Updated Simulation of Tropospheric Ozone and Its Radiative Forcing over the Globe and China Based on a Newly Developed Chemistry–Climate Model

Tropospheric ozone (O₃), a major pollutant in the atmosphere, causes damage to human health and the ecosystem (Zeng et al., 2008; Stevenson et al., 2013; Lou et al., 2014, 2015; Zhu and Liao, 2016; Revell et al., 2018; Young et al., 2018). It is a prominent greenhouse gas (Zhu and Liao, 2016; Young et al., 2018) with a strong infrared absorption band at 9.6 μm (Wang, 1999; Geng et al., 2012). As such, tropospheric O₃ contributes significantly to the radiation budget of the earth–atmosphere system and thus has a large effect on climate. Tropospheric O₃ is also considered a short-lived climate forcer (SLCF), as its lifetime in the atmosphere is relatively short (Li et al., 2016; Zhang et al., 2018). Previous studies have indicated that tropospheric O₃ forms mainly from photochemical reactions among its precursors that include nitrogen oxides (NO_x), carbon monoxide (CO), methane (CH₄), and nonmethane volatile organic compounds (NMVOCs) (Stevenson et al., 2013; Liao and Shang, 2015; Li et al., 2018; Revell et al., 2018). The temporal and spatial distributions of these precursors greatly affect the distribution of tropospheric O₃.

Recent observations and studies have demonstrated an increasing trend in tropospheric O₃, mainly in the Northern Hemisphere (NH) and especially in industrial areas. In regions with heavy industries, there has been a steady increase in O₃ precursor emissions from anthropogenic sources since the pre-industrial era (Logan, 1985; Volz and Kley, 1988; Ji and Qin, 1998; Wang and Sun, 1999; Lamarque et al., 2005; Stevenson et al., 2013). The steady increases in tropospheric O₃ precursor emissions are now having a significant effect in remote rural areas (Jacob et al., 1993; Bakwin et al., 1994). Additionally, many studies have shown a shift in high-emitting areas of anthropogenic precursors of O₃ from North America and Europe to Asia (Granier et al., 2011; Cooper et al., 2014; Zhang et al., 2016) and developing countries. Therefore, it is necessary to investigate changes in tropospheric O₃ in China and its long-term effects.

There have been numerous studies on the radiative forcing (RF) of tropospheric O₃. Shindell et al. (2003) used the GISS (Goddard Institute for Space Studies) chemistry–climate general circulation model (GCM) to obtain the present global annual mean RF of tropospheric O₃ of 0.30 W m⁻². Skeie et al. (2011) estimated an RF of 0.44 W m⁻² for tropospheric O₃ from 1850 to 2010 using OsloCTM2 (chemical transport model); additionally, their data showed an increasing trend. Søvde et al. (2011) revealed RF values of tropospheric O₃ and its precursors of 0.38 and 0.44 W m⁻², respectively, from 1750 to 2010 with OsloCTM2. Stevenson et al. (2013) assessed the RF of tropospheric O₃ from 1750 to 2010 and obtained a value of 410 mW m⁻² with 17 atmospheric chemistry models as part of the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP); future tropospheric O₃ RF values for the year 2030 (relative to 1750) were projected to be 350, 420, 370, and 460 mW m⁻² under Representative Concentration Pathways (RCPs) 2.6, 4.5, 6.0, and 8.5 scenarios, respectively. Researchers have also analyzed tropospheric O₃ RF in China and its adjacent areas. Li et al. (2018) estimated the tropospheric O₃ RF as 0.69 W m⁻² over China and its adjacent areas in summer, which the data were obtained by using RegCM4 (Regional Climate Model version 4.0) from 2005 to 2010. Zhu and Liao (2016) estimated that the tropospheric O₃ RF over China was 0.48 W m⁻² in 2000 (relative to 1850) using GEOS-Chem (Goddard Earth Observing System-Chemistry).

The Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC AR5) introduced the concept of effective radiative forcing (ERF). Boucher et al. (2013) pointed out that the calculation of ERF requires longer simulations, with more complex models than that used for RF calculation; however, the inclusion of the additional rapid adjustments makes ERF a better indicator of the eventual global mean temperature response (Smith et al., 2020). Zhao et al. (2019) performed time-slice experiments to investigate the ERF of tropospheric O₃ with CESM1 (Community Earth System Model) and O₃ concentrations derived from WACCM (Whole Atmosphere Community Climate Model) simulations (Marsh et al., 2013) from 1970 to 2010; an ERF value of 0.24 ± 0.01 W m⁻² was obtained. Xie et al. (2016) used the BCC-AGCM_CUACE/Aero (Beijing Climate Center Atmospheric General Circulation Model_ China Meteorological Administration Unified Atmospheric Chemistry Environment for Aerosols) model and off-line O₃ profile data from Ozone Monitoring Instrument (OMI) to calculate the global annual mean ERF of tropospheric O₃; a value of 0.46 W m⁻² was obtained from 1850 to 2013. MacIntosh et al. (2016) estimated a tropospheric O₃ ERF of 0.26 W m⁻² using the atmosphere-only version of the HadGEM3 (Hadley Centre Global Environment Model version 3) climate model and the zonal-mean O₃ distribution taken from ACCMIP. However, relatively few tropospheric O₃ ERF studies have been performed to date, and attempts have been limited by various factors, including off-line O₃ data used in previous studies. In addition, some of these earlier investigations did not consider atmospheric chemistry–radiation–climate interactions (MacIntosh et al., 2016; Xie et al., 2016). As mentioned above, tropospheric O₃ is formed mainly by photochemical reactions; thus, it is necessary to study the ERF of tropospheric O₃ while considering the emissions of its precursors and chemical processes (not considering the stratosphere–troposphere transport) using a chemistry–climate online coupled model.

In this study, the tropospheric O₃ burden, RF, and ERF of the past are investigated in an attempt to quantify the changes in tropospheric O₃ and its RF and ERF for the year 2050 at global and regional scales, using our newly developed atmospheric chemistry–climate model, BCC-AGCM_CUACE2.0. Here, we focus on regions in China.

In this study, we simulated the burden of tropospheric O₃ and the associated RF and ERF due to changes in the O₃ burden, using the newly developed atmospheric chemistry–climate model BCC-AGCM_CUACE2.0, which couples an atmospheric GCM (BCC-AGCM2.0) and a chemistry model (CUACE2.0).

BCC-AGCM2.0 was developed by the National Climate Center of the China Meteorological Administration (NCC/CMA; Wu et al., 2010). As part of the IPCC Coupled Model Intercomparison Project Phase 5, BCC-AGCM2.0 has been validated as state-of-the-art representation against satellite observations (e.g., Zhang et al., 2012a; Jiang et al., 2015). The model adopts a horizontal resolution of 2.8° latitude × 2.8° longitude, with terrain-following hybrid vertical coordinates having 26 layers and a rigid lid at 2.9 hPa. Additional details, including a description of the detailed dynamics, physical processes, and validation procedure for BCC-AGCM2.0, can be found in Wu et al. (2010). The definition of tropopause in BCC_AGCM2.0 is the same as in the NCAR Community Atmosphere Model (Collins et al., 2004):

where p_t (hPa) refers to the pressure at the tropopause and φ refers to latitude.

To improve the simulation of radiation fluxes at the top of atmosphere (TOA) and the surface, the cloud overlap scheme of a Monte–Carlo independent column approximation (Pincus et al., 2003; Jing and Zhang, 2012, 2013) and the new Beijing Climate Center radiation transfer model (BCC_RAD; Zhang et al., 2014; Zhang, 2016) were implemented. BCC_RAD divides the wavelength range (0.204–1000 μm) into 17 bands (8 longwave and 9 shortwave bands) and 5 main greenhouse gasses (H₂O, CO₂, O₃, N₂O, and CH₄). In addition, three types of chlorofluorocarbons are included in the model. Gas absorption ranges and their overlaps are calculated by using the correlated K-distribution method (Zhang et al., 2003).

CUACE2.0 is a unified atmospheric chemistry environment that includes an aerosol module (CUACE/Aero), a gas chemistry module (CUACE/Gas), and a thermodynamic equilibrium module (CUACE/ISO). CUACE/Aero is a size-segregated multicomponent aerosol module used to calculate the mass concentrations of seven aerosol species [sea salt, sand/dust, black carbon (BC), organic carbon (OC), sulfate, nitrate, and ammonium salt]. The hygroscopic growth of soluble aerosol particles, such as nitrates, sulfates, OC, and sea salt, was taken into account. The performance of CUACE/Aero has been evaluated by Wang et al. (2014), Zhang et al. (2014), Wang et al. (2015), and An et al. (2019) in detail; these studies demonstrate the accuracy of this model in simulating the mass concentrations of aerosols. These information in detailed transport, physical processes, chemical transformation, cloud interactions, and the deposition of atmospheric aerosols are described in Gong et al. (2002, 2003), Zhang et al. (2012b), and Zhou et al. (2012). CUACE adopts ISORROPIA to calculate the thermodynamic equilibrium between the aerosols and their gas precursors (Nenes et al., 1998; West et al., 1998; Yu et al., 2005) and is known as CUACE/ISO in our model. A more detailed description of ISORROPIA can be found in Nenes et al. (1998).

