Satellite-based estimates of daily NO2 exposure in urban agglomerations of China and application to spatio-temporal characteristics of hotspots

Abstract

The analysis of the daily spatial patterns of near-surface Nitrogen dioxide (NO2) concentrations can assist decision makers mitigate this common air pollutant in urban areas. However, comparative analysis of NO2 estimates in different urban agglomerations of China is limited. In this study, a new linear mixed effect model (LME) with multi-source spatiotemporal data is proposed to estimate daily NO2 concentrations at high accuracy based on the land-use regression (LUR) model and Ozone Monitoring Instrument (OMI) and TROPOspheric Monitoring Instrument (TROPOMI) products. In addition, three models for NO2 concentration estimation were evaluated and compared in four Chinese urban agglomerations from 2018 to 2020, including the COVID-19 closed management period. Each model included a unique combination of methods and satellite NO2 products: ModelⅠ: LUR model with OMI products; Model Ⅱ: LUR model with TropOMI products; Model Ⅱ: LME model with TropOMI products. The results show that the LME model outperformed the LUR model in all four urban agglomerations as the average RMSE decreased by 16.09% due to the consideration of atmospheric dispersion random effects, and using TropOMI instead of OMI products can improve the accuracy. Based on our NO2 estimations, pollution hotspots were identified, and pollution anomalies during the COVID-19 period were explored for two periods; the lockdown and revenge pollution periods. The largest NO2 pollution difference between the hotspot and non-hotspot areas occurred in the second period, especially in the heavy industrial urban agglomerations.