China has serious air pollution problems, and fine particulate matter (PM2.5) and ozone (O3) are the two major air pollutants (Lin et al., 2010; Hu et al., 2016; Lu et al., 2019; Wu et al., 2019). The annual average PM2.5 concentrations were higher than 50 µg m-3 in 26 out of the total 31 provincial capital cities in mainland China during 2013–2014 (Wang et al., 2014a), and the national fourth highest daily maximum 8 h average O3 (O3-8 h) was 86.0 ppb during the warm seasons (April–September) in 2013–2017, which is 6.3 %–30 % higher than that in other industrialized regions of the world (Lu et al., 2018). PM2.5 alone caused 0.87–1.36 million deaths every year in China, and long-term exposure to O3 was responsible for an extra 254 000 deaths (Cohen et al., 2017; Silver et al., 2018; Apte et al., 2015; Hu et al., 2017b). China has made remarkable improvement in air quality during recent years (Zhang et al., 2017; Zheng et al., 2018; Zhao et al., 2017); however, air pollution is still severe, making it the fourth-ranked health risk factor (Stanaway et al., 2018).

Surface PM2.5 and O3 concentrations are determined by atmospheric processes of emissions, transport and dispersion, chemical transformation (due to gas-phase, aqueous-phase, and aerosol chemistry), and dry and wet deposition. These processes are affected by meteorological conditions. Studies have shown that the surface O3 and PM2.5 concentrations are sensitive to different meteorological parameters. For example, Dawson et al. (2007b) investigated the sensitivity of surface O3 to different meteorological parameters in the eastern United States (US) using the comprehensive air quality model with extensions (CAMX). The results showed that temperature (T) had the greatest influence on daily O3-8 h of 0.34 ppb K-1, followed by absolute humidity (AH) of 0.025 ppb %-1. Bernard et al. (2001) also confirmed that T presented a notable positive correlation with the surface O3 concentration. The effects of meteorological parameters on PM2.5 are even more complicated. Tran and Mölders (2011) showed that elevated PM2.5 concentrations tended to occur under the conditions of calm wind, low T, and relative humidity in Fairbanks, Alaska. Olvera Alvarez et al. (2018) used a land use regression model to analyze the effects of different meteorological parameters on PM2.5 in El Paso, Texas, and obtained the same conclusion in winter, but in spring, the high PM2.5 level was associated with high wind speed (WS) and low humidity. Dawson et al. (2007a) studied the effects of individual meteorological parameters in the eastern US and found that PM2.5 concentration decreased markedly as the precipitation increased (PCP) in winter, but in summer, the main meteorological factors affecting the PM2.5 concentration were T, WS, and planetary boundary layer height (PBLH). Dawson et al. (2009) simulated the effects of climate change on regional and urban air quality in the eastern US and found that PM2.5 concentration decreased by 0.3 µg m-3 in January, mostly due to increasing PCP, and increased by 2.5 µg m-3 in July, largely due to decreasing PBLH and WS. Horne and Dabdub (2017) altered various meteorological parameters to investigate their effects on O3, PM2.5, and secondary organic aerosols (SOA) and found that the T predominated the effects of meteorology in California.

Many studies have proved that meteorological conditions play very important roles in air pollution events in China. Studies found that the pollutant concentrations could vary by up to several times, due to meteorological changes with the same emission sources (Zhang et al., 2010; Zheng et al., 2015; Cai et al., 2017; Ning et al., 2018; Yang et al., 2018; Li et al., 2019b; Xing et al., 2011; Liu et al., 2017). For example, Xing et al. (2011) studied the difference between the effects of 2007 and 2008 meteorological conditions on air quality during the 2008 Beijing Olympics. They found that higher humidity in August 2008 was beneficial to the formation of SO42- by up to ∼60 %, and lower T prevented the evaporation of NO3- by up to ∼60 %. Liu et al. (2017) reported that the monthly mean PM2.5 concentrations in the Jing–Jin–Ji (JJJ) area in December 2015 increased by 5 %–137 % due to the unfavorable weather conditions such as low WS and high humidity.