CUACE/Gas is based on the second-generation Regional Acid Deposition Model (RADM2; Stockwell et al., 1990; Zhou et al., 2012). The model consists of 63 gaseous species generated through 21 photochemical reactions and 121 gas phase reactions and is applicable under a wide variety of environmental conditions. In CUACE/Gas, OH, H₂O₂, O₃, NO₃, HNO₃, and other important gas species can be simulated online. As mentioned above, NO_x, CO, CH₄, and NMVOCs are important precursors of O₃, as they can form or consume O₃ (e.g., NO) through complex chemical reactions in the atmosphere. Some of the chemical reactions in which these precursors participate are described below.

NO_x:

Reactions (1)–(3) are fast cycle processes; the net effect of these reactions cannot promote O₃ formation. Reactants that can consume NO during the photolysis of NO₂ will be conducive to O₃ production; these reactants are CO, CH₄, and NMVOCs. The associated reactions are given below.

CO reacts with HO to produce HO₂, which can consume NO:

Combining reactions (1) and (2), the net reaction of CO to O₃ is as follows:

CH₄:

where ALD is acetaldehyde, MO₂ is a methyl peroxy radical, and ACO₃ is an acetylperoxide radical. ACO₃ can consume NO, which is conducive to O₃ formation:

NMVOCs:

Given reactions (12)–(15), the net reaction of NMVOCs to O₃ is given below:

where R is a hydrocarbon radical and CARB is macromolecular carbon. More detailed species, parameterization schemes, and reactions can be found in Stockwell et al. (1990).

In this study, the year 1850 was chosen to represent the O₃ concentration status of the pre-industrial era (PI), and the year 2010 was chosen to denote the beginning of the 21st century (BG). Choosing these particular years allowed for an RF comparison with the best estimate of RF from the IPCC AR5, in which 2010 represents the present day (PD). The year 2050 was chosen to represent the mid-term future (MT) assessment of tropospheric O₃.

As mentioned above, NO_x, CO, CH₄, and NMVOCs are important precursors of tropospheric O₃. In this study, NMVOCs include formaldehyde (HCHO), ethane (C2H6), ethylene (C2H4), isoprene (C5H8), toluene (C7H8), and xylene (C8H10), as these six species are abundant in the atmosphere and they are also the key species used in CUACE2.0. Emission data for these gases from the RCP database version 2.0 (Moss et al., 2010; van Vuuren et al., 2011) were used in our model. The emission data of O₃ precursors in 1850 (RCP Hist) are shown in Table 1. To investigate the variation in tropospheric O₃ under different emission scenarios, emission data from RCP2.6, RCP4.5, RCP6.0, and RCP8.5 for 2050 were applied.

Figures 1 and 2 show the time series of annual mean emissions of O₃ precursors in the globe and China from 2010 to 2050. Emissions of the four species increased significantly from 1850 to 2010, which was closely related to the large amount of anthropogenic pollutants produced after the Industrial Revolution. We also found that the anthropogenic emissions of O₃ precursors are much higher than the natural emissions in the future (supplement Figs. s1, s2), which means that anthropogenic emissions make a great contribution to the generation of tropospheric O₃. By contrast, RCP data revealed decreasing emission trends for the four species from 2010 to 2050, indicating that tropospheric O₃ pollution may improve in the future. From a global perspective, it should be noted that the burden of CH₄ exhibits an increasing trend under RCP8.5, and CO remains basically unchanged under RCP6.0. The burden of NMVOCs exhibits a slowly increasing trend under RCP6.0; however, NO_x, CO, and NMVOCs emissions in China exhibit an increasing trend under RCP6.0, significantly different from those at the global scale. This suggests that tropospheric O₃ pollution in China will be more severe in the near future.

Figure 1.  Changes in emissions of ozone (O₃) precursors from 2010 to 2050.

Figure 2.  As in Fig. 1, but for China.

In this study, six experiments are conducted (Table 2). The RF and ERF values of tropospheric O₃ are estimated as the differences between a specified year and the PI. Based on the emission data of the period, calculations are performed by using the BCC-AGCM_CUACE2.0 model. Sea surface temperature (SST) and sea-ice cover (SI) are fixed at values specified in 21-yr (1981−2001) climatology data from the Hadley Centre for the calculations (Hurrell et al., 2008), and the stratospheric O₃ is prescribed as fixed value and obtained from the OMI observational data (see http://aura.gsfc.nasa.gov/). The model integration time is 30 yr (Pincus et al., 2016; Smith et al., 2020).

