资源环境科技发展态势分析平台

Although the ground-level NO₂ measurement from air quality monitoring sites is relatively accurate, it is a challenge to obtain continuous spatial coverage due to the discrete distribution of sites. Thus, the tropospheric column NO₂ amount from satellites with wide spatial and temporal coverage and higher resolution has been increasingly used to estimate ground-level NO₂. However, most estimation methods were performed spontaneously using a simple linear model throughout the study period. These simplified models improve the efficiency of model development and enhance the generality of the model application, but they ignore the fact that contributors to changes of ground-level NO₂ are not always consistent with time. This study considered the fixed and random effects of influencing factors and developed a mixed effect model (MEM) to estimate the ground-level NO₂. By using the data of tropospheric NO₂ in China from January 1, 2014 to June 30, 2020 and other multivariate auxiliary data such as meteorological elements and terrain elevation, the reliability of daily ground-level NO₂ in typical populated areas of China estimated by the MEM was evaluated. The average of monthly R² of 10-fold CV in each study area during 2014–2020 is greater than 0.60 and the proportion of R² greater than 0.7 is about 71%, suggesting the reliability of MEM. It is found that the ground-level NO₂ distribution characteristics of each study area are more distinct, and the influential factors are also different. In addition, associated with the air quality control policies and emission reduction measures in various regions, the ground-level NO₂ in each study area has shown an overall downward trend during 2014–2019. The uncertainty of daily-scale meteorological elements and boundary layer conditions can lead to varying degrees of deviations in daily-scale predictions of ground-level NO₂. Validation with the station NO₂ observations demonstrates that the ground-level NO₂ prediction at seasonal time scale (R² = 0.81, RMSE = 3.86 μg/m³) performs better than those at time scales of daily and monthly (R² = 0.65, and 0.75, RMSE = 7.92, and 6.24 μg/m³). Therefore, the method of averaging can be used to improve the accuracy of ground-level NO₂ predictions on individual dates. In summary, this study shows that MEM is a promising ground-level NO₂ modeling method, and is effective for air pollution mapping in a large geographic region.