Direct Retrieval of NO ₂ Vertical Columns from UV-Vis (390-495 nm) Spectral Radiances Using a Neural Network

Abstract

Satellite retrievals of columnar nitrogen dioxide (NO2) are essential for the characterization of nitrogen oxides (NOx) processes and impacts. The requirements of modeled a priori profiles present an outstanding bottleneck in operational satellite NO2 retrievals. In this work, we instead use neural network (NN) models trained from over 360,000 radiative transfer (RT) simulations to translate satellite radiances across 390-495 nm to total NO2 vertical column (NO2C). Despite the wide variability of the many input parameters in the RT simulations, only a small number of key variables were found essential to the accurate prediction of NO2C, including observing angles, surface reflectivity and altitude, and several key principal component scores of the radiances. In addition to the NO2C, the NN training and cross-validation experiments show that the wider retrieval window allows some information about the vertical distribution to be retrieved (e.g., extending the rightmost wavelength from 465 to 495 nm decreases the root-mean-square-error by 0.75%) under high-NO2C conditions. Applying to four months of TROPOMI data, the trained NN model shows strong ability to reproduce the NO2C observed by the ground-based Pandonia Global Network. The coefficient of determination (R2, 0.75) and normalized mean bias (NMB, -33%) are competitive with the level 2 operational TROPOMI product (R2=0.77, NMB=−29%) over clear (geometric cloud fraction<0.2) and polluted (NO2C≥7.5×1015 molecules/cm2) regions. The NN retrieval approach is ~12 times faster than predictions using high spatial resolution (~3 km) a priori profiles from chemical transport modeling, which is especially attractive to the handling of large volume satellite data.

Use of machine learning to retrieve nitrogen dioxide with hyperspectral imagers in the ultraviolet and blue spectral range

“…For example, the range 400-500 nm slightly degrades the results, while further reducing the range to 400-470 nm leads to a bit more degradation. This is consistent with the full spectral resolution results of Li et al (2022) who found an optional retrieval window of 390-495 nm for estimating NO 2 vertical columns. Our results indicate that most of the information for NO 2 at OCI spectral resolution is provided by the high frequency structure of the radiances produced by NO 2 absorption within the fitting window currently used in the OMNO2 product.…”

“…Spectral fitting algorithms can be computationally intensive and it may still be desirable to use machine learning to speed up the processing of dense imager data as shown in Figure 8. For example, Li et al (2022) found that a NN implementation for NO 2 vertical columns using TROPOMI spectra was about 12 times faster than a full implementation using a priori profiles from a high spatial resolution chemistry-transport model.…”

“…Algorithm efficiency is particularly important for current and future atmospheric composition instruments, including those in geostationary orbit, that have very large data volumes. Machine learning has been shown to be an efficient means of estimating NO 2 vertical columns from satellite spectra (Li et al, 2022) as well as for other applications in remote sensing and geoscience (Maxwell et al, 2018;Lary et al, 2016).…”

Use of machine learning and principal component analysis to retrieve nitrogen dioxide (NO2) with hyperspectral imagers and reduce noise in spectral fitting

“…Algorithm efficiency is particularly important for current and future atmospheric composition instruments, including those in geostationary orbit, that have very large data volumes. Machine learning has been shown to be an efficient means of estimating NO 2 vertical columns from satellite spectra (Li et al, 2022) as well as for other applications in remote sensing and geoscience (Maxwell et al, 2018;Lary et al, 2016). In addition, machine learning combined with principal component analysis may be able to reduce noise in the spectral fitting compared with the more traditional approaches.…”

Background nitrogen dioxide (NO2) over the United States and its implications for satellite observations and trends: effects of nitrate photolysis, aircraft, and open fires

“…The importance of characterizing the free tropospheric NO2 background was brought to the fore by the use of OMI NO2 data to infer NOx emission trends in the contiguous United States (CONUS) (Jiang et al, 2018;Silvern et al, 2019;Qu et al, 2021;He et al, 2022;Jiang et al, 2022). The OMI NO2 data over CONUS show a steady decrease from 2005 to 2009 (Russell et al, 2012;Duncan et al, 2013;Duncan et al, 2016;Krotkov et al, 2016), consistent with the decreases of NOx emissions reported in the EPA National Emission Inventory (NEI), but Jiang et al (2018) found that the trend flattened after 2009 despite sustained decreases in NOx emissions according to the NEI and supported by NO2 surface data (Silvern et al, 2019).…”