Observations of tropospheric O₃ (DU) for model validation in Section 3 are provided by the NASA’s Goddard Space Flight Center. These data with a resolution of 1.25° × 1° (valid from October 2004 to October 2010) are derived from ozone measurements obtained by OMI and Microwave Limb Sounder (MLS) onboard on the Aura satellite using the tropospheric O₃ residual method (Ziemke et al., 2006, 2011). These data are verified by using ozonesonde data and exhibited good reliability (Li et al., 2018). We also evaluated the performance of the model for the seasonal cycle of surface O₃ using data from 2014 to 2016 derived from the China National Environmental Monitoring Center (http://www.cnemc.cn/).

To calculate RF, the “double radiation call” method was used, which calls the radiation scheme twice each time step, such that the meteorological conditions remain unchanged. We used the real-time tropospheric O₃ concentration calculated online by the model in one call of the radiation scheme and doubled the tropospheric O₃ concentration in the other. During the same time step, tropospheric O₃ affected only the radiative process, without effect on any other climate process. RF due to tropospheric O₃ refers to the difference in radiative flux at the tropopause during the two calls. RF during a specified period due to tropospheric O₃ changes refers to the difference in RF between two time slices. The Radiative Forcing Model Intercomparison Project states that the model ERF can be diagnosed by suppressing the response, that is, by specifying SST and SI concentrations (Pincus et al., 2016). In this work, ERF values are calculated by using the method proposed by Hansen et al. (2005), which the SST and SI coverage are fixed at climatological values while allowing all other parts of the system to respond until an equilibrium state is reached.

Figure 3 presents the global distributions of the column concentration of tropospheric O₃ derived from simulations and OMI/MLS and the difference between them for 2010. BCC-AGCM_CUACE2.0 captured the tropospheric O₃ patterns in high-emission areas of O₃ precursors (mainly industrial areas), such as in North America, Middle East, South Asia, and East Asia. Simulations of the Southern Hemisphere (SH) produced values that were lower than the observations, which was mainly due to the low emissions in the SH used in this study. Moreover, the underestimates in the SH are common in most chemistry–climate models due to the inaccurate emissions (Young et al., 2013; Parrish et al., 2014; 2018). Revell et al. (2018) also found that multi-model results overestimated the concentration, on average by up to 40%–50% in the NH, compared with OMI/MLS observations, and underestimated the concentrations by approximately 30% in the SH; in the study, the annual mean tropospheric column O₃ was analyzed by using the 15 models applied in the stratosphere–troposphere processes and their role in SPARC-IGAC(Stratosphere–troposphere Processes And their Role in Climate-International Global Atmospheric Chemistry) Chemistry–Climate Model Initiative (CCMI). Recently, Lu et al. (2020) has also developed BCC-GEOS-Chem v1.0 that can simulate the evolution of atmospheric chemical interactive constituents and capture well the spatial distributions and seasonal cycles of tropospheric O₃; and this study shows that the low STE (stratosphere–troposphere exchange) in BCC-GEOS-Chem v1.0 appears to be the main factor causing ozone underestimates in the upper troposphere.

Figure 3.  Global distributions of tropospheric O₃ (DU). (a) Simulations of BCC-AGCM_CUACE2.0, (b) observations from Ozone Monitoring Instrument (OMI)/Microwave Limb Sounder (MLS) onboard the Aura satellite, and (c) differences between simulation and observation.

In this study, the simulated column concentration of tropospheric O₃ was approximately 4.9 DU lower than the observations, which is within the bias range between multi-model mean results and observations (Young et al., 2013). In addition, a different definition of tropopause in a particular model can also cause some simulation biases. However, we obtained results that are fairly consistent among regions for the tropospheric O₃ distribution. It should be noted that the distribution of the simulated column concentration of tropospheric O₃ reproduced the pattern well, compared with OMI/MLS observations for China, with a difference in the annual mean value of only 0.1 DU (Fig. 4). The lower simulation values for the Qinghai–Tibetan Plateau were mainly due to the fact that the model did not consider O₃ stratosphere–troposphere transport.

Figure 4.  As in Fig. 3, but for China.

