Classification, Estimation, and Prediction of Unfavourable Boundary-Layer Meteorological Conditions in Beijing for Pm2.5 Concentration Changes Using Vertical Meteorological Profiles

Abstract

PM2.5 pollution is closely associated with vertical boundary-layer (BL) meteorological conditions. However, the vertical profiles of major meteorological elements are rarely classified and used to quantify unfavourable meteorological conditions and surface PM2.5 concentration changes. This study uses time-series clustering to classify radiosonde observations in the winter 2012~2021 in Beijing. Temperature stratification classification was found to quantify the extent of unfavourable meteorological conditions well. The worst meteorological conditions often corresponded to an extremely stable boundary layer that became relatively warmer in the upper part and cooled in the near-ground region. Temperature inversions move downwards and force vertically-distributed PM2.5 to accumulate towards the ground. This upper BL warming in the Beijing area originated from southerly warm air flows, while the lower BL cooling partly resulted from the cooling effects of accumulated aerosols. From 2012~2016 to 2017~2021, the most significant difference appeared in the proportions of severely unfavourable vertical conditions, which dropped by 47% from 3.2% in 2017 to 1.7% in 2021. This proportional decline in severely unfavourable vertical conditions was estimated to contribute to ~23% of the PM2.5 concentration decline (16.8 μg m-3) from 2017 to 2021. The interaction between aerosol pollution and the vertical meteorological structure was an essential contributing factor to near-surface pollution. We found that the degree of modification of unfavourable meteorological conditions and the corresponding increase in the surface PM2.5 concentration could be estimated by the proportional change in the stable BL vertical structure. These changes could also be inferred from an autoregressive integrated moving average (ARIMA) model based on long-term balloon observations since 1991. The proportion of heavily unfavourable conditions were projected to increase from close to 0% in winter 2021 to 2.5% in winter 2022. Therefore, we suggest that the PM2.5 reduction target for the following winter should consider the effect of unfavourable meteorological conditions.

Keywords: time-series clustering, vertical meteorological conditions, PM2.5 pollution, unfavourable meteorological condition prediction