Effect of Meteorological Characteristics on Particulate Matter and Ozone Concentrations in Coastal Cities

Abstract:

In this study， based on particulate matter and ozone data from air quality monitoring stations in Shanghai near-sea， offshore and central urban area in 2023， the Pearson’s correlation coefficient and XGBoost model are used to investigate the effect of meteorological characteristics and sea-land winds on PM_2.5， PM₁₀ and O₃. The results indicate， pollutants exhibit clear seasonal variation， with the highest concentration of PM_2.5 in winter， while O₃ concentration peaks occurr from May to August， PM₁₀ concentrations are the highest in April in coastal areas and in December in the central urban area. The occurrence of sea-land wind days at each site is central city （17 d） < offshore （47 d） < near-sea （73 d）， sea-land wind days in offshore and near-sea areas are the most in summer， and the central urban area shows no significant seasonal variation. When the sea-land wind day occurs， particulate matter concentrations in coastal areas decrease， and O₃ concentrations increase. The impact of sea-land winds on pollutants in the central urban area is relatively minor. Meteorological characteristics significantly influence pollutants in different seasons， the precipitation is significantly negatively correlated with PM_2.5 and PM₁₀ concentrations in spring and winter， the temperature and sunshine duration are positively correlated with O₃ concentration， and the humidity is significantly negatively correlated with O₃ concentration. The feature importance analysis of the XGBoost model indicates that the wind speed， near-surface air pressure， and temperature are the dominant meteorological characteristics for PM_2.5 concentrations， the near-surface pressure contributes the most to PM₁₀ concentration at each site， with the highest contribution of 74.7% to the central city， and the temperature has the most significant effect on O₃ concentration at all sites. The simulation results of the XGBoost model show that PM_2.5 and O₃ are better， and the overall prediction accuracy shows the central city < offshore < near-sea.