A few studies investigated the relationships between air quality and meteorological conditions in China. Zhang et al. (2015) conducted a correlation analysis between air quality and meteorology in three megacities: Beijing, Shanghai, and Guangzhou in China. The result showed that air pollutants were significantly negatively correlated with WS, and O3 had a positive correlation with T. Yin et al. (2016) found that the relationship between WS and PM2.5 has a complicated influence, with higher PM at low and high WS than in light to moderate winds in Beijing from 2008 to 2014. Xu et al. (2018) examined the variations of PM2.5 concentration in January 2017 in China compared to that in January 2016 and found that meteorological conditions of low WS, high humidity, low PBLH, and low PCP contributed to PM2.5 concentration worsening by 29.7 %, 42.6 %, and 7.9 % in the JJJ region, the Pearl River delta (PRD) region, and the Cheng–Yu Basin (CYB) region, respectively. Ma et al. (2019) analyzed the effects of meteorology on air pollution in the Yangtze River delta (YRD) region during 2014–2016 and found that PM2.5 was highly negatively correlated with WS, while O3 concentration was positively correlated with T but negatively related to relative humidity. Zhu et al. (2017) reported that the surface concentrations of O3 increased by 2–6 in January and 8–12 ppb in July 2014 in the PRD, mainly due to the increase in T and the decrease in NOx emissions.

These studies have investigated the impacts of meteorological conditions on PM2.5 and O3 in certain regions of China; however, quantitative sensitivity of PM2.5 and O3 to meteorological parameters has not been examined. The objective of this study is to quantify the sensitivity of O3 and PM2.5 to different meteorological parameters in winter and summer in different regions of China. The paper is constructed as follows: Sect. 2 describes the method used to estimate the sensitivity, and Sect. 3 presents the effects of each meteorological variable on O3 and PM2.5 in China and in five representative cities. Conclusions are then summarized in Sect. 4.

The sensitivity of O3 and PM2.5 associated with changes in meteorological parameters was quantified using the Community Multiscale Air Quality (CMAQ) model version 5.0.2. The meteorological parameters include T, WS, AH, PBLH, PCP, and cloud liquid water content (CLW). A base case was firstly simulated with meteorological fields predicted by the Weather Research and Forecasting (WRF) model v3.7.1 (https://www2.mmm.ucar.edu/wrf/users/) using the NCEP FNL Operational Model Global Tropospheric Analyses dataset as the initial and boundary conditions. The base case has been described in a previous study and the model configurations of the base case were reported there (Hu et al., 2015). The WRF-predicted meteorological parameters and the CMAQ-predicted surface O3 and PM2.5 have been evaluated against observations at 422 sites in 60 major cities in China, and the accuracy of the model performance has been validated (Hu et al., 2016).

A suite of perturbation scenarios was created, and in each scenario, a certain meteorological parameter was perturbed to a certain extent. The details of the perturbation scenarios are listed in Table 1. Among those changes, T was absolute changes and other parameters were relative variations. The magnitude ranges of perturbations are based on the IPCC AR5 report and the study of Dawson et al. (2007a, 2007b), and the references therein. For each parameter, three positive and three negative perturbations were then designed within its range to have a more comprehensive examination of the sensitivity of PM2.5 and O3 to this parameter. All perturbations were implemented uniformly in space on the modeling domain and in time through the modeling periods. The perturbations on temperature, wind speed, and absolute humidity were made in all layers. To separate the effects of individual meteorological parameters, only one parameter was changed in each case, while all the other parameters were kept unchanged. Therefore, cloud dissipating or forming in response to changing temperature was not considered in the simulations. Please note that these types of perturbations are not what happens in the real world, where meteorological parameters are inter-linked. When perturbing horizontal wind speed, to avoid unphysical situations that mass would not be conserved, the vertical wind speed was adjusted in the vertical transport calculation based on the air density changes to conserve mass.