To further evaluate the performance of the model for the seasonal cycle of surface O₃ in China, we first divided China into the following five geographic areas: Northeast China (NEC; 38°–55°N, 121°–135°E), North China (NC; 33°–50°N, 110°–121°E), Northwest China (NWC; 34°–50°N, 73°–110°E), Southwest China (SWC; 21°–34°N, 78°–110°E), and Southeast China (SEC; 17°–33°N, 110°–123°E). In each area, surface O₃ concentrations were observed from four sites distributed evenly throughout the area; the sites are shown in Fig. 5 (blue dots). The observations showed that the maximum surface O₃ concentrations occurred during the summer, whereas the minimum concentrations occurred during the winter in most areas (Fig. 6). This is mainly due to the strong sunlight and high temperatures in summer, which are conducive to O₃ formation. By interpolating the simulations over the observation site using bilinear interpolation, we found that the simulated surface O₃ concentrations matched the seasonal evolution of the observed concentrations in China well, although some sites showed a slight bias. The small bias in NEC might have been brought about by the emission data (An et al., 2019). This indicates that extensive efforts are still required to improve emission data and simulations in future studies. In addition, the model resolution and method of interpolation can also introduce some biases between simulations and observations (Wild and Prather, 2006; Hodnebrog et al., 2011; Stock et al., 2014). The simulated surface O₃ concentrations were lower than the observations during spring in Lhasa, which is mainly due to the high altitude of Lhasa; thus, the downward transport of O₃ from the stratosphere can have a great influence on surface O₃ concentration. Notably, we did not consider the O₃ transported via stratosphere–troposphere transport, which might have resulted in an underestimation for Lhasa. However, the simulations captured the seasonal cycle at most sites in China well. Therefore, overall, our results demonstrate that the BCC-AGCM_CUACE2.0 model can reasonably reproduce the distribution and annual mean value of tropospheric O₃ globally and for China. As such, we applied the model to further analyze the RF and ERF of tropospheric O₃.

Figure 5.  Distribution of observation sites in China.

Figure 6.  Seasonal cycles of simulations (interpolated over observation sites) and observations from the China National Environmental Monitoring Center of surface O₃ concentration (ppbv) in China. The blue curves are the simulated values for 2010. The red curves are the mean observation values from 2014 to 2016.

Figure 7 shows the changes in O₃ precursor emissions from 1850 to 2010, with areas with high concentrations of NO_x and CO located mainly in eastern North America, Europe, and eastern China (Fig. 7a). The East USA, Southeast Asia (especially Malaysia and Indonesia), Indian Peninsula (especially northern India), and East China emit large amounts of CO, which is closely related to local industrial and domestic emissions. Most high-CH₄ areas were located in northern India, Malaysia, NC, and SWC. Compared with NO_x, CO, and CH₄, the magnitudes of increased NMVOCs were larger and their distribution coverage more widely spread. Besides the industrial areas, NMVOCs also exhibited substantial increases over the central part of South America and the central part of southern Africa, which is mainly due to biogenic VOCs (BVOCs) such as isoprene.

Figure 7.  Distributions of O₃ precursors (kg m⁻² yr⁻¹) in 2010 (relative to 1850).

The simulated global annual column burden of tropospheric O₃ increased by 14.1 DU from the PI to the BG, with the greatest increase in tropospheric O₃ over the industrialized NH, including China, India, Southeast Asia, and the Arabian Peninsula (Fig. 8a). The value of tropospheric O₃ over China increased by approximately 21.1 DU, which is much higher than the global annual mean value. For China, the greatest increase in tropospheric O₃ is located over SEC and some parts of SWC; in this case, the increase is closely related to O₃ precursor emissions. Additionally, lower O₃ precursor emission is the main cause for the low column burden of tropospheric O₃ over the Tibetan Plateau; here, a higher altitude and lower tropopause are the two main contributing factors.

Figure 8.  Distributions of tropospheric O₃ (DU) over (a) the globe and (b) China in 2010 (relative to 1850).

Table 3 compares calculated changes in the column burden of tropospheric O₃ from the PI to the BG in this work with those reported by other studies. The value obtained in this study is close to the multi-model mean value of Gauss et al. (2003), but 2.7 DU higher than the value cited by Skeie et al. (2011). This is because Skeie et al. (2011) used the BVOCs emissions in 2000 to represent the emissions in 2010 in their study, which might have introduced some underestimation. The value in Xie et al. (2016) is much higher, using off-line O₃ profile data from OMI in 2013; thus, the interaction between radiation and O₃ chemistry was not considered in their simulation. In addition, different definitions of the tropopause may also explain the discrepancies among the abovementioned study results (Young et al., 2013).

