*2.3. ERA5 Reanalysis Data and Objective Synoptic Classification*

ERA5 dataset contains the fifth-generation atmospheric reanalysis data released by the European Center for Medium-Range Weather Forecasting (ECMWF). These data were obtained based on the 4D-Var data assimilation and model prediction of CY41R2 in the Integrated Forecasting System (IFS). Compared with its previous generation, namely ERAinterim, the horizontal and vertical resolutions of ERA5 were significantly improved, and the performance of ERA5 was better in the evaluation of tropospheric temperature, wind and humidity. Since the first release, ERA5 was widely used in research fields of atmospheric sciences, environmental issues [33–36], etc. ERA5 data at both upper-level isobaric surfaces and several single levels from December 1st 2015 to November 30th 2019 were collected. The data at 850 hPa, including potential height, relative humidity, temperature, vertical velocity and horizontal wind speed and direction, were used in this study. Besides, data at single levels, including surface pressure, 2 m temperature, 10 m

horizontal wind speed and direction at ground level and PBLH, were used too. The data in full horizontal resolution (0.25◦ × 0.25◦) and temporal resolution (1 h) were used.

The synoptic classification was conducted to explore the possible meteorological causes of the variation in PM2.5 spatial distribution. Synoptic classification could be carried out by subjective and objective methods, among which the subjective methods were mainly based on artificially defined a priori criteria and had great uncertainties [37]. On the contrary, objective methods were based on the maximization of similarity and variance; hence, these methods were appropriate for processing mass data without relying on a priori experiences [38]. The Cost733class package, a software jointly developed by Earth System Science and Environmental Management (ESSEM) and European Cooperation in Science and Technology (COST), focused on creating and evaluating weather and circulationtype classifications utilizing various different methods, including PCT (t-mode principal component analysis using oblique rotation), PTT (t-mode principal component analysis using orthogonal rotation), SOM (self-organizing maps) [39,40], etc. In this study, the PTT method was used to classify the synoptic patterns in SCB and its surrounding areas. The input data were the daily average geopotential heights at 850 hPa isobaric surface from ERA5 reanalysis data of ECMWF with a spatial resolution of 0.25◦ × 0.25◦. The region implementing the synoptic classification was the area in 95◦ E~120◦ E and 20◦ N~40◦ N (Figure 1).
