP is also expected to reduce mortality risks caused by exposure to highly polluted environments. Second, the government should follow the CN scenario to promote further air-quality improvement. Emissions of major air pollutants (BC, OC, SO 2 , NOx, and PM 2.5 ) will decline sharply under the CN pathway (figure S7), and the national mean PM 2.5 concentration would be below 10 µg m −3 by 2060 as a consequence of clean energy structures and strict clean-air policies (Cheng et al 2021), with HL variations barely affecting air quality and allowing longer heating periods for public welfare.

Conclusion
During the heating season of the boreal winter, increasing coal consumption due to central heating is one of the primary causes of severe PM 2.5 pollution in North China. Given the causal relationship between air temperature and HL, increased nearsurface air temperatures during climate change is likely to shorten HL in North China. It is therefore essential to quantify the air quality benefit under different climate scenarios in order to guide future central heating policy. Based on daily temperature observations during 1961-2020 and outputs of the CMIP6 climate model for 2020-2050, historical and future changes in three heating indices (HS, HE, and HL) were assessed. The GOES-Chem model was used to evaluate the benefits to air quality of promoting FHPs under various climate-change scenarios.
The results indicate that HL over North China decreased during 1961-2020 owing to delayed HS and advanced HE. Changes in HE made the greatest contribution because of the increasing spring warming rate. For the most polluted NCP, the regional mean HL decreased by 15.8 d during 2011-2020 relative to 1961-1970. Interannual variations in HL in the NCP ranged from 8 to 12 d, indicating potential for pollution abatement through FHPs. A GEOS-Chem simulation for 2019 with minimum HL indicated a 24% reduction in SPPDs over the NCP during FHP implementation periods.
With future climate warming, HL in the NCP will continue to decrease during 2020-2050. Relative to 2010, changes in HS and HE during 2050-2055 are projected to respectively be delayed and advanced by 5 and 8 d under the CN scenario, and by 9 and 12 d under the SSP5-85 scenario. Adoption of these changes into the heating period for 2050 yielded simulated mean PM 2.5 concentrations of 8.1-10.0 and 18.4-22.6 µg m −3 under the under CN and SSP5-85 scenarios, respectively. High risks of pollution days remain with daily PM 2.5 of >100 µg m −3 under SSP5-85, but improved air quality should be realized on more than 60% of days throughout the heating season with daily PM 2.5 concentrations of <10 µg m −3 , and 95% with <35 µg m −3 , under the CN scenario. We recommend adoption of the future CN pathway for achievement of both air-quality improvement and heating welfare, despite the warming scenario of SSP5-85 reducing more HL. Our findings may guide further implementation of heating policy in North China.
The data that support the findings of this study are openly available at the following URL/DOI: https:// esgf-node.ipsl.upmc.fr/search/cmip6-ipsl/.

Code availability
The GEOS-Chem model code can be downloaded from http://acmg.seas.harvard.edu/geos/.