With the reduction of precursor emissions, global tropospheric O₃ is predicted to decline under different scenarios in 2050 (Fig. 9). The global annual mean column burden of tropospheric O₃ is predicted to decrease by 2.28, 1.23, 2.49, and 0.30 DU by 2050 (relative to 2010) under RCP2.6, RCP4.5, RCP6.0, and RCP8.5, respectively. The largest reduction, 11%, occurs under RCP6.0. It should be noted that tropospheric O₃ is predicted to increase over India and some parts of Africa under RCP4.5 and RCP8.5. Especially over India, the maximum increase in tropospheric O₃ is expected to exceed 4 DU. Over Africa, the area of tropospheric O₃ increase is broader under RCP8.5, which may be related to the large amount of BVOCs emissions locally. From Fig. 1, we can see that global CH₄ emissions exhibit an obvious increasing trend, whereas NMVOCs emissions remain nearly unchanged; this may explain why global tropospheric O₃ does not decrease significantly under RCP8.5. For China, the largest reduction in tropospheric O₃ occurs under RCP2.6, with an annual mean value of 3.97 DU by 2050 (relative to 2010), which would be 12.6% lower than the level in 2010, especially over NC and SEC (Fig. 10). The annual mean value of tropospheric O₃ is predicted to decrease by 2.28 DU (7.3%) by 2050. However, a positive value is predicted for the Qinghai–Tibet Plateau, mainly due to the transportation of high concentrations of O₃ from northeast India. Similarly, SWC will also be affected by high concentrations of O₃ from northeast India under RCP8.5, resulting in an increase of 1.05 DU of tropospheric O₃ by 2050. In addition, NC will likely see a slight increase in tropospheric O₃ over the same period.

Figure 9.  Global distributions of tropospheric O₃ (DU) under different scenarios in 2050 (relative to 2010).

Figure 10.  As in Fig. 9, but for China.

Figure 11 presents the global distribution of tropospheric O₃ RF at the tropopause at the BG relative to the PI. The RF distribution pattern is basically consistent with the distribution of tropospheric O₃ column burden, such that high values appear over China, India, and Southeast Asia. Thus, this suggests that tropospheric O₃ is the main factor affecting the magnitude of RF. The global annual mean RF value of tropospheric O₃ was 0.48 W m⁻², within the estimation range reported in IPCC AR5 of 0.4 (0.2–0.6) W m⁻² (Boucher et al., 2013) and IPCC AR6 of 0.47 (0.24–0.70) W m⁻² (Forster et al., 2021). The normalized RF (NRF) of tropospheric O₃ was 34 mW m⁻², which is very close to the value of 36 mW m⁻² obtained by Gauss et al. (2003) using multi-model results (Table 4).

Figure 11.  Tropospheric O₃ radiative forcing (RF; W m⁻²) at the tropopauseover over (a) the globe and (b) China in 2010 relative to the pre-industrial era (PI).

The annual mean RF of tropospheric O₃ over China was 0.59 W m⁻². Higher values appeared over SEC (0.76 W m⁻²) and NC (0.62 W m⁻²), followed by NEC, NWC, and SWC (0.56, 0.56, and 0.57 W m⁻², respectively). Table 5 compares the RF values in this work with those of other studies.

Corresponding to the changes in tropospheric O₃ burden, global tropospheric O₃ RF is predicted to decline in 2050 relative to 2010, and the RF distributions are predicted to be similar among the four scenarios, as shown in Fig. 12. The projected global annual mean values of tropospheric O₃ RF are 0.42, 0.45, 0.41, and 0.47 W m⁻² in 2050 under the RCP2.6, RCP4.5, RCP6.0, and RCP8.5 scenarios relative to the PI, respectively (Table 6). The O₃ RFs for 2050 are expected to decrease by 12.5%, 6.3%, 14.6%, and 2.1% relative to 2010 under RCP2.6, RCP4.5, RCP6.0, and RCP8.5, respectively. The largest reduction in RF occurs under RCP6.0, with a value of −0.07 W m⁻². The annual mean tropospheric O₃ RF over China is predicted to decrease by 0.08, 0.03, and 0.04 W m⁻² under RCP2.6, RCP4.5, and RCP6.0 from 2010 to 2050, respectively (Fig. 13). It should be noted that RF is estimated to increase by 0.01 W m⁻² under RCP8.5, consistent with the changes in the burden of tropospheric O₃.

Figure 12.  Global distributions of tropospheric O₃ RF (W m⁻²) at the tropopause in 2050 (relative to the PI).

Figure 13.  As in Fig. 12, but for China.

