The control of anthropogenic emissions contributed to 80 % of the 1 decrease in PM 2 . 5 concentrations in Beijing from 2013 to 2017

19 With the completion of the Beijing Five-year Clean Air Action Plan by the end of 2

In this study, hourly PM2.5 concentration data were acquired from the website PM25.in, which collects 84 official data provided by China National Environmental Monitoring Center (CNEMC). Beijing has 85 established an advanced air quality monitoring network with 35 ground stations across the city. 86 Considering the major contribution of industry and traffic-induced emissions in urban areas, we 87 selected all twelve urban stations to analyze the variation of PM2.5 concentrations and quantify their 88 influencing factors. In addition to these urban stations, we also selected two background stations, the 89 DingLing Station located in the suburb and the MiYun Reservoir Station located in the outer suburb,

106
For this research, we employed both regional and local emission inventories for running model 107 simulation. Multi-resolution Emission Inventory for China, MEIC, (http://meicmodel.org/) provided by 108 Tsinghua University, were employed as the regional emission inventories. MEIC has been widely 109

121
A key step for quantifying the relative contribution of anthropogenic emissions to the decrease of PM2.5 122 concentrations is to properly filter meteorological influences on PM2.5 concentrations, which is highly 123 challenging and rarely investigated by previous studies. Therefore, we employed both a statistical 124 method and a chemical transport model in this study to comprehensively evaluate the role of 125 anthropogenic emissions and meteorological conditions in the decrease of PM2. 5

concentrations in 126
Beijing during the past five years. The raw time-series data of airborne pollutants can be decomposed as: 147

Where X (t) is the original time series of airborne pollutants, E(t) is the long-term trend component, S(t) 153
is the seasonal variation, W(t) is the residue or synoptic-scale (short-term) variations. KZi, j(X) 154 indicates a KZ filtering on the original dataset X with a moving wind size of i and j iterations. We conducted correlation analysis between PM2.5 concentrations and a series of meteorological 165 factors, including temperature, wind speed, wind direction, precipitation, relative humidity, solar 166 radiation, evaporation and air pressure. The correlation analysis revealed that wind speed, relative 167 humidity, temperature, solar radiation and air pressure were strongly and significantly correlated with 168 PM2.5 concentrations in Beijing, which was consistent with the findings from previous studies (Sun et  temperature and solar radiation as follows. 172 Next, KZ filtering is conducted on the ε(t) for its long-term component ( ). After the 180 variation of meteorological influences was filtered, the reconstructed time series of airborne pollutants 181 XLT(t) was calculated as the sum of ( ) and the average value of E(t) , ( ).
After KZ filtering, the relative contribution of meteorological conditions to the variation in PM2.5 184 concentrations can be calculated as follows: 185 Where Pcontrib is the relative contribution of meteorological conditions to the variation of PM2.5 Therefore, we acquired the influence of the relative contribution of each emission source on PM2.5 247 reduction in Beijing (Table 1). For each station, the original time series of PM2.5 data was processed by the KZ filter and the relative 253 contribution of the long-term trend, seasonal variation and short-term variation to the total variance 254 was shown as Table 2. The sum of the long-term trend, seasonal variation and short-term variation 255 contributed to more than 93.6~95.3% of the total variance for different stations respectively. The larger 256 the total variance, the three components are more independent to each other. According to Table 2, the 257 large value of the total variation for each station indicated a satisfactory result from the KZ filtering. 258 The relative contribution of short-term variation was much larger than that of the seasonal and

265
We employed the emission inventory and meteorological data for 2013 to verify the accuracy of the 266 WRF-CMAQ model. For three different stations (the DingLing background station, the Yufa rural 267 station and the Olympic Center urban station), we compared the observed and estimated PM2.5 268 concentrations (Fig 2). According to Fig 2, the general trend of the simulated PM2.5 concentrations was 269 similar to that of the observed value. A general agreement was found between the simulated and 270 observed data with more than 85% of data points falling into the siege area of 1:2 and 2:1 lines. The original time series of PM2.5 concentrations and adjusted time series of PM2.5 concentrations 289 processed using KZ filtering were illustrated using one urban station, one rural station, one 290 transportation station, and two background stations (Fig 3). As shown in Fig 3,

Estimation based on WRF-CMAQ model
318 In addition to the KZ filter, we also employed the WRF-CMAQ model to estimate the relative 319 contribution of emission-reduction measures and meteorological conditions to the decrease of PM2.5 320 concentrations in Beijing. The result is shown in Table 4. 321 As Table 4 shows, and based on the WRF-CMAQ model, the relative contribution of meteorological 324