Retrieval of NO₂ profiles from three years of Pandora MAX-DOAS measurements in Toronto, Canada

Abstract. The purpose of this work is to derive new NO₂ vertical profiling data products from Pandora spectrometers and investigate the factors contributing to the bias of this dataset relative to established ground-based and spaceborne datasets. Possible applications of the NO₂ vertical profile dataset include air quality monitoring and satellite validation studies. We explore the application of the optimal estimation method to Pandora multi-axis differential optical absorption spectroscopy (MAX-DOAS) measurements to retrieve vertical profile information for nitrogen dioxide (NO₂). We use the Heidelberg Profile (HeiPro) retrieval algorithm to derive, for the first time, NO₂ profiles and partial columns (0–4 km) from Pandora MAX-DOAS measurements from 2018–2020 at Downsview, a suburban neighbourhood in the north end of Toronto, Canada that is subject to local traffic emissions and urban influences. Validation of the new dataset was done via comparison with official Pandora direct-Sun measurements, in situ observations, satellite data, and an air quality forecasting model. We find that, for tropospheric partial column comparisons, the HeiPro dataset has a positive mean relative bias to Pandora direct-Sun (61 % ± 9.7 %) and TROPOMI (41 % ± 47 %) observations, as well as the GEM-MACH model output (61 % ± 7.5 %), with similar seasonal and diurnal cycles in the bias with Pandora direct-Sun and GEM-MACH. Contributing factors to the large bias of HeiPro to Pandora direct-Sun were investigated, and NO₂ heterogeneity, combined with differences between direct-Sun and multi-axis viewing geometries, was found to contribute a maximum of 52 % of the total relative bias during morning measurement times. For surface NO₂ comparisons, we find that HeiPro measurements capture the magnitude and diurnal variability of surface NO₂ reasonably well (mean relative bias to in situ surface NO₂: −8.9 % ± 7.6 %) but are low-biased compared to GEM-MACH (mean relative bias: −36 % ± 2.4 %). Compared to HeiPro, the GEM-MACH model profiles are high-biased in the lower boundary layer and low-biased in the free troposphere.