Boucher et al. (2013) in IPCC AR5 proposed the ERF concept, stating that it is a better indicator of the eventual global mean temperature response. Our results show that the ERF is evenly distributed, with no obvious patterns (not closely correlated with the change in burden of tropospheric O₃), and many places are associated with negative values (Fig. 14). Thus, the discussion focuses on annual mean values (Zhang et al., 2018). The global annual mean ERF of tropospheric O₃ is 0.25 W m⁻² in 2010 relative to the PI (Fig. 14), which is close to the value of 0.26 W m⁻² calculated by MacIntosh et al. (2016). The annual mean ERF value of tropospheric O₃ over China is 0.50 W m⁻², twice the global mean value, indicating that the effect of tropospheric O₃ is significant for China. Although the ERF distribution exhibits no obvious patterns and is positive in many places, ERF values still have some dependence on cloud distribution. Figure 15 indicates that positive ERF values occur where the total cloud cover (TCC) is reduced, such as over western Europe, southern Australia, NEC, and SEC. Meanwhile, negative values are obtained in areas where TCC increases, e.g., over the East European Plain, India, and SWC. This is because rapid adjustments of cloud cover can significantly affect local radiation flux (Takemura et al., 2002; Zhang et al., 2012a; An et al., 2019).

Figure 14.  Tropospheric O₃ effective RF (ERF; W m⁻²) at the top of atmosphere over (a) the globe and (b) China in 2010 (relative to the PI).

Figure 15.  Distributions of clouds (%) over (a) the globe and (b) China in 2010 (relative to the PI).

To further investigate the impact of cloud cover on ERF values, the relationship between changes in longwave ERF (LW ERF) and changes in high cloud cover (HCC) from the PI to the PD are investigated (Fig. 16), as well as the relationship between changes in shortwave ERF (SW ERF) and changes in low cloud cover (LCC) (Fig. 17). Figure 16 shows that LW ERF is positive with increased HCC over northern South America, Europe, Central Africa, South Asia, and SWC. This is because an increase in HCC leads to greater absorption of LWR (Liu and Ye, 1991; Lelli et al., 2014), thus reducing outgoing longwave radiation at the TOA and resulting in a positive shift in LW ERF. More LWR will be emitted from the TOA in areas where the cloud cover is reduced, resulting in enhanced negative LW ERF values. Thus, negative LW ERF values appeared with a reduction in HCC over parts of the United States, the Arabian Peninsula, Australia, NWC, and SEC. Conversely, an increase in LCC leads to additional reflection of shortwave radiation out of the TOA due to the increased albedo, which enhances negative SW ERF values. Less shortwave radiation is emitted from the TOA in areas where there is less LCC, resulting in a positive shift in SW ERF values. Hence, areas with more LCC were associated with negative SW ERF values, such as over the northern part of South America, the Arabian Peninsula, South Asia, Southeast Asia, and SWC; areas with less LCC were associated with positive SW ERF, such as over the northeast part of South America, northern Africa, East China, and SEC.

Figure 16.  Distributions of tropospheric O₃ longwave ERF (top row; W m⁻²) and high cloud cover (bottom row; %) over the globe and China in 2010 (relative to the PI).

Figure 17.  Distributions of tropospheric O₃ shortwave ERF (top row; W m⁻²) and low cloud cover (bottom row; %) over the globe and China in 2010 (relative to the PI).

Table 7 compares current studies on tropospheric O₃ ERF. The ERF value of the current study is very close to that of MacIntosh et al. (2016), but much smaller than that given by Xie et al. (2016). It should be noted that Xie et al. (2016) used off-line O₃ profile data from OMI observations; in this case, there may be contributions from the stratosphere. In addition, feedback between chemical processes and the climate in Xie et al. (2016) was not considered. In contrast to the earlier studies, our model calculates only tropospheric O₃ formed by chemical reactions due to its precursors; thus, the transportation of O₃ from the stratosphere is not a contributing factor.

Future global annual mean ERF values of tropospheric O₃ were predicted to be 0.29, 0.18, 0.23, and 0.25 W m⁻² under RCP2.6, RCP4.5, RCP6.0, and RCP8.5 in 2050, respectively. As for China, the corresponding values were 0.24, 0.32, 0.03, and 0.01 W m⁻² for the year 2050, respectively. Table 8 shows the changes in the ERF of tropospheric O₃ for 2050 relative to 2010. With the implementation of pollutant reduction measures (Huang, 2006; Zhu and Liao, 2016; Saikawa et al., 2017; Geng et al., 2019), the ERF of tropospheric O₃ over China is projected to decrease significantly in the future.

