**2. Materials and Methods**

The following data are used in this study. (1) Hourly PM2.5 mass concentration from 2013 to 2020 in autumn and winter (October of the current year to February of the following year) is obtained by controlled stations of the Ministry of Environmental Protection in Northeast China (including Harbin (HEB), Changchun (CC) and Shenyang (SY)), Beijing-Tianjin-Hebei and surrounding areas (including Beijing (BJ), Tangshan (TS), Shijiazhuang (SJZ), Xingtai (XT), Jinan (JN) and Zhengzhou (ZZ)), Feiwei Plain (LinFen (LF), BaoJi (BaJ), XiAn (XA)), Yangtze River Delta (Nanjing (NJ), Hefei (HF), Shanghai (SH), Hangzhou (HZ)), Sichuan-Chongqing (Chengdu (CD) and Chongqing (CQ)), Central China (Wuhan (WH), Huangshi (HS), Nanchang (NC) and Changsha (CS)) and Pearl River Delta (Guangzhou (GZ), Shenzhen (SZ), and Zhuhai (ZH)). The data used were the mean values of urban observation stations. The geographical location of the relevant stations is shown in Figure 1. (2) To match the above area, the hourly conventional meteorological element data, including wind speed, relative humidity, wind direction, pressure, temperature etc., are provided by National Meteorological Information Center of the China Meteorological Administration (urban stations average). The time resolution of meteorological elements is 1 h. In order to more intuitively show the growth periods of PM2.5, taking Beijing as an example, Figure A1 in Appendix A shows the time series of PM2.5 and meteorological elements in January 2015 in Beijing. Furthermore, the average diurnal of PM2.5 and meteorological variables from 2013 to 2020 in autumn and winter in Beijing is shown in Figure A2. (3) The chemical composition data of submicron PM (PM1) were sampled from Institute of Atmospheric Physics (IAP, 39◦58 28 N, 116◦22 16 E), an urban site located between the north 3rd and 4th ring road in Beijing (Jiang et al., 2015). The sampling time was October 2012 to February 2013. The sampling instrument was the Aerosol Chemical Speciation Monitor (ACSM), with a time resolution of ~15 min. ACSM mainly detects particles below 1 μm, which can realize real-time online determination of OA, sulfate (SO4 <sup>2</sup>−), nitrate (NO3 <sup>−</sup>), ammonium salt (NH4 +), and chloride (Chl). The detailed description of the relevant instrument principles and parameters of the ACSMs have been to the references of Sun et al. [28] and Ng et al. [29]. At vaporizer temperature of ~600 ◦C, the ACSM cannot detect refractory materials, e.g., mineral dust and black carbon (BC). Thus, an aethalometer (Model AE22, Magee Scientific Corporation, Berkeley CA, USA) is used to simultaneously measure BC in PM2.5. The PM2.5 and gaseous species (including CO, SO2, NO, NOx, and Ox) were measured by a heated Tapered Element Oscillating Microbalance (1400a, Thermo Scientific, Waltham, MA, USA) and various gas analyzers (Thermo Scientific). (4) The data of aerosol number spectrum from 27 November 2014 to 28 February 2015, with a time resolution of 3 min and a measurement range of 14.6 nm to 661.2 nm, were measured by an scanning mobility particle sizer (SMPS), provided by the Beijing Meteorological Bureau, China. (5) The PM2.5 grid distribution data come from a 1-km-resolution PM2.5 dataset, called China High Air Pollutants (CHAP, https://weijing-rs.github.io/product.html (accessed on 18 September 2021)) from 2013 to 2020 across China, generated by the Moderate Resolution Imaging Spectroradiometer (MODIS, MODIS Collection 6 MAIAC AOD product (MCD19A2)) and multi-angle implementation of the atmospheric correction (MAIAC) algorithm (Wei et al., 2021). The inversion method is the space–time extra-trees (STET) model with high accuracies (i.e., cross-validation coefficient of determination, CV-R<sup>2</sup> = 0.86–0.90) and strong predictive powers (i.e., R2 = 0.80–0.82) [30].

The Roche method is used to calculate the height of the atmospheric mixing layer height (MLH). It is a method proposed by Nozaki et al. [31] in 1973 to estimate the height of the mixed layer using ground meteorological data. This method takes into account that the atmospheric mixing layer is the result of the combined action of thermal and dynamic turbulence. Moreover, the movement of the atmosphere in the upper boundary layer often interconnect and feedback with ground meteorological elements, so ground meteorological parameters can be used to estimate the height of the mixed layer. In addition, positive matrix factorization (PMF) [32] was preformed to resolve distinct OA factors from specific sources on ACSM mass spectra. The related principles and steps of this method are

detailed in Ulbrich et al. [33] and Decarlo et al. [34]. In this study, we limit PMF analysis to *m/z* 12–125 considering the low contribution of *m/z* 125–150 to the total signal. An Igor Pro-based PMF evaluation tool (PET, v 2.04) is used to further evaluated the results of PMF.

**Figure 1.** Average PM2.5 mass concentration distribution during the autumn and winter of 2013–2020 in the central and eastern of China. The circle is the approximate position of the region.
