Satellite-based estimation of surface NO2 concentrations over east-central China: A comparison of POMINO and OMNO2d data

Abstract

The OMI NO2 standard product, OMNO2d, has been widely used in estimating surface NO2 concentrations. The Peking University Ozone Monitoring Instrument NO2 product (POMINO) is claimed to provide an improved quality over east-central China. This study estimated one year (Dec.2016–Nov.2017) of surface NO2 concentrations at satellite overpass time based on OMNO2d data and POMINO data, respectively. We used an extra-trees (ET) regression model to convey the non-linear relationship between surface NO2 and predictors, and compared the prediction accuracy with that of random forests (RF) regression model. The ET model showed a better estimation performance than the RF model, with the cross-validation R2 of 0.72 (RMSE = 9.20 μg/m3) and R2 of 0.70 (RMSE = 9.42 μg/m3) based on POMINO and OMNO2d data, respectively. The POMINO-derived monthly mean surface NO2 concentrations were closer to ground NO2 measurements than that OMNO2d-derived. Although the estimations from both satellite products were underestimated in polluted situations, the use of POMINO reduced the underestimation as compared to the use of OMNO2d data.