Abstract. Satellite observations of NO2 are crucial for tracking air pollution and its impacts on health and climate on the global scale. However, these measurements are affected by uncertainties arising from instrumental limitations, retrieval assumptions, and representation errors, making quantification of uncertainties critical for reliable data use. In this study, we assess key sources of uncertainty in tropospheric NO2 columns from the TROPOMI satellite instrument by studying the retrieval steps, and by comparing with Pandora and MAX-DOAS ground-based observations. For this assessment, we make use of high-resolution model simulations available for Europe and the Netherlands. Systematic errors in the stratosphere–troposphere partitioning of NO2 are identified, with TROPOMI overestimating stratospheric columns by up to 0.15 Pmolec cm−2 at high northern latitudes during winter, corresponding to tropospheric biases of up to 1.5 Pmolec cm−2, linked to limitations in the TM5-MP assimilation and magnified by large air-mass factor ratios in winter. In comparing satellite and ground-based observations, representation errors due to sub-pixel horizontal gradients are assessed using high-resolution LOTOS-EUROS simulations, resulting in uncertainties of approximately 6 % at polluted locations. Furthermore, major differences in vertical sensitivity between TROPOMI and MAX-DOAS lead to smoothing errors reaching up to 20 %. Comparisons of TROPOMI with Pandora direct sun measurements show a good seasonal agreement. The negative bias obtained when using the default TM5-MP a-priori profiles is partly mitigated with high-resolution CAMS-European a-priori profiles. A further reduction of this comparison bias is obtained when kilometer-scale simulations over the Netherlands are used, indicating the crucial role of the a-priori spatial resolution in the comparisons. Significant differences in absolute value and seasonality are observed between the MAX-DOAS FRM4DOAS, Pandora direct-sun, and Pandora sky-scan, indicative of the uncertainties in the ground-based remote sensing observations. Finally, uncertainties derived from the histogram of differences between TROPOMI and ground-based measurements generally still exceed expectations from the combination of all estimated uncertainty contributions, indicating that current estimates are likely still optimistic.
Cifuentes et al. (Tue,) studied this question.