- 1Max Planck Institute for Chemistry, Satellite Group, Mainz, Germany (l.kuhn@mpic.de)
- 2Institute for Environmental Physics, University of Heidelberg, Germany
Nitrogen dioxide (NO2) plays a key role in the formation of urban smog and its adverse impact on human health. However, the routinely deployed measurements of NO2 are essentially limited to:
- tropospheric NO2 columns with global daily coverage, measured by the TROPOMI satellite instrument with a horizontal resolution of up to 3.5 x 5.5 km2
- near-surface NO2 concentrations, measured by in situ instruments at typically 0-8 m above ground with sparse spatial coverage
- NO2 profiles from MAX-DOAS measurements, with extremely sparse coverage and significant retrieval uncertainties
Regional chemistry and transport models (CTMs) facilitate the prediction of air pollutants with dense spatial coverage at urban-scale resolutions (e.g. 10 x 10 km2 or better), thereby yielding a valuable extension to the available observational data. Furthermore, significant progress has recently been made in developing neural network surrogate models with the ability to partially replace the computationally expensive CTM simulations. The NitroNet model, for example, can predict tropospheric NO2 profiles based on TROPOMI satellite observations and other ancillary variables, such as emission data and meteorological information.
A particular challenge is the validation of such models at altitudes within the boundary layer due to the lack of suitable observations, but data from measurement campaigns can partially fill these gaps. The CINDI-3 measurement campaign took place in May 2024 and comprised diverse spectroscopic measurements of NO2 concentrations in the lowest few hundred meters above ground. Moreover, CINDI-3 is the first CINDI campaign since the launch of the TROPOMI instrument, whose measurements are the main input to the NitroNet model.
We present an intercomparison of lower-tropospheric NO2 profiles from CINDI-3 long-path DOAS measurements, the CAMS CTM, and the NitroNet neural network. We provide a comprehensive overview of the near-surface NO2 profile shapes, as well as the level of agreement between measurements and simulation results obtained with different modelling approaches (here: a classic CTM simulation and a neural network based surrogate model).
How to cite: Kuhn, L., Beirle, S., Ziegler, S., Pozzer, A., and Wagner, T.: Intercomparison of lower-tropospheric NO2 profiles from the CINDI-3 measurement campaign, the CAMS regional model, and the NitroNet neural network, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2882, https://doi.org/10.5194/egusphere-egu25-2882, 2025.