Chung and Soden (2015a, b) pointed out that ERF is composed of RF and rapid adjustments (RAs). RAs can be further divided into RAs of the atmosphere (RA_atm) and RAs of clouds (RA_cld) (Smith et al., 2018, 2020; Zhao and Suzuki, 2019). In this section, the RF and ERF values refer to changes in the net radiation flux at the TOA caused by tropospheric O₃ (unless specified, the calculation of the radiation process is under all-sky conditions in this work). According to IPCC AR5 (Boucher et al., 2013), RAs caused by a forcing agent (here, tropospheric O₃) can be obtained by subtracting its RF from the corresponding ERF. RA_atm can be obtained by using the same approach as that for determining RAs; however, clear-sky conditions are applied (i.e., the residual of ERF_clr and RF_clr in which the subscript “clr” refers to clear-sky conditions). RA_cld is determined based on the residuals of RA and RA_atm. An RA flowchart for the calculations is provided by Zhao and Suzuki (2019).

Table 9 shows the global annual mean tropospheric O₃ values of ERF, RF, and their differences (namely, RAs) at the TOA in 2010 relative to the PI. The total RAs (RA_tot) can be expressed as

which is 0.05 W m⁻². RA_atm can be expressed as

with the value of 0.02 W m⁻². Thus, RA_cld can be calculated as

producing a value of 0.03 W m⁻². Table 10 lists the RA_atm and RA_cld values due to tropospheric O₃ for the year 2050 relative to the PI. The largest RA_cld of 0.07 W m⁻² is obtained under RCP2.6.

In this study, a newly developed atmosphere chemistry–climate model, BCC-AGCM_CUACE2.0, was used to simulate the past and future burden, RF, and ERF of tropospheric O₃, and the corresponding RA for the atmosphere (RA_atm) and clouds (RA_cld). The model reproduced the geographical distribution and seasonal changes in tropospheric O₃ well, especially over China. However, biases might have been introduced by inaccuracies in the emission inventory of the SH compared with the observations derived from OMI/MLS, which is a common problem in current chemistry–climate models, as pointed out by Young et al. (2013).

The global annual mean burden of tropospheric O₃ was predicted to have increased by 14.1 DU from the PI to 2010, with China, South Asia, and the Arabian Peninsula seeing more significant increases. As a high-emission area in 2010, the burden of tropospheric O₃ over China increased by as much as 21.1 DU, with the largest value of 32.2 DU occurring over SEC. By 2050, global tropospheric O₃ is expected to decrease by 0.30–2.49 DU globally compared with 2010 levels, whereas the corresponding burden over China should decrease by 3.97, 2.28, and 1.60 DU under the RCP2.6, RCP4.5, and RCP6.0 scenarios, respectively. However, tropospheric O₃ over China exhibited a different trend under RCP8.5, with its burden predicted to increase by 0.09 DU, which may be due to the influence of a high tropospheric O₃ burden over South Asia.

The annual mean tropospheric O₃ RF between the PI and the BG at the tropopause was estimated to be 0.48 globally and 0.59 W m⁻² over China. The RF distribution was consistent with the distribution of the column burden of tropospheric O₃. With a reduction in tropospheric O₃, the global annual mean RF value was predicted to be 0.41–0.47 W m⁻² according to different scenarios by 2050, whereas the RF over China was expected to reach 0.51–0.60 W m⁻²; thus, China will still experience severe ozone pollution in the future.

The global annual mean ERF for tropospheric O₃ at the TOA was estimated to be 0.25 W m⁻² for the year 2010. Meanwhile, the RAs of atmosphere and clouds were 0.02 and 0.03 W m⁻², respectively. By 2050, global ERF values are expected to reach 0.29, 0.18, 0.23, and 0.25 W m⁻² under the RCP2.6, RCP4.5, RCP6.0, and RCP8.0 scenarios, respectively. As for the ERF over China, an estimate of 0.50 W m⁻² was obtained for 2010, with lower values of 0.24, 0.32, 0.03, and 0.01 W m⁻² by 2050. The ERF distribution was consistent with the cloud cover distribution. Positive (negative) ERF occurred in areas with an increase (decrease) in HCC. In addition, an increase in LCC can cause a negative shift in ERF, and vice versa.

Additionally, we will validate our simulation of O₃ concentration in other areas outside China by getting higher resolution observational data in the future.

Acknowledgments. We thank the anonymous reviewers and editors for their valuable and stimulating comments, which have greatly improved our paper.