Validation Summary

The Sentinel-5P (S5P) L2_NO₂ products—covering tropospheric, stratospheric, and total columns from versions 02.04.00 (RPRO) and 02.04.00–02.09.01 (NRTI, OFFL)—show overall good agreement with independent reference measurements. Validation is based on ground-based networks, including MAX-DOAS (tropospheric), NDACC ZSL-DOAS/SAOZ (stratospheric), and the Pandora Global Network (total columns), as well as comparisons with OMI satellite data. Across products, a systematic negative bias is observed that generally increases with higher NO₂ column amounts in the tropospheric and total columns. Similar bias characteristics and uncertainty estimates are found for both NRTI and OFFL/RPRO datasets.

For tropospheric NO₂ columns, comparisons with MAX-DOAS data from 34 stations yield a median bias of −27.7% (−1.5 Pmolec/cm²), which meets the mission requirement of 50%. The bias is strongly dependent on pollution levels: it is positive (+12%) in clean conditions (< 2 Pmolec/cm²) and negative (−42%) in highly polluted environments (> 15 Pmolec/cm²). Applying TROPOMI averaging kernels to vertically smooth MAX-DOAS profiles reduces the absolute bias by 20%. The median dispersion is approximately 2.9 Pmolec/cm², exceeding the mission precision requirement (0.7 Pmolec/cm²), though it remains within acceptable limits at clean sites.

For stratospheric NO₂ columns, validation against ZSL-DOAS UV-visible measurements from 25 NDACC stations (globally distributed) accounts for horizontal smoothing effects and the NO₂ diurnal cycle. The data show a small negative offset of about −0.1 Pmolec/cm², corresponding to a median bias of −2.5%, well within the 10% mission requirement. The dispersion (0.3 Pmolec/cm²) also meets the requirement of 0.5 Pmolec/cm² when considering combined random errors and co-location mismatches. Comparisons with FTIR measurements at 26 NDACC stations indicate a positive median bias of +5.2% and a similar dispersion (0.3 Pmolec/cm²). Regional deviations are larger, with biases of 10–12% at Northern Hemisphere high latitudes and 13–16% in the tropics.

For total NO₂ columns, comparisons with Pandora observations from 81 stations show a median bias of −10.8% (−0.7 Pmolec/cm²) and a dispersion of 2.0 Pmolec/cm². As with tropospheric columns, the bias depends on pollution levels: it is slightly positive (+3.3%) at low column amounts (< 6 Pmolec/cm²), typically at clean or high-altitude sites, and becomes negative (−16.6%) in more polluted conditions.

Overall, the S5P NO₂ products meet mission accuracy requirements for most applications, with performance varying systematically as a function of NO₂ abundance and environmental conditions.

The figure below visualizes, as a box-whisher graph, the bias and dispersion between TROPOMI and MAX-DOAS tropospheric NO2 VCDs per ground-based station, ordered by mean MAX-DOAS tropospheric NO2 VCD (clean stations at the bottom, polluted stations at the top). The pink shaded area represents the mission requirement of at most 50% bias. The light and dark green shaded areas represent the 1- and 2-sigma uncertainty on the differences (combined uncertainty of satellite and ground-based measurements). 

A detailed description of the method and a comprehensive discussion of validation results can be found in the Sentinel-5p Quarterly Validation Report #29: April 2018 - November 2025