Then the CMAQ model was re-run to predict the air quality under the perturbed meteorological condition. The emissions and other inputs were kept unchanged in each perturbed meteorological scenario; therefore, the difference of O3 and PM2.5 concentrations between each of the perturbation case and the base case was due to the change in the specific meteorological parameter, and the sensitivity of O3 and PM2.5 to individual meteorological parameters could be quantitatively determined. It is worthwhile noting that some meteorological parameters could have significant impacts on emissions, such as the effect of T on biogenic VOC and soil NOx emissions, the cloud cover/convection on lightning NOx emissions, and the effect of T on power plant NOx emissions (high T leads to higher electricity demand in summer), which would affect air quality. Therefore, the sensitivities in this study only include the “direct” effects of individual meteorological parameters on air quality. A full evaluation of the impacts of climate/weather changes on air quality should consider effects of the emissions changes.

The modeling domain covers East Asia, including all of China, with a horizontal resolution of 36×36 km2. The base and perturbation cases were conducted in January and July in 2013, representing the winter and summer conditions, respectively. In addition to the regional analysis, five representative megacities were selected, i.e., Beijing, Shanghai, Guangzhou, Chongqing, and Xi'an (Fig. 1). These cities are located in the North China Plain (NCP), YRD, PRD, CYB, and Guanzhong Plain, respectively, where serious air pollution problems often occur. In this study, O3-8 h was used in the O3 analyses, and 24 h average PM2.5 was used in the PM2.5 analyses, if not specifically stated. The O3–T relationship is examined using the method in Rasmussen et al. (2012) in the five cities. Observed and predicted O3–T relationships were estimated using the daily observed and predicted O3-8 h concentrations and daily maximum T in July (O3 observations became available from March 2013 in China, so no O3 observations in January). The results are shown in Fig. S1 (in the Supplement). CMAQ predicts a positive O3–T relationship in most cities except in Beijing, and the model tends to underestimate the daily O3–T relationship except in Shanghai. The underestimation of O3–T by the CMAQ model in this study is consistent with the findings in Rasmussen et al. (2012). Please note that we only have 1-month data and that we use daily O3-8 h and daily maximum temperature in the evaluation, while a much more meaningful evaluation should be performed to use monthly average O3 and monthly average temperature over a long-term period (Rasmussen et al., 2012).

Meteorological conditions can have a great influence on surface O3 and PM2.5 concentrations. In this study, the sensitivities of O3-8 h and PM2.5 to T, WS, AH, PBLH, PCP, and CLW are quantitatively estimated in January and July, respectively, in China. The response of O3-8 h to T is important and the sensitivity can be up to +2 ppb K-1 in both January and July, and the sensitivity is dependent on the O3 chemistry formation or loss regime, i.e., positive in the net O3 formation areas and negative in the O3 consumption areas. In general, PM2.5 sensitivities are negative to T, WS, PBLH, and PCP and positive to AH and CLW in most regions of China in both January and July. The sensitivities in January are much larger than in July. PM2.5 sensitivities to T, WS, AH, and PBLH are important. The PM2.5 sensitivities to these meteorological parameters are through major effects on SO42-, NO3-, NH4+, and SOA. The sensitivities of O3 and PM2.5 to CLW and PCP are negligible. The results show that O3 and PM2.5 concentrations in China are greatly affected by meteorological conditions; therefore, changes in these meteorological parameters due to climate change or inter-annual meteorological variations could potentially alter O3 and PM2.5 concentrations significantly, and it should consider these effects when developing emission control strategies. The results also show that the O3 and PM2.5 sensitivities to meteorological parameters have substantial spatial variations. Future studies can further investigate how the changes in meteorological conditions affect the effectiveness of emission control plans in reaching the designed air quality objectives in the different